716
OPTICS LETTERS / Vol. 26, No. 10 / May 15, 2001
Ultraviolet transparent silicon oxynitride waveguides for biochemical microsystems Klaus B. Mogensen and Peter Friis Mikroelektronik Centret (MIC), Building 345 East, Technical University of Denmark, DK-2800 Lyngby, Denmark
Jörg Hübner Research Center for Communications, Optics and Materials (COM), Building 345 West, Technical University of Denmark, DK-2800 Lyngby, Denmark
Nickolaj Petersen, Anders M. Jørgensen, Pieter Telleman, and Jörg P. Kutter Mikroelektronik Centret (MIC), Building 345 East, Technical University of Denmark, DK-2800 Lyngby, Denmark Received November 28, 2000 The UV wavelength region is of great interest in absorption spectroscopy, which is employed for chemical analysis, since many organic compounds absorb in only this region. Germanium-doped silica, which is often preferred as the waveguide core material in optical devices for telecommunication, cannot accommodate guidance below 400 nm, owing to the presence of UV-absorbing centers. We show that silicon oxynitride 共SiOx Ny 兲 waveguides exhibit very good UV performance. The propagation loss for 24-mm-wide SiOx Ny waveguides was found to be ⬃1.0 dB兾cm in the wavelength range 220 – 550 nm. The applicability of these waveguides was demonstrated in a biochemical microsystem consisting of multimode buried-channel SiOx Ny waveguides that were monolithically integrated with microf luidic channels. Absorption measurements of a b-blocking agent, propranolol, at 212 – 215 nm were performed. The detection limit was reached at a concentration of 13 mM, with an optical path length of 500 mm (signal兾noise ratio, 2). © 2001 Optical Society of America OCIS codes: 170.3890, 230.7380, 300.6540, 310.6860, 300.1030.
Interest in miniaturization of optochemical sensors has steadily increased because of the possibility of fabricating entirely integrated systems, so-called micrototal analysis systems, on a single substrate.1 These devices are expected to become useful instruments for obtaining and assessing analytical data in many different situations and environments, because of the advantages of low reagent consumption, inherent stability, and portability. Furthermore, implementation of automated and mass-fabrication methods will lower the price of such devices signif icantly compared with those of existing systems. In an effort to achieve this goal, researchers have adapted well-established microfabrication techniques such as deposition of silica glass and reactive-ion etching from the microelectronics industry.2 – 4 However, extending these technologies to integration of waveguides in biochemical microsystems imposes new requirements upon the dopants that are used. In telecommunications, germanium-doped silica is often favored as the core material because of its superior transmission properties at ⬃1550 nm. In chemical-absorption spectroscopy, however, the UV region is of importance, because most organic compounds absorb in only this region.5 This condition limits the applicability of germanium as a core dopant, because highly UV-absorbing centers are present in germanium-doped silica.6 Nitrogen is a promising alternative as a core dopant for operation in the UV range, because the position of the optical bandgap of silicon oxynitride 共SiOx Ny 兲 is de0146-9592/01/100716-03$15.00/0
pendent on the relative nitrogen and oxygen content in the silica network and is hence located between the silicon oxide limit at ⬃160 nm 共⬃8.0 eV兲 and the silicon nitride limit at ⬃250 nm 共⬃5.0 eV兲.7 The ultimate limit in chemical-absorption spectroscopy in which aqueous solutions are employed is 190 nm, owing to the onset of absorption by H2 O molecules. This limit obviously cannot be reached with silicon nitride. SiOx Ny , however, is promising because it is necessary to incorporate only a slight amount of nitrogen into the silicon oxide network to obtain a higher index of refraction and therefore guidance of light in the waveguide core. The small content of nitrogen will, in the best case, result in only a minor shift of the optical bandgap from the silicon dioxide limit at ⬃160 nm toward the silicon nitride limit at ⬃250 nm.7 The low absorption will limit UV-induced color-center formation and thereby potentially increase the lifetime of the device.6 To the best of our knowledge, the UV transmission properties of SiOx Ny -based buried-channel waveguides have not been presented for biochemical microsystems before. The transmission properties of 24-mm-wide multimode buried-channel waveguides fabricated by two different sets of dopants were compared. The fabrication schemes are outlined in Table 1. The applied plasma-enhanced chemical-vapor deposition reactor and the dependence of the mechanical and thermal properties of the waveguides on deposition conditions are described in Ref. 8. The buffer layer consisted of 10 mm of thermally grown oxide. The core and © 2001 Optical Society of America
May 15, 2001 / Vol. 26, No. 10 / OPTICS LETTERS Table 1.
Component Buffer Core Cladding a
Glass Composition of the Tested Waveguides a Type A (Nitrogen Doped)
Type B (Germanium Doped)
10 mm SiO2 共n 苷 1.458兲 4.0 mm SiOx Ny 共n 苷 1.483兲 7.0 mm SiO2 共n 苷 1.460兲
10 mm SiO2 共n 苷 1.458兲 4.0 mm SiGex1 Oy1 共n 苷 1.474兲 5.5 mm Bx2 Py2 SiOz2 共n 苷 1.459兲
The waveguide width was 24 mm in all cases.
cladding layers were deposited by plasma-enhanced chemical-vapor deposition by reacting silane, nitrogen, nitrous oxide, germane, ammonia, diborane, and phosphine. The cores were etched by reactive-ion etching with f luorine chemistry. Annealing of the core and cladding layers was performed at 1000 ±C for 4 h. The refractive indices were measured at 633 nm. The type A waveguides consisted of a 4.0-mm-thick SiOx Ny core 共n 苷 1.483兲 and a 7.0-mm-thick undoped silica cladding 共n 苷 1.460兲. We investigated undoped silica as the cladding material to avoid losses as a result of absorption in the cladding. The type B waveguides had a 4.0-mm-thick germanium-doped core 共n 苷 1.474兲 and a 5.5-mm-thick index-matched boron–phosphorous-doped silica cladding. Figure 1 shows the spectrally resolved propagation losses of the two types of waveguide. The losses were calculated by the cutback method,9 by measurement of the spectrally resolved transmitted optical power for five different waveguide lengths (ranging from 7 to 1 cm). The dependence of the propagation loss on the wavelength for the germanium-doped waveguide (type B) is as expected. At 800 nm the propagation loss is reasonably low, owing to the absence of absorption states associated with germanium in the glass network and to the superior step coverage of boron – phosphorous-doped silica compared with undoped silica. The major loss contribution in this region is thus believed to be attributable to Rayleigh scattering at the core– cladding interface.10 The loss increases steadily for shorter wavelengths, and essentially no light is transmitted below 330 nm. This result is in accordance with theory, because the germanium oxygen-deficient center absorbs at 330 nm.6 The SiOx Ny waveguide (type A) is superior for wavelengths below 550 nm. Below 220 nm the tail of the optical bandgap of the nitrogen-doped silica leads to strong absorption and hence a high propagation loss. However, the drawback of using an undoped silica cladding can also be clearly seen. The scattering loss is significantly higher because of the poorer step coverage of undoped silica. The absorption peaks observed above 550 nm for the SiOx Ny waveguide are attributed to vibrational overtones of N–H and Si– H bridging in the silica network and thus indicate insuff icient annealing of the core.11 A maximum annealing temperature of 1000 ±C was chosen because higher temperatures result in enhanced introduction of recombination centers in the silicon crystal, thereby hampering future integration
717
of photodiodes because the diffusion length of the minority carriers will decrease.12 The SiOx Ny waveguides were tested in a microsystem consisting of monolithically integrated f luidic channels and buried-channel waveguides (Fig. 2). The device had a size of 1 cm 3 2 cm. The waveguides had the same composition and dimensions as type A in Table 1. After fabrication of the waveguides the microchannels were etched into the silica glass by reactive-ion etching. The microchannels were hermetically sealed by anodic bonding of a Borof loat glass lid on top of the structure. The fabrication procedure is described in more detail in Ref. 4. The waveguides were arranged perpendicularly to a 500-mm-wide microchannel. The depth of the microchannel was 20 mm. A combined deuterium – halogen lamp was used as a light source (Instrument Systems, Munich, Germany), and an optical spectrum analyzer equipped with a Peltier-cooled photomultiplier tube 共25 ±C兲 from Instrument Systems served as the detection instrument. UV-enhanced fibers with a 50-mm pure silica core were employed for light transmission. The losses in such systems are severe. At 212 nm the total insertion loss was 21 dB; 1.5 dB originated from the propagation loss in the waveguide (1 cm) and was hence only a minor fraction of the total loss.
Fig. 1. Spectrally resolved propagation loss in the wavelength range 200– 800 nm. The propagation loss was calculated by the cutback method. The error bars on the linear f its are included every 100 nm. The type A waveguide had a nitrogen-doped core, and type B was germanium doped (see Table 1).
Fig. 2. Schematic of the simple single-channel setup used for the chemical-absorption measurements.
718
OPTICS LETTERS / Vol. 26, No. 10 / May 15, 2001
Fig. 3. Absorbance versus concentration of propranolol averaged from 212 to 215 nm with a scan time of 400 ms. The lowest detected concentration was 13 mM (signal兾noise ratio, 2).
Approximately 8 dB was insertion loss that was due to geometrical mismatch between the circular fiber and the rectangular waveguide, and the loss across the 500-mm-wide f luidic channel was 11 dB. The applicability of the microsystem for biochemical purposes was investigated by absorption measurements of a b-blocking agent, propranolol. A strong absorption band at 212 –215 nm was chosen for the measurement. The absorbance 共A兲 values were calculated from the Lambert–Beer law: A共c兲 苷 log关P0 兾P 共c兲兴 苷 ebc, where e is the molar absorptivity, b is the optical path length of 500 mm, and c is the concentration. P0 is the reference power measured with water in the channel, and P is the power with the channel that contains the solute (propranolol) dissolved in water. The scan time was 400 ms. Linearity between the absorbance and the concentration can clearly be seen from Fig. 3. The offset of the calibration from the origin is believed to be due to drift in the baseline during the measurements. The molar absorptivity from 212 to 215 nm was measured to be e 苷 6 3 104 M21 cm21 , and the lowest detected propranolol concentration was 13 mM (signal兾noise ratio, 2). Similar measurements could not be done in a microsystem containing germanium-doped planar waveguides. In such a system the extremely high absorption in the waveguides would cause the transmitted power in the 212 –215-nm region to be far below the noise level of the detector. The coupling loss across the f luidic channel can be reduced by a decrease of the refractive-index step between the core and the buffer–cladding, because the numerical aperture and hence the divergence of light over the microchannel will decrease. Lowering of the nitro-
gen content of the core might have the additional benefit of moving the optical bandgap toward shorter wavelengths, because the position of the bandgap is dependent on the relative oxygen and nitrogen content of the doped silica.7 In conclusion, silicon oxynitride 共SiOx Ny 兲 buriedchannel waveguides have been shown to display significantly better UV transparency than germanium-doped silica. This is due to the presence of highly UV-absorbing centers in the latter material. The SiOx Ny waveguides were further tested in a microsystem consisting of monolithically integrated waveguides and f luidic channels by measurements of absorption of the drug propranolol at 212 – 215 nm. The lowest detected concentration was 13 mM, with an absorption length of 500 mm (signal兾noise ratio, 2). It is expected that these investigations will open up a new area of applications for SiOx Ny waveguides in microsystems for biochemical analysis. This research was supported by the Danish Technical Research Council, Statens Teknisk-Videnskabelige Forskningsråd (contract 9900683). K. B. Mogensen’s e-mail address is
[email protected]. References 1. S. C. Jakeway, A. J. de Mello, and E. L. Russel, Fresenius J. Anal. Chem. 366, 525 (2000). 2. O. Leistiko and P. F. Jensen, J. Micromech. Microeng. 8, 148 (1998). 3. J. M. Ruano, V. Benoit, J. S. Aitchison, and J. M. Cooper, Anal. Chem. 72, 1093 (2000). 4. J. Hübner, K. B. Mogensen, A. M. Jørgensen, P. Friis, P. Telleman, and J. P. Kutter, Rev. Sci. Instrum. 72, 229 (2001). 5. D. A. Skoog, D. M. West, and F. J. Holler, Fundamental of Analytical Chemistry, 7th ed. (Saunders, Philadelphia, Pa., 1997). 6. A. V. Amossov and A. O. Rybaltovsky, J. Non-Cryst. Solids 179, 75 (1994). 7. C. Ance, F. de Chelle, J. P. Ferraton, and G. Leveque, Appl. Phys. Lett. 60, 1399 (1992). 8. K. E. Mattsson, J. Appl. Phys. 77, 6616 (1995). 9. R. G. Hunsperger, Integrated Optics: Theory and Technology, Vol. 33 of Springer Series in Optical Sciences (Springer, New York, 1991). 10. F. Landouceur and J. D. Love, Silica-Based Buried Channel Waveguides and Devices (Chapman & Hall, New York, 1996). 11. F. Bruno, M. del Guidice, R. Recca, and F. Testa, Appl. Opt. 30, 4560 (1991). 12. T. Warabisako, T. Uematsu, S. Muramatsu, K. Tsutsui, H. Ohtsuka, Y. Nagata, and M. Sakamoto, Sol. Energy Mater. Sol. Cells 48, 137 (1997).
