Micromachined Fourier transform spectrometer on silicon ... - CiteSeerX

0 downloads 0 Views 889KB Size Report
Fourier transform spectrometer has a measured spectral resolution of approximately 45 nm near ... machining process and integrated on a silicon optical bench.
Sensors and Actuators A 130–131 (2006) 523–530

Micromachined Fourier transform spectrometer on silicon optical bench platform Kyoungsik Yu a,∗ , Daesung Lee b , Uma Krishnamoorthy c , Namkyoo Park d , Olav Solgaard b,1 a

Korea Electrical Engineering & Science Research Institute, Seoul National University, San 56-1, Sillim-dong, Gwanak-gu, Seoul, 151-742, Republic of Korea b Department of Electrical Engineering, Stanford University, E.L. Ginzton Laboratory, 450 Via Palou, Stanford, CA 94305-4085, USA c Sandia National Laboratory, Albuquerque, NM 87185-1080, USA d School of Electrical Engineering and Computer Science, Seoul National University, San 56-1, Sillim-dong, Gwanak-gu, Seoul, 151-742, Republic of Korea Received 9 June 2005; received in revised form 3 December 2005; accepted 5 December 2005 Available online 20 January 2006

Abstract We present a miniaturized Fourier transform spectrometer implemented on a silicon optical bench platform. Both optical and opto-mechanical components of a Michelson interferometer, including a silicon beam splitter, micromirrors, MEMS actuators, and fiber U-grooves, are simultaneously fabricated by micromachining of the device layer of a silicon-on-insulator wafer. Our specialized bulk micromachining process combines the flexible definition capability of deep reactive ion etching with the good surface quality provided by anisotropic KOH wet etching. This integrated Fourier transform spectrometer has a measured spectral resolution of approximately 45 nm near 1500 nm wavelength. © 2005 Elsevier B.V. All rights reserved. Keywords: Fourier transform spectroscopy; DRIE; KOH; Lateral combdrive actuator; Optical MEMS

1. Introduction Near-infrared spectroscopy is becoming increasingly important in a number of applications such as environmental monitoring, chemical analysis, and biomedical diagnostics. Compared to conventional table-top spectrometers, miniaturized spectrometers offer great advantages in cost and portability [1,2], and have the potential to enable new applications in distributed spectral sensing and monitoring. Advances in microelectromechanical systems (MEMS) technology have facilitated the development of several types of microspectrometers, including micromachined gratings [3,4], tunable Fabry–Perot filters [5], and miniaturized Fourier transform spectrometers [6–10]. Among them, Fourier transform spectroscopy (FTS) is favored in a number of applications because of its high optical efficiency, accuracy, and simplicity [11].



Corresponding author. Tel.: +82 2 885 9443; fax: +82 2 883 0827. E-mail addresses: ks [email protected] (K. Yu), [email protected] (O. Solgaard). 1 Tel.: +1 650 724 2765; fax: +1 650 725 2533. 0924-4247/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.sna.2005.12.022

In this work, the FTS hardware, including the beam splitter, reference mirror, MEMS-actuated micromirror, and fiber U-grooves are simultaneously fabricated by a simple bulk micromachining process and integrated on a silicon optical bench occupying less than 4 mm × 8 mm × 0.6 mm. When integrated with microfluidic channels or miniaturized gas cells, the proposed device can be used for lab-on-a-chip applications that require real-time spectral monitoring and analysis. 2. Device design Fourier transform spectrometers use Michelson interferometers to obtain interferograms corresponding to a range of optical path length differences (OPLD). The spectral information is generated by performing a Fourier transform of the measured interferograms. In our micromachined Michelson interferometer, it is assumed that a fiber-coupled light source is used as the input port to the interferometer, and that the sample is located in the path of the combined beams for power FTS, or in the optical path of one arm for amplitude FTS [11]. The schematic diagram of the device is shown in Fig. 1. The optical beam from the input fiber is partially transmitted and partially reflected by the silicon

524

K. Yu et al. / Sensors and Actuators A 130–131 (2006) 523–530

Fig. 1. Schematic diagram of the integrated FTS. The operation of the FTS is determined by the sample locations. (a) Power FTS and (b) amplitude FTS.

free-space beam splitter. The beam splitting ratio depends on many factors such as the input wavelength, the input polarization, the surface roughness, and the thickness of the beam splitter. After the beam splitter, the two divided beams are reflected by two micromirrors, and subsequently recombined by the beam splitter to generate an interference fringe at the output arm. Using an electrostatic MEMS actuator, we change the position of one micromirror to modulate the OPLD between the two optical paths. The intensity of the interference signal varies according to the OPLD, and this signal is captured by an output multimode fiber connected to an optical detector. 3. Fabrication As shown in Fig. 2, our fabrication process combines the flexible definition capability of deep reactive ion etching (DRIE) with the good surface quality provided by wet anisotropic etching in aqueous KOH [12,13]. The process starts with the preparation of silicon-on-insulator (SOI) wafers with (1 1 0) device layers and (1 0 0) substrates (Fig. 2a). The thicknesses of the device silicon and buried oxide layer are 100 and 2.3 ␮m, respectively. The 500 ␮m thick (1 0 0) substrate layer is used to support the finished device structures. In the current fabrication process, the crystallographic orientation of the substrate silicon layer is not important because it is not etched by DRIE or KOH. However, although not demonstrated in this work, it is possible to further etch the substrate layer to define other useful structures such as alignment structures for ball lenses. More details about the fabrication process can be found in [12,14]. Optical components, such as beam splitters and micromirrors, as well as the free-space optical paths within the interferometer, are defined by etching the device layer. U-grooves for passive alignment of optical fibers are also defined. The sidewalls of the

Fig. 2. Process flow for the micromachined Fourier transform spectrometer. (a) Two-mask nitride film patterning for KOH etching and DRIE. (b) Wet anisotropic KOH etching of the (1 1 0) device layer to create smooth sidewalls. (c) Oxidation for sidewall protection followed by partial nitride removal. (d) DRIE. (e) Nitride film removal, sacrificial oxide release, and gold-deposition for wire-bonding. (f) Removal of protection structures and assembly of optical fibers. (g) Top view layout indicating the DRIE and KOH etching areas and the position of the crosssection for figures (a–f).

device layer are used as beam splitting and reflecting surfaces, so the surface roughness of the sidewalls is very important for good optical characteristics. In our bulk micromachining process, the sidewalls that require good surface quality for optical purposes are defined by KOH etching (Fig. 2b), whereas other features are etched by DRIE for flexible structure definition (Fig. 2d). Due to the high aspect ratio structures made possible by wet KOH etching, silicon beam splitters and micromirrors with good verticality and smooth optical surfaces can be fabricated. As demonstrated in [12], typical RMS surface roughness of the KOH etched region is less than 20 nm. The crystallographic (1 1 0) silicon device layer does, however, only allow two possible directions for smooth vertical planes and the angle between them is 70.53◦ . This is the reason that the incident angle to the beam splitter is

K. Yu et al. / Sensors and Actuators A 130–131 (2006) 523–530

525

Fig. 3. (a) A scanning electron microscope (SEM) image of the micromachined Fourier transform spectrometer. (b) Microscope image and SEMs of the device showing details of the fabricated beam splitter with protection structure, fixed and movable micromirrors, and fiber U-grooves. Sidewalls obtained by KOH etching (the two surfaces of the beam splitter and the movable micromirror) show much better surface roughness than surfaces defined by DRIE.

526

K. Yu et al. / Sensors and Actuators A 130–131 (2006) 523–530

not the typical 45◦ (incident angle θ i = 70.53◦ in Fig. 1), and that the reference mirror plane is defined by DRIE etching with worse surface roughness (Fig. 3). For future generations of devices, a turning mirror will be used to solve this problem. After the etching process for the definition of device structures, the sacrificial oxide layer is removed in 49% hydrofluoric acid (between Fig. 2d and e). Etch holes are used to release the movable structures while the buried oxide under the anchors without etch holes are not removed. For wire-bonding and enhancement of mirror reflectivity, a thin Au layer (150 nm) is subsequently deposited (Fig. 2e). During gold-coating, the beam splitter surfaces are shielded by protection structures as shown in Fig. 3a and b, to avoid metal coating of the beam splitter surfaces. The distance between the protection structure and the beam splitter surface is ∼6 ␮m. After the metal coating step, the protection structures are removed manually, and the cleaved input and output fibers are assembled within the fiber U-grooves (Fig. 2f and Fig. 3b). The two-mask layout and the position of the cross-section for Fig. 2a–f are shown in Fig. 2g. The device thickness is comparable to the diameters of standard optical fibers, and this opens up possibilities such as active fiber positioning and manipulation. Important examples of this are the U-grooves and fiber integration of our silicon optical bench platform that create the interface between the free-space optical bench module and input light sources such as fiber lasers and fiber pig-tailed semiconductor lasers and light emitting diodes. It is also possible to integrate microfluidic channels on chip at the sample position [15]. The optical fibers used in the experiment are typical communication-grade multimode fibers, and their core and outer diameter are 62.5 and 125 ␮m, respectively. The measured divergence angle from the multimode fiber is ∼10◦ , which results in large insertion loss from the input to the output fiber. In the current device layout, the total optical path length from the input to the output fiber is approximately 1.8 mm. In future devices, fiber collimators or lensed fibers will be used to reduce the optical insertion loss due to beam divergence.

4. Experiments and results Our FTS is designed for operation in the near-infrared, which makes the use of a silicon beam splitter convenient. In our experiment, we used beam splitters with thicknesses ranging from 5 to 25 ␮m. The length of the beam splitters is ∼700 ␮m to cover the whole optical beam from the multimode fiber at the incident angle of 70.53◦ . To measure the spectral variation of the beam splitter, the output wavelength of a tunable laser is swept from 1515 to 1590 nm. A polarization controller is used to maintain the input polarization in the TE polarization. Since the Brewster angle of air–silicon interface at near-infrared wavelengths (∼74◦ ) is close to the incident angle (θ i = 70.53◦ ) defined by the crystallographic orientation, the reflection of TM-polarized light is very small. The calculated power reflectance, R, and transmittance, T, of the TE-polarized beam at the silicon–air inter◦ face  with the incident angle of θ i = 70.53 are R = (cos θi −

n2Si − (1/ sin2 θi ))/(cos θi + n2Si − (1/ sin2 θi )) = 0.67 and T = 1 − R = 0.33, respectively, assuming the refractive index of silicon material is nSi = 3.52. Fig. 4 shows the fiber to fiber loss as a function of input wavelength when a 15 ␮m-thick silicon beam splitter and multimode fibers are used for the input and the transmitted and reflected outputs. The total optical path length, i.e. the distance from the input fiber via the beam splitter to one of the mirrors and back to the output fiber is 1.8 mm in our MEMS FTS implementation. The transmission and reflection profile in Fig. 4b include ∼3.3 dB free-space propagation loss due to the beam divergence (see Fig. 4a for the experimental setup). This contribution to the loss can be significantly reduced when fiber collimators are used at the fiber ends. As shown in Fig. 4b, the beam splitting ratio is not constant over the spectral range of interest because of multiple reflections inside the beam splitter. The spectral period is inversely proportional to the thickness of the beam splitter, so thinner beam splitters exhibit slower spectral variation of the beam splitting ratio. The total insertion loss

Fig. 4. (a) Experimental setup for the characterization of a silicon beam splitter. Output fibers 1 and 2 are for reflection and transmission measurement, respectively. (b) Fiber-to-fiber loss as a function of input wavelength when the silicon beam splitter reflects and transmits the incident optical beam.

K. Yu et al. / Sensors and Actuators A 130–131 (2006) 523–530

from the input fiber to the output fiber in the FTS configuration of Fig. 1 is the sum of the reflection and transmission loss shown in Fig. 4, and is approximately 13 ± 1 dB for the combined light beam at the output fiber. The spectral variation of our micromachined silicon beam splitters can be compensated to yield accurate spectral measurement. In practice, most spectrometers obtain spectral information by comparing measurement and reference scans with and without the actual sample. In this mode of operation, the spectral variations of the beam splitter throughput are automatically compensated, at the cost of somewhat reduced signal-to-noise in the spectral regions where the beam splitter throughput is low. In general, the throughput of a beam splitter in the FTS is defined as 4RT [11], and the optimal condition occurs when the optical reflectance, R, and transmittance, T, are equal (R = T = 0.5). Although the integrated silicon beam splitters do not have this optimal beam splitting ratio, they provide reasonable throughput and signal intensity that can be easily detected by photo detectors. In practical situations where the input polarization is random and unknown, a polarization scrambler can be employed to ensure that at least half of the optical power is TE-polarized regardless of the input polarization state. Due to the large difference between the reflectance and transmittance, the beam splitting throughput, or the intensity of the interference signal at the detector, of TM-polarized light is negligible compared to the TE polarization. For optical path length modulation, a micromirror with optically flat sidewall is actuated by an electrostatic lateral combdrive actuator [16]. The width and length of the comb fingers are 11 and 130 ␮m, respectively, and the gap distance between comb fingers is 8 ␮m. The total footprint of the combdriveactuated micromirror is approximately 4 mm × 4 mm. The rest of chip area is occupied by the beam splitter and fiber U-grooves. The combdrive forces are inversely proportional to the gap distance between the comb fingers, so DRIE with higher aspect ratio can reduce the actuator size by reducing the gap distance. Micromirror deflection as a function of input voltage is shown in Fig. 5, showing the expected square dependence on voltage. The resonant frequency of the movable micromirror structure is ∼320 Hz. When a sinusoidal electric signal is applied to the combdrive actuator, the velocity of the micromirror varies nonlinearly with time, and therefore uniform sampling of the output interferogram signal in the time domain results in nonuniform sampling in the OPLD domain. Unlike conventional schemes, in which the OPLD is made to vary linearly as a function of time, we allow the micromirror to move with the sinusoidal input signal and correct the resulting distortion with signal processing [7]. If we use the interferogram data from the constant velocity region of the micromirror motion only, the spectral resolution is compromised due to the reduction in the maximum OPLD. Instead, the interferogram data are first sampled at a higher-than-required sampling rate, and then the over-sampled data are re-sampled to obtain a sequence with constant OPLD spacing. After this process, we use a standard Fast Fourier Transform (FFT) routine to obtain the spectral information from the re-sampled interferogram data.

527

Fig. 5. Micromirror deflection as a function of input voltage to the combdrive actuator.

To verify this procedure, we recorded interferogram data of an external-cavity tunable laser in the wavelength range around 1550 nm. The micromirror was driven slowly at the rate of 5 Hz for minimum delay between the applied electric signal and the actual device motion. As the driving frequency approaches the resonance frequency of a mechanical system, the phase lag between the input signal and the mechanical response becomes larger. When the input frequency is 5 Hz, the micromirror device operates with negligible phase shift between the driving signal and the micromirror movement. If the micromirror is fully characterized so that the effect of the phase lag can be taken into account, it is possible to drive the micromirror at higher speeds. Fig. 6 shows measured interferograms with the laser tuned to three different wavelengths (1500, 1520, and 1580 nm). In

Fig. 6. Normalized single-sided interferogram for three different input wavelengths (1500, 1520, and 1580 nm). Only one-half of the scanning period is shown.

528

K. Yu et al. / Sensors and Actuators A 130–131 (2006) 523–530

these measurements the micromirror moves by ∼25 ␮m, which corresponds to an OPLD variation of ∼50 ␮m. A 50 kS/s analogto-digital converter is used to produce a 10,000-point output waveform per 200 ms period, and the first half period (when the micromirror moves outward in Fig. 1) is used to generate the 5000-point single-sided interferogram sequence shown in Fig. 6. The other half of the output waveform is the mirror image of Fig. 6 (not shown). If the output intensity of the interferogram signal in the time domain is I(t), its sampled version after analogto-digital conversion can be written as I[n] = I(Ts n), where Ts is the sampling interval. For a 50 kHz sampling rate, the sampling interval Ts is equivalent to 20 ␮s. The sample index, n, for the interferogram data is indicated on the upper x-axis of Fig. 6. The measured signals were very repeatable, and the major source of noise in the interferogram data is not the micro mechanical structure, but the photodetector. Assuming that the electrostatic force from the combdrive actuator is proportional to the square of the applied voltage (Fig. 7), we calculate the micromirror position at each sampling point. Using this information, a new sequence with a constant OPLD sampling interval is constructed by using samples only at correct OPLD locations. An example of the sample selection procedure is illustrated in Fig. 7. In this example, we divide the micromirror motion into 40 equally spaced points in the OPLD domain (N = 40), and find the corresponding sampling indices, n, in the time domain. If d(t) is the amount of micromirror motion in the time domain, we assume that d(t) is a monotonically increasing function in 0 ≤ t ≤ T/2, and d(t = 0) = 0 and d(t = T/2) = dmax , where T is the period of the driving signal. To obtain the correct time-domain re-sampling indices, nk , for equally spaced OPLD sampling, we find the solutions of nk that satisfy d(Ts nk ) = dmax × (k/N), 1 ≤ k ≤ N. We indicate all solutions of nk (k = 1, 2, . . ., 40) in Fig. 7 to illustrate this process. The re-sampled sequence I[nk ] is now equally spaced in the OPLD domain with nonuniform sampling time intervals.

Fig. 7. Input voltage as well as the square of the input voltage, to the MEMS actuator. An example of the sample selection procedure for 40 equally spaced intervals is illustrated.

Fig. 8. Re-sampled interferograms compensated for nonlinear micromirror motion.

In our experiment, the total number of samples in the resampled interferograms (Fig. 8) is reduced from 5000 to 500 (N = 500). Although the OPLD sampling interval of the new resampled sequences is increased to ∼100 nm (i.e. ∼50 ␮m/500), it is sufficiently small to reliably sample the infrared wavelengths where the silicon beam splitter is partly transparent. According to the Nyquist theorem, the sampling interval should be smaller than half of the smallest wavelength in the spectrum to avoid distortion due to the aliasing effect [17]. As expected, the re-sampled interferograms in Fig. 8 show sinusoidal variations with the same OPLD period across the whole interferogram. From this observation, we conclude that the actual micromirror motion agrees very well with the simple electromechanical model of an electrostatic combdrive actuator. After the re-sampling process, a numerical apodization operation with a Gaussian envelope and zero padding is performed, followed by a Fourier transform. The final optical spectra for three different wavelengths are shown in Fig. 9. The minimum wavelength resolution with a rectangular envelope is approximately given by λ2 /L, where λ and L are the input wavelength and the maximum OPLD, respectively. In our experiment with a maximum OPLD of ∼50 ␮m, this theoretical minimum wavelength resolution varies from 45 for 1500 nm input wavelength to 50 nm for 1580 nm input wavelength. Although the rectangular envelope without numerical apodization produces the smallest spectral resolution, the sidelobe amplitude is relatively high and additional apodization process is usually required to obtain cleaner spectra [17]. For the spectra shown in Fig. 9, we used a Gaussian envelope to reduce the peak amplitudes of undesired sidelobes by sacrificing the spectral resolution. The measured spectral resolution with and without the numerical apodization process is approximately 100 and 50 nm, respectively. As can be seen from Fig. 9, this amount of resolution is not sufficient to clearly resolve the 1500 and 1520 nm laser lines.

K. Yu et al. / Sensors and Actuators A 130–131 (2006) 523–530

Fig. 9. Optical spectra obtained by Fourier transform of the resampled interferograms.

5. Conclusion We have demonstrated a miniaturized Fourier transform spectrometer with optical and opto-mechanical components, including beam splitters and movable micromirrors, that are fabricated on a silicon optical bench platform by a simple bulk micromachining fabrication process. Sidewalls of the etched SOI wafer are used as optical surfaces, and good surface quality as well as flexible patterning is made possible by the combination of wet KOH etching and DRIE. Due to its compactness, simplicity and flexibility, we believe that this type of device will find a number of applications such as lab-on-a-chip spectral monitoring. Moreover, the demonstrated fabrication and integration of miniaturized Michelson interferometers are directly applicable to other important areas such as optical coherence tomography. Acknowledgements This work was partly supported by the Birdseed Fund of Stanford University Office of Technology Licensing. References [1] R.F. Wolffenbuttel, State-of-the-art in integrated optical microspectrometers, IEEE Trans. Instrum. Meas. 53 (2004) 197–202. [2] T. Kiyokura, T. Ito, R. Sawada, Prism interferometer for a compact Fourier-transform spectroscope, Opt. Lett. 25 (2000) 893–895. [3] G.M. Yee, N.I. Maluf, P.A. Hing, M. Albin, G.T.A. Kovacs, Miniature spectrometers for biochemical analysis, Sens. Actuator A Phys. 58 (1997) 61–66. [4] O. Manzardo, R. Michaely, F. Schadelin, W. Noell, T. Overstolz, N. De Rooij, H.P. Herzig, Minature Lamellar grating interferometer based on silicon technology, Opt. Lett. 29 (2004) 1437–1439. [5] A.T.T.D. Tran, Y.H. Lo, Z.H. Zhu, D. Haronian, E. Mozdy, Surface micromachined Fabry–Perot tunable filter, IEEE Photon. Technol. Lett. 8 (1996) 393–395. [6] S.D. Collins, R.L. Smith, C. Gonzalez, K.P. Stewart, J.G. Hagopian, J.M. Sirota, Fourier-transform optical microsystems, Opt. Lett. 24 (1999) 844–846.

529

[7] O. Manzardo, H.P. Herzig, C.R. Marxer, N.F. De Rooij, Miniaturized time-scanning Fourier transform spectrometer based on silicon technology, Opt. Lett. 24 (1999) 1705–1707. [8] C. Solf, J. Mohr, U. Wallrabe, Miniaturized LIGA Fourier Transformation Spectrometer, Presented at Sensors, Toronto, Canada, 2003. [9] S.R. Bhalotra, Adaptive optical microspectrometers and spectra-selective sensing, Ph.D. Thesis, Stanford University, 2004. [10] D. Knipp, H. Stiebig, S.R. Bhalotra, E. Bunte, H.L. Kung, D.A.B. Miller, Silicon-based micro-Fourier spectrometer, IEEE Trans. Electron Devices 52 (2005) 419–426. [11] R.J. Bell, Introductory Fourier Transform Spectroscopy, Academic Press, New York, 1972. [12] D. Lee, K. Yu, O. Solgaard, Vertical micromirror fabricated in (1 1 0) silicon device layer by combination of KOH and DRIE etch, in: Presented at IEEE/LEOS International Conference on Optical MEMS, Takamatsu, Japan, 2004. [13] Y. Uenishi, M. Tsugai, M. Mehregany, Micro-opto-mechanical devices fabricated by anisotropic etching of (1 1 0) silicon, J. Micromech. Microeng. 5 (1995) 305–312. [14] U. Krishnamoorthy, D. Lee, O. Solgaard, K. Yu, Integrated Optical MEMS Devices, US Patent Pending (2005). [15] P. Mach, M. Dolinski, K.W. Baldwin, J.A. Rogers, C. Kerbage, R.S. Windeler, B.J. Eggleton, Tunable microfluidic optical fiber, Appl. Phys. Lett. 80 (2002) 4294–4296. [16] W. Noell, P.A. Clerc, L. Dellmann, B. Guldimann, H.P. Herzig, O. Manzardo, C.R. Marxer, K.J. Weible, R. Dandliker, N.F. Derooij, Applications of SOI-based optical MEMS, IEEE J. Select. Top. Quant. Electron. 8 (2002) 148–154. [17] J. Kauppinen, J. Partanen, Fourier Transforms in Spectroscopy, WileyVCH Verlag GmbH, Berlin, 2001.

Biographies Kyoungsik Yu received the BS degree from Seoul National University, Korea, in 1999, and the MS and PhD degrees in electrical engineering from Stanford University, Stanford, CA, in 2001 and 2004, respectively. Since 2004, he has been with Korea Electrical Engineering & Science Research Institute in Seoul National University, Korea. His current research interests are on the applications of microphotonic devices and techniques, including tunable devices for optical communication networks and optical sensors for imaging and monitoring. He is the author or coauthor of more than 30 technical publications. Daesung Lee received the BS degree from Korea Advanced Institute of Science and Technology (KAIST), Korea, in 1998 and the MS degree in electrical engineering from Stanford University in 2000. He is currently working toward the PhD Degree in electrical engineering at Stanford University. His research interests include the fabrication processes and designs of two types of silicon-on-insulator (SOI)-based micromirrors for optical applications: single-axis, two-axis scanning mirrors actuated by self-aligned vertical combdrives in double SOI layers and MEMS actuated vertical mirrors for optical bench technology. Uma Krishnamoorthy received PhD degree in electrical engineering from University of California, Davis, in 2002. She was a postdoctoral researcher at E.L. Ginzton Laboratory, Stanford University, CA from 2002 to 2004. In the course of her research at UC Davis and Stanford University, she developed enabling technologies for design of reliable optical micro-scale components in optical switching, scanning and spectroscopy applications. She has authored/co-authored 1 book chapter and >25 journal and conference papers. She holds two current and one pending U.S. patents. Currently, she is a senior member of technical staff at Sandia National Laboratory in Albuquerque, NM where she works on design of low-g optical acceleration sensors and large area photonic lattices. Namkyoo Park received his BS and MS degrees in Physics from Seoul National University and Brown University, respectively. After receiving his PhD in applied physics from Caltech in 1994, he worked in the specialty opti-

530

K. Yu et al. / Sensors and Actuators A 130–131 (2006) 523–530

cal fiber division of Lucent Bell Laboratories (MH) before his short stay at Samsung Electronics (1996–1997). He joined the School of EECS of Seoul National University in 1997, where he is currently an associate professor leading his research group, which was selected by the Korean Ministry of Science and Technology as a National Research Laboratory for next generation optical amplifiers. With over 15 years of research experiences in the area of fiber optics, he has authored/coauthored more than 180 international journals and conference publications, patents, book chapters and invited articles. During the past few years, his research efforts have been focused to Raman amplifier, TDFA, OCDMA, PMD, multi-level transmission, nanocluster Si sensitized EDWA and smart applications of OTDR. He is currently an associate editor for IEEE Photonics Technology Letters and Optical Fiber Technology (Elsevier).

Olav Solgaard received the BS degree in electrical engineering from the Norwegian Institute of Technology and his MS and PhD degrees in electrical engineering from Stanford University, California. He was a postdoctoral researcher at the University of California at Berkeley, before joining the University of California at Davis as an Assistant Professor in 1995. In 1999, he joined Stanford University where he is now an associate professor of electrical engineering. His research interests are optical devices and systems for communication and measurements with an emphasis on semiconductor fabrication and MEMS technology. He has authored more than 150 technical publications, and holds 20 patents. He is a co-founder of Silicon Light Machines, Sunnyvale, CA, and an active consultant in the MEMS industry.