silicon-based optical mems for visible to mid-wave...

Silicon-based optical MEMS for visible to mid-wave infraredwavelengthsTripathi, D. K. (2016). Silicon-based optical MEMS for visible to mid-wave infrared wavelengths

Link to publication in the UWA Research Repository

Rights statementThis work is protected by Copyright. You may print or download ONE copy of this document for the purposeof your own non-commercial research or study. Any other use requires permission from the copyright owner.The Copyright Act requires you to attribute any copyright works you quote or paraphrase.

General rightsCopyright owners retain the copyright for their material stored in the UWA Research Repository. The University grants no end-userrights beyond those which are provided by the Australian Copyright Act 1968. Users may make use of the material in the Repositoryproviding due attribution is given and the use is in accordance with the Copyright Act 1968.

Take down policyIf you believe this document infringes copyright, raise a complaint by contacting [email protected]. The document will beimmediately withdrawn from public access while the complaint is being investigated.

Download date: 28. May. 2018

http://research-repository.uwa.edu.au/en/publications/siliconbased-optical-mems-for-visible-to-midwave-infrared-wavelengths(845939a2-0b40-443f-9c06-840d3bce9b29).html

Silicon-based optical MEMS forvisible to mid-wave infrared

wavelengths

Dhirendra Kumar TripathiB.Tech. (UPTU), M.Tech. (NIT)

The University of Western Australia

This thesis is submitted in fulfilment of the requirementsfor the degree of Doctor of Philosophy

School of Electrical, Electronic and Computer EngineeringFaculty of Engineering, Computing and Mathematics

The University of Western Australia

May 16, 2016

Acknowledgements

At first I offer all my gratitude to my teachers who motivated me topursue this important phase of my academic life. I am very thank-ful to my thesis supervisors Prof. Lorenzo Faraone, Prof. K.K.M.B.Dilusha Silva, Associate Prof. Mariusz Martyniuk and Associate Prof.Jarek Antoszewski for their unconditional support. I am very thankfulto Prof. Lorenzo Faraone for sacrificing a great chunk of his personaltime for reviewing this thesis and papers and providing several turningpoint directions during the whole research work. I gratefully acknowl-edge Prof. Dilusha’s unconditional, and friendly help throughout thisresearch work and express many thanks for his many constructive in-puts to bring out the best possible outcome from this research work.Dr. John Bumgarner, with all his expertise in MEMS, always helpedme to overcome some of the critical difficulties in design and processingof MEMS devices. His wife Dr. Susan Bumgarner provided valuableinputs on my writing work. Dr. Ramin Rafiei helped me on opticalmeasurements of the mirrors and filters. Many thanks to him for spend-ing many hours with me. I am very obliged to my close friend HaifengMao for being with me during the clean room processing, machine re-pairs, long meetings and tedious measurement runs. We shared similarproblems in our research work, so discussion with him was very fruit-ful to identify and resolve issues in this research work. My other cleanroom mate was Dr. Fei Jiang, with whom I worked on silicon-siliconoxide based filters and acknowledge his help during my initial trainingin clean-room procedures and tools. My MRG friends Won Jae Lee,N. Radha Krishnan and Dr. Venkatesh Chinappan always extendedtheir friendship, gave encouragement and support, I am very thankfulto them. My many MRG friends Hemendra, Amit, Farah, Ben Cheah,Sun Xiao, Moe Susli, Michal, Dr. Yongling Ren and Kirsten Brookshiresacrificed many of their experimental plans in order to allow this workto move forward. I am indebted to them for their special contributions.Outside my PhD research work, Jegathisan Guruswamy and Jorge R.Silva remained my good friends and supporters; many thanks to themfor being cordial to me. I am very thankful to my wife for being a selflessfriend during my PhD work. My parents and family members gave a lotof love and kind words during this final phase of my education. I owe alot to them. I also seek forgiveness from all my friends and well wishers,since, during this research work I neglected many of my duties towardsthem.

i

Abstract

Optical microelectromechanical systems (Optical MEMS) offer a myr-iad of advantages, such as system miniaturization, robustness, cost ef-fectiveness and high portability. Microspectrometers are optical MEMSbased low cost, portable and robust spectroscopic systems which canbe employed for soil characterization, food quality inspection, chemicalmapping, and spectral imaging. The Microelectronics Research Groupat The University of Western Australia has been involved in the designand development of various microspectrometers technologies for wellover a decade. The heart of a microspectrometer is a tunable Fabry-Perot filter, which usually consists of two distributed Bragg reflectors(DBRs), one of which is fixed and the other is actuated to modulate thewavelength transmitted through the filter.

The objective of this work is to fabricate and characterize silicon basedDBRs and filters as building blocks for microspectrometers operating invisible to mid-infrared wavelengths (450-5000 nm). In order to achievethis objective, first, thin films of inductively coupled chemical vapordeposited (ICPCVD) silicon were optimized for their residual stress andoptical properties. Second, the optimal structural design and fabricationprocess for DBRs and filters was identified and, third, the performanceof fabricated structures was assessed through optical and mechanicalcharacterizations.

It was found that higher deposition temperature leads to better qualityof ICPCVD silicon films. Furthermore, at higher deposition tempera-tures, the decrease in the inductively coupled power results in films withlow tensile stress, higher refractive index and low extinction coefficients.The experimental work shows that formation of stable monohydride(Si-H) bonds between the hydrogen and silicon leads to improved op-tical and mechanical quality of ICPCVD silicon thin films. Thus, it isconcluded that the hydrogen concentration and the specific hydrogen-silicon bonding nature together play a vital role in improving opticaland mechanical quality of the silicon thin films.

This thesis presents a simple process to fabricate quarter wavelengthsilicon-air-silicon based surface micromachined DBRs. A precise controlof stress in the silicon thin films by in-situ and pre-release annealingwas highly effective in fabricating DBRs ranging in area from 200 µm ×200 µm to 5000 µm × 5000 µm size with only a 5-15 nm variation in flat-ness across the surface of the suspended membrane. DBRs from visible

to MWIR wavelengths show near theoretical reflectance properties. Inthe visible and near infrared (NIR) wavelengths, silicon-air-silicon basedDBRs demonstrate higher than 90% reflectance, despite the relativelyhigh absorbance of silicon in these wavelength ranges.

A new notch based actuation structure is presented. It is demonstratedthat this structure allows a suspended membrane to actuate beyond50% of the initial tunable air-gap. This planer actuation structure is ro-bust against the effects of out of plane stress on the suspended mirrors,and can be applied across a wide array of mirror materials and sizes. Atunable multi-spectral MWIR filter was fabricated and optically charac-terized. The filter shows near ideal performance with close to 70% peaktransmittance and 380 nm full width at half maximum (FWHM), anda 800 nm tuning range. This thesis also present the fabrication processand optical characterization of an air-gap mirror based short-wave in-frared (SWIR) filter. The optical transmittance measurement indicatesa first order peak with a peak transmission of 70% and 90 nm FWHMat a wavelength of 2240 nm. The second order peak of the measuredspectrum shows 53% peak transmission and 35 nm FWHM at 1550 nm,which compares very well with the predicted 44% peak transmissionand 30 nm FWHM. In the NIR wavelength range a silicon-silicon oxide-silicon DBR based filter was fabricated. The peak transmission of thefilter at 1140 nm center wavelength is around 54% with 40 nm FWHM,which is in close agreement with simulation. The fabricated filter showsan excellent out of band extinction. In order to assess optical unifor-mity, transmission measurements of the optical MEMS devices wereundertaken as a two dimensional optical transmission map of fabricateddevices.

iii

Publications arising from this thesis

Journal Articles1. D. K. Tripathi, Fei Jiang, Ramin Rafiei, K. K. M. B. Dilusha Silva, Jarek An-toszewski, Mariusz Martyniuk, John M. Dell, “Suspended large area MEMS-basedoptical filters for multispectral shortwave infrared imaging applications” , IEEEJournal of Microelectromechanical Systems, Vol. 24, No. 4, Pages (1102–1110),2015.Summary: The design, fabrication, and optical and mechanical characterization,of freely suspended silicon/silicon-oxide based optical filters and distributed Braggreflectors, in sizes ranging from 500 µm × 500 µm to 5 mm × 5 mm is presented,inconjunction with focal plane array, this opens pathway for fabricating large areaadaptive focal plane array.2. D. K. Tripathi, F. Jiang,M. Martyniuk, J. Antoszewski,K.K.M.B.D. Silva, J.Dell, L. Faraone, “Optimization of ICPCVD silicon for the optical MEMS appli-cations”, IEEE Journal of Microelectromechanical Systems, Vol. 24, No. 6, Pages(1998–2007), 2015.Summary: The optimization of optical and mechanical properties of inductivelycoupled plasma chemical vapor deposited (ICPCVD) amorphous silicon thin filmsfor fabrication of high quality optical microelectromechanical systems based de-vices operating from visible to short-wave infrared wavelengths (450-3000 nm) ispresented. This article make a strong case in favor of silicon as the choice of mate-rial for fabricating MEMS based mirrors and optical filters operating in visible toshort-wave infrared wavelengths.3. Dhirendra K. Tripathi, H. Mao, K. K. M. B. Dilusha Silva, J. Bumgar-ner, M. Martyniuk, John M. Dell and Lorenzo Faraone,“Large-area MEMS baseddistributed Bragg reflectors for short-wave and mid-wave infrared hyperspectralimaging applications”. IEEE Journal of Microelectromechanical Systems, Vol. 24,No. 6, Pages (2136–2144), 2015.Summary: The design, fabrication, and optical characterization, of silicon-air-silicon based distributed Bragg reflectors or quarter wavelength mirrors, in sizesranging from 200 µm × 200 µm to 5 mm × 5 mm, with very low surface pro-file variation, is presented. Such mirrors can be used in conjunction with eithersingle-element photodetectors or large-area focal plane arrays to realise tunablemultispectral sensors or adaptive focal plane arrays from the short-wave infraredwavelength ranges (1500-3000 nm) to mid-wave infrared wavelength (3000-6000 nm)ranges.4. D. K. Tripathi, K. K. M. B. D. Silva, J. W. Bumgarner, R. Rafiei, M. Mar-tyniuk,J. M. Dell, L. Faraone, “Silicon-Air-Silicon Distributed Bragg Reflectors forVisible and Near Infrared Optical MEMS”, IEEE Journal of Microelectromechani-

cal Systems Letters, Vol. 24, No. 5, Pages (1245–1247), 2015.Summary: This article presents the design, fabrication and optical characteriza-tion of silicon-air-silicon based surface micro–machined distributed Bragg reflectors(DBRs), for the visible to near infrared wavelength range (540 nm – 960 nm). It’smain contrbution is to show the feasibility of application of ultra-thin silicon filmsfor silicon-air-silicon based DBRs.

Conference Proceedings1. Dilusha Silva, Dhirendra Tripathi, Haifeng Mao, Jarek Antoszewski, BrettD Nener, John M Dell, Lorenzo Faraone, “Recent developments towards low-costMEMS spectrometer”, Proc. SPIE 9101, Next-Generation Spectroscopic Technolo-gies VII, 910108 (May 21, 2014).Summary: This work presented silicon and silicon oxide based mirrors and filtersas strong candidates for fabricating low-cost MEMS spectrometer.2. D.K. Tripathi, K.K.M.D. Silva, R. Rafiei, R., J. Bumgarner, J., M. Martyniuk,M., T.H. Nguyen, J. Antoszewski, J. M. Dell,L. Faraone, “A silicon based surfacemicro-machined distributed Bragg reflector for MEMS spectroscopic applications”,Proceeding of IEEE 8th International Conference on Industrial and InformationSystems, (ICIIS), Kandy,Sri Lanka (2013) 289-293.3. Dhirendra K. Tripathi, Haifeng Mao, K.K.M.B.D. Silva and Lorenzo Faraone“Compositional and mechanical properties of PECVD silicon for thin-film opticalapplications”, IEEE Proceedings on Conference on Optoelectronic and Microelec-tronic Materials and Devices (COMMAD), Melbourne, Australia (2012) 185-186.Summary: The results of our first investigation on inductively coupled plasmachemical vapour deposited silicon thin film was presented. The main contributionof this article was to show the controllability of stress and optical properties ofICPCVD silicon thin films by RF power and substrate temperature variation.

Conference Presentations1. Dilusha Silva, Dhirendra Tripathi, Haifeng Mao, Jarek Antoszewski, BrettD Nener, John M Dell, Lorenzo Faraone, “MEMS for multispectral imaging in thevisible to the long wave infrared”, SPIE sensing technology + Application, 2015,Balitmore, USA.Summary: This work presented a recent contributions of microelectronics researchgroup in extending the multispectral imaging to the visible and long wave infrared,by fabricating DBRs and filters operable in these wavelengths.2. D.K.Tripathi, K. K. M. B. D. Silva, M .Martyniuk, J. Antoszewski and L.Faraone, “An investigation in the structural and mechanical properties of induc-tively coupled plasma chemical vapour deposited silicon dioxide for MEMS andNEMS applications”, International conference on nanotechonology and nonosciene(ICONN), Adelaide, Australia (2014).Summary: An investigation in the structural and mechanical properties of induc-tively coupled plasma chemical vapour deposited silicon dioxide was presented. Itwas shown that very low residual compressive stressed silicon oxide thin films can

v

be obtained by increasing deposition pressure.Summary: The results of our first investigation of silicon based, solid cavity DBRswas presented. The main contribution of this article was to show fabrication processfor high reflectivity silicon based mirrors.

Patent1. K.K.M.D. Silva, D.K. Tripathi, J. Bumgarner, M. Martyniuk, K. Brookshire, J.Antoszewski, L. Faraone. “Microelectomechanical Systems (MEMS) and Methods”.Summary: This patent present the novel method to fabricate very large area aircavity based distributed Bragg reflectors and optical filters, which can be used fromvisible to long wave infrared wavelength range. Another novel contribution is anotch based actuation structure for optical filters, which actuate more than 50%of initial cavity gap and can be used to generate highly flat MEMS based opticalfilters. Patent Application Number AU2015901600, 5 May 2015.

vi

Contents

1 Introduction 1

1.1 Spectroscopy and applications . . . . . . . . . . . . . . . . . . . . . 1

1.1.1 History of spectroscopy and spectrometers . . . . . . . . . . 1

1.1.2 Portable spectrometers . . . . . . . . . . . . . . . . . . . . . 2

1.2 Dispersive spectrometers . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2.1 Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2.2 Available systems . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Interference based (Fourier Transform) spectrometers . . . . . . . . 4

1.3.1 Theory of operation . . . . . . . . . . . . . . . . . . . . . . . 4

1.3.2 Available systems . . . . . . . . . . . . . . . . . . . . . . . . 6

1.4 Microelectromechanical systems . . . . . . . . . . . . . . . . . . . . 6

1.5 MEMS spectrometers . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.5.1 Microelectromechanical systems in spectrometry . . . . . . . 6

1.5.2 MEMS based standing-wave spectrometers . . . . . . . . . . 7

1.5.3 Linear Variable Filters . . . . . . . . . . . . . . . . . . . . . 8

1.5.4 MEMS FTIR . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.5.5 Fabry-Perot spectrometers . . . . . . . . . . . . . . . . . . . 10

1.6 Development of MEMS based FP filters at UWA . . . . . . . . . . . 12

1.7 Applications and Spectral Requirements for silicon based microspec-trometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1.8 Silicon based process development . . . . . . . . . . . . . . . . . . . 16

1.8.1 Silicon thin film deposition method . . . . . . . . . . . . . . 16

1.8.2 The structure of the DBRs and filters . . . . . . . . . . . . . 17

vii

CONTENTS

1.9 Thesis Outline and Scope . . . . . . . . . . . . . . . . . . . . . . . 18

2 Thin-film optics and Fabry-Perot filters 20

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.2 Maxwell’s equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.3 Fresnel coefficients for single planar interface . . . . . . . . . . . . . 21

2.4 Thin Film interference . . . . . . . . . . . . . . . . . . . . . . . . . 22

2.5 Transfer matrix model . . . . . . . . . . . . . . . . . . . . . . . . . 25

2.5.1 Transfer matrix modelling of a thin film . . . . . . . . . . . 25

2.5.2 Formation of Transfer matrix . . . . . . . . . . . . . . . . . 26

2.5.3 Transfer matrix for a stack of thin films . . . . . . . . . . . . 27

2.5.4 Distributed Bragg Reflector . . . . . . . . . . . . . . . . . . 28

2.6 Design of Fabry-Perot filters using DBRs . . . . . . . . . . . . . . 31

2.7 Anti-reflective coating . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.8 Review of Fabry-Perot Filter Technologies . . . . . . . . . . . . . . 34

2.8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.8.2 Fabry-Perot filter technologies . . . . . . . . . . . . . . . . 35

2.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

3 Material selection and deposition process development 47

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

3.2 Material Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

3.2.1 Modelling studies . . . . . . . . . . . . . . . . . . . . . . . . 51

3.2.2 Simulation of DBRs . . . . . . . . . . . . . . . . . . . . . . 51

3.2.3 Simulation of silicon based filters . . . . . . . . . . . . . . . 51

3.2.4 Final material choice . . . . . . . . . . . . . . . . . . . . . . 56

3.3 Materials deposition process development . . . . . . . . . . . . . . . 58

3.3.1 Instruments and Methodologies . . . . . . . . . . . . . . . . 58

3.3.2 Biased target ion beam deposition . . . . . . . . . . . . . . . 59

3.3.3 Electron beam evaporated silicon thin films . . . . . . . . . 63

3.4 ICPCVD silicon deposition . . . . . . . . . . . . . . . . . . . . . . . 65

viii

CONTENTS

3.4.1 Experimental approach . . . . . . . . . . . . . . . . . . . . . 66

3.4.2 Effect of deposition temperature . . . . . . . . . . . . . . . . 66

3.4.3 Effect of annealing . . . . . . . . . . . . . . . . . . . . . . . 72

3.4.4 Effect of ICP power . . . . . . . . . . . . . . . . . . . . . . . 76

3.5 ICPCVD silicon oxide deposition . . . . . . . . . . . . . . . . . . . 80

3.5.1 Experimental Approach . . . . . . . . . . . . . . . . . . . . 81

3.5.2 Optimization of deposition temperature . . . . . . . . . . . . 81

3.5.3 Optimization of deposition pressure . . . . . . . . . . . . . . 81

3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

4 Silicon-air-silicon DBRs for the SWIR and MWIR wavelengths 86

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

4.1.1 Towards air-spaced tunable filters . . . . . . . . . . . . . . . 86

4.1.2 Filter structures for large area imaging . . . . . . . . . . . . 87

4.2 DBR support structures . . . . . . . . . . . . . . . . . . . . . . . . 88

4.2.1 Support structure with arm based support . . . . . . . . . . 89

4.2.2 Support structure with full peripheral support . . . . . . . . 90

4.3 Si-air-Si based mirrors for SWIR wavelengths . . . . . . . . . . . . 94

4.3.1 Fabrication process . . . . . . . . . . . . . . . . . . . . . . . 94

4.3.2 Optical characterisation of mirror array . . . . . . . . . . . . 96

4.4 Si-air-Si mirrors for MWIR wavelengths . . . . . . . . . . . . . . . . 99

4.4.1 Device layout . . . . . . . . . . . . . . . . . . . . . . . . . . 99


4.4.3 Optical characterisation of MWIR DBRs . . . . . . . . . . . 101

4.5 Silicon-silicon oxide based freely suspended mirrors and fixed-filters 103

4.5.1 Experimental Steps . . . . . . . . . . . . . . . . . . . . . . . 105

4.5.2 Fabrication Process . . . . . . . . . . . . . . . . . . . . . . . 105

4.5.3 Optical characterization of SWIR devices . . . . . . . . . . . 108

4.5.4 Dynamic properties of SWIR DBR . . . . . . . . . . . . . . 111

4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

ix

CONTENTS

5 Fabry-Perot filters for the SWIR and MWIR wavelength ranges 115

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

5.2 Tether and beam based Fabry-Perot filters . . . . . . . . . . . . . . 115

5.2.1 Tether and beam based structures . . . . . . . . . . . . . . . 115

5.2.2 Hex cell based structures . . . . . . . . . . . . . . . . . . . . 117

5.2.3 Fabrication process of silicon-air-silicon based filters . . . . . 117

5.2.4 Optical Surface profilometry of filters . . . . . . . . . . . . . 119

5.2.5 Tether and beam based structures in visible and NIR wave-length range . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

5.3 Stacked layer based cake structures . . . . . . . . . . . . . . . . . . 124

5.4 The etch back process . . . . . . . . . . . . . . . . . . . . . . . . . 125

5.5 Notch based suspension structure . . . . . . . . . . . . . . . . . . . 128

5.6 Ultra-thin silicon single membrane suspended structures . . . . . . 130


5.6.2 Optical surface profile . . . . . . . . . . . . . . . . . . . . . 131

5.6.3 Actuation characteristics . . . . . . . . . . . . . . . . . . . . 132

5.7 Multi-spectral MWIR tunable filter . . . . . . . . . . . . . . . . . . 134

5.7.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134


5.7.3 Optical surface profile . . . . . . . . . . . . . . . . . . . . . 136

5.7.4 Optical transmission characteristics . . . . . . . . . . . . . . 136

5.7.5 Wavelength tuning of fabricated filter . . . . . . . . . . . . . 138

5.8 Fixed cavity SWIR Fabry-Perot filters with silicon-air-silicon top mirror140

5.8.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

5.8.2 Fabrication Process . . . . . . . . . . . . . . . . . . . . . . . 141

5.8.3 Optical Characterization . . . . . . . . . . . . . . . . . . . . 142

5.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

6 Visible and NIR wavelengths DBRs and filters 148

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

6.2 Visible and NIR wavelengths DBRs . . . . . . . . . . . . . . . . . 149

x

CONTENTS

6.2.1 DBR architectures . . . . . . . . . . . . . . . . . . . . . . . 149

6.2.2 Fabrication of silicon oxide cavity DBRs . . . . . . . . . . . 149

6.2.3 Fabrication of air-cavity DBRs . . . . . . . . . . . . . . . . . 150

6.2.4 Final process for fabricating air cavity DBRs . . . . . . . . . 153

6.2.5 Optical characterization of silicon-silicon oxide-silicon DBRs 155

6.2.6 Optical characterization of visible wavelength silicon-air-siliconDBRs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

6.2.7 Optical characterization of NIR silicon-air-silicon DBRs . . . 157

6.3 NIR Fabry-Perot filter using silicon-silicon oxide-silicon based DBRs 161


6.3.2 Optical characterization . . . . . . . . . . . . . . . . . . . . 162

6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

7 Conclusions and Future Work 167

7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

7.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

Bibliography 171

xi

List of Figures

1.1 Basic layout of (a) bench-top grating-based dispersive spectrometer, consisting

of two slits, two curved mirrors, and a grating; and (b) miniaturized grating

spectrometer, where the input slit is replaced by an optical fiber and the output

slit is replaced by a CCD array. . . . . . . . . . . . . . . . . . . . . . . . 4

1.2 Simplified layout of FT spectrometer, showing the source, sample and the core

interferometer. Note, although a transmission setup is shown here, the inter-

action of the light with the sample may be transmission, specular reflection, or

diffuse reflection [14]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.3 Layout of standing-wave MEMS transform spectrometer [19]. If and Id are,

respectively, the forward (incident) and backward (reflected) intensities of light. 7

1.4 (a) Design of a set of linear variable Fabry-Perot filters; and (b) modeled (lines)

and measured (dots) performance of a fabricated set of linear variable Fabry-

Perot filters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.5 Basic operating principle of Fabry-Perot filter. . . . . . . . . . . . . . . . . 10

1.6 Illustration of changing the mirror separation in an ideal Fabry-Perot filter, re-

sulting in a shift in the wavelength of the transmission peaks. Colours are used

for illustration purposes only. . . . . . . . . . . . . . . . . . . . . . . . . 11

1.7 Schematic diagram of Infratec Fabry-Perot microspectrometer construction, show-

ing the mirrors and supports, suspension arms, actuation electrodes, and the

spacer [26]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

1.8 MEMS-based Fabry-Perot filter fabricated on top of detector [27]. . . . . . . 13

1.9 Illustration of the MWIR Fabry-Perot devices and SEM image of array of filters

with improved actuation structure [28, 29]. . . . . . . . . . . . . . . . . . 14

1.10 Measurements demonstrating operation of two separate microspec-trometer devices in the shortwave infrared and midwave infraredspectral regions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

1.11 The plot shows the classic three most common iron minerals; jarosite,hematite and goethite [30]. . . . . . . . . . . . . . . . . . . . . . . . 16

xii

LIST OF FIGURES

2.1 Reflection and refraction at an optical interface between the two lossless dielec-

tric media of different refractive index. Incident (Ei, ki), reflected (Er, kr) and

refracted (Et, kt) electric field are decomposed in orthogonal (E⊥) and parallel

(E‖) components with respect to the plane of incidence. . . . . . . . . . . . 21

2.2 Thin film interference showing overall reflectance and transmittance of a thin

film is a superposition of a multitude of beams with different optical paths.rpq

and tpq correspond to the Fresnel coefficients at the optical interface between

medium p and medium q. p denotes the medium of incidence (i.e. the direction)

of the ray. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.3 Airy function for different reflection coefficients R. . . . . . . . . . . . . . 25

2.4 Propagation of optical wave through a thin film of thickness d. . . . 26

2.5 Layer thicknesses of silicon-air-silicon based DBR at 2000 nm designwavelength on sapphire substrate. . . . . . . . . . . . . . . . . . . . 28

2.6 The reflectance of silicon-air-silicon based DBR for different DBRperiods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.7 Definition of terms used in the text for an ideal Fabry-Perot cavity. The re-

flectance is drawn at an angle for clarity; only normal incidence light is considered

in this thesis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

2.8 Reflectance of glass substrate with and without ARC. . . . . . . . . . . . . 35

2.9 SEM image of the side of a Fabry-Perot filter with InP-Air mirrors 36

2.10 SEM of InP-Air mirror based filter . . . . . . . . . . . . . . . . . . 37

2.11 SEM micrographs of fully processed filter structures . . . . . . . . . 39

2.12 SEM image of Si3N4 -air mirror based filter showing vertical crosssection of the practical silicon nitride/air stack . . . . . . . . . . . . 40

2.13 IR transmission images of a 4 × 4 mirror array (left) and a 14 × 14mirror array (right) . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.14 Spring suspended Bulk micromachined FP filter. Left side showsschemetic and right hand side shows SEM image . . . . . . . . . . . 43

2.15 Dual band Fabry-Perot filter . . . . . . . . . . . . . . . . . . . . . . 43

2.16 1-D DBR based filter on Si substrate. . . . . . . . . . . . . . . . . . 44

2.17 SEM photos of Fabry-Perot cavity with Bragg mirrors of cylindricalshape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.1 Modelled absorption from a quarter-wave thick layer of (a) a quarter–wave thick

layer of germanium; and (b) a quarter–wave thick layer of silicon. Note, these

models used refractive index and absorption data for crystalline materials, from

the Handbook of Optical Constants of Solids by Edward D. Palik [92]. . . . . 49

xiii

LIST OF FIGURES

3.2 Band gap of semiconductor materials (a) germanium; and (b) silicon. . . . . . 50

3.3 Modelled reflectivity of QWMs made of different dielectric materials. . . . . . 52

3.4 Transmittance of Fabry-Perot filter consisting of DBRs of different dielectric

materials. Note that there is an offset of 50 nm among the plots for clarity. . . 53

3.5 Transmittance plot of Fabry-Perot filters in visible wavelength; (a) Ta2O5-MgF2-

Ta2O5 DBR based filter; (b) Si-SiOx-Si DBR based filter; (c) Si-air-Si DBR based

filter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

3.6 Transmittance plot of Fabry-Perot filters in NIR wavelength; (a) Si-SiOx-Si DBR

based filter; (b) Si-air-Si DBR based filter. . . . . . . . . . . . . . . . . . 55

3.7 Transmittance plot of Fabry-Perot filters in SWIR wavelength; (a) Si-SiOx-Si

DBR based filter; (b) Si-air-Si DBR based filter. . . . . . . . . . . . . . . . 56

3.8 Transmittance plot of Fabry-Perot filters in MWIR wavelength; (a) Si-SiOx-Si

DBR based filter; (b) Si-air-Si DBR based filter. . . . . . . . . . . . . . . . 57

3.9 Schematic diagram of a three target BTIBD system (courtesy: 4WAVE Inc USA.) 60

3.10 Refractive index and extinction coefficent of silicon thin film deposited using

BTIBD system.) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

3.11 A microgrph of formation of micro-crystallites on annealed silicon film. . . . . 61

3.12 XRD plot of the BTIBD depoisted silicon film which is annealed at 600 ◦Cfor

30 minutes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

3.13 EDS spectra of the silicon film deposited from BTIBD system. . . . . . . . . 62

3.14 Schemtaic diagram of BOC Edwards system 500 E-beam system. . . . . . . . 63

3.15 Refractive index and extinction coefficient of the e-beam deposited silicon film. 64

3.16 Measured tensile stress in the E-beam deposited silicon film. . . . . . . . . . 65

3.17 A picture of the SENTECH SI500D reactor. . . . . . . . . . . . . . . . . . 67

3.18 Measured optical constants of ICPCVD silicon thin films at different deposition

temperature; (a) shows refractive index of thin films deposited at different tem-

peratures, (b) shows extinction coefficient of thin films deposited at different

temperatures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

3.19 FTIR spectra of the films deposited at different deposition temperature. . . . 70

3.20 Residual stress in the silicon films as a function of annealing temperature, using

the deposition conditions of sample 1 from Table 3.5. . . . . . . . . . . . . 73

3.21 Measured optical constants of silicon thin films annealed at different annealing

temperatures; (a) shows refractive index, (b) shows extinction coefficient. The

films were deposited using the deposition conditions of sample 1 from Table 3.5. 74

xiv

LIST OF FIGURES

3.22 FTIR spectra of silicon films annealed at different annealing temperatures. The

films were deposited using the deposition conditions of sample 1 from Table 3.5. 75

3.23 Hydrogen concentration of silicon films as a function of annealing temperature.

The films were deposited using the deposition conditions of sample 1 from Table

3.5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

3.24 Residual stress and deposition rate of the Si films as a function of ICP power. . 77

3.25 Measured optical constants of thin films deposited at different ICP power; (a)

shows refractive index, (b) shows extinction coefficient (see Table 3.5). . . . . 78

3.26 FTIR spectra of silicon films deposited at different ICP powers (see Table 3.5). 79

3.27 Hydrogen concentration in the films as a function of ICP power (see Table 3.7). 80

3.28 Deposition rate and etch rate of SiOx films as a function of deposition tempera-

ture see Table ??. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

3.29 Residual stress in the SiOx films as a function of deposition temperature (see

Table 3.8). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

3.30 Deposition rate and etch rate of SiOx films as a function of deposition pressure

(see Table 3.8). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

3.31 Residual stress in the SiOx films as a function of deposition pressure (see Table

3.8). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

4.1 Optical arrangement of layers in a MEMS-based tuneable Fabry Perot filter con-

sisting of two air-spaced mirrors. . . . . . . . . . . . . . . . . . . . . . . 87

4.2 Working principle of proposed module consisting of a tunable Fabry-Perot cavity

hybrid bonded to a detector or imaging FPA. . . . . . . . . . . . . . . . . 88

4.3 Arm based DBR structure with an optional anti-stiction bump at centre. . . . 89

4.4 3-D Optical surface profiles of fabricated circular suspended structures: (a)

100 µm unannealed structure; (b) 100 µm structure annealed at 310 ◦C for

45 minutes; and (c) 200 µm structure with anti-stiction bump annealed at 350 ◦C

for 60 minutes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

4.5 Single line scan across the surface profiles of fabricated circular suspended struc-

tures shown in Fig. 4.4: (a) 100 µm unannealed structure; (b) 100 µm structure

annealed at 310 ◦C for 45 minutes; and (c) 200 µm structure with anti-stiction

bump annealed at 350 ◦C for 60 minutes. . . . . . . . . . . . . . . . . . . 92

4.6 Optical surface profiles of fabricated 270 µm diameter circular suspended struc-

ture, (a) 3-D surface profile, (b) extracted line scan across the suspending mem-

brane avoiding the etch holes, and (c) profile of fabricated structure on an ex-

panded vertical scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

xv

LIST OF FIGURES

4.7 The mask layout for circular mirrors ranging from 270 µm to 420 µm diameter

in size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

4.8 Fabrication process of SWIR mirrors. (a) Deposition of Si thin film on sapphire

substrate; (b) Spinning, hard-baking and patterning of PI 2610 polyimide ; (c)

Deposition of the second Si thin film on top of polyimide; (d) Patterning the etch

holes through the top Si membrane; (e) Annealing and dry release of mirror in

an O2 plasma. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

4.9 Micrograph of the fabricated SWIR DBRs (a) A 2×2 array of 270 µm diameter

DBRs (b) A magnified view of a single DBR. . . . . . . . . . . . . . . . . 97

4.10 Optical surface profiles of the fabricated DBRs (a) Shows 3 D optical profile of

a 2 × 2 array of 270 µm diameter DBRs, and (b) shows a magnified image of a

3 D surface profile of the one of DBRs of the array shown in (a). . . . . . . . 97

4.11 Surface profile of the quarter wave mirror in Fig. 4.10, showing the step height. 98

4.12 Measured (squares) and simulated (line) optical transmission of a SWIR quarter

wave mirror. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

4.13 SWIR mirror transmission and reflection. Measured Transmittance: measured

transmittance data; Simulated Transmittance: expected transmittance based on

a transmission matrix model using estimated thickness of layers and initial silicon

characterisation; Simulated Reflectance: expected reflectance of mirror based

on simulated transmission; Calculated Reflectance: mirror reflectance based on

measured transmission. . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

4.14 Structural design of large area DBRs with conformal support around the entire

periphery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

4.15 Fabrication process of MWIR mirrors. (a) Deposition of 350 nm Si thin film on

silicon oxide coated substrate; (b) Deposition and etching of metal optical shield;

(c) Spinning, hard-baking and patterning of PL100–16 polyimide ; (d) Conformal

deposition of the second Si thin film on polyimide; (e) Patterning the etch holes

through the top Si membrane; (f) Dry release of mirror in an O2 plasma. . . . 101

4.16 Micrograph of fabricated 2 mm × 2 mm MWIR DBR. . . . . . . . . . . . . 102

4.17 3-D optical surface profile of 2 mm × 2 mm MWIR DBR. . . . . . . . . . . 102

4.18 A line scan across the 3-D optical surface profile of a 2 mm × 2 mm mirror

avoiding the etch holes. . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

4.19 MWIR mirror transmission and reflection. Measured Transmittance: measured

transmittance data; Simulated Transmittance: expected transmittance based on

a transmission matrix model using estimated thickness of layers and initial silicon

recipe characterisation; Simulated Reflectance: expected reflectance of mirror;

Calculated Reflectance: mirror reflectance based on measured transmission. . . 104

xvi

LIST OF FIGURES

4.20 Fabrication process flow of freely suspended Fabry-Perot optical filters and op-

tical mirrors: (a) deposition of SiOx etch stop layer (b) deposition of the optical

layers (c) patterning of the back-side of the handle wafer with AZ 5214 resist (d)

cryogenic back-side etching of handle wafer (e) removal of SiOx etch stop layer

in HF to leave the final suspended Fabry-Perot optical filters and DBR mirrors. 106

4.21 Fabricated SWIR devices before annealing showing significant distortions in the

optical layers due to high compressive stress. The coin is an Australian one dollar

piece of diameter 25 mm. . . . . . . . . . . . . . . . . . . . . . . . . . . 107

4.22 Optical surface profile of a 2 mm × 1 mm SWIR DBR after a 40 minute anneal

at 400 ◦C in N2 atmosphere. The central 1.5 mm section is flat to within 10 nm. 108

4.23 Optical surface profile of a 2 mm × 1 mm Fabry-Perot filter after a 1 hour anneal

at 420 ◦C in a N2 atmosphere. The central 1.25 mm section is flat to within 25 nm.109

4.24 Measured and simulated optical transmission spectra of the SWIR filter shown

in Fig. 4.23. Figure shows a close match between calculated and simulated

transmittance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

4.25 Measured and modelled optical transmission and reflection spectra of the sus-

pended SWIR mirror shown in Fig. 4.22. Measured Tx: measured transmittance

data; Model Tx: mirror transmittance based on a fit to the mirror structure us-

ing a transmission matrix model; Model Rx: expected reflectance of mirror;

Estimated Rx: mirror reflectance based on fitted transmission. . . . . . . . . 110

4.26 Spatially resolved measurements of calibrated minimum transmission of a 3 mm

× 3 mm quarter-wave SWIR DBR. . . . . . . . . . . . . . . . . . . . . . 111

4.27 Spread of minimum spectral transmission from Figure 4.26. . . . . . . . . . 112

4.28 Flexural modes in 1 mm× 1 mm suspended SWIR DBR showing the fundamental

and harmonic flexural modes. . . . . . . . . . . . . . . . . . . . . . . . . 113

5.1 Tether and beam based structures used by earlier works performed at micro-

electronics research group: (a) Angeled tether based filter for SWIR and MWIR

wavelength range operation [28]; (b) straight tether based filter for LWIR wave-

length range operation [164]; (c) straight tether based structure used in this

thesis for fabricating Si-air-Si suspended mirror. Note that the etch holes size is

exaggerated for clarity. . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

5.2 Hexagonal cells based structures for circular and square shaped mirrors. . . . . 117

xvii

LIST OF FIGURES

5.3 Fabrication process of SWIR tunable Filter. (a) Deposition of 141 nm silicon thin

film on sapphire substrate; (b) deposition and definition of bottom electrode; (c)

spinning, hard-baking and patterning of PI2611 polyimide; (d) deposition of

141 nm silicon thin film for top membrane of bottom mirror; (e) spinning, hard-

baking and patterning of PI2610 polyimide; (f) deposition of 141 nm silicon

thin film for bottom membrane of top mirror; (g) spinning, hard-baking and

patterning of PI2611 polyimide; (h) deposition of 141 nm silicon thin film for

top membrane of top mirror; (i) deposition and definition of top electrode; (j)

patterning top mirror and etch holes; (k) deposition of the optical metal shield

on the back side of substrate to define the optically active area; (l) pre-release

annealing and dry release in an O2 plasma. . . . . . . . . . . . . . . . . . 120

5.4 Optical surface profile and a line scan across the fabricated silicon-air-silicon

DBR based Fabry-Perot filters; (a) Circular shaped filter; (b) Square shaped

filter; (c) Hex-cell based filter. . . . . . . . . . . . . . . . . . . . . . . . 121

5.5 A line scan across the width of a beam of fabricated silicon-air-silicon DBR based

filters showing gutter-like profile. . . . . . . . . . . . . . . . . . . . . . . 122

5.6 Fabrication process of suspended top mirror for NIR and visible wavelength

range. (a) Spinning, hard-baking and patterning of 20% diluted PI2611 poly-

imide; (b) deposition of quarterwave thick silicon thin film for bottom membrane

of top mirror; (c) spinning, hard-baking and patterning of 50% diluted PI2610

polyimide; (d) deposition of second quarterwave thick silicon thin film for top

membrane of top mirror; (e) patterning top mirror and etch holes; (f) pre-release


5.7 Collapsed suspended top mirror structures and line scan across their optical

surface profile (a) NIR wavelength range structure (b) visible wavelength range

structure (c) thicker beam, tether and rim based structure for visible wavelength

range. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

5.8 A conceptual cross section image of a circular cake structure. . . . . . . . . . 125

5.9 Optical surface profile and a line scan across the fabricated silicon-air-silicon

cake structures; (a) Circular shaped cake structures; (b) Square shaped cake

structures; (c) Hex-cell based cake structures. . . . . . . . . . . . . . . . . 126

5.10 Etch back process steps (a) deposition and patterning of sacrificial layer; (b)

deposition of thick silicon layer; (c) etch back of the optically active area; (d)

formation of etch holes; (e) removal of sacrificial layer. . . . . . . . . . . . . 127

5.11 (a) Optical surface profile of 200 µm diameter circular structure; (b) a line scan

across surface profile of circular structure; (c) optical surface profile of 200 µm ×200 µm square structure; (d) a line scan across surface profile of square structure. 128

5.12 Layout of the fabricated filters using the notch based suspensionstructure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

xviii

LIST OF FIGURES

5.13 Fabrication process for single membrane suspended structure (a) deposition of

bottom electrode on sapphire substrate; (b) spinning, hard-baking and patterning

of Prolift100-16 polyimide; (c) deposition of thick silicon film for top membrane;

(d) deposition and defining top electrodes; (e) etch back of optical area to 150 nm

thick; (f) patterning of etch holes and notches; (g) pre-release annealing and dry

release in O2 plasma; (h) etch back of additional 80 nm in CF4 based etch recipe. 132

5.14 Fabricated 1 mm × 1 mm area 70 nm thick suspended membrane. . . . . . . 132

5.15 (a) Optical surface profile of fabricated 1 mm × 1 mm area 70 nm thick suspended

membrane shown in Fig. 5.14; (b) a line scan across the surface profile; and (c)

a close up of the central 600 µm area of the membrane shown in (b). . . . . . 133

5.16 DC actuation characteristic of fabricated 1 mm × 1 mm area 70 nm thick sus-

pended silicon membrane. . . . . . . . . . . . . . . . . . . . . . . . . . 134

5.17 Fabrication process of MWIR tunable Filter. (a) Deposition of silicon oxide–

silicon–silicon oxide–silicon bottom mirror stack on silicon substrate; (b) deposi-

tion and definition of bottom electrode; (c) spinning, hard-baking and patterning

of Prolift100-16 polyimide; (d) deposition of silicon film for top membrane mirror,

deposition of top electrodes, and etch back of silicon membrane; (e) patterning

of etch holes and notches; (f) deposition and definition of anti-reflective layer;

(g) deposition of the optical metal shield on the back side of substrate to define

the optically active area; (h) pre-release annealing and dry release in an O2 plasma.137

5.18 A micro-graph of fabricated MWIR filter. . . . . . . . . . . . . . . . . . . 137

5.19 (a) Optical surface profile of fabricated 1 mm × 1 mm area MWIR filter; (b) Line

scan across the surface profile shown in (a); (c) Expanded scale surface profile of

center 800 µm area of top mirror. . . . . . . . . . . . . . . . . . . . . . . 138

5.20 Optical transmittance of 1 mm × 1 mm size fabricated MWIR filter using the

fabrication process outlined in Figure 5.17. Measured transmittance is shown as

open squares against the simulated performance shown as a dashed line. . . . . 139

5.21 Displacement vs voltage characteristics of 1 mm × 1 mm MWIR filter. . . . . 140

5.22 Spectra showing wavelength tuning of 1 mm × 1 mm MWIR Fabry-Perot filter

with a suspended silicon membrane mirror. Measured transmittance is shown as

open squares against the simulated performance shown as dashed lines. Applied

actuation voltage is shown on top of each transmittance peak. . . . . . . . . 141

xix

LIST OF FIGURES

5.23 Fabrication process of SWIR Fabry-Perot filter. (a) Deposition of silicon oxide–

silicon –silicon oxide–silicon bottom mirror stack on silicon substrate; (b) depo-

sition of bottom electrode; (c) spinning, and hard-baking of Prolift100-16 poly-

imide; (d) deposition of first three quarter wave thick silicon film for top mirror;

(e) spinning, hard-baking and wet patterning of Prolift100-16 polyimide for the

top mirror air cavity; (f) deposition of second three quarter wave thick silicon

film for top mirror and deposition and definition of top electrodes; (g) patterning

the etch holes and notches; (h) deposition of back-side anti-reflective layer; (i)

deposition of the optical metal shield on the back side; (j) pre-release annealing

and dry release in O2 plasma. . . . . . . . . . . . . . . . . . . . . . . . . 144

5.24 A micro-graph of fabricated SWIR filter. . . . . . . . . . . . . . . . 145

5.25 (a) Optical surface profile of fabricated 500 µm × 500 µm SWIR Fabry-Perot

filter; (b) A close up of the top mirror; (c) line scan across the surface profile;

(d) line scan on top mirror on an expanded scale avoiding the etch holes. . . . 145

5.26 Experimental and simulated transmittance spectra of Fabry-Perot SWIR filter.

The points represent measurements, while the dashed line represents model pre-

dictions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

5.27 Measured and simulated reflectivity of silicon-silicon oxide-silicon bottom mir-

ror of SWIR filter shown in Figure 5.24. Also shown is the transmittance and

absorbance through the mirror. . . . . . . . . . . . . . . . . . . . . . . . 146

6.1 Structure of visible and NIR wavelengths DBRs (a) visible wavelength DBR with

SiOx as the low index medium ; (b) visible wavelength DBR with air as the low

index medium; (c) NIR wavelength DBR with SiOx as the low index medium ;

(d) NIR wavelength DBR with air as the low index medium. . . . . . . . . . 150

6.2 Optical modeling of NIR and visible wavelength DBRs shown in Figure 6.1 (a)

Reflectance and absorbance of visible wavelength DBR with SiOxand air as the

low index medium ; (b) Reflectance of NIR wavelength DBR with SiOx and air

as the low index medium. . . . . . . . . . . . . . . . . . . . . . . . . . 151

6.3 De-lamination of membrane at the edges during dry-release process due to ex-

cessive residual tensile stress. . . . . . . . . . . . . . . . . . . . . . . . . 152

6.4 Mask Layout for the visible and NIR DBRs. . . . . . . . . . . . . . . . . . 153

6.5 Fabrication process of NIR and visible wavelength DBRs: (a) deposition of Si

thin film on sapphire substrate; (b) spinning, hard-baking and patterning of PI

2610 polyimide; (c) deposition of the second Si thin film on top of polyimide

and patterning of the etch holes through the top Si membrane; (d) pre-release


6.6 270 µm diameter fabricated DBRs: (a) visible wavelength mirror; (b) NIR wave-

length mirror. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154

xx

LIST OF FIGURES

6.7 Measured transmittance, reflectance and absorbance of fabricated silicon-silicon

oxide-silicon DBRs. Also shown is the simulated reflectance for each DBR: (a)

NIR wavelength DBR; (b) visible wavelength DBR. . . . . . . . . . . . . . 156

6.8 Measured and simulated optical transmisttance, reflectance and absorbance spec-

tra of a silicon-air-silicon quarter wave mirror at visible wavelengths. . . . . . 156

6.9 High resolution, spatially resolved map of the calibrated minimum transmission

measuremnt on an array of quarter wave silicon-air-silicon mirrors at a wave-

lengths of 680 nm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

6.10 High resolution, 3-D plot of a spatially resolved map of calibrated minimum

transmission measurements on an array of silicon-air-silicon DBRs at 680 nm

wavelength. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

6.11 Transmission profile of 320 µm diameter visible wavelength range one-quarter

wave silicon-air-silicon mirror at a wavelength of 680 nm. . . . . . . . . . . . 159

6.12 Measured and simulated optical transmission spectra of the quarter wave silicon-

air-silicon NIR wavelength mirror. Measured Tx: measured transmittance data;

Simulated Tx: expected transmittance based on a transmission matrix model

using estimated thickness of layers and initial silicon recipe characterisation;

Simulated Rx: expected reflectance of mirror based on simulated transmission;

Fitted Tx: mirror transmittance based on a fit to the mirror structure using a

transmission matrix model; Calculated Rx: mirror reflectance based on fitted

transmission. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

6.13 High resolution, spatially resolved maps of the calibrated transmission mea-

suremnt of an array of one-quarter wave silicon-air-silicon NIR mirrors. . . . . 160

6.14 Transmission profile of 370 µm diameter NIR wavelength range one-quarter wave

silicon-air-silicon mirror at 1000 nm wavelength. . . . . . . . . . . . . . . . 161

6.15 Fabrication process of NIR Filter. (a) Deposition of Silicon-Silicon Oxide-Silicon

bottom mirror stack on BK7 glass substrate; (b) deposition of bottom elec-

trode; (c) spinning, hard-baking and patterning of Prolift100-16 polyimide; (d)

deposition of silicon-silicon oxide-silicon top mirror stack; (e) deposition of top

electrode; (f) patterning of etch holes; (g) pre-release annealing and dry release

in O2 plasma. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

6.16 3-D surface profile of 500 µm × 500 µm NIR Fabry-Perot filter. . . . . . . . 163

6.17 Line scan of the surface profile of 500 µm × 500 µm NIR filter shown in Fig. 6.16.163

6.18 Single point optical transmission spectra of 500 µm × 500 µm NIR Fabry-Perot

filter with silicon-oxide-silicon DBRs. Measured data is shown by markers and

dashed line shows simulated transmittance. . . . . . . . . . . . . . . . . . 164

6.19 High resolution 3-D spatial map of the peak wavelength for a 500 µm × 500 µm

NIR Fabry-Perot filter with silicon-silicon oxide-silicon DBRs. . . . . . . . . 165

xxi

List of Tables

3.1 Optical refractive index, n, extinction coefficient, k, quarter wave thickness, t,

and absorption, A, in a quarter wave thick layer, of various materials at visible

(560 nm), NIR (1000 nm) and SWIR (2000 nm) wavelengths [92]. “ngl” stands

for negligible. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3.2 Layer thickness and cavity gap used in modelling Fabry-Perot filters. . . . . . 52

3.3 Summary of FWHM, FSR and Transmittance of modelled Fabry-Perot filters. . 53

3.4 Elemental analysis of BTIBD deposited Si thin films using EDS. . . 63

3.5 ICPCVD Process parameters for deposition of silicon thin films. . . . . . . . 68

3.6 Summary of results of optimization of silicon by annealing in nitrogen for 30

minutes. All the films were deposited at 4 Pascal deposition pressure, and 100 ◦C

deposition temperature. The flow rates of silane and helium were kept at 5 sccm

and 95 sccm, respectively. . . . . . . . . . . . . . . . . . . . . . . . . . 71

3.7 Summary of results of in situ optimization of silicon. All the films were deposited

at 4 Pascal deposition pressure. The flow rate of silane and helium were kept at

5 sccm and 95 sccm, respectively. . . . . . . . . . . . . . . . . . . . . . . 71

3.8 Summary of results of the optimization process for silicon oxide films. . . . . . 85

4.1 Silicon deposition recipe used for mirrors. . . . . . . . . . . . . . . . . . . 89

4.2 Recipe for etching PI2610. . . . . . . . . . . . . . . . . . . . . . . . . . 96

4.3 Silicon etch recipe used in fabricating mirrors. . . . . . . . . . . . . . . . . 96

4.4 ICPCVD SiO2 deposition parameters. . . . . . . . . . . . . . . . . . . . . 105

4.5 ICPCVD deposition parameters for Si thin films. . . . . . . . . . . . . . . . 105

4.6 Targeted and actual thicknesses of the SWIR optical layers. . . . . . . . . . 107

4.7 Resonant frequencies of 1 mm × 1 mm square shaped suspended SWIR DBR. . 113

5.1 Silicon deposition recipe. . . . . . . . . . . . . . . . . . . . . . . . . . . 118

5.2 Recipe for etching PI2610. . . . . . . . . . . . . . . . . . . . . . . . . . 118

xxii

LIST OF TABLES

5.3 Silicon etch recipe used in etch back. . . . . . . . . . . . . . . . . . . . . 127

5.4 ICPCVD silicon oxide deposition parameters. . . . . . . . . . . . . . . . . 136

5.5 Targeted vs actual thickness of the individual MWIR filter layers (from bottom

to top). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

5.6 Targeted vs actual thickness of the SWIR filter layers (from bottom to top). . . 143

6.1 Silicon deposition recipe. . . . . . . . . . . . . . . . . . . . . . . . . . . 150

6.2 Process parameters for deposition of silicon oxide films. . . . . . . . . . . . 150

6.3 Summary results of fabricated silicon-silicon oxide-silicon DBRs. . . . . . . . 155

6.4 Summary results of fabricated silicon-air-silicon DBRs. . . . . . . . . . . . . 157

xxiii

LIST OF TABLES

xxiv

Chapter 1

Introduction

This chapter defines spectroscopy, and presents a brief introduction of bulk opticsand MEMS based spectroscopic technologies. It gives a brief account of the theorybehind Fabry-Perot tunable filters. This chapter describes how MEMS technologyis being used to develop miniaturized Fabry-Perot tunable filters. The chapterconcludes by presenting the contributions to knowledge born of this thesis and thelayout of the thesis.

1.1 Spectroscopy and applications

Applications of spectroscopy

Spectroscopy is a scientific method to study the nature of interaction of the elec-tromagnetic spectrum with materials. It is a non-destructive method. It decodesthe information about the sample in question through its absorption, emission, re-flection or transmission spectra. It is extensively used for characterizing materials,identification of presence of elements, food quality inspection, drug quality con-trol, soil analysis, crop disease identification, thin film analysis, water stress mea-surement, chemical and physicochemical characterization of chemical compounds[1, 2, 3, 4, 5].

1.1.1 History of spectroscopy and spectrometers

The two dominant spectrometer types used for laboratory-based spectrometry are(1) the grating based spectrometer and (2) the Fourier Transform (FT) spectrome-ter. Grating spectrometers have been in existence for quite some time, being demon-strated as a useful spectroscopic tool after a ruling engine was first demonstratedby Rowland in 1882 [6]. The core engine of a FT spectrometer, the Michelson inter-ferometer, has been in existence since the 1890s. The concept of a FT spectrometer

1

CHAPTER 1. INTRODUCTION

has been in existence since Rayleigh proposed using the interference pattern to pro-duce a spectrum [6]. However, a viable FT spectrometer was not realized until theconvergence of a number of technologies, including sampling theory, instrumenta-tion, computational hardware, and the Fast Fourier Transform (FFT) algorithm.In year 1949, P. B. Fellgett made a remarkable contribution by proposing that for asignal, dominated by detector noise, Fourier transform spectrometers can producea relative improvement of the order of the square root of m in signal-to-noise ratio(SNR), compared to an equivalent scanning monochromator, where m is the num-ber of sample points comprising the spectrum [7]. Two years after this discovery,an infrared spectrometer was first demonstrated by Kaye et al. [8], in the 1950’s, atBeckman Instruments. Kaye modified an existing spectrometer to measure spectrawith wavelengths ranging up to 2.7 µm. Subsequently, more instrument manu-facturers entered the market, with various designs of infrared spectrometers. Keyplayers in this area today include Foss, Bruker, Perkin Elmer, Thermo Scientific,and Agilent. However, many of these manufacturers cater for as broad as possi-ble a market with their instruments. As a result, their systems tend to be largebench-based units with specifications far exceeding that required for any single ap-plication. The benefit of offering such systems is that, due to the low initial uptakeof spectroscopy in industrial applications, the manufacturers do not have to producecustom solutions. However, this all-in-one solution also results in rather expensivesystems. These bench-top systems are too large and costly for widely distributedapplication. Their size makes it difficult to incorporate into many field environmentsor existing industrial production lines. The cost of these systems places them outof reach for all but the largest operators in most industries. Regular servicing andcalibration is required to ensure that these systems are operating to specification,which further increases the cost associated with these systems.

1.1.2 Portable spectrometers

Bench-top spectrometers are unsuitable for field applications primarily due to theirfragility and bulk. Additionally, because of the high cost, someone purchasing oneof these units will likely be reluctant to risk their investment in a field environment.Although some ruggedized systems exist [9], this is clearly a non-optimal solutionbecause of the increased cost and weight associated with the ruggedization. For thisreason, a great deal of research effort is being expended world-wide in the pursuitof miniature spectrometers that are suitable for the field. It should be noted,however, that many of these miniaturized systems involve a trade-off in terms ofperformance, which, in turn, limits the scope of application for these systems. Thekey factors hindering the application of traditional spectrometers, pervasively inthe field, are (1) bulk (weight and volume), (2) robustness, and (3) cost. Varioussolutions have been developed, addressing some or all of these issues, in the drive tomove spectroscopy out of the laboratory. The portable spectrometer technologiesexamined here will be discussed in terms of either dispersive systems (grating/prismbased); interference based systems; or MEMS systems, which is a broad categorywith a number of different implementations. There may sometimes also be some

2


overlap between the MEMS systems and the other categories.

1.2 Dispersive spectrometers

1.2.1 Operation

Most miniature dispersive spectrometers operate in a very similar fashion to theirbench-top counterparts. Figure 1.1 depicts the internal operation of a common(a) bench-top and (b) miniature dispersive spectrometer. In a benchtop gratingspectrometer, the input slit is located at the focal point of the mirror M1. Externallight will be focused on the input slit, to diverge towards M1. Light reflected fromM1 will be collimated and directed towards the grating. Light dispersed by thegrating is re-focused at the output slit by the mirror M2. However, the light at theoutput slit will be a set of images of the input slit, dispersed in wavelength. Byrotating the grating, the wavelength of the image incident on the output slit canbe varied. In many portable dispersive spectrometers, there is no explicitly definedinput or output slit. Instead, the core of the input optical fiber forms the inputslit, and a linear CCD or photodiode array is placed in the plane of the outputslit. Consequently, the width of the CCD pixel defines the width of the output slitand, because there are numerous pixels in the plane of the output slit, the gratingdoes not need to be rotated to obtain a spectral measurement. A very desiredfeature of this type of spectrometer is that it contains no moving parts. Therefore,ruggedization of the unit requires much less effort than for a FTIR system. Thedisadvantage of this type of system is the restricted light throughput due to therequirement for coupling into optical fibre. The cost of this type of system is usuallyin the range of a $ 1,000 - $ 5,000. Due to the componentry required, it is unlikelythese systems will get much cheaper.

1.2.2 Available systems

Numerous vendors have miniaturized grating spectrometers commercially availablefor well over a decade. Some examples include the systems by Fibre photonics [10],Ocean Optics [11], Thorlabs [12], and Hamamatsu [13]. Fibre photonics offers anumber of fiber-optic grating-based spectrometers, in both single- and dual-channelform, and wavelengths ranging from 200 nm to 1,750 nm. The Thorlabs system hasmeasurement wavelength range of 200 nm to 1000 nm. The spectrometer moduleitself weighs less than 400 g. The spectral resolution of the unit is as good as 0.4 nmat 365 nm. Hamamatsu uses an arrangement based on lenses and a transmissiongrating, rather than mirrors and a reflection grating. However, the basic principleis the same. This range of systems has operational wavelength ranges in the 200 nmto 2,550 nm window, with spectral resolution as good as 8 nm for some models.

3


(a) (b)

Figure 1.1: Basic layout of (a) bench-top grating-based dispersive spec-trometer, consisting of two slits, two curved mirrors, and a grating; and (b)miniaturized grating spectrometer, where the input slit is replaced by an op-tical fiber and the output slit is replaced by a CCD array.

1.3 Interference based (Fourier Transform) spec-

trometers

1.3.1 Theory of operation

Most non-MEMS miniaturized FTIR systems tend to be simply smaller, sometimesmore rugged, versions of the bench-top FTIR. Although some elegant engineeringmay be included to overcome such problems as vibrations, the core operationalprinciple is the same. The basic operational principal of a Fourier Transform spec-trometer is shown in Figure 1.2. As with any other spectrometer, there is a source,and an interaction segment where light from the source interacts with the sampleof interest. Note, for simplicity of explanation, Figure 1.2 depicts the spectrometerwith the sample between the source and interferometer. However, it is very com-mon, to have the sample between the interferometer and the photodetector. Thisconfiguration, in fact, will improve signal at the detector, because coupling lightscattered from a sample into an interferometer is very difficult. The core engine ofthe FT spectrometer is the interferometer. Figure 1.2 shows a Michelson configu-ration interferometer, consisting of a beam splitter, a stationary mirror MS, and atranslating mirror MT. The input light to the interferometer is the light from thesample. It can be shown [14] that the signal at the photo-detector is given by thefollowing equation:

Vd(∆z) = kdg2(1− g2)((ρ2

τ + ρ2s) + 2ρτρsRe[C(2∆z)]) (1.1)

where kd is a proportionality constant, g is the splitting ratio of the beam splitter,ρτ and ρs are the amplitude reflectivities of each of the mirrors, ∆z is the opticalpath length difference between the two arms of the interferometer, C is the corre-

4


Figure 1.2: Simplified layout of FT spectrometer, showing the source, sam-ple and the core interferometer. Note, although a transmission setup is shownhere, the interaction of the light with the sample may be transmission, specularreflection, or diffuse reflection [14].

lation function of the normalized optical amplitude entering from the sample, and“Re” denotes taking the real part. Note, differential polarization effects have beenignored in this treatment. By the Wiener-Khinchin theorem [15], C(∆z) is the in-verse Fourier transform of the two-sided optical power spectral density S(ν), whereν is the optical frequency. Therefore, the AC component of the detector signal in(1.1) can be inverse-Fourier transformed to obtain the optical power spectral den-sity. Although grating spectrometers can be ruggedized with much less effort dueto no moving parts, the multiplex and throughput advantages of FT instrumentsmake it attractive to pursue portable FT spectrometers. The multiplex advantageof the FT instrument arises because the energy of the light is not split into the dif-ferent wavelength channels. Each point in the interferogram has information fromall wavelengths of light. As a result, the signal to noise ratio of the photodetectedsignal is much higher, for a given scan time. The throughput advantage of FT sys-tems arises from the fact that no slits are present in a FT instrument to restrict thelight throughput of the system. In a grating based system, the spectral resolutionis inversely related to the slit-width. In other words, there is a trade-off betweensignal-to-noise ratio and the optical resolution. No such limitation exists for a FTinstrument. The tradeoff encountered by miniaturized FT spectrometers is in res-olution where, due to the shorter scan range of the miniaturized interferometer,the spectral resolution is degraded. However, the resolutions achieved by commer-cial FT spectrometers are sufficient for the targeted applications. Unfortunately,bulk-optic miniature FT spectrometer systems cannot become truly low cost. Thisis for two reasons. Firstly, the required componentry is very specialized and ex-pensive. Additionally, the expertise required to assemble, calibrate, and performmaintenance on one of these units is also significant.

5


1.3.2 Available systems

A highly miniaturized and highly ruggedized FTIR system is commercially avail-able in the Exoscan system from Agilent Technologies. At the core of the systemis a ruggedized ultra-miniaturized bulk-optic interferometer. The unit comprises ahand-held spectrometer with in-built source and sample presentation optics, and arear-mounted PDA for data processing/storage. The unit has a spectral measure-ment range of 2.5 µm–15.4 µm, and a spectral minimum spectral resolution of 15nm at 2.5 µm. The unit weighs 2.9 kg. Ocean Optics also produces a miniatur-ized FTIR spectrometer range named “ANIR” [16]. This range has three models,servicing the wavelength ranges 0.9–1.7 µm, 0.9-2.6 µm, and 2.0-4.5 µm. The spec-tral resolution of these devices at 2.5 µm is 9 nm. These are fibre-coupled models,weighing around 850g. This weight, however, does not include the source, probe,power supply, or processing hardware, which are individually separate components.Therefore, this system is not marketed as a system for single-hand operation. Thesystem is also not as rugged as the Agilent Exoscan.

1.4 Microelectromechanical systems

Microelectromechanical systems (MEMS) is a miniaturization technology which haspenetrated our daily life in some way or another. It adopts fabrication techniquesfrom the mature semiconductor microelectronics industry to create moving com-ponents on the scale of microns to millimeters. It has revolutionized the sensingand actuation domain by replacing bulky, heavy mechanical components with tiny,low priced, very controllable and precise components. The outstanding exampleof this are MEMS based accelerometers [17], which have rendered obsolete manyprevious technologies and are now enjoying unrivalled dominance in air bag sensorsautomobiles. MEMS are used as tactical grade gyroscopes and as systems for air-plane guidance, stability and pressure sensing during landing and takeoff. In theera of mobile computing, they have been integrated into millions of smartphonesand tablet computers as motion sensors [18].

1.5 MEMS spectrometers

1.5.1 Microelectromechanical systems in spectrometry

The application of MEMS technologies to spectrometry is a logical next-step inminiaturization. With bulk-optic devices nearing their limits in terms of how smallthey can be made, MEMS offers a new exciting opportunity for further miniaturiza-tion. Development of spectrometric tools using MEMS technologies has not takena single uniform pathway. In fact, numerous spectrometer designs have been inves-tigated and developed using MEMS. A number of these pathways will be discussed

6


Figure 1.3: Layout of standing-wave MEMS transform spectrometer [19].If and Id are, respectively, the forward (incident) and backward (reflected)intensities of light.

below. A key advantage of MEMS devices is that by leveraging the high-volumefabrication capabilities of the microelectronics industry, the unit costs of the MEMScomponents can be made exceedingly small, even down to the range of dollars peritem. This places MEMS devices in a special cost category because, if successful,they can truly be employed in a pervasive manner in any field of application.

1.5.2 MEMS based standing-wave spectrometers

This is a very elegant design, despite the fact that little work has been reportedin the field in the last few years. This in-line Fourier-transform spectrometer, pro-posed by Kung et al.[19] at Stanford University, works on the principle of samplinga portion of the interferogram, and generating the spectrum via transform tech-niques. Figure 1.3 gives a basic outline of the function of the device. Reportson these types of devices to date [20] indicate that the mirror scan range can beapproximately 30 µm. It is also envisaged that MEMS fabrication will allow themirror to be located at a distance of the order of 1 µm from the detector, althoughactual implementations have the mirror much further away.

The fact that the in-line Fourier-transform-based microspectrometer cannotsample the interferogram at zero path-length difference and that the detector re-quires finite thickness prevents this type of device being used with light that ishighly incoherent. As a result, such a device is unlikely to be of use in generalmulti or hyperspectral imaging. However, it may be able to be used to identifythe presence and wavelength of coherent illumination provided the illumination issustained for enough time that the mirror can be completely scanned.

Such devices have the ability to control the spectral resolution by changing the

7


distance over which the mirror is scanned. This may be used to reduce the timerequired to detect the presence of coherent radiation by only scanning over a smalldistance but with the mirror located a long way from the detector.

Fourier-transform-based microspectrometers do not make multiple passes (i.e. noetalon effects), the sensitivity of the system to mirror and actuation imperfectionsis reduced, as compared to a Fabry-Perot implementation. However, an accurateknowledge of the mirror position is required for this implementation to be of use.Unfortunately, little has been reported on this design since 2003.

1.5.3 Linear Variable Filters

Although these technically do not have a mechanical (i.e. moving) component,they have been included with MEMS devices because they are manufactured usingthe same technologies. Linear variable filters are a simple spectrometer technology,requiring no moving parts. The linear variable filter technology developed by theUniversity of Western Australia is shown in Figure 1.4 (a). This consists of a lineararray of fixed (non-tunable) Fabry-Perot filters, fabricated optically ahead of anarray of infrared photodiodes. The inter-mirror gap in the filters is created with anoptically transparent silicon monoxide material. To create a set of linear variablefilers, the inter-mirror gap spacing is graded across the array. It is evident fromFigure 1.4 (b) that the filters in the array produce a set of spectral transmissionpeaks at uniformly separated wavelengths. It is evident from the modeling andmeasurements that there is some scope for improvement of the actual devices, interms of spectral resolution and peak transmission.

The MicroNIR 1700 Spectrometer by JDSU [21] is a commercially available spec-trometer based on linear variable filters. This is a very portable USB powereddevice that requires minimal maintenance and weighs less than 60 g. It can operateover a wavelength range from 950 nm–1650 nm, has a spectral resolution of 1.5nm at 1000 nm, provides 128 points in its spectrum, and has typically a 0.25 smeasurement time.

1.5.4 MEMS FTIR

In order to benefit from the advantages of FTIR spectrometers (FT-spectrometers),there has also been a strong push to implement FTIR spectrometers in MEMS. Thevery large MEMFIS project [22] launched in September 2008, heavily funded by theEuropean Commission, and contributed to by a consortium of research institutesand industrial partners. The goal of the MEMFIS project is to develop a minia-turized FT-spectrometer technology, making use of MEMS/MOEMS technologies.Available information on the project demonstrates it has examined all aspects ofdeveloping the spectrometer, from MEMS fabrication, to packaging, electronics de-velopment, and system integration. In terms of the spectrometer design, variousoptions have been examined, including an ultra-miniaturized bulk-optic interfer-

8


(a)

MCT

Substrate

Detector

Fabry‐Pérot filterGe

SiO

(b)

Figure 1.4: (a) Design of a set of linear variable Fabry-Perot filters; and (b)modeled (lines) and measured (dots) performance of a fabricated set of linearvariable Fabry-Perot filters.

ometer as well as a MEMS option based on a Lamellar grating interferometer.Tortschanoff, et al. [23], from Carinthian Tech Research (CTR) in Austria, whichis part of the MEMFIS consortium, demonstrated a FT spectrometer based on aMEMS translatory mirror. Note, that the mirror of this system was MEMS based,while the rest was miniaturized bulk optics. Other research into MEMS FT spec-trometers includes the 2008 work of Yu et al. [24], who demonstrated a MEMSFT spectrometer based on an electromagnetically driven Lamellar grating. Thisarrangement produced an actuation of 63 µm with a electromagnetic drive cur-rent of 130 mA. The spectrometer successfully demonstrated measurement of twolaser lines at 633 nm and 523 nm. The spectral resolution of the spectrometer was3.8 nm and 3.4 nm, respectively, at the two wavelengths. In 2009, Deutsch et al.at Block MEMS LLC, demonstrated a Michelson interferometer built solely withMEMS technologies [25]. The beam splitter and mirrors in this system were 1 mm

9


Figure 1.5: Basic operating principle of Fabry-Perot filter.

in diameter. The beam splitter in this interferometer was designed to operate overthe 2–14 µm spectral range. This interferometer demonstrated a travel range of600 µm which, if the interferometer was built into a FT spectrometer, would yielda resolution of 8 cm-1 (20 nm at a wavelength of 2 µm).

1.5.5 Fabry-Perot spectrometers

MEMS-based Fabry-Perot filters are plane Fabry-Perot interferometers, consistingof mirrors which are lossless, perfectly parallel and flat, and have identical highreflectance, which together form a narrow bandpass resonating filter. Only thewavelengths that resonate within the cavity between the mirrors are transmitted,while all other wavelengths are reflected. The maximum transmittance is obtainedat the resonating wavelength, and transmittance falls off rapidly away from thiscritical wavelength. In effect, we obtain narrow bands of wavelength transmittedthrough the filter, known as the transmission peak. Each band has its own peakvalue at a certain wavelength, known as the pass wavelength. The order of eachtransmission peak refers to the number of half-wavelengths that fit into the cavity.

Figure 1.5 depicts the first and second order transmission peaks of an idealFabry-Perot filter, using spatially separated visible wavelengths for illustration. As

10


done in the third chapter.

Figure 1.1 : Illustration of changing the mirror separation in an ideal Fabry-Pérot filter, resulting in a shift in the wavelength of the transmission peaks, colors are used for the illustration purpose only.

Full width at half maximum signifies the width of transmission peak at the 50%

transmittance. Spectral resolution(ℜ) or the quality factor is one of the very important

parameter to quantify the quality of the FPF ,and is defined as :

0

Free spectral range

Wavelength tuning range

Maximum transmittance

Half maximum

Full-width at half-maximum

Initial position

Final position

d=400nm

d=650nm

Initial wavelength

Final wavelength

Actuation in top mirror

Tuning of output wavelength

Incident Spectrum

Wavelength (nm)

Tran

smitt

ance

(%)

Figure 1.6: Illustration of changing the mirror separation in an ideal Fabry-Perot filter, resulting in a shift in the wavelength of the transmission peaks.Colours are used for illustration purposes only.

the actuation mechanism of the MEMS-based Fabry-Perot filter changes the mirrorseparation, the transmitted wavelength can be controlled. The parameters usedfor spectral characterization of the filter are, the wavelength tuning range, the full-width at half-maximum (FWHM) of the transmission peak, the peak transmittance,and the out-of-band rejection. These characteristics are depicted in Figure 1.6,which also illustrates how changing the mirror separation tunes the pass wavelengthof the transmission peak. More elaboration on the spectral characterization will beprovided in Chapter 2.

Full width at half maximum signifies the width of the transmission peak at 50%transmittance of the peak transmittance. Spectral resolution or quality factor ofthe Fabry-Perot filter is defined as the ratio of the peak pass wavelength to FWHMof a transmission peak. The free spectral range is initial wavelength separationbetween orders, and is illustrated in Figure 1.6. In the absence of reflectance phasechanges, the mirror separation needs to be reduced to 50% of the initial mirror

11


Figure 1.7: Schematic diagram of Infratec Fabry-Perot microspectrometerconstruction, showing the mirrors and supports, suspension arms, actuationelectrodes, and the spacer [26].

separation to tune over the complete first order. Infratec has two commerciallyavailable Fabry-Perot microspectrometer devices, combining a bulk micromachinedFabry-Perot filter and a Pyroelectric detector. A cross section of the Infratec designis depicted in Figure 1.7 [26]. These two devices operate in the wavelength rangesof 3 µm to 4.1 µm and 3.9 µm to 5 µm, respectively. The spectral resolution ofthe filters is 80–100 nm, with a peak transmission of above 50%. The low tuningvoltage, of approximately 30 V, makes this system compatible with low voltagereadout technologies.

Axsun Technologies Inc., also produce a near infrared spectrometer using a Fabry-Perot MEMS filter. This consists of a bulk-optic top mirror, and a bulk-micromachinedbottom mirror fabricated on a silicon-on-insulator substrate. The unit produces aspectrum with a wavelength range from 1,350 nm to 1,800 nm, and a 0.1 cm-1

spectral resolution ( 0.2 nm at 1,500 nm). Only half of the Axsun device is MEMSfabricated, however, and this can restrict the cost reductions afforded by MEMStechnologies.

1.6 Development of MEMS based FP filters at

UWA

Figure 1.8 illustrates the first MEMS Fabry-Perot prototype tunable filter fabri-cated and demonstrated at The University of Western Australia for miniaturizationof hyper spectral imaging systems. In this case monolithic integration of a low-temperature fabricated MEMS filter and HgCdTe infrared detector technology hasbeen implemented and characterized [27]. The MEMS-based tunable optical filter,integrated with the infrared detector, is tunable in the wavelength range of 1.85 µmto 2.2 µm within the short-wavelength infrared (SWIR) region of the electromag-netic spectrum. The full range actuation voltage of this filter is 7.5 volt with fullwidth at half maximum (FWHM) of 95-105 nm.

In subsequent developments at The University of Western Australia, the design wasimproved and an array of Fabry-Perot filters was integrated with an array of infrareddetectors to create an array of on-pixel wavelength-tunable detectors [28, 29]. Anarray of such filters is shown in Figure 1.9. In this case by using MEMS actuators,

12


Figure 1.8: MEMS-based Fabry-Perot filter fabricated on top of detector[27].

the wavelength of transmitted light is selected and subsequently detected by theindividual detectors. Each filter is made up of a fixed dielectric mirror, whichalso acts as an isolation layer, a bottom electrode for the electrostatic actuationmechanism, and a moveable dielectric mirror suspended above the fixed mirror.The suspension arms are flexible, and fabricated from ultra-low stress silicon nitridebeams that also support the conductors and actuation electrodes. The mirrorsare composed of alternating layers of high refractive index (germanium) and lowrefractive index (silicon monoxide) materials. Actuation is achieved electrostaticallyby applying a voltage to the arms of the doubly-supported beam structures. Byusing doubly supported beams, the usual snap-down limitation of such structuresthat typically limits the wavelength tuning range can be extended to greater thanhalf of the relaxed separation between the top and bottom mirrors. This is criticalin achieving the theoretical maximum one-octave tuning range that can be achievedusing Fabry-Perot filters in first order.

Figure 1.10 shows the measured spectrum of two such filters designed to operatein two different wavelength bands. Since mirror separation is directly related to thetransmitted wavelength of the Fabry-Perot filter, scanning the top movable mirrortunes the microspectrometer wavelength. As evident in Figure 1.10, the spectralcharacteristics of the SWIR device include a 50 nm spectral resolution at a centrewavelength of 2 µm, and a spectral tuning range from 1.6 µm to 2.5 µm. This isachieved with a low drive voltage of 25 V. With appropriate changes to mirror layerthicknesses and optical cavity length, the spectral tuning range can also be designedto fit anywhere in the short-wave infrared (SWIR) and mid-wave infrared (MWIR)spectral range. Furthermore, the microspectrometer is robust, portable, and wellsuited to continuous operation.

Due to absorption in the Ge layers, Ge-SiO based filter designs have a lowestoperational wavelength of approximately 1.5 µm. Since a large amount of important

13


Figure 1.9: Illustration of the MWIR Fabry-Perot devices and SEM imageof array of filters with improved actuation structure [28, 29].

Figure 1.10: Measurements demonstrating operation of two sepa-rate microspectrometer devices in the shortwave infrared and mid-wave infrared spectral regions.

14


spectroscopic information is available in the visible (VIS) and near infrared (NIR)bands, it makes good sense to extend the operational wavelength range into thesebands. This limitation results from the strong absorption of germanium below1.5 µm. As such, microspectrometer operation in the VIS and NIR bands is notpossible using the existing Ge-SiO mirror technology. For this reason, an alternativematerial to germanium must be selected for the high-refractive index medium ofthe mirrors. Silicon can be an alternative material of choice. There are two keyadvantages to using silicon as the high index medium in the mirror. First, theabsorption characteristics of silicon allows operation of the mirrors well into thenear infrared NIR and VIS regions of the optical spectrum. Second, as silicon isalready used as a MEMS structural material, the need for silicon nitride in thesupport structures is eliminated. As a result, the fabrication process is simplifiedby the removal of one material system from the microspectrometer structure.

1.7 Applications and Spectral Requirements for

silicon based microspectrometers

The goal of this PhD was to optimize size, weight and power (SWaP) of spec-trometers and to develop materials and technologies for a silicon-air-silicon mirrorbased microspectrometer which can offer a higher spectral resolution and is oper-able at room temperature. In this work the prime focus was to fabricate siliconand air based mirrors and filters from visible wavelengths to short wave infraredwavelengths. In these wavelengths uncooled detectors are readily available in themarket hence the fabricated filters can be easily chip bonded with commercial detec-tors such as InGaAs for operating from visible wavelengths to short wave infraredwavelengths. This ensures the high portability and reduced size of silicon basedfilters.

As part of an extension to the upper limit of operating wavelengths, silicon basedfilters have also been fabricated to operate at mid-wave infrared (MWIR) wave-lengths. To operate silicon based filters at room temperature in the MWIR wave-length uncooled detectors such as triglycine sulfate (TGS) detectors can be used.The fabricated microspectrometer can be extensively used in SWIR-NIR-VIS spec-troscopy of soil for analysis of important soil attributes such as soil organic matter(SOM), minerals, texture, nutrients, water, pH, and heavy metals [1, 2]. Absorptionin the visible region 400-780 nm is primarily associated with minerals that containiron (e.g. haematite, goethite). The common molecules detected by SWIR spec-trometers are H2O, carbonates and NH4. Selected sulfates can be detected when anOH bond is present, as are phosphates and a variety of minerals in the soil such asclays, all pyllosilicates, carbonates, micas, chlorites (Mg, Fe, Al), garnets, silicateslike talc, topaz and scapolite [30]. The application of SWIR-NIR spectroscopy canbe extended to crop growth and tissue analysis to track damage of crops [31], wa-ter content analysis of crops [3], characteristics of tree wood [32], disease damageestimation in trees [4] and food quality analysis [33].

15


Figure 1.11: The plot shows the classic three most common ironminerals; jarosite, hematite and goethite [30].

1.8 Silicon based process development

1.8.1 Silicon thin film deposition method

At The University of Western Australia a silicon based process for fabricating mi-crosepctrometers did not exist. Hence, a large portion of the research in this thesiswas focused on creating a silicon thin film based technology for the fabrication ofmicrospectrometers. At The University of Western Australia we had tried threemethods for the deposition of silicon thin films: biased target ion beam deposition(BTIBD), electron beam (E-beam) evaporation, and inductively coupled plasmachemical vapor deposition (ICPCVD). We identified pros and cons of each of thesethree deposition methods in terms of optical and mechanical properties and envi-ronmental stability of the silicon thin films. It was found that silicon thin filmsdeposited from the BTIBD system had a refractive index higher than 4 for theNIR and SWIR wavelength range but a prohibitively high extinction coefficient.Through the experimental investigation, it was found that iron cross-contaminationin the BTIBD was affecting the optical properties of the silicon thin films. It wasalso found that the presence of iron contaminants rapidly increases the etch rateof the silicon films in 50% hydrofluoric acid (HF). A longer pre-deposition cleaningtime of the silicon target was attempted as a solution for removing the iron crosscontamination, yet only a marginal improvement in the properties of the silicon

16


thin films was observed. Due to the poor properties of silicon thin films depositedfrom the BTIBD system, E-beam evaporation method was attempted as the sec-ond choice for the silicon thin film deposition. In the E-beam 99.999% pure P-type3 mm-6 mm size silicon pieces were used as the source material. E-beam evaporatedsilicon thin films were highly environmentally stable and had negligible etch rate inHF. The silicon thin films were free from any cross contamination since only oneevaporation boat is exposed to the electron beam. It was found that silicon thinfilms deposited in the E-beam had a refractive index close to that of crystallinesilicon for the NIR and SWIR wavelength range but a higher extinction coefficientin the visible and NIR wavelength range. Since E-beam deposition is not conformal,another material such as silicon nitride is required as a structural material for theDBRs and filters. However, the stress in the silicon thin films deposited from theE-beam was highly tensile which would have resulted in structural failure in theDBRs and filters fabricated using e-beam silicon thin films. The only parameteravailable to control the stress in E-beam deposited silicon thin films was the depo-sition rate. However, the deposition rate window was very narrow and there wasvery little control on reproducibility of stress value, hence, this deposition methodwas abandoned. Towards the end of the first year of this PhD work, the SEN-TECH SI500D ICPCVD reactor was commissioned at The University of WesternAustralia. This new machine proved a turning point for this research work. TheSI500D provided many options for controlling the mechanical and optical propertiesof thin films, such as a change in the ICP power, deposition temperature, depositionpressure, substrate bias, and gas flow rates. The chemical vapor deposition can bea conformal deposition, thus, the need for an extra material for the structural layerof the DBRs and filters can be eliminated. Through the optimization of depositionparameters, high quality amorphous silicon thin films could be deposited. Thesefilms showed refractive index values close to the refractive index values of crystallinesilicon and a relatively lower extinction coefficient in the visible to MWIR wave-length range. The stress in the film could be controlled in-situ or post-deposition.Hence, throughout this thesis, ICPCVD technology is used for depositing siliconand silicon oxide thin films.

1.8.2 The structure of the DBRs and filters

The second major part of this research work was to identify MEMS structures,which are to be used for the silicon based DBRs and filters. At first the experimen-tal work involved fabrication of single layer freely suspended membranes. Theseare the fundamental MEMS structure to provide a very good idea of the effect ofstress on the bowing of the freely suspended membranes. With the help of thesestructures, pre-release and post-release annealing processes were developed. Theannealing process became a very crucial step in fabricating ultra-flat suspendedmembranes. It was identified that a conformal support around the periphery of thesuspended membrane gives it rigidity and a very high degree of surface flatness ascompared to other structures. As a demonstration vehicle for the optimization ofoptical properties, three layer silicon-silicon oxide-silicon based distributed Bragg

17


reflectors were fabricated for the visible (500-700 nm), near infrared (700-1000 nm)and short-wave infrared (2000-3000 nm) wavelength ranges using an optimized sili-con fabrication recipe. After the confirmation of optical properties of silicon throughthe DBRs and developing the fabrication process of silicon membrane based struc-tures, silicon-air-silicon based DBRs were fabricated from the visible to MWIRrange. One of the major contributions of this thesis is to develop this technology offabricating high optical quality DBRs with a low surface variation. This researchwork also demonstrated that the fabrication process of these DBRs enables scalingthe area of these DBRs from a few hundreds microns to several millimeters in scale.This work identifies the challenges in fabricating silicon based Fabry-Perot filtersusing traditional tether-beam based structures and shows it’s difficulty to fabricatesilicon based filters. This thesis proposes solutions to them by introducing a newnotch based actuation structure. Through these silicon based MEMS DBRs andfilters this work opens a pathway for the technology for the next generation siliconbased microspectrometers.

1.9 Thesis Outline and Scope

This PhD work presents many enhancements to the existing microspectrometertechnology in terms of improved structural design, tunable material propertiesand the demonstration of silicon based MEMS-based distributed Bragg reflectors(DBRs) capable of operating in the visible to MWIR wavelength range, a total op-erable spectral range spanning from 500-5000 nm. Finally, the thesis demonstratesa pathway towards fabrication of silicon-air-silicon mirror based Fabry-Perot filters.In the process of achieving this aim, contributions to scientific knowledge has beenmade in three key areas:

1. Optimization of residual stress and optical properties of inductively coupledplasma enhanced chemical vapor deposited silicon to ensure flat mirrors, nar-row FWHM and large wavelength tuning range.

2. Design of novel MEMS structures to simplify the fabrication process andminimize the influence of thin films stress gradients on the suspended mirrorflatness.

3. Finding strategies to fabricate large area suspended optical filters and mirrorsfor future focal plane array applications such as hyperspectral imaging.

The layout of this thesis, is as follows:Chapter 2 briefly introduces the theory behind the operation of Fabry-Perot filters.It also presents a comprehensive review of MEMS-based Fabry-Perot filters, withan emphasis on Si based DBRs and filters. A comparison of published optical trans-mission characteristics is presented.Chapter 3 gives a comparative analysis of possible candidate MEMS material to

18


build a microspectrometer and gives optical transmission modelling of silicon-air-silicon based ideal Fabry-Perot filters. It discusses optimization of residual stressand optical properties of inductively coupled plasma chemical vapor deposited(ICPCVD) silicon. The change in the properties of silicon by changing the de-position and annealing parameters is corroborated by FTIR studies. This chapteralso presents the optimization process of ICPCVD silicon oxide which is used as anadditional optical material for the mirrors and filters.Chapter 4 details the fabrication and optical characterization of silicon-air-siliconbased quarter wave mirrors at SWIR and MWIR wavelengths. This chapter alsopresents the fabrication process and optical characterization of large area suspendedfixed cavity mirrors and filters.Chapter 5 presents limitations of the tether-beam based structures in fabricatingsilicon based filters. It presents a novel notch based actuation structure for tunablefilters to overcome the limitation of the tether-beam structure. Subsequently, anetch back technique based fabrication process for ultra-thin suspended structuresis presented. Finally, this chapter presents the fabrication and optical characteri-zation of silicon based Fabry-Perot filters in the SWIR and MWIR spectral ranges.Chapter 6 presents the fabrication and optical characterization of silicon-air-siliconbased quarter wave mirrors. It also describes the fabrication and characterizationof fixed cavity silicon based Fabry-Perot filters at NIR wavelengths.Chapter 7 concludes the thesis by summarising this work, and proposes possibledirections for future research.

19

Chapter 2

Thin-film optics and Fabry-Perotfilters

2.1 Introduction

This chapter first introduces electromagnetic wave propagation theory which is thebasis of optical signal propagation and manipulation. Next, this chapter introducesoptical matrix modelling as a mathematical tool to analyse the reflectance, trans-mittance and phase change through a stack of thin films. Based on this theory,the optical characteristics of an ideal distributed Bragg reflector and Fabry-Perotfilter are presented. The chapter concludes with a review of published literature onMEMS-based tunable Fabry-Perot filters, which are either silicon based and/or useair as one of the mirror materials.

2.2 Maxwell’s equations

Light can be considered an electromagnetic wave phenomenon, and electromagneticphenomena can be described by Maxewell’s laws [34]. Consider a wave of lightthat is incident upon a planar interface between two homogeneous isotropic losslessdielectric media, as shown in Figure 2.1.

Light is incident on the interface from medium “1” with permittivity ε1, and per-meability µ1. On the other side of the interface is medium “2” with permittivity ε2,and permeability µ2. Using the fundamental Maxwell’s laws, a set of relationshipscan be derived which describe the behaviour of an electromagnetic wave at the inter-face between two dielectric media. These relationships can be considered Maxwell’slaw’s at a dielectric boundary, and describe the boundary conditions for the parallel(‖) and perpendicular (⊥) field components on either side of the interface

ε1E⊥1 = ε2E

⊥2 , (2.1)

20

CHAPTER 2. THIN-FILM OPTICS AND FABRY-PEROT FILTERS

Ei

Eiǁ Er

Et

y

x z

θ1 θ1

θ2

T

Erǁ

T

Etǁ

.

.

T

. kr

ki

kt

ε1, µ1, n1

ε2, µ2, n2

Figure 2.1: Reflection and refraction at an optical interface between the twolossless dielectric media of different refractive index. Incident (Ei, ki), reflected(Er, kr) and refracted (Et, kt) electric field are decomposed in orthogonal (E⊥)and parallel (E‖) components with respect to the plane of incidence.

B⊥1 = B⊥2 , (2.2)

E‖1 = E

‖2 , (2.3)

and

B‖1/µ1 = B

‖2/µ1. (2.4)

A monochromatic planar wave can be represented as Ei = E0,i ei(kirωit), where k is

wave vector , r is position vector and ω is angular frequency of wave. After incidenceon such a planar interface between two media, part of the wave is transmitted, andpart is reflected. Using Maxwell’s laws for this wave the reflection law can be givenas

θr = θi = θ1. (2.5)

The transmitted component also undergoes refraction, as described by Snell’s law,

n1 sin θ1 = n2 sin θ2. (2.6)

2.3 Fresnel coefficients for single planar interface

Snell’s law (Equation (2.6)) provides only a partial picture in that it only conveys arelationship between the incident and refracted angles. No information is provided

21


on the field magnitudes. Application of Maxwell’s laws at a dielectric boundaryallows a set of more detailed equations to be derived, including both field magnitudeand angle. These are called Fresnel’s coefficients, and represent the field amplitudetransmission and reflection coefficients for each polarization.

Assuming that E⊥ and E‖ are the orthogonal and parallel component E-field of theincident optical wave, the Fresnel coefficients for each polarization can be given as

r⊥ =Er,⊥Ei,⊥

=n1 cos θ1 − n2 cos θ2

n1 cos θ1 + n2 cos θ2

, (2.7)

t⊥ =Er,⊥Ei,⊥

=2n1 cos θ1


, (2.8)

r‖ =Er,‖Ei,‖

=n2 cos θ1 − n1 cos θ2


, (2.9)

tp =Er,‖Ei,‖

=2n1 cos θ2


. (2.10)

It is to be noted that t⊥ = 1 + r⊥ but t‖ 6= 1 + r‖. Since the incident and reflectedwave travels in the same medium the intensity (or power) reflectance R can be givenas

R⊥,‖ =E2r

E2i

= |r|2⊥,‖. (2.11)

The transmitted wave travels in two different media. Hence the intensity (or power)transmittance T is given as

T⊥,‖ =n2 cos θ2E

2t

n1 cos θ1E2i

=n2 cos θ2

n1 cos θ1

|t|2⊥,‖. (2.12)

2.4 Thin Film interference

As shown in Figure 2.2 a thin film has two optical interfaces, each with its char-acteristic Fresnel coefficients for reflection and transmission. An incident wave issplit by the two interfaces, as shown in Figure 2.2 into a multitude of coherentwaves. The total reflection from the thin film is the field (phasor) sum of all thefield amplitudes travelling back into medium 0, while the total transmission fromthe film is the field sum of all field amplitudes travelling forward into medium 2.The phasor sum of each optical amplitude depends on the phase accrued by thewave within the film, which in turn depends on the optical path difference ∆ of twoadjacent rays given by

∆ = 2dn1 cos θ1, (2.13)

where n1 is the refractive index in medium 1.

22


E0 E1 E2 E3 E4

t01t12 t01r12r10t12 t01 (r12r10)2t12

r01 t01r12t10 t01r12 r10 r12t10 t01r12 (r10 r12)2t10

t01r12 t01r12 r10 r12 t01r12 (r10 r12)2 t10

θ1

θ2

d

medium 0

n0

medium 1

n1

medium 2

n2

Figure 2.2: Thin film interference showing overall reflectance and transmit-tance of a thin film is a superposition of a multitude of beams with differentoptical paths.rpq and tpq correspond to the Fresnel coefficients at the opticalinterface between medium p and medium q. p denotes the medium of incidence(i.e. the direction) of the ray.

A corresponding phase shift arises, of φm = (m+1)(2π/λ)∆, where m is the numberof total internal reflections.

Figure 2.2 shows first reflected ray E1 and other reflected rays E2, E3, E4..... inmedium 0.

Since each reflection provides e−2iπ∆λ phase change, the Fresnel reflection coefficient

for the total reflected optical signal can be given as:

r =ErEi

= r01 + t01t10r12e−2iπ∆λ + t01t10r12e

−2iπ∆λ (r10r12e

−2iπ∆λ )

+ t01t10r12e−2iπ∆λ (r10r12e

−2iπ∆λ )2 + .........

(2.14)

r = −r10 + t01t10r12e−2iπ∆λ

k=∞∑k=0

(r10r12e

−2iπ∆λ

). (2.15)

r = −r10 + t01t10r12e−2iπ∆λ

k=∞∑k=0

(r10r12e

−2iπ∆λ

). (2.16)

For a lossless dielectric film, t01t10 =√

1− r201

√1− r2

10.

Hence, it can be shown that

r =r12e

−2iπ∆λ − r10

1− r10r12e−2iπ∆λ

. (2.17)

23


The above relation is equally valid for s and p polarization. Similarly the transmis-sion Fresnel coefficient for s and p polarization can be calculated as

t = t01t12 + t01t12r10r12e−2iπ∆λ + t01t12(r10r12e

−2iπ∆λ )2

+ t01t12(r10r12e−2iπ∆λ )3 + .....

(2.18)

t = t01t12

k=∞∑k=0

(r10r12e

−2iπ∆λ

)(2.19)

t =t01t12

1− r10r12e−2iπ∆λ

. (2.20)

Assuming the thin film is freely suspened in air (i.e. n0 = n2 = 1), it can be shownthat r10 = r12 and t10 = t12. The intensity (power) transmittance and reflectancecan be given as T = tt∗ and R = rr∗, respectively. The transmittance can then beexpanded as

T =T10T12

(1−R10)2 + 4R10 sin2 φ/2(2.21)

where T10 = t210, T12 = t212, R10 = r10r12 = r210 = r2

12 = R and φ = 2π∆λ

.

Since, t210 = (1− r210), t212 = (1− r2

12), equation 2.21 can written as

T =1

1 + 4R(1−R)2 sin2 φ/2

. (2.22)

Equation 2.22 is known as an Airy function. The maximum value of this function,Tmax = 1, occurs for φ = 2mπ or ∆ = 2mλ. Minima of this function occur atφ = (2m− 1)π or ∆ = (2m− 1)λ/2. The minimum values are given as

Tmin =(1−R)2

(1 +R)2. (2.23)

The normalized Airy function as a function φ is depicted in Figure 2.3. It is clearlyevident from Figure 2.3 that with increasing the reflectivity of thin film the outof band rejection significantly increases and the FWHM narrows. This proves theneed of high reflective layers for narrow FWHM and high out of band rejection.

The full-width at half-maximum (FWHM) of the Airy function, dφ/2 is obtainedfrom

1

2=

1

1 +Msin2(

12dφ

2

) (2.24)

with M = 4R(1−R2)

as

24


0

0.2

0.4

0.6

0.8

1

R = 0.4

R = 0.6

R = 0.8

R = 0.95

Tran

smitt

ance

FSR

FWHM

0 2 3

Figure 2.3: Airy function for different reflection coefficients R.

dφ

2≈ 2√

M=

1−R10√R10

. (2.25)

The free spectral range (FSR), i.e., ∆σ = σmax − σmin, between two successivemaxima of Airy function can be given as

∆φ

2= 2πn1d cos θ1∆σ = π (2.26)

∆σ =1

2n1d cos θ1

. (2.27)

Thus, it can be seen that even a single thin film can be wavelength selective.

2.5 Transfer matrix model

2.5.1 Transfer matrix modelling of a thin film

The optical transfer matrix model of a multilayer thin-film stack relates the electricand magnetic field at all the thin-film interfaces via a characteristics matrix. Detailsof the presented theory can be found in References [35, 36, 37]. The optical transfermatrix model has been developed for normal and angled incidence on transparentand weakly absorbing substrates. However, in this thesis only lossless substratesare considered, which simplifies the general modelling of mirrors and filters.

25


E01 Ei0

Em1

Er0

En1

H

E E

H

H

E

Ei1

n2

n1

Er1 E12

n0

d

y

x z

θ0 θ0

θ1

Rs

Ts

T2

T1

TFP

TFPS

R2, ϕr2

R1, ϕ-r1

d

Top Mirror

Bottom Mirror

Round Trip Phase change:

1 24

r rdπ ϕ ϕ

λ−Ψ = + −

Substrate

Figure 2.4: Propagation of optical wave through a thin film ofthickness d.

2.5.2 Formation of Transfer matrix

Figure 2.4 shows a ray of light incident on a film with refractive index n1. Em1 andEn1 show the left and right traveling waves, respectively, at the first interface ofthe thin film. Ei1 and Er1 shows the left and right traveling wave, respectively, atthe second interface of thin film. Free space is assumed as the medium of incidencewith refractive index n0. The permittivity and permeability of free space are εo =8.854 × 10−12 Fm−1, and µo = 4π × 10−7 Hm−1, respectively. The admittance offree space is given by

η0 =√ε0/µ0. (2.28)

Using Maxwell’s boundary conditions [38] the tangential electric fields E12 andcorresponding TE mode magnetic fields at the first and second interface H12 of thethin film can be give as

E01 = Em1 + En1, (2.29)

η0H01 = (Em1 − En1)γ, (2.30)

E12 = Ei1 + Er1, (2.31)

η0H12 = (Ei1 − Er1)γ, (2.32)

where γ = n1cosθ2. The left and right traveling wave, at the first and secondinterface, are related through their path difference as

Ei1 = Em1e−iδ, (2.33)

andEn1 = Er1e

−iδ, (2.34)

26


where δ is the path difference created by the propagation through the thin film andcan be given as

δ =2π

λn1d cos θ2. (2.35)

Using Equations 2.33 and 2.34, Equations 2.29 and 2.34 can be written as

E01 =Ei1e−iδ

+ Er1e−iδ, (2.36)

η0H01 =Ei1e−iδ− Er1e−iδγ. (2.37)

These Equations can be represented in a matrix form as[E01

η0H01

]=

[cos δ (i/γ ) sin δ

iγ sin δ cos δ

] [E12

η0H12

]. (2.38)

The matrix

M =

[p11 p12

p21 p22

]=

[cos δ (i/γ ) sin δ

iγ sin δ cos δ

](2.39)

is known as the transfer matrix.

2.5.3 Transfer matrix for a stack of thin films

Assuming that the thin films are arranged in a multilayer stack where each filmhas thickness tf , a refractive index nf and an extinction coefficient kf , the transfermatrix model reduces this entire assembly into a single pseudo-layer with a singletransfer matrix. The model assigns a characteristic matrix M to each layer at eachwavelength λ, and by multiplying these matrices together in the order of theiroccurrence in the stack, the optical transfer matrix of the entire stack is obtained.The characteristic matrix of layer f is

[Mf ] =

[cos δf (i/γf ) sin δf

iγf sin δf cos δf

]. (2.40)

where δf is the optical phase advance on a single traversal of the layer, expressedas

δf = 2πdf (nf − ikf )/λ. (2.41)

The wavelength λ represents the free space incident wavelength. The transfer matrixMtotal of the multilayer stack for light incident from free space can be obtained from[

E01

η0H01

]= [M1] [M2] · · ·

[Mq

] [ EN,N+1

η0HN,N+1

] [E01

η0H01

]= [Mtotal]

[EN,N+1

η0HN,N+1

](2.42)

and[E01

η0H01

]=

[Ei0 + Er0

γ0(Ei0 − Er0)

]= [Mtotal]

[Et

γt(Et)

]= [Mtotal]

[EN,N+1

η0HN,N+1

]. (2.43)

27


Si‐145 nm (tH=λo\4nH)

Air‐500 nm (tL =λo\4nL)

Si‐145nm (tH =λo\4nH)

Sapphire Substrate

Figure 2.5: Layer thicknesses of silicon-air-silicon based DBR at2000 nm design wavelength on sapphire substrate.

This can be rearranged to extract the reflection coefficient r and transmission co-efficient t for a stack of thin films, as[

1γ0

]+

[1−γ0

]r = [Mtotal]

[1γt

]t (2.44)[

1γ0

]+

[1−γ0

]r =

[p11total p12total

p21total p22total

] [1γt

]t. (2.45)

The reflection and transmission coefficients from this transfer matrix model can begiven as

r =γ0p11total + γ0γtp12total − p21total − γtp22total

γ0p11total + γ0γtp12total + p21total + γtp22total

(2.46)

and

t =2γ0

γ0p11total + γ0γtp12total + p21total + γtp22total

. (2.47)

2.5.4 Distributed Bragg Reflector

A distributed Bragg reflector (DBR) or quarter wave mirror (QWM), is a multilayerthin film stack which consists of a periodic sequence of high and low refractiveindex layers, each one-quarter wavelength thick. In other words, tH = λ0/4nH andtL = λ0/4nL, where λ0, is the central design wavelength and n is the refractive indexof the material used in the mirror. Figure 2.5 shows the structure of a silicon-air-silicon DBR on a sapphire substrate. The central design wavelength of this DBRis chosen as 2000 nm, and the thickness of each layer corresponding to the designwavelength. At λ0 the reflections from each interface add constructively to producea very high reflectance. A pair of high index, low index layer constitute a singleorder, and the number of such high-low pairs decide the final order of the DBR.For a quarter wave thick layer the path difference δ will be equal to π/2. Hence fornormal incidence the transfer matrix of the single quarter wave thick layer of lowrefractive index can be given as

M =

[0 i/nLinL 0

](2.48)

28


The transfer matrix for a combination of two quarter wave thick layers of higherand lower refractive index can be give as

M =

[−(nH/nL) 0

0 −(nL/nH)

](2.49)

and correspondingly for an N -period DBR with 2N layers the transfer matrix canbe given as

M =

[(−(nH/nL))N 0

0 (−(nL/nH))N

]. (2.50)

Using equation 2.46 the reflectance of the stack can be given as

R =

[(nH/nL)2N − 1

(nH/nL)2N + 1

]2

. (2.51)

Garmire [39] calculated that the highest reflectivity for an N -order DBR can beachieved by having the high refractive index layer as the first layer receiving theincident wave from free space. Garmire calculated that at the central design wave-length λ0 the reflectivity of a lossless N-period DBR deposited on a substrate withrefractive index nS with a high-index first layer is given by

R(N) =

(1− nS(nH/nL )2N

1 + nS(nH/nL )2N

)2

(2.52)

and for a N+1/2 period QWM,

R(N + 1/2) =

(nS − n2

H(nH/nL )2N

nS + n2H(nH/nL )2N

)2

. (2.53)

Equations (2.52) and (2.53) show that the reflectivity of a DBR is also dependenton the refractive index of the substrate. This refractive index determines the reflec-tivity of the interface between the last layer of the mirror and the substrate. Thetop mirror of a typical MEMS-based Fabry-Perot filter is usually suspended in airso that nS = 1, in which case the highest reflectivity is achieved with an N+1/2period mirror, compared to either an N+1 period mirror or an N-1 period mirror.

The design bandwidth of a lossless DBR, is the range of wavelengths where themirror reflectivity approaches 100% as N becomes large [40]. In a DBR maximumreflectivity is obtained at the central design wavelength of λ0, with the reflectiv-ity gradually falling as the wavelength moves away from this central wavelength.The DBR design bandwidth extends between short wavelength, λS, to a long wave-length, λL [35]. Figure 2.6 shows modeled reflectance for a silicon-air-silicon DBRof different orders on sapphire substrate for 2000 nm design wavelength. At thisdesign wavelength the refractive indexes of sapphire, silicon and air,are 1.76, 3.45and 1, respectively. As shown in Figure 2.6 as the order of DBR increases, the

29


0

20

40

60

80

100

1000 1500 2000 2500 3000 3500

1.5 - period

2.5 - period

3.5 - period

4.5 - period

Ref

lect

ance

(%)

Wavelength (nm)

DBR design bandwidth

λLs

Figure 2.6: The reflectance of silicon-air-silicon based DBR fordifferent DBR periods.

reflectance increases to 100%. This figure also illustrates the design bandwidth ofthe DBR between λS and λL.

λ0

λS− λ0

λL=

4

πsin−1

(nH − nLnH + nL

). (2.54)

The relationship of λ0 to λS and λL is given by Babic et al. [40] as

1

λ0

=1

2

(1

λS+

1

λL

). (2.55)

This relationship is useful when designing a DBR for a Fabry-Perot filter for aspecified wavelength tuning range.

The fractional bandwidth, ΛB can be derived from Equation (2.54);

ΛB =λL − λSλL

=4sin−1

(nH/nL−1nH/nL+1

)π + 2sin−1

(nH/nL−1nH/nL+1

) . (2.56)

In general, the fractional bandwidth is useful to directly compare the wavelengthtuning limitation imposed by the actuation mechanism with the finite mirror band-width.

In a DBR the reflectance phase change ϕ is zero at λ0, and is odd symmetric withinverse wavelength. Garmire used regions close to λ0 to calculate ϕ with respect to

30


wavelength [39]. The reflectance phase change in a DBR in the close range of λ0 isgiven as

ϕ = λ

(1− λ

λ0

)∂ϕ

∂λ. (2.57)

The reflectance phase change with respect to wavelength for both N period andN+1/2-period QWMs, is given as

∂ϕ

∂λ= −λC

λ2

π

(nH − nL). (2.58)

for large N or large nH/nL. Equations (2.57) and (2.58) are a good approximationsfor modelling the influence of nH and nL on the transmittance of Fabry-Perot filtersthat use DBRs.

2.6 Design of Fabry-Perot filters using DBRs

This section presents an analytical model to calculate the FWHM, FSR and crit-ical wavelengths at which the transmittance maxima occur. Figure 2.7 shows thevariables which are going to be used in deriving the transmittance characteristicsof Fabry-Perot filters. T1, T2, TFP and TFPS shows the transmittance through topmirror, bottom mirror, air cavity, and substrate, respectively. R1 and R2 representreflectance from the top and bottom mirror, respectively. The cavity material is as-sumed to be free space (n = 1) in all calculations. Each reflected wave goes througha phase change.

The round-trip phase change, Ψ, is given by [41, 42]

Ψ =4πd

λ− ϕ−r1 + ϕr2 (2.59)

Ψ =4πd

λ− ϕ (2.60)

where ϕ = −ϕ−r1 + ϕr2 represents the sum of the reflectance phase changes frombottom and top mirrors, and the negative sign in ϕ−r1 is added for consistency withthe calculation of phase. At phase

Ψ = 2πm (2.61)

maximum transmittance occurs, where m is the order of the maximum. The trans-mittance for the filters is given in [41] and can be expressed as

TFP (Ψ) =

T 2

(1−R)2

1 +4F 2R

π2 sin2(

Ψ2

) (2.62)

31


Rs

Ts

T2

T1

TFP

TFPS

R2, ϕr2

R1, ϕ-r1

d

Mirror1

Mirror2

Round Trip Phase change:

1 24

r rdπ ϕ ϕ

λ−Ψ = + −

Substrate

Figure 2.7: Definition of terms used in the text for an ideal Fabry-Perot

cavity. The reflectance is drawn at an angle for clarity; only normal incidence

light is considered in this thesis.

TFP (Ψ) =Tmax

1 +4F 2R

π2 sin2(

Ψ2

) (2.63)

where Tmax is the peak transmittance.

The reflectance finesse of the filter FR, is given by

FR =π√R

1−R. (2.64)

In the above equations R and T represent reflectance and transmittance of the filterand can be given as

R =√R−1 R2; (2.65)

T =√T1T2. (2.66)

FWHM (δλ) is associated with the TF through the relationship [43]

δλ =λ

FR(m+ ϕ

2π+ λ

2π∂ϕ∂λ

) . (2.67)

Atherton [43] reported the free spectral range ∆λ between order m and order m+1as

∆λ =λ(

m+ 1 + ϕ2π

+ λ2π

∂ϕ∂λ

) . (2.68)

The above mentioned modelling equations are suitable for ideal Fabry-Perot filterswith no imperfections in the mirrors.

32


Imperfections in a mirror of a filter reduce the peak transmittance and increase theFWHM [28]. The increase in FWHM results in the finesse of the filter decreasing,from FE = FR:

1

F 2E

=1

F 2R

+1

F 2D

(2.69)

where FD is termed the imperfection finesse. A detailed discussion on Fabry-Perotfilters with imperfections can be found in References [42, 43, 44, 45].

The equation for the FWHM (δλ) and free spectral range (∆λ) of a Fabry-Perotfilter can be simplified using equations (2.57) and (2.58), and the term

ϕ/2π + (∂ϕ/∂λ)× λ/2π (2.70)

can be reduced to

ϕ

2π+

λ

2π

∂ϕ

∂λ=

1

2 (nH − nL). (2.71)

The approximate FWHM, δλ, and the free spectral range, ∆λ, can be obtained bysubstituting the simplified Equation (2.71) in equations (2.67) and (2.68). In thissubstitution the phase terms must be doubled, since equation (2.71) represents thecontribution from a single mirror. The resulting equations are

δλ =λ

FE

(m+ 1

nH−nL

) ; (2.72)

and

∆λ =λ(

m+ 1 + 1nH−nL

) . (2.73)

These two equations give the direct relationship between the mirror material pa-rameters, and δλ and ∆λ. The effective finesse FE is used here instead of thereflectance finesse FR, since equation (2.67) is valid with FE or FR, but FE is usedfor the generalized result.

In the presence of imperfections in the mirrors, the FR = FD criterion can be appliedby targeting a particular mirror reflectivity. This provides the analytical expressionof FWHM to be achieved. Setting FR = FD in Equation (2.69) leads to an effectivefinesse of

FE = FR/√

2 FE = FR/√

2 (2.74)

and combining this equation with the equation for the FWHM, (Equation (2.67))and the analytical expression of reflectance finesse (Equation (2.64)), the practicaldesign equation for the Fabry-Perot filter can be given as

δλ =

√2 (1−R)λ

π√R(m+ 1

nH−nL

) . (2.75)

33


For a given FWHM this equation yields the desired mirror reflectivity, R, which inturn can be used to determine the required number of layers in the mirrors usingEquation (2.53).

In an electrostatically tunable MEMS-based Fabry-Perot filter, actuators changethe mirror separation of a filter from dL to dS and the transmittance peak movesfrom a long wavelength of λL to a short wavelength of λS. The percentage travelrange of the mirrors is then given by

D =dL − dSdL

; (2.76)

and the percentage wavelength tuning range is

Λ =λL − λSλL

. (2.77)

2.7 Anti-reflective coating

An anti-reflective coating (ARC) is a single quarter wave thick layer or a stack ofquarter wave thick layers which is aimed to reduce the reflection from the substrateor surface on which it is applied. A single layer of quarter wave thick dielectricmaterial is the simplest form of ARC. As shown in Equation 2.48 the transfermatrix of such layer is given as:

M =

[0 i/nLinL 0

]. (2.78)

From Equation 2.46 the corresponding reflection coefficient can be give as

r =nS − nL2

nS + nL2. (2.79)

The reflection coefficient will be zero when nL =√nS. However in practice this con-

dition is difficult to achieve, hence there will be some reflections from the substrate.Figure 2.8 shows the reflectance from a glass substrate before and after applicationof quarter wave thick layer of MgF2 as the ARC. The application of ARC reducesthe minimum reflectance by 2%. As shown in Fig. 2.8 a single layer ARC has abroadband response. By increasing the number of dielectric layers a narrow bandlow reflectance ARC can be achieved.

2.8 Review of Fabry-Perot Filter Technologies

This section reviews published literature on MEMS-based tunable Fabry-Perot fil-ters, which are either silicon based and/or use air as one of the mirror materials.

34


4.5

5

5.5

6

6.5

7

7.5

8

300 400 500 600 700 800 900 1000

Glass with no ARCGlass with ARC

Ref

lect

ance

(%)

Wavelength (nm)

Figure 2.8: Reflectance of glass substrate with and without ARC.

It also provides examples of Fabry-Perot filters which use silicon based oxides ornitrides as structural materials. This chapter highlights the actuation technolo-gies, the mirror structures, the achieved spectral characteristics and the differenttechnological challenges faced in order to fabricate these Fabry-Perot filters.

2.8.1 Introduction

Fabry-Perot filters reported in the literature vary in the material used in the reflec-tors (mirrors), the structure and size of the filter, the actuation mechanism, opticalcharacteristics and applications. Some researchers have utilized III-V semiconduc-tors; others preferred silicon. With the rapid advancement of the semiconductor in-dustry, micro machining of silicon and associated nitrides and oxides is now matureand inexpensive. Thus, silicon based MEMS devices have dominated the industryand are now deployed in millions of instruments. However, each approach offersa number of innovative features and this review will attempt to highlight some ofthese innovative features.

2.8.2 Fabry-Perot filter technologies

The optical characteristics of a Fabry-Perot filter are a function of the mirror ma-terials and design. As discussed in chapter 1, the ideal mirrors for a symmetricFabry-Perot filter are identical, lossless, completely smooth, perfectly parallel, andfree of curvature. In order to achieve these characteristics some researchers haveused metal mirror based technologies [46, 47]. Metal mirrors are simple to fabricateand metals such as aluminum, gold and silver, are deposited on a lossless dielectricusing sputtering or evaporation [48]. However, maximum transmission at resonance

35


is limited by absorption in the metal. Since more metal layers are needed to im-prove reflectivity, the thickness of the metal layer determines the trade-off betweenFWHM and maximum transmission in these filters [49].

The solution to the absorption vs spectral resolution trade-off problem of metalmirrors is to use dielectric mirrors, where a high reflectivity can be obtained byconstructive interference of the optical signal in a half wavelength cavity formedbetween two dielectric mirrors. The dielectric mirrors can be fabricated as dis-tributed Bragg reflectors to achieve constructive interference by the multiple layersof dielectric materials. In a distributed Bragg reflector, quarter wavelength thick (orother odd multiple of quarter wavelength thickness), high refractive index dielectricmaterial layers are separated by a quarter wavelength thick low refractive indexdielectric material. A high reflectance can be obtained over a range of wavelengthsthat define the mirror bandwidth.

Compound semiconductor based filters and mirrors

Using the above described concept a horizontally moving Fabry-Perot filter wasreported by Datta et al. [50] using InP and air as the mirror material . This filerwas monolithically integrated on an InP substrate, and reported to have tunabilityof 400 nm, covering the coarse wavelength division multiplexing (WDM) spectralrange. It was actuated to it’s full range using less than 7 V .

Figure 2.9: SEM image of the side of a Fabry-Perot filter with InP-Airmirrors, the air gaps in the mirrors are one-quarter-wave thickness, while theInP layers are three-quarter-wave thickness. Taken from Reference [51]

.

InP and air were further used as mirror materials by Hasse et al. [51]. With WDMapplication in mind, this vertical cavity filter was reported to have a continuoustuning range of 221 nm with an actuation voltage of 28 V. An SEM cross sectionof this filter is shown in Figure 2.9.

36


Figure 2.10: The left side image shows the SEM of InP-Air mirror basedfilter. The right side images show a schematic of the vertical cross sectionof the practical InP/air optoelectro-mechanical stack for a demonstrator. Allthicknesses are given in optical units. Taken from Reference [52].

As shown in Figure 2.10, Garrigues et al. [52] chose InP as the optical cavitymaterial, and five and seven quarter wave thick InP layers were used in the mirrors.A tuning range of 60 nm with a supply voltage of 6 V was achieved. Another groupreported building a tunable InP-air filter where wavelength tuning was achieved byelectrostatic actuation of two DBRs which were p-doped and n-doped, respectively.The cavity and the DBRs consisted of a stress compensated InP/airgap structurewhich was fabricated by wet etching of an InGaAs sacrificial layer. The authorsaddressed the need for stress tuning of the InP layers by changing the ratio of Asand Ga in the InGaAs sacrificial layer.

For an InP/air based filter an extended tuning range of 127 nm was attained byDaleiden et al. [53]. The authors designed filters with variable cavity lengths anda number of suspensions for the main membrane. The optimized suspension lengthwas found to be 30 µm. This enhanced tuning range was the result of strain tuningin the InP layers. However, the width of membrane was only 40 µm. The measuredFWHM of the filter resonance was between 3.7 nm and 7.7 nm.

Spisser et al. [54] used InP and air as the mirror material to create a tunablefilter centred at 1.57 µm. In order to give good mechanical strength to InP theykept the thickness of the InP layers to seven to nine quarter wavelengths. InGaAswas used as the sacrificial layer, which was wet etched to form the air cavities.The structure was grown by low-pressure metal-organic chemical vapor deposition(LP-MOCVD) lattice matched to the InP <001> substrate. The tuning range ofthe filter was reported as 62 nm with 14 V maximum actuation voltage and theFWHM was measured as 0.4 nm. The authors identified the problem of residual

37


stress in the InP layers when using longer actuating arms and suggested use ofhighly controllable epitaxially grown InP as a solution for it.

Ochoa et al. [55] utilized the good optical properties of AlGaAs based materials andthe good mechanical properties of silicon to fabricate a hybrid MEMS filter via flip-chip bonding an AlGaAs-based distributed Bragg reflector to a polysilicon MEMS.This electrostatically actuated structure had a built-in gold reflector. They usedSU-8 bond pads to flip-chip bond the DBR to the actuation structures. The SU-8provided enough flexibility and chemical resistance required for the wet release ofthe sacrificial layers. The device demonstrated 53 nm (936.5-989.5 nm) of resonantwavelength tuning over the electrostatic actuation voltage range of 0-10 V.

Kim et al. [56] used GaAs and AlGaAs based DBRs to fabricate electrothermallyactuated filters. The maximum voltage for 69 nm wavelength tuning using a 5.2 µmthick filter structure was 1.7 V, and a high tuning efficiency of 75 nm/mW wasreported. The filter was designed at 1538 nm wavelength.

Amano et al. [57] fabricated a GaAlAs-GaAs thermally tunable vertical-cavityfilter with a thermal strain control layer and heating element on a micromachinedcantilever structure. They demonstrated a large negative temperature dependenceof 0.79 nm/K for the vertical-cavity filter by using a GaAs strain control layer. Thetuning voltage was as low as 4.9 V for wavelength tuning of 18.1 nm. With theproposed structure they could exhibit both rise time and fall time below 100 µs.

In a similar work, Quack et al. [58] fabricated a mid-infrared filter with integrateddetectors that were sensitive in a tunable narrow spectral band. A Fabry-Perotcavity was formed by two DBRs made of IV-VI narrow gap semiconductor layers.Using electrostatically actuated MEMS micromirrors, a very compact tunable de-tector system was fabricated. Mirror movements of more than 3 µm at 30 V wasobtained. With these mirrors, detectors with a wavelength tuning range of about0.7 µm were realized.

A tunable optical filter using a torsional micromechanical structure was presentedby Mateus et al. [59]. The device had a continuous tuning range over 100 nm, from1500 to 1600 nm, and the structure eliminated the potential catastrophic dischargeproblem associated with electrostatic micromachines. This device was used witha 1550 nm vertical-cavity surface-emitting laser to directly modulate the laser andachieve a 2.5-Gb/s data rate.The travel range for the filter was 100 nm. The authorsused AlGaAs based DBRs having 99.4% mirror reflectivity.

Silicon, silicon nitride and silicon oxide based filters and mirrors

Despite these many successes in compound semiconductor-air based mirror tech-nology, the deposition of compound semiconductor mirrors and sacrificial layers isvery expensive as it uses molecular beam epitaxy (MBE) and metal organic chemicalvapor deposition (MOCVD) systems. Stress tuning of these materials requires veryspecific considerations as well. As a result, mainstream MEMS fabrication relies

38


Figure 2.11: SEM micrographs of fully processed filter structures. On theleft is shown a Fabry-Perot filter with six membranes, four suspensions andcorresponding supporting posts. The image on the right gives insight into thedetails of the InP/airgap structure within the released filter area. Taken fromReference [53].

mostly on mature silicon technology, which with low cost deposition processes,produces inexpensive MEMS devices.

Considering this, it is not a surprise that the first MEMS-based tunable Fabry-Perotfilter was built using bulk micromachining of silicon [60]. Many research groupshave fabricated Fabry-Perot filters based on silicon or its oxide and/or nitride.Irmer et al. [61] presented a low-cost surface micromachining approach based onplasma-enhanced chemical vapour deposited (PECVD) Si3N4 and Si layers. Thefilms were stress-optimized, and reactive ion etching was used to pattern the filter,with wet chemical etching of the sacrificial silicon layers using KOH to create thefinal suspended structures. The surface micromachined filter consisted of two DBRmirrors, each having five 590 nm thick Si3N4 membranes separated by 390 nm wideair gaps. With the 710 nm distance between the mirrors (cavity) the filter shownin Figure 2.12 was designed for a center wavelength of 1490 nm. The full width athalf maximum (FWHM) of the filter was reported as 1.5 nm.

LPCVD poly-Si and air were used as the long wave infrared mirror and filter mate-rial to operate in the 7 µm to 12 µm spectral range [62]. In [63] a bulk micromachinedvertical FPF was realized by Saadany et al. on silicon on insulator (SOI) for SWIRwavelengths, FWHM of the order of 1.2 nm was demonstrated. This lower FWHMwas the result of using two quarter wave Si-air stacks for the left and right mirrorsof the filter. However, the relative tuning range was limited to 0.1% owing to useof an electrostatic parallel plate actuator and the WDM application requirements.In a similar way Pruessner et al. [64] fabricated an in-plane Fabry-Perot filter atSWIR wavelengths which achieved a FWHM of 0.38 nm, which is the lowest everreported in the SWIR region. This was used for detection of thermal noise wheresuch a low FWHM was a pre-requisite of the application. Another SOI based in-plane Si-air-Si Fabry-Perot filter with reported FWHM and finesse of 3 nm and513, respectively, and was targeted to be employed in densely multiplexed optical

39


Figure 2.12: SEM image of Si3N4 -air mirror based filter showing verticalcross section of the practical silicon nitride/air stack. Taken from Reference[61].

communications in the wavelength range of 1530 nm–1560 nm [65].

There have been many attempts to integrate filters with detectors where the lowerdeposition temperature of silicon based PECVD materials is of particular benefit.One such attempt was by Keating et al. [66]. In their work a dielectric mirrorstack made of Ge and SiO was deposited using electron beam deposition at roomtemperature. The structural layers and suspensions were made of silicon nitride,which was deposited at 100 ◦C. This low temperature allowed them to monolithi-cally integrate the filter to a mercury cadmium telluride (MCT) detector. The filterwas designed to operate from 1.7 µm-2.5 µm, with just 9 V tuning voltage.

Cristea etal. [47] presented a MEMS voltage tunable Fabry-Perot interferometerstructure with a silicon p-n photodiode in order to obtain a tunable optical sen-sor. The Fabry-Perot interferometer was used as a voltage tunable filter for theinput radiation or as a voltage controlled attenuator to regulate the light from amonochromatic source. The top mirror of the Fabry-Perot cavity was a Au/SiO2

movable membrane, formed by anisotropic etching of <111> oriented Si wafers.The air-silicon surface acted as the lower mirror. The designed peak wavelengthwas 790 nm with a FWHM of 5 nm, 73 V actuation voltage, although the tuningrange achieved was limited to 54 nm.

By a combination of bulk and surface micromachining Russin et al. [67] fabricatedDBRs and filters using a Si, SiO2 stack for NIR wavelengths. They reported 97.3%reflectivity of DBRs at 1550 nm, compared to 99.1% for an ideal dielectric stack.The Fabry-Perot filter constructed using these mirrors had a measured finesse of115, a natural spectral line width of 35 nm, and a free spectral range of 165 nm.They used arrays and smaller mirror elements to decrease overall mirror stress andcurvature while maximizing mirror aperture area. Such a mirror array is shown inFigure 2.13.

40


Figure 2.13: IR transmission images of a 4 × 4 mirror array (left) and a 14× 14 mirror array (right). Taken from Reference [67].

In Ref [68], Tuohiniemi et al. reported a surface-MEMS tunable Fabry-Perot filterfor thermal infrared applications. The transmission-peak FWHM was 140 nm,which indicates a finesse of 37. The transmission wavelength was shown to betunable up to 30 % below the zero-voltage wavelength. Maximum tuning range wasachieved by an applied voltage of 27 V. The surface-micromachined tunable Fabry-Perot interferometer was designed for the thermal infrared spectral range of 7-12 µmwavelength. Winchester et al. [46] used PECVD silicon nitride as the structurallayer for MEMS structures, and tuned the intrinsic stress through variations inthe PECVD deposition parameters. A MEMS-based Fabry-Perot cavity was thusfabricated using PECVD silicon nitride as the membrane layer with ZnS as thesacrificial material. The fabricated devices had an initial 1 µm cavity length. Theycould attain 240 nm displacement across a 380 µm membrane span for an appliedbias of only 1.6 V. Noro et al. [69] used LPCVD deposited poly-Si and siliconnitride as the mirror material to fabricate a MWIR range filter. They used simpleannealing to tune the stress in the poly-Si layer from a highly compressive value totensile. The filter was integrated with an IR source and a wideband pass filter tofinally yield a gas sensor. Note that the use of CVD deposited materials providesa low cost method to tune the stress and optical properties to obtain the desiredperformance.

Taking advantage of the good optical and mechanical properties of crystalline Si,Lee et al. [70] demonstrated a low-voltage magnetic actuation based tunable opticalfilter, having a wide linear tuning range. They fabricated and tested three differ-ent types of spring structures (straight, corrugated, and meander) to optimize theproposed magnetic actuation for high linearity and low power consumption. Thisnovel approach increased the linear tuning range to over 200 nm in wavelength.The fabricated tunable filter was measured to have a maximum static power con-sumption of less than 25 W for wavelength tuning up to 200 nm with an externalmagnetic field of 0.28 T. Unlike electrostatic force, the direction of the Lorentz forceis dependent on the direction of the applied current, and so bi-directional tuningwas possible.

41


Halbritter et al. [71] fabricated Si3N4 and SiO2 based DBRs and used an elec-trothermal tunable strategy to tune the filter. The filter exhibited a full widthhalf maximum (FWHM) of 0.16 nm and a free spectral range (FSR) of 34 nm. AGaussian lens was coupled with the filter. Due to the use of Si based material theFabry-Perot filter was a low cost solution for WDM applications. A surface micro-machined electrothermally actuated filter for WDM applications was fabricated byHohlfeld et.al. [72]. Focusing on a 1575 nm center wavelength, they used LPCVDpolycrystalline silicon and silicon oxide to fabricate DBR mirrors. A range of 5.3 nmwas covered using external heating with a Peltier element with a temperature dif-ference of 85 K. The tuning efficiency was measured to be 0.071 nm K−1. Thesame group extended their work to fabricate a Fabry-Perot filter centered around1530 nm and integrated with a fiber optic cable. The improved design had a tuningefficiency up to 113 pm K−1. A tuning range of 3.5 nm was measured throughexternal temperature modulation [73].

Electrothermal tuning was combined with bulk micromachining of <100> Siliconto fabricate a Fabry-Perot interferometer for WDM application [74]. The actuatorwas able to displace 1 µm with 15 oC temperature difference. The authors observedthat convection currents affect the actuation temperature range. Their researchunderscores one challenge faced by electrothermal actuation.

Ebermann et al. [26] fabricated a Fabry-Perot filter with bulk micromachiningtechnology. Two optical and mechanical structured wafers were bonded together byan intermediate SU-8 layer. The stiffness of the optimized thick reflector carriersallowed them to realize a comparatively large aperture of 1.9 mm and a finesseof 40–60. The reflectors were made from a double-stack of silicon dioxide andpolycrystalline silicon (n=3.33), and a broad high reflective zone was achieved forMWIR applications. Figure 2.14 shows the fabricated filter and its schematic. Thefilter was used in the 3 to 5 µm wavelength range with reported FWHM of 80-100 nm.

The same group has reported a tunable dual band Fabry-Perot filter which can beused in the LWIR and MWIR wavelength bands [75]. The designed spectral rangeswere from 8 to 10.5 µm and from 4 to 5 µm. The maximum control voltage was 41 V.The FWHM was measured as 115 nm to 200 nm at LWIR wavelengths, and 57 nmto 97 nm at MWIR wavelengths. At LWIR wavelengths peak transmission was 75%and at MWIR wavelengths peak transmission was 80%. The dielectric stack mirrorswere made of Ge, ZnS and a stress compensating layer. With stress compensationthe filter had an extremely low curvature of less than 15 nm over an aperture of 1.9mm diameter. Figure 2.15 shows the fabricated dual band Fabry-Perot filter.

The idea of wafer bonding two reflectors was utilized by Stupar et al. [76] tofabricated a MEMS Fabry-Perot filter array that is capable of tuning over the7.9-11.8 µm spectral range. They solved the challenge of providing antireflectivecoating and at the same time protecting the thin actuating membrane by using twosided processing on a SOI wafer. The two-sided process consists of a processed SOIdevice layer that is transferred onto a processed silicon substrate using a Au-Au

42


(a) (b)

Figure 2.14: Spring suspended Bulk micromachined FP filter. Left sideshows schemetic and right hand side shows SEM image. Taken from Reference[26].

Figure 2.15: Dual band Fabry-Perot filter. Taken from Reference [75].

thermocompression bond. The achieved FWHM was 128 nm.

Lipson et al. [65] fabricated an electrostatically tunable optical bandpass filter on<110> silicon. Deep reactive-ion etching was the main process used to fabricatethe overall device structure. Wet etching of the vertical <111> planes was carriedout using KOH to create highly parallel surfaces that were needed for the photonicband gap elements. Back etching was used to release the moving parts. Authors

43


Figure 2.16: 1-D DBR based filter on Si substrate. Taken from Reference[65] .

attached fiber pigtails in etched alignment grooves. The fiber-fiber insertion losswas measured to be below 11 dB. The measured passband width was 3 nm with atuning range of 8 nm. The target application for this filter was wavelength divisionmultiplexing. Figure 2.16 shows the SEM image of such a fabricated filter.

Jerman et al. [77] used bulk micromachining of silicon to fabricate a Fabry-Perotcavity. The authors deposited dielectric mirrors with alternating layers of high andlow refractive index, on two separate silicon wafers. The mirrors had a reflectiv-ity which ranged from 95% at 1.3 µm to 97.5% at 1.55 µm. The electrodes weredeposited on these two wafers and finally, they were wafer bonded to form theFabry-Perot cavity. At a wavelength of 1.3 µm, the measured FWHM optical band-width and the free spectral range of the device were 0.9 nm and 38 nm, respectively.At a wavelength of 1.5 µm, the FWHM bandwidth was 1.2 nm with a free spectralrange of 49 nm. The peak wavelength transmission of the devices was around 45%.With the application of 0-70 V, on a pair of control electrodes, the device couldbe tuned over the full free spectral range. Bulk micromachining of Si was usedby Sihua Li et al. [78] to create a thermally tunable Fabry-Perot interferometer.They used a boron doped silicon resistor as the heater and thermal isolation wasensured using a cantilever structure. Tunability was realized by changing the re-fractive index of the bulk silicon cavity through heating the resistor. They achievedmaximum wavelength tuning of about 17 nm with a driving voltage less than 14 V.The filter linewidth was measured as 0.29 nm, with a measured FSR of 21.9 nm.The authors reported less than 50 ms heating and cooling response times . In Ref[79] a micromachined tunable Fabry-Perot filter was fabricated using the anodicbonding method. The filter consisted of an SOI wafer and a glass substrate, andhad multilayer dielectric mirrors. The wafer and substrate were bonded togetherusing the anodic bonding method leaving an optical cavity between them. The filterwas electrostatically tunable by applying voltage to the SOI wafer and electrodeson the glass substrate. The achieved FWHM and FSR were around 0.5 nm and

44


Figure 2.17: SEM photos of Fabry-Perot cavity with Bragg mirrors ofcylindrical shape. Taken from Reference [80].

35 nm, respectively, and the driving voltage was 29.1 V.

Malak et al. [80] used a novel idea of designing a cylindrical shaped resonant cavityas shown in Figure 2.17. They combined bulk micromachined multilayered silicon-air Bragg mirrors of a fiber rod lens (FRL) to achieve a quality factor close to 9000.Design dimensions were chosen to keep the cavity stable. This 1-D resonator hada large cavity with a cavity length of 250 µm. The reported FWHM of the devicewas 0.1765 nm while FSR was limited to less than 4 nm.

In conclusion, it can be said that from a structural perspective Fabry-Perot filterscan be fabricated to be horizontal cavity 1-D filters or vertical cavity filters. While1-D filters can be applied to very specific applications such as WDM, the verticalcavity filter is more versatile and can be used in many applications such as WDM,spectroscopy and imaging. Among all actuation mechanism, electrostatic actuationis clearly the leading mechanism used by researchers. Although III-V semicon-ductor based filters offer high finesse, technological and economical considerationsrender silicon based filters as the preferred choice for commercial applications. Sincethe work presented in this thesis targets spectroscopic application from the visibleto MWIR wavelength range, the focus has been on silicon based vertical cavity,electrostatically actuatable Fabry-Perot filters.

45


2.9 Summary

This chapter optical matrix modelling as a mathematical tool to analyse the re-flectance, transmittance and phase change through a stack of thin films. This isthe basis of designing and optimizing optical filters and mirrors. Next this chapterpresented a comprehensive review of published MEMS-based Fabry-Perot filters,including the actuator types, mirror types and spectral characteristics of devicesdescribed in the literature. A specific attention was given for the silicon basedelectrostatically actuatable Fabry-Perot filters employing air as one of the cavitymaterial.

46

Chapter 3

Material selection and depositionprocess development

3.1 Introduction

This chapter initially presents the selection of materials for the DBRs and filters.Next, it details on the three methods for silicon thin film deposition attempted inthis research work. Finally, it presents the characterization of silicon oxide films.

3.2 Material Selection

Compound semiconductors, such as InGaN, InAlP, GaAs, AlGaInP and InP havebeen used extensively for realizing DBRs and filters operating from the visible toshort-wave infrared wavelengths (SWIR) [77, 81, 82, 83, 52]. Deposition of com-pound semiconductors generally requires either molecular beam epitaxy (MBE) ormetal organic chemical vapor deposition (MOCVD). Both these methods are high-cost processes in both set up and running costs, and tuning of the stress of thesematerials requires precise control of the ratio of elements in the deposited materi-als. Due to these issues, compound semiconductor based materials have not beenadopted in mainstream MEMS manufacturing, which has moved to lower-cost thinfilm deposition processes.

Silicon and silicon-based nitrides and oxides are widely used as alternatives tocompound semiconductor materials in optical MEMS technology, and are com-mon in bulk micromachined and surface micromachined MEMS based DBRs andfilters operating from SWIR wavelengths to long-wave infrared wavelengths (LWIR)[65, 77, 84, 63, 62]. Silicon has attractive mechanical properties, relatively inexpen-sive processing technologies, and a range of thin film stress tuning methods. Theseadvantages have made silicon the mainstream MEMS material [85, 86, 87]. For de-positing silicon based materials, many standard chemical vapor deposition (CVD)

47

CHAPTER 3. MATERIAL SELECTION AND DEPOSITION PROCESSDEVELOPMENT

Table 3.1: Optical refractive index, n, extinction coefficient, k, quarter wave

thickness, t, and absorption, A, in a quarter wave thick layer, of various ma-

terials at visible (560 nm), NIR (1000 nm) and SWIR (2000 nm) wavelengths

[92]. “ngl” stands for negligible.

Material 560 nm 1000 nm

n k t (nm) A (%) n k t (nm) A (%)

Si 4.06 0.038 34.8 1 3.67 4.0×10-4 70 1 × 10-4

Ge 5.38 2.049 26.7 29 4.62 1.7 × 10-1 56 4InP 3.63 0.380 38.0 19 3.32 1.1× 10-5 75 5 × 10-4

GaAs 4.01 0.276 26.7 10 3.51 1.3 × 10-4 71 nglSiNx 2.0 10-6 67.0 ngl 1.99 10-6 125 ngl

Ta2O5 2.15 10-6 62.0 ngl 2.09 ngl 120 ngl

Material 2000 nm

n k t (nm) A (%)

Si 3.45 1.5 × 10-6 145 nglGe 4.12 1.4 × 10-6 122 nglInP 3.13 1.1 × 10-5 160 ngl

GaAs 3.34 1.3 × 10-4 150 nglSiNx 1.99 10-6 250 ngl

Ta2O5 2.09 ngl 240 ngl

processes such as atmospheric pressure chemical vapor deposition (APCVD), lowpressure chemical vapor deposition (LPCVD), plasma enhanced CVD (PECVD)and high density plasma CVD (HDPECVD) and inductively coupled plasma CVD(ICPCVD) have been used. The mechanical and optical properties of the depositedfilms are generally optimized using deposition parameters or subsequent annealing[88, 89, 90, 91].

In order to consider the suitability of various materials for optical MEMS appli-cations extending from visible, through the NIR and to the SWIR bands, it isuseful to consider the data in Table 3.1, which lists the optical constants of somecommonly used dielectrics for optical MEMS [92]. Also shown are correspondingcalculated absorption in a quarter wavelength thick optical layer for wavelengthsof 560 nm, 1000 nm and 2000 nm. It is to be noted that SiNx and Ta2O5 havenegligible absorption at these wavelengths. However, due to their lower refractiveindex in comparison to other materials shown in Table 3.1, creating a high finesseFabry-Perot filter using these two materials would require a five-layer mirror stackfor each of the top and bottom DBR of the filter. Although this increases the re-flectivity of the mirror, it comes at the cost of reduction in the free spectral rangeand a more complex fabrication process. InP, GaAs and Ge have high refractiveindices, but their absorption in the visible wavelength range is prohibitively high. Incomparison to InP, GaAs, and Ge, silicon has a relatively low extinction coefficientwith a relatively high refractive index in the visible to SWIR wavelength range.

Silicon is commonly the material of choice for MEMS processes. There are two key

48


advantages of using silicon as the high index medium in the mirror. First, the ab-sorption characteristics of silicon allows operation of the mirrors well into the visibleand NIR regions of the optical spectrum. Second, as silicon is already commonlyused as a MEMS structural material, the need for an additional material in the sup-port structures is eliminated and the fabrication process is significantly simplified.Since the earlier Fabry-Perot filter technology developed at The University of West-ern Australia used Ge as the high refractive index material, it is worth comparingGe with silicon. The value of silicon as opposed to germanium in the visible andNIR spectral regions is illustrated in Figure 3.1, which plots the absorption loss inan isolated quarter–wave layer of the high index medium in the mirror. The highindex medium is considered to be either silicon or germanium. The significance ofFigure 3.1, as opposed to simply examining the extinction coefficient of germaniumor silicon, is that Figure 3.1 takes into account the effect of multiple coherent reflec-tions from the two interfaces of the thin films, and the refractive–index dependenceof the number of passes made by the light through the film. It is evident fromFigure 3.1 that as the mirror centre wavelength is shifted to shorter wavelengths,the absorption of the layer increases. However, this increase is different for the twomaterials. In germanium, the absorption cuts on around 1550 nm, and increasesalmost exponentially towards shorter wavelengths. In silicon, the absorption cutson around 1000 nm, but increases approximately linearly towards shorter wave-lengths. In other words, for silicon, in the spectral region of interest, the thicknessreduction required for shorter wavelength operation sufficiently keeps track with theabsorption coefficient increase, thus preventing a sharp increase in layer absorption.

(a) (b)

0

2

4

6

8

10

12

14

16

500 1000 1500 2000

Abso

rptio

n at

pea

k (%

)

Wavelength (nm)

0

0.2

0.4

0.6

0.8

1

1.2

500 1000 1500 2000

Abso

rptio

n at

pea

k (%

)

Wavelength (nm)

Figure 3.1: Modelled absorption from a quarter-wave thick layer of (a) a

quarter–wave thick layer of germanium; and (b) a quarter–wave thick layer

of silicon. Note, these models used refractive index and absorption data for

crystalline materials, from the Handbook of Optical Constants of Solids by

Edward D. Palik [92].

The explanation for the absorption behavior in Figure 3.1 lies in the band structureof the two materials as shown in Figure 3.2. The shortest direct band transitionin germanium has an energy of 0.8 eV [93], which corresponds to a wavelength of

49


1550 nm. Thus, germanium will begin to absorb at wavelengths shorter than thisvalue. In comparison, the shortest direct band transition in silicon, in the absenceof phonon coupling, has an energy of 3.4 eV [93], which corresponds to a wavelengthof 365 nm. However, because the density of states in silicon is so high, phonon–assisted transitions can also take place, albeit with low efficiency. These indirecttransitions correspond to an energy of 1.12 eV [93], or wavelengths below 1110 nm.In other words, germanium will be an efficient (strong) absorber below wavelengthsof 1550 nm, while silicon will be an inefficient (week) absorber below wavelengthsof 1110 nm and an efficient (strong) absorber below wavelengths of 365 nm.

(a) (b)

Figure 3.2: Band gap of semiconductor materials (a) germanium; and (b)

silicon.

Therefore, a mirror using silicon as the high-index medium will allow operation wellinto the visible region of the optical spectrum. Fabricating a surface micro machinedsilicon based reflector entails a multitude of challenges, including optimization ofthe mechanical properties of the silicon to allow for its use as a structural sup-port material, and optimization of the optical properties to allow use as an opticalmaterial.

As silicon has a lower refractive index than germanium, it will degrade the reflectiv-ity of the mirror if no other change is made to the mirror. Despite this degradation,the mirror will be capable of operating down to a wavelength of 600 nm or shorter.Silicon oxide and air, with refractive indices of 1.44 and 1, respectively, have beenconsidered here as low index mirror media.

By selecting a low refractive index material as the second dielectric layer for theDBRs, a large refractive index contrast between the two dielectric layers can beachieved, thus resulting in a very high finesse Fabry-Perot filter. In this way, siliconcan be used as the high-index layer to fabricate Fabry-Perot filters operating fromvisible to SWIR wavelengths. However, note that the deposited silicon thin filmsmust meet a range of criteria, including low tensile stress, low surface roughness,high refractive index, and low extinction coefficient. Satisfying all of these condi-tions poses a significant challenge in using deposited silicon films for this purpose.

50


3.2.1 Modelling studies

Optical modelling of the DBRs and filters was undertaken to ascertain that theoptical properties of materials are optimum for the fabrication of DBRs and filters,operating from visible to MWIR wavelength range. The theoretical models used inthe simulation have been explained in the previous chapter of this thesis. Since theadvantages of silicon over other materials have been highlighted in earlier sections,the optical modelling was extensively focused on silicon based DBRs and filters. Thenext two section describes the simulated spectra of DBRs and filters, which givetheoretical estimations of reflectivity and bandwidth of DBRs, and transmittance,FWHM, and FSR of filters.

3.2.2 Simulation of DBRs

As discussed in the previous chapter, for increasing reflectivity of the mirrors, theFWHM of a Fabry-Perot filter becomes narrower and the extinction (or out-of-band rejection) of the filter improves. As given by Equations (2.52) and (2.53), fora given substrate or incidence medium, the reflectivity of a quarter wave mirror fora given order is a function of the refractive indices of the DBR layer materials. Itcan also be concluded that for a given order by increasing the difference betweenthe high refractive index nH and low refractive index nL, the reflectivity of themirror can be increased. Figure 3.3 plots the modelled reflectivity of the four firstorder quarter wavelength mirrors built with different material combinations. Therefractive indices of materials are taken from the Handbook of Optical Constants[92]. As can be seen in Fig. 3.3 in the SWIR wavelength (1500 nm – 2500 nm)range germanium-air-germanium (Ge-Air-Ge) mirrors have the highest reflectivitydue to high contrast between the refractive index of Ge and air. The next highestreflectivity is indicated by silicon-air-silicon (Si-Air-Si) mirrors followed by silicon-silicon dioxide- silicon (Si-SiO2-Si) and germanium-silicon monooxide-germanium(Ge-SiO-Ge). Note that the material systems included in this simulation are theoptical thin film materials that are readily available within the University of WesternAustralia micro-fabrication facility. This thesis is focused on Si-air-Si DBRs as highreflectivity and wide bandwidth substitutions for the Ge-SiO-Ge based DBRs,whichwere used in earlier work at The University of Western Australia.

3.2.3 Simulation of silicon based filters

No natural substance has a lower refractive index than air. Replacing silicon monox-ide with air, as the low-index medium in a 3-layer DBR, greatly improves the reflec-tivity of the mirror. Modelling indicates that using a 3-layer MEMS mirror madeof silicon and air will result in a Fabry-Perot filter with a spectral resolution of ap-proximately 20 nm at a centre wavelength of 2000 nm; an improvement by a factorof 2.5 over the previous germanium/silicon-monoxide designs undertaken at UWA[27, 28]. Similarly, modelling indicates that using a 3-layer MEMS mirror made of

51


0

20

40

60

80

100

1000 1500 2000 2500 3000 3500

Ge-SiO-GeSi-Air-SiGe-Air-GeSi-SiO

2-Si

Ref

lect

ivity

(%)

Wavelength (nm)

Figure 3.3: Modelled reflectivity of QWMs made of different dielectric ma-

terials.

Table 3.2: Layer thickness and cavity gap used in modelling Fabry-Perotfilters.

DBR Layers Wavelength Range Substrate Layer thickness Tunable cavity-gap(nm) (nm) (nm)

Ta2O5-MgF2-Ta2O5 Visible Fused Silica 84-130-84 350Si-SiOx-Si Visible Fused Silica 46-121-46 350Si-SiOx-Si NIR Fused Silica 100-241-100 700Si-SiOx-Si SWIR Silicon 185-430-185 1250Si-SiOx-Si MWIR Silicon 365-865-365 2500Si-air-Si Visible Fused Silica 46-175-46 350Si-air-Si NIR Fused Silica 100-350-100 700Si-air-Si SWIR Silicon 185-625-185 1250Si-air-Si MWIR Silicon 365-1250-365 2500

silicon and silicon oxide will result in a Fabry-Perot filters with a spectral resolutionof approximately 30 nm at a centre wavelength of 2000 nm; an improvement by afactor of 2. Modelled optical transmission for all three types of filter are shown inFigure 3.4. Note that peak positions have been offset on the wavelength-axis forclarity. These simulations prove that silicon can be used as the high refractive indexmaterial to fabricate DBRs and filters. Table 3.2 shows the thicknesses of the layersof materials and initial tunable air cavity gap used in the filter simulations.

Visible wavelength filters

Three types of filters are modelled in the visible wavelength range. The filters arebased on Ta2O5-MgF2-Ta2O5, Si-SiOx-Si and Si-air-Si DBRs. As shown in Figure3.5 the Ta2O5-MgF2-Ta2O5 based filter shows 95% peak transmittance. However,the FWHM and FSR are limited to 50 nm and 100 nm, respectively. The Si-SiOx-SiDBR based filter shows 57-70% transmittance and approximately 10 nm FWHM.This filter shows higher transmittance than the Si-air-Si filter, and comparableFWHM and FSR. Hence, in the visible wavelength range Si-SiOx-Si based filter are

52


Table 3.3: Summary of FWHM, FSR and Transmittance of modelled Fabry-Perot filters.

DBR Material Wavelength Range FWHM FSR Transmittance(nm) (nm) (nm) (%)

Ta2O5-MgF2-Ta2O5 Visible 50 100 95Si-SiOx-Si Visible 10 160 57-70Si-SiOx-Si NIR 20 400 90-92Si-SiOx-Si SWIR 30 800 94-95Si-SiOx-Si MWIR 110 1500 76-85Si-air-Si Visible 10 175 40-55Si-air-Si NIR 15 450 85-90Si-air-Si SWIR 20 840 84-95Si-air-Si MWIR 60 1600 74-83

0

20

40

60

80

100

1500 2000 2500

Ge-SiO-GeSi-Air-SiSi-SiO

2-Si

Tran

smitt

ance

(%)

Wavelength (nm)

Figure 3.4: Transmittance of Fabry-Perot filter consisting of DBRs of dif-

ferent dielectric materials. Note that there is an offset of 50 nm among the

plots for clarity.

a good choice for meeting the high finesse and moderate transmittance requirements.

Near infrared wavelength filters

In the NIR wavelength range, filter based on Si-SiOx-Si DBRs shows 90-95% trans-mittance, 20 nm FWHM and 400 nm FSR (see Figure 3.6). The Si-air-Si DBRsbased filter shows 15 nm FWHM and 450 nm FSR. Thus, filters based on Si-air-SiDBRs exhibit narrower FWHM and wider FSR than filters based on Si-SiOx-SiDBRs. Thus, in the NIR wavelength range, both silicon based filters can be usedfor high performance.

53


0

20

40

60

80

100

500 550 600 650 700 750 800 850

Gap - 250 nmGap - 275 nmGap - 300 nm

Gap - 325 nmGap - 350 nm

Tran

smitt

ance

(%)

Wavelength (nm)

(a)

0

20

40

60

80

500 550 600 650 700 750

Gap - 225 nm Gap - 250 nm

Gap - 300 nmGap - 350 nm

Tran

smitt

ance

(%)

Wavelength (nm)

(b)

0

10

20

30

40

50

60

70

500 550 600 650 700 750

Gap - 200 nm

Gap - 250 nm

Gap - 300 nm

Gap - 350 nm

Tran

smitt

ance

(%)

Wavelength (nm)

(c)

Figure 3.5: Transmittance plot of Fabry-Perot filters in visible wavelength;

(a) Ta2O5-MgF2-Ta2O5 DBR based filter; (b) Si-SiOx-Si DBR based filter; (c)

Si-air-Si DBR based filter.

Short-wave infrared wavelength filters

Figure 3.7 shows that in the SWIR wavelength range Si-SiOx-Si and Si-air-Si DBRsbased filters show 95% peak transmittance. Si-SiOx-Si DBRs based filter shows54


0

20

40

60

80

100

900 1000 1100 1200 1300 1400 1500

Gap-400 nmGap-500 nm

Gap-600 nmGap-700 nm

Tran

smitt

ance

(%)

Wavelength (nm)

(a)

0

20

40

60

80

100

900 1000 1100 1200 1300 1400 1500

Gap-400 nm

Gap-500 nm

Gap-600 nm

Gap-700 nm

Tran

smitt

ance

(%)

Wavelength (nm)

(b)

Figure 3.6: Transmittance plot of Fabry-Perot filters in NIR wavelength;

(a) Si-SiOx-Si DBR based filter; (b) Si-air-Si DBR based filter.

30 nm FWHM and 800 nm FSR. The Si-air-Si DBRs based filter shows 20 nmFWHM and 850 nm FSR. Hence, in SWIR wavelength range both silicon basedfilters have high optical performance.

Mid-wave infrared wavelength filters

In the MWIR wavelength range Si-SiOx-Si and Si-air-Si DBRs based filters show83% and 85% peak transmittance, respectively. Si-SiOx-Si DBRs based filter shows110 nm FWHM and 1500 nm FSR. The Si-air-Si DBRs based filter shows 60 nmFWHM and 1600 nm FSR. Hence, in MWIR wavelength range Si-air-Si DBRsbased filters provide a far better FWHM and FSR than Si-SiOx-Si DBR basedfilters. Table 3.3 Summarizes these optical modelling results.

55


0

20

40

60

80

100

1600 1800 2000 2200 2400 2600



Tran

smitt

ance

(%)

Wavelength (nm)

(a)

0

20

40

60

80

100

1600 1800 2000 2200 2400 2600 2800

Gap - 650 nmGap -750 nmGap - 850 nmGap - 950 nm


Tran

smitt

ance

(%)

Wavelength (nm)

(b)

Figure 3.7: Transmittance plot of Fabry-Perot filters in SWIR wavelength;


3.2.4 Final material choice

From the optical simulations of transmittance response of filters, it can be concludedthat silicon based filters are suitable for applications from the visible to MWIRwavelength range. In the visible wavelength range tantalum oxide and magnesiumfluoride based filter shows acceptable response, fabrication of this filter requireseither e-beam or thermal evaporation of tantalum oxide and magnesium fluoride.Additionally, use of this material for optical MEMS structures will require an extrastructural material such as silicon nitride. On the other hand, silicon is the mostprominently used semiconductor for MEMS fabrication and has mature fabricationprocess technologies, making it a cheap solution for high volume MEMS devicefabrication, in addition to that, silicon itself can work as the structural material forthe optical MEMS devices, thus eliminating need of an extra structural material.As an added advantage, silicon thin film deposition processes provide a large choiceof variation of deposition parameters for optimization of optical and mechanical

56


0

20

40

60

80

100

3500 4000 4500 5000

Gap - 1300 nmGap - 1500 nmGap - 1700 nmGap - 1900 nm


Tran

smitt

ance

(%)

Wavelength (nm)

(a)

0

20

40

60

80

100

3500 4000 4500 5000

Gap - 1300 nmGap - 1500 nmGap - 1700 nmGap - 1900 nm


Tran

smitt

ance

(%)

Wavelength (nm)

(b)

Figure 3.8: Transmittance plot of Fabry-Perot filters in MWIR wavelength;


57


qualities of silicon thin films. Considering these advantages, this thesis work willfocus on fabrication and characterization of silicon based DBRs and filters.

3.3 Materials deposition process development

With the modelling work complete and materials selected, we next have to develop asuitable method of fabricating thin films of these materials. Given the available toolsin our microfabrication facility, three material deposition methods were seen to havepotential: (1) biased target ion-beam deposition, (2) electron-beam evaporation,and (3) plasma enhanced chemical vapour deposition. However, as no silicon processhad previously been developed on these tools, a process development exercise wasneeded. The remainder of this chapter follows a systematic process to ascertain thebest tool for process development, and development of the process on that tool.

3.3.1 Instruments and Methodologies

The investigated thin films were deposited on three substrates: (1) 320 µm thick<100> oriented silicon substrates for film thickness measurements and FTIR spec-troscopic analysis; (2) 70 µm thick <100> silicon wafers for stress characterization;and (3) 280 µm thick sapphire substrates for characterization of optical proper-ties. Huang et al. [94, 95] had given many other methods to measure rmechanicalproperties of thin films such as nanoindentation method, micrometric cantileverbeams,cross-membrane structures and mechanical simulation. This thesis will adoptsubstrate bowing method to estimate residual stress in thin films. A Zygo Newviewoptical surface profiler was used for measuring radius of curvature of 70 µm thicksubstrates before and after the deposition of thin films. Stoney’s formula was usedto extract the residual stress in silicon thin films from stress induced bowing of the70 µm thick substrates [96, 97, 98].

An in-house bench-top optical transmission measurement system was used to deter-mine the optical transmission spectra of the layers on sapphire substrates at roomtemperature. This system allows single point optical measurement across a wave-length range of 550 nm to 2200 nm. We used Cauchy’s dispersion equations in theNKD Matl software to extract the refractive index and the extinction coefficientof the silicon thin films from the measured optical data [99]. Back side reflectionsfrom the sapphire substrate were included in our calculations. The thfickness of thedeposited films was measured with a Dektak surface profiler.

Fourier transform infrared (FTIR) spectroscopy was carried out using a PerkinElmer Spectrum One FTIR spectrometer. The hydrogen concentration in the siliconfilms was calculated using a standard method published in the literature [100, 101],as follows. In a thin film the concentration of the oscillating species is proportionalto the integrated intensity of the absorption band. In amorphous silicon the peakaround 640 cm -1 corresponds to rocking mode vibrations of the Si and hydrogen

58


bond, and peaks near 2000 cm -1 and 2100 cm -1 correspond to the stretching modevibration of the monohydride bond (Si-H) and the stretching mode of dihydridebond (Si-H2), respectively [102]. The hydrogen bond concentration, NH , is thengiven by

NH = AI = A

∫α(ω)

ωdω, (3.1)

where α is the absorption coefficient, ω is the angular frequency, A is a propor-tionality constant (oscillator strength), and I is the integrated intensity over theGaussian fit curve of interest. The Si-H rocking mode has a peak at 640 cm-1

and its bond concentration can be calculated using NH = A640I640, with oscillatorstrength A640 of 1.6 × 1019 cm-2 . The total hydrogen concentration, measured bythe integrated intensity of the peak at 640 cm-1 wavenumber, is the best measureof hydrogen concentration, since other peaks are less reliable [103, 104, 105]. There-fore, in this work the peak at 640 cm-1 wavenumbers is used to calculate hydrogenconcentration. However, other peaks at 2100 cm-1 and 2000 cm-1 wavenumber areused to extract vital bonding information regarding the nature of silicon hydrogenbonding in the amorphous silicon thin film.

3.3.2 Biased target ion beam deposition

Biased target ion beam deposition (BTIBD) is a reactive sputtering technique,where an ion source is used to provide low-energy ions that are accelerated via anelectrostatic potential toward a set of biased targets to generate sputtering. BTIBDcan simultaneously sputter from up to four targets for thin film deposition. Theduty cycle of the applied bias controls the rate of sputtering from each target.BTIBD systems have been successfully used for depositing magnetic and mageto-optic materials [106, 107, 108].

A BTIBD system from the 4Wave inc., USA, is used at The University of WesternAustralia. A schematic diagram of a typical BTIBD system is shown in Figure3.9. A low-energy argon ion source is directed at the negatively biased sputteringtargets. The bias voltage and the duty cycle are used to control the deposition ratefrom each target. Used over multiple targets, these two parameters allow controlof the composition and other parameters of thin films. It was observed that, for afixed negative bias of -800 V, varying the pulsewidth alone for each target providessufficient control over deposition rate. The frequency of the pulse cycle used is 71.43kHz. For each 14 µs cycle, the sputtering “on-time” is 3 µs. The plasma sheaththat develops on the surface of the negatively biased targets accelerates positiveplasma ions toward the targets and results in sputtering of the target material.If not accelerated, the low-energy ions do not possess sufficient energy to producesputtering. This system also allows controlled flow of oxygen and argon gases inthe plasma chamber. The oxygen gas can be used to form oxide thin films.

The initial aim of this research was to examine the applicability of a newly-installed4Wave BTIBD system for fabrication of silicon based optical MEMS structures. The

59


Figure 3.9: Schematic diagram of a three target BTIBD system (courtesy:4WAVE Inc USA.)

BTIBD system was already generating high quality oxide thin films, that promptedus to assess its suitability for deposition of high quality silicon based optical films.Since BTIBD is not a conformal deposition technique, we progressed on this pathwith the intention of using silicon nitride as a structural material.

For assessing suitability of BTIBD silicon, a number of deposition trials were con-ducted on sapphire substrates. The deposited thin films were used to assess theoptical performance of the material, in terms of refractive index (n) and absorp-tion coefficient (k). A Woollam VASE Ellipsometer was used to characterize theoptical constants. The measured optical constants are shown in Figure 3.10. TheBTIBD silicon shows a refractive index close to 4.24 and an extinction coefficientclose to 0.7 in the SWIR wavelengths. The extinction coefficient of the silicon thinfilms steeply increases towards shorter wavelengths. In the visible wavelength range(450-700 nm) the extinction coefficient is prohibitively high (1–3). A very high ex-tinction coefficient indicates that BTIBD silicon thin films were highly absorbingand unsuitable for fabricating DBRs and filters in the visible to SWIR wavelengthrange.

In an attempt to reduce the absorption coefficient, amorphous BTIBD depositedsilicon thin films were annealed at 600 ◦Cfor 30 minutes in a quartz tube furnace inAr ambient. It was observed that the texture of the smooth film became very rough.Figure 3.11 shows a microscopic image of the film after annealing. It can be clearlyseen that there is formation of micro crystallites, several microns in height, whichincreases the roughness of the film making it impractical for optical applications.

Figure 3.12 shows an Xray-Diffraction (XRD) spectrum of the annealed silicon thin

60


Figure 3.10: Refractive index and extinction coefficent of silicon thin filmdeposited using BTIBD system.)

Figure 3.11: A microgrph of formation of micro-crystallites on annealedsilicon film.

film obtained via grazing angle XRD using a PANalytical EMPYREAN system.In Figure 3.12 the smooth blue curve indicates the peak matching with the XRDmaterial database, the smooth green curve is a fit to the XRD spectrum of BTIBD Si, and the blue spikes that coincide with the peaks in the blue curve indicate XRDpeaks. There are also some orange spikes above the plot frame which indicateswhere the peaks were identified by the analytical software. The 2θ peaks at 28.4 ◦,47.3 ◦, 56.1 ◦, corresponds to <110>, <220> and <311> oriented micro-crystallinestructures in the annealed silicon thin film.

To determine the quality of deposited and annealed silicon films a basic etch ratetest of as deposited and annealed silicon films was carried out in hydrofluoric acid(HF). It is well known that silicon has a negligible etch rate in HF. However, theBTIBD silicon films were quickly dissolved in HF. Given this high susceptibilityto etching, and the high optical absorption, it was suspected that the depositionprocess resulted in metallic contamination from the other targets. Further analysisusing Energy Dispersive X-ray Spectroscopy (EDS) indicated the presence of ironin the films (see Fig.3.13 and Table 3.4).

It is possible that cleaning of the chamber could have removed the metal contam-

61


Count

3000

1000

0

Position ( 2 θ ) 20 30 40 50 60

Figure 3.12: XRD plot of the BTIBD depoisted silicon film which is an-nealed at 600 ◦Cfor 30 minutes.

Figure 3.13: EDS spectra of the silicon film deposited from BTIBD system.

ination. However, the physical geometry of the BTIBD machine required a verylong and impractical cleaning time. Hence, it was decided to find an alternative toBTIBD for depositing amorphous silicon thin films.

62


Table 3.4: Elemental analysis of BTIBD deposited Si thin filmsusing EDS.

Element W (%) Error (%)

Silicon 96.35 3.23Argon 3.35 .15Iron 0.30 .04

Figure 3.14: Schemtaic diagram of BOC Edwards system 500 E-beam sys-tem.

3.3.3 Electron beam evaporated silicon thin films

Electron-beam (E-beam) deposition was examined as a second choice for the fabri-cation of silicon thin films. E-beam deposited silicon thin films have been used inthe past for applications such as solar cells and laser diodes [109, 110, 111]. At theUniversity of Western Australia, we have a BOC Edwards Auto 500 electron beamevaporation system, which can deposit films of materials with high melting points.Figure 3.14 shows a schematic diagram of the BOC Edwards Auto 500 E-beamsystem. This E-beam system can deposit silicon, germanium, silicon monoxide,gold, titanium and platinum thin films. Very fast deposition rates can be achievedusing electron beam evaporation. The Auto 500 is supplied with quartz crystalfilm-thickness monitors and deposition rate controllers. It allows front loading foreasy access into the chamber. The ultimate vacuum is 5 × 10-7 mbar. It uses twoshutters, one for the crucible and another for the substrate. The substrate holderis a rotary holder.

The investigated thin films were deposited on three substrates: (1) 320 µm thick<100> oriented silicon substrates for film thickness measurements and FTIR spec-troscopic analysis; (2) 70 µm thick <100> silicon wafers for stress characterization;and (3) 280 µm thick sapphire substrates for characterization of optical properties.

63


3.4

3.6

3.8

4

4.2

4.4

4.6

0

0.5

1

1.5

2

2.5

0 500 1000 1500 2000 2500

Refractive Index

Extinction Coefficient

Ref

ract

ive

Inde

x

Extinction C

oefficient

Wavelength (nm)

Figure 3.15: Refractive index and extinction coefficient of the e-beam de-posited silicon film.

Stoney’s formula was used to extract the residual stress in silicon thin films fromstress induced bowing of the 70 µm thick substrates [96, 97, 98].

In the E-beam, 99.999% pure P-type 3 mm-6 mm size silicon pieces were used assource material. Since only one crucible was exposed to the electron beam, thedeposited silicon thin films were found to be free from contamination from othermaterials. These films showed negligible etch rate in HF.

Figure 3.15 shows the optical constants for the E-beam deposited silicon films.We used Cauchy’s dispersion equations in the NKD Matl software to extract therefractive index and the extinction coefficient of the silicon thin films from themeasured optical data. In the visible and NIR wavelength range the refractive indexis close to or above 4.0. In the visible wavelength range (450-700 nm) the extinctioncoefficient is in the range of 1.2 to 0.19. Thus, the silicon films deposited with theE-beam show high absorption in the visible wavelength range and can not be usedfor fabricating DBRs and filters in the visible wavelength range. However, over therange of 950-2250 nm wavelength, the silicon films show negligible absorption anda high refractive index. Thus, their optical properties are suitable for fabricatingfilters and DBRs in this wavelength range.

Many research groups have reported that the stress of E-beam deposited thin filmsshows a dependency on the deposition rate [112, 113]. To confirm this, siliconthin films were deposited at 0.1, 0.2 and 0.3 nm/min average deposition rate, to athickness of 150 nm. The stress generated by each deposition rate was measuredusing silicon thin films deposited on strips of 5 cm × 5 mm in size, cleaved froma 70 µm thick <100> silicon substrate. Figure 3.16 shows the tensile stress in thesilicon thin films as a function of deposition rate, which increases with an increasein average deposition rate. At 0.1 nm/min the films show almost zero stress. At0.2 nm/min the films show a tensile stress in the 20–60 MPa range . At 0.3 nm/minaverage deposition rate, we see a rapid increase in the tensile stress in the siliconthin films, which reaches the 330–380 MPa range. The stress range in silicon thinfilms at 0.2 nm/sec average stress is suitable for fabricating MEMS structures, how-

64


ever, absorption in the 450-950 nm wavelength range is prohibitively high. Since,the E-beam deposition process is not conformal in nature, the fabrication processwill require an additional structural material, such as silicon nitride, which will in-crease fabrication time and cost. In addition, the use of two different materials willcertainly cause stress induced deformations. Hence, additional development workwill be required to compensate for these deformations. These issues have led to theconclusion to abandon E-beam deposition for fabrication of the silicon thin films.

0

50

100

150

200

250

300

350

400

0.05 0.1 0.15 0.2 0.25 0.3 0.35

Stre

ss (M

Pa)

Avg. Deposition rate (nm/min)

Figure 3.16: Measured tensile stress in the E-beam deposited silicon film.

3.4 ICPCVD silicon deposition

This section discusses the optimization of optical and mechanical properties of theinductively coupled plasma chemical vapor deposited (ICPCVD) silicon thin films,which have allowed us to fabricate the mirrors and filters required for a wide range ofoperating wavelengths. The optimization process involves the tuning of depositionparameters and post-deposition annealing in an inert gas environment.

Silicon-based materials are commonly deposited using standard chemical vapor de-position (CVD) processes such as atmospheric pressure chemical vapor pressuredeposition (APCVD), low pressure chemical vapor deposition (LPCVD), plasmaenhanced CVD (PECVD) and high density plasma CVD (HDPECVD) or ICPCVD.The mechanical and optical properties of these deposited films are readily optimizedvia control of deposition parameters or via annealing [88, 89, 90, 91]. ICPCVD pos-sesses unique characteristics such as high plasma density, low electron temperature,low plasma potential, and simplicity of configuration, which allows deposition ofhigh quality Si films. Some research groups have reported on the use of ICPCVD tooptimize the optical properties of Si in applications such as diffraction gratings andsolar cells [114, 91]. Optimization of mechanical properties of poly-silicon has beenproposed for making microphones and bimorphs [115, 116]. Silicon based DBRs andfilters raise a new challenge of simultaneous optimization of both the optical andmechanical properties of ultra-thin silicon layers. This work has simultaneously op-

65


timized both the optical and mechanical properties of ICPCVD amorphous silicon(a-Si), which was undertaken by changing deposition temperature and ICP power.Annealing at relatively low temperatures was used to modify the properties of sili-con thin films post-deposition. Fourier transform infrared spectroscopy (FTIR) wasused to investigate the relationship between changes in the optical and mechanicalproperties of the Si films and the nature of hydrogen–silicon bonding and hydrogenconcentration in the films.

3.4.1 Experimental approach

The deposition of amorphous silicon and silicon oxide thin films was performedusing a SI500D ICPCVD system manufactured by SENTECH Instruments. Thissystem uses a planar triple spiral antenna (PTSA) as the ICP source. With a 60 Acurrent supply at a frequency of 13.56 MHz, this antenna can generate 6 × 1011cm−3

of ion plasma density. The permissible deposition temperature range extends fromroom temperature to 300 ◦C. Figure 3.17 shows a picture of the SENTECH SI500Dreactor with its important components.

Table 3.5 lists all the process parameters used to deposit the silicon thin films. Eightsamples were fabricated to examine the effects of various process parameters. Flowrates of silane and helium were selected based on the manufacturer recommenda-tions of using 5% diluted process gas for optimal quality films. The pressure wasfixed at 4 Pa after a separate experiment indicated that higher pressures led to anunstable plasma and the formation of polysilane powder on the films. Such powderformation in a certain pressure range has been reported previously for PECVD re-actors [117, 118]. For samples 1 – 3, the deposition temperature was varied whileall other parameters were kept constant. For samples 4 – 8, the ICP power wasvaried while all other parameters were kept constant. Sample 1 was used to charac-terize the effects of annealing, which was performed in a quartz tube furnace in N2

environment. As will be discussed later, all the samples followed similar trends onannealing. Thus characterizing the effects of annealing on one sample gave sufficientinformation on the variation of silicon films properties with annealing temperature.All samples were tested for chemical stability against common chemicals used inmicrofabrication, such as photoresist developers and HF.

3.4.2 Effect of deposition temperature

Samples 1-3 were used to determine the effect of varying the deposition temperature,using deposition temperatures of 100, 200 and 300 ◦C. The deposition rate wasfound to be independent of deposition temperature and was stable at 16 nm/min.Table 3.7 shows the measured film stress versus deposition temperature (Negativevalues denote compressive stress). Note that the films were compressively stressedand the value of compressive stress of deposited thin films decreased monotonicallywith increasing deposition temperature. The thin films deposited at 300 ◦C are

66


Figure 3.17: A picture of the SENTECH SI500D reactor.

approximately 200 MPa less compressively stressed in comparison to films depositedat 100 ◦C. Therefore, deposition of silicon thin films at higher temperatures appearsbeneficial in order to minimize stress in thin films.

Many authors have reported that the higher the deposition temperature the betterthe optical properties of the PECVD deposited amorphous silicon thin films [91,119]. We also see a similar trend in our ICPCVD films. Figure 3.18 shows themeasured optical constants for samples 1-3. Note that the refractive index increaseswith increasing deposition temperature, and is substantially higher at 300 ◦C incomparison to silicon deposited at temperatures of 100 ◦C and 200 ◦C. However,the extinction coefficient is positively correlated with deposition temperature in thevisible (500-700 nm) and negatively correlated with temperature in the short waveinfrared wavelength range (1500-2250 nm). The transition region is in the nearinfrared wavelength region (700-1400 nm) . Within the SWIR wavelength region,the extinction coefficient of films deposited at 300 ◦C is more than an order of

67


Table 3.5: ICPCVD Process parameters for deposition of silicon thin films.

Sample No. ICP power Dep. temperature Pressure SiH4 Heflow rate flow rate

(W) (◦C) (Pa) (sccm) (sccm)

1 150 100 4 5 952 150 200 4 5 953 150 300 4 5 954 20 300 4 5 955 26 300 4 5 956 35 300 4 5 957 75 300 4 5 958 300 300 4 5 95

magnitude lower than that of films deposited at 100 or 200 ◦C.

Figure 3.19 shows FTIR spectra for samples 1-3 (offset on the absorption axis forclarity). Note that with increasing deposition temperature, the area under all thepeaks decreases. However, while the spectrum of the film grown at 300 ◦C showsthe presence of the stretching mode of both monohydride and dihydride bonds, thespectra of films deposited at 100 and 200 ◦C do not exhibit the stretching mode ofthe monohydride bond near 2000 cm-1.

68


(a)

2.4

2.6

2.8

3

3.2

3.4

3.6

3.8

4

500 1000 1500 2000 2500

100 oC200 oC 300 oC

Ref

ract

ive

Inde

x

Wavelength (nm)(b)

10-5

10-4

10-3

10-2

10-1

500 1000 1500 2000 2500

100 oC200 oC300 oC

Ext

inct

ion

Coe

ffici

ent

Wavelength (nm)

Figure 3.18: Measured optical constants of ICPCVD silicon thin films

at different deposition temperature; (a) shows refractive index of thin films

deposited at different temperatures, (b) shows extinction coefficient of thin

films deposited at different temperatures.

69


2200 2100 2000 1900 800 700 600 5000.00

0.02

0.04

0.06

Si-H2rSi-H

100 o C 200 o C 300 o C

Ab

sorb

ance

(%)

Wavenumber (cm-1)

Si-H

Figure 3.19: FTIR spectra of the films deposited at different deposition

temperature.

70


Table

3.6

:S

um

mar

yof

resu

lts

ofop

tim

izat

ion

ofsi

lico

nby

ann

eali

ng

in

nit

roge

nfo

r30

min

ute

s.A

llth

efi

lms

wer

ed

epos

ited

at4

Pas

cal

dep

osit

ion

pre

ssu

re,

and

100◦ C

dep

osit

ion

tem

per

atu

re.

Th

efl

owra

tes

ofsi

lan

ean

d

hel

ium

wer

eke

pt

at5

sccm

and

95sc

cm,

resp

ecti

vely

.

Sam

ple

no.

150y

show

sth

eas

dep

osit

edb

ase

sam

ple

use

dfo

rth

ean

nea

lin

gan

dal

lot

her

sam

ple

sar

eth

ean

nea

led

sam

ple

s.S

amp

leIC

PA

nn

eali

ng

Str

ess

560

nm

1000

nm

2000

nm

Hyd

roge

nN

o.p

ower

tem

per

atu

reC

once

ntr

atio

n(W

)(◦

C)

(MP

a)n

kn

kn

kcm

-3

115

0yA

sd

epos

ited

-331

2.89

0.01

22.

575.

6×

10-3

2.46

3.4×

10-3

6.31×

1020

1a15

025

014

72.

95.0

52.

598×

10-3

2.47

2.71×

10-3

4.48×

1020

1b15

035

064

73.

040.

142.

686.

6×

10-3

2.57

1×

10-3

3.47×

1020

1c15

040

011

833.

180.

312.

737.

5×

10-3

2.59

7.5×

10-4

1.06

6×10

20

Table

3.7

:S

um

mar

yof

resu

lts

ofin

situ

opti

miz

atio

nof

sili

con

.A

llth

e

film

sw

ere

dep

osit

edat

4P

asca

ld

epos

itio

np

ress

ure

.T

he

flow

rate

ofsi

lan

e

and

hel

ium

wer

eke

pt

at5

sccm

and

95sc

cm,

resp

ecti

vely

.

Sam

ple

ICP

Dep

./A

nn

eal

Dep

.S

tres

s56

0n

m10

00n

m20

00n

mH

yd

roge

nN

o.p

ower

tem

per

atu

rera

teC

once

ntr

atio

n(W

)(◦

C)

(nm

/m

in)

(MP

a)n

kn

kn

kcm

-3

115

010

016

-331

2.89

0.01

22.

575.

6×10

-32.

463.

4×

10-3

6.31×

1020

215

020

016

-180

2.96

0.04

82.

637.

0×10

-32.

522.

0×

10-3

3.56×

1020

315

030

016

-120

3.89

0.07

53.

321.

3×10

-33.

369.

4×

10-5

3.43×

1020

420

300

7.2

200

4.91

0.45

63.

911.

4×

10-4

3.58

1.34×

10-6

1.38×

1021

526

300

10.

665

4.36

0.03

73.

773.

2×10

-43.

554.

05×

10-6

1.18×

1021

635

300

11.

67-5

4.24

0.16

53.

653.

6×10

-43.

517.

61×

10-6

7.4×

1020

775

300

13.

33-8

54.

230.

128

3.6

5.0×

10-4

3.47

1.59×

10-5

6.34×

1020

830

030

016.

5-5

503.

440.

051

2.95

4.1×

10-3

2.90

8.7×

10-4

3.07×

1020

71


Table 3.7 lists the hydrogen concentration for the films deposited at 100, 200 and300 ◦C, as calculated from analysis of the FTIR peaks at 640 cm-1. The hydrogenconcentration in the film deposited at 100 ◦C is significantly higher than that offilms deposited at higher temperatures. This decrease in hydrogen concentrationin amorphous silicon with increasing deposition temperature is well known in theliterature [119].

With an increase in deposition temperature, the diffusion rate of hydrogen atoms inthe silicon thin films also increases. Hence, the atomic hydrogen (H) diffuses inter-stitially in the films and inserts itself into weak Si-Si bonds, resulting in breakage ofthese bonds. After the bond-breaking charge is removed, either by recombinationor thermal re-emission, the broken Si-Si bond reconstructs as a strong Si-Si bondor as a stable Si-H bond. This leads to an increase in density of the silicon films,which correlates with the refractive index of films deposited at 300 ◦C being higherthan the refractive index of films deposited at lower temperatures. These findingsare consistent with the generally accepted explanation of the effect of temperatureon the properties of Si films as reported in the literature [120, 102, 121].

It is also well known that amorphous hydrogenated silicon has many defect stateswithin the bandgap which contribute to an increased absorption in the SWIR wave-length range [120, 122]. The increase of the diffusion rate of atomic hydrogen dueto an increase in deposition temperature reduces the defect density in the amor-phous hydrogenated silicon, which reduces the absorption coefficient in the SWIRwavelength region [123]. The increase in mass density of films at higher depositiontemperatures also reduces the optical band gap between the conduction and valenceband [124], which increases absorption at visible wavelengths. This explains the ob-servations presented in Fig. 3.18(b), of positive correlation of extinction coefficientwith temperature in the visible wavelength range (500-700 nm), and negative cor-relation of extinction coefficient SWIR wavelength range (1500-2250 nm), with acrossover in the near infrared region.

3.4.3 Effect of annealing

The samples chosen for the post deposition annealing studies were prepared withthe recipe used for sample 1, which exhibited an as-deposited compressive stressof 331 MPa. Each of the samples listed in Table 3.6 correspond to separatelyprepared films that were annealed for 30 minutes in a N2 environment at differenttemperatures. The effect of annealing temperature on the residual stress of thethin films is shown in Figure 3.20. For annealing below 225 ◦C, the films remainedcompressive, whereas annealing at temperatures above 225 ◦C resulted in tensilestressed films with a positive correlation between the values of tensile stress andannealing temperature. There is a steep rise in the tensile stress above 300 ◦Cannealing temperature, such that for annealing temperature of 400 ◦C the tensilestress reaches 1.2 GPa. Since such a high stress in the films will most likely lead tostructural failure, annealing at temperatures higher than 400 ◦C was not examined.

72


-400

-200

0

200

400

600

800

1000

1200

50 100 150 200 250 300 350 400 450

Stre

ss (M

Pa)

Annealing Temperature (oC)

Tensile

Compressive

Figure 3.20: Residual stress in the silicon films as a function of annealing

temperature, using the deposition conditions of sample 1 from Table 3.5.

Figure 3.21 shows the measured optical constants of the annealed samples. Anincrease in annealing temperature increased the refractive index for all the samplesregardless of the wavelength range. The extinction coefficient was positively cor-related with annealing temperature in the visible wavelength range (500-700 nm).The extinction coefficient was negatively correlated with the annealing temperaturein SWIR wavelength range (1500-2250 nm). The crossover in extinction coefficientoccurred near a wavelength value of 1000 nm.

Figure 3.22 shows the FTIR spectra of the samples annealed at different tempera-tures, and Figure 3.23 shows the hydrogen concentration remaining in the samplesas a function of annealing temperature. Certain trends can be noted from these twofigures. With an increase in annealing temperature, absorbance in the thin filmsdecreases for all types of bonds. This indicates that the concentration of hydrogenin the thin films is reduced as the annealing temperature is raised. It can be clearlyestablished that for the same change in the annealing temperature, the percentagereduction of absorbance in the stretching mode is more than the percentage reduc-tion of absorbance in the rocking mode. This is due to a lower hydrogen evolutiontemperature for the stretching mode Si-H2 bond than for the rocking mode Si-Hbonds [125, 90]. We also observe a shift in peak position of the Si-H rocking bondat 400 ◦C. The removal of hydrogen relieves the compressive stress in the film andincreases their density through the formation of strong Si-Si bonds. The higher theannealing temperature, the more hydrogen is released, resulting in a higher densityof Si-Si bonds; thus, the films become denser and more tensile.

73


(a)

2.2

2.4

2.6

2.8

3

3.2

3.4

500 1000 1500 2000 2500

100 oC250 oC350 oC400 oC

Ref

ract

ive

Inde

x

Wavelength (nm)(b)

10-4

10-3

10-2

10-1

100

500 1000 1500 2000 2500

100 oC250 oC350 oC400 oC

Ext

inct

ion

Coe

ffici

ent

Wavelngth (nm)

Figure 3.21: Measured optical constants of silicon thin films annealed at

different annealing temperatures; (a) shows refractive index, (b) shows extinc-

tion coefficient. The films were deposited using the deposition conditions of

sample 1 from Table 3.5.

As reported in the literature, the density of defect states within the bandgap ofamorphous silicon thin films can be reduced by annealing [126]. These defect stateslead to higher absorption in the SWIR wavelength range. Thus, a reduction of thedensity of defect states, via annealing, renders the amorphous silicon more transpar-ent in the SWIR wavelength region. It is also well reported that the bandgap of thehydrogenated silicon is a function of hydrogen concentration in the films [124], such

74


2200 2100 2000 1900 800 700 600 5000.00

0.01

0.02

0.03

0.04

0.05

0.06

Si-H2

Si-H

Si-H

Abso

rban

ce (%

)

Wavenumber (cm-1)

As deposited @ 100 oC 250 oC 350 oC 300 oC 400 oC

Figure 3.22: FTIR spectra of silicon films annealed at different annealing

temperatures. The films were deposited using the deposition conditions of

sample 1 from Table 3.5.

1 x 1020

2 x 1020

3 x 1020

4 x 1020

5 x 1020

6 x 1020

7 x 1020

50 100 150 200 250 300 350 400 450

Hyd

roge

n C

once

ntra

tion

(cm

-3)

Annealing Temperature (oC)

Figure 3.23: Hydrogen concentration of silicon films as a function of an-

nealing temperature. The films were deposited using the deposition conditions

of sample 1 from Table 3.5.

75


that a decrease in the hydrogen concentration reduces the bandgap. This bandgapnarrowing makes amorphous silicon more absorbing in the UV-visible wavelengthrange of 300-700 nm.

Therefore, as observed in Fig. 3.21 (b), there is positive correlation between therise in annealing temperature and the extinction coefficient in the visible wavelengthrange (500-700 nm). For SWIR wavelengths (1500-2250 nm) there is negative cor-relation between the rise in annealing temperature and the extinction coefficient,and can be explained as follows. With an increase in annealing temperature, thedensity of dangling or defect states in the amorphous silicon decreases, which ren-ders it more transparent in the SWIR wavelength region. At the same time, thereduction in hydrogen concentration in the film reduces the optical bandgap whichmakes it more absorbing at visible wavelengths. Annealing at a higher temperaturefacilitates formation of strong Si-Si bonds, which increases the atomic density in thefilm, reduces the volume of the film, and increases the refractive index. This pat-tern of reasoning is somewhat parallel to that observed as a function of depositiontemperature in Fig. 3.18 (b).

3.4.4 Effect of ICP power

As shown in Table 3.5, to study the effect of ICP power on the quality of siliconfilms, the ICP power was varied from 20 W to 300 W, while all other processparameters were kept unchanged. As shown in Figure 3.24, a positive correlationis evident between the deposition rate and ICP power. In the low ICP power range(below 50 W) the deposition rate increases steeply with increasing ICP power, andeventually saturates in the high ICP power range (above 150 W). Figure 3.24 alsoshows the variation in the residual stress in the deposited silicon films as a functionof ICP power. Films deposited in the ICP power range 20 W to 25 W are tensile innature. The value of the residual tensile stress decreases with increasing ICP power,and then crosses over towards compressive residual stress values near an ICP powervalue of 35 W. Subsequent increases in ICP power results in increasing values ofcompressive stress, with Si thin films deposited at 300 W ICP power exhibiting550 MPa of residual compressive stress.

Figure 3.25 shows the optical constants of the silicon thin films as a function ofwavelength at various ICP powers. It is evident from Figure 3.25 (a) that witha decrease in ICP power the refractive index increases for all samples regardlessof wavelength range. In Figure 3.25 (b) ICP power is negatively correlated withthe extinction coefficient in the visible wavelength range (500-700 nm), whereas ICPpower is positively correlated with the extinction coefficient in the SWIR wavelengthrange (1500-2250 nm). The crossover occurs near the wavelength of 700 nm. Atvisible wavelengths (500-700 nm) the high extinction coefficient (0.012–0.456) ofthe films is detrimental for making Fabry-Perot filters. However, since the filmsrequired in this wavelength range are very thin, the effect of this higher extinctioncoefficient is somewhat mitigated. Therefore, filters and DBRs designed with these

76


6

8

10

12

14

16

18

-600

-400

-200

0

200

400

0 50 100 150 200 250 300 350

Deposition rateStress

Dep

ositi

on ra

te (n

m/s

ec)

Stress (M

Pa)

ICP power (W)

Figure 3.24: Residual stress and deposition rate of the Si films as a function

of ICP power.

thin films can still produce adequate optical performance at these wavelengths.

Figure 3.26 shows the FTIR spectra measured for thin film samples No. 3 to 8 (seeTable 3.5) deposited with ICP power varying from 20 W to 300 W. The spectra havebeen offset for clarity. Note that with an increase in ICP power the absorbance ofthe hydrogen bonds decreases both for the rocking mode ( 640 cm-1) and stretchingmode ( 2000–2200 cm-1). Note that in Figure 3.26, for the films deposited at an ICPpower of 35 W and below, the peaks at ( 2000 cm-1) associated with the stretchingmode of the monohydride bond (Si-H) are dominant over the stretching mode di-hydride bond peaks ( 2100 cm-1). At 75 W ICP power, the (Si-H2) stretching bond(peak at 2100 cm-1) dominates over monohydride stretching mode vibration peaks.For ICP powers of 150 W and 300 W, we note the presence of Si-H and Si-H2 peaks.

Figure 3.27 shows the hydrogen concentration in the films as a function of ICPpower. As ICP power increases, the hydrogen concentration in the films significantlyreduces. This is an interesting observation to note. As discussed in Section 3.4.2 andSection 3.4.3, a higher hydrogen concentration renders the films more compressive,and has a detrimental effect on the optical properties. But as presented in thissection, the increase in hydrogen concentration, due to a decrease in the ICP power,improves the optical quality of the films and tends to make the films more tensile.This apparent contradiction can be resolved by an examination of the physics behindincorporation of hydrogen in the amorphous silicon films. In the SiH4 plasma avariety of neutral radicals, ionic and emissive species are generated by the ionizationof SiH4. The neutral radicals SiH2 and SiH3 are actual contributors to the filmgrowth rate. Among all the radicals, the SiH3 radical has the longest lifetime andis considered to be the major contributor to growth of the film [127, 102] due tosticking of the neutral radical SiH3 to the surface of the growing film. Thus, thegrowth rate of the film depends on available locations on the growing surface, aswell as deposition temperature.

77


(a)

2.5

3

3.5

4

4.5

5

500 1000 1500 2000 2500

20 W

26 W

35 W

75 W

150 W

300 W

Ref

ract

ive

Inde

x

Wavelength (nm)(b)

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

500 1000 1500 2000 2500

20 W

26 W

35 W

75 W

150 W

300 W

Extin

ctio

n C

oeffi

cien

t

Wavelength (nm)

Figure 3.25: Measured optical constants of thin films deposited at different

ICP power; (a) shows refractive index, (b) shows extinction coefficient (see

Table 3.5).

The density of ionic species (such as H+, H2+, SiHx

+ and Si2Hy+) on the growing

surface is much lower than neutral radicals and, therefore, their contribution to thedeposition rate can be neglected. However, their higher kinetic energy as comparedto neutral ions causes a significant change in the properties of the film. Matsuda etal. [128] reported that impinging ionic species on the growing surface give rise tocompressive stress in the films.

78


2200 2100 2000 1900 800 700 600 5000.00

0.05

0.10

0.15

0.20

0.25

Si-H2 Si-H

300 W

75 W

150 W

35 W

26 W

Abso

rban

ce (%

)

Wavenumber (cm-1)

20 W

Si-H

Figure 3.26: FTIR spectra of silicon films deposited at different ICP powers

(see Table 3.5).

As shown in Figure 3.24, at 20 W ICP power the deposition rate is just 7.2 nm/sec.Due to low ICP power, the density of neutral and ionic species is low, which leadsto lower growth rate. The low density of neutral radicals and high depositiontemperature (300 ◦C), assists neutral radicals such as SiH3 to diffuse over a longerdistance and to form a denser network of strong Si-Si and Si-H bonds. In suchcases the dangling bond density will be low and it is plausible that most of thehydrogen used in the passivation of dangling bonds ends up forming stable Si-Hbonds. Hence, for the films deposited at 20 W the FTIR spectra in Figure 3.26shows the highest absorption for the Si-H rocking bond at 640 cm-1 wavenumber.As a result of formation of stable Si-H bonds, we find that despite having thehighest hydrogen concentration, the films deposited with 20 W of ICP power havethe highest refractive index and highest tensile stress among all deposited films.

As the ICP power is increased, the density of neutral and ionic species in the silaneplasma also increases, which increases the growth rate of the film. However, thisalso reduces the diffusion length of neutral radicals. Therefore, many sites in thegrowing film remain unpassivated, which leads to an increase in the defect densitywithin the film. The increased defect density finally results in a decreased densityof the films, which is reflected in a decrease of refractive index with increasing ICPpower. The optical bandgap increases with an increase in ICP power leading to adecrease in the extinction coefficient in the visible wavelength range. In addition,the increase in defect density states due to increase in ICP power also leads tohigher absorption in the SWIR wavelength range (1500-2250 nm).

With a continued increase in the ICP power, the density of the ionic species also

79


2 x 1020

4 x 1020

6 x 1020

8 x 1020

1 x 1021

1.2 x 1021

1.4 x 1021

1.6 x 1021

0 50 100 150 200 250 300 350

Hyd

roge

n C

once

ntra

tion

(cm

-3)

ICP power (W)

Figure 3.27: Hydrogen concentration in the films as a function of ICP

power (see Table 3.7).

continues to increase. At first, this reduces the tensile stress and, finally, at an ICPpower of 35 W the stress in the films tends to become compressive. With furtherincreases in ICP power the compressive stress in the films also increases. Theseresults contradict the results reported in References [129, 121, 130, 131, 132], wherethe authors suggest that a higher concentration of hydrogen is the sole factor leadingto a low refractive index and high compressive stress in silicon films. Through theseexperimental observations, it can be proposed that if hydrogen in the film forms astable bond with the silicon, then even a high concentrations of hydrogen can leadto a denser and more tensile film.

A summary of the results of the optimization process based on annealing and de-position parameters are shown in Table 3.6 and Table 3.7.

3.5 ICPCVD silicon oxide deposition

Silicon oxide (SiOx) is widely used as a sacrificial layer in MEMS, optical layer inthin-film optics [133, 134], passivation layer in solar cells[135], a structural materialin MEMS, and as an insulating layer [136]. There are many reported methods toform silicon oxide thin films such as thermal oxidation [137, 138], LPCVD [139],PECVD [140, 141, 142], APCVD [143], ICPCVD [144, 145], and in recent timesatomic layer deposition (ALD) [146]. This thesis has used ICPCVD deposited SiOx

thin films for fabricating optical layers for filters and DBR mirrors. Silicon oxidewas optimized for mechanical properties only, since its optical properties , in visibleto SWIR wavelengths, do not vary significantly in comparison to those of silicon

80


[145, 92].

3.5.1 Experimental Approach

This aspect of work focused mainly on variations in deposition temperature anddeposition pressure in order to optimize the stress in deposited SiOx layers. Theinvestigated thin films were deposited on two types of substrates: (1) 320 µm thick<100> oriented silicon wafers that were used for film thickness measurements andetch rate measurements; and (2) 70 µm thick <100> silicon wafers that were usedfor stress characterization based on Stoney’s formula [96] for the wafer bowingmeasurements. Table 3.8 shows the process parameters used for characterizingthe properties of silicon oxide films. It is to be noted that sample numbers startfrom 9 to make a clear distinction between the sample numbers used in Table 3.5for optimization of amorphous silicon. The wet etch rate of SiOx is a very usefulparameter if it is to be employed as a sacrificial layer. The wet etch rate also givesan indication of the density of the SiOx films. The higher the etch rate, the loweris the density of SiOx thin films. The wet etch rate of the SiOx was determined ina buffered oxide etch (BOE) or buffered HF (BHF).

3.5.2 Optimization of deposition temperature

Figure 3.28 shows the effect of variation of deposition temperature on the ICPdeposition rate and wet etch rate. With an increase in the deposition temperaturethe deposition rate of silicon oxides first decreases from 20 nm min−1 at 50 ◦Cto 17 nm min−1 at 130 ◦C and than remains constant up to 275 ◦C. The etchrate of the silicon oxide films was found to decrease monotonically with increasingdeposition temperature, hence silicon oxide films deposited at higher temperatureare denser than films deposited at lower temperature. Figure 3.29 shows that withan increase in the deposition temperature the silicon oxide films become increasinglycompressive. The measured compressive stress of 214 MPa at 50 ◦C and 580 MPa at275 ◦C indicates that lower stress silicon oxide films are obtained at lower depositiontemperatures.

3.5.3 Optimization of deposition pressure

Figure 3.30 shows the effect of ICP pressure on the deposition rate and etch rate.The deposition temperature was kept fixed at 50 ◦C to obtain low stress SiOx. Withan increase in the deposition pressure, the deposition rate monotonically decreases.From 20 nm/min at 2 Pa to 14 nm/min at 6 Pa, and eventually to 13 nm/min at14 Pa. It was found that any further increase in the pressure leads to formation ofpowder in the chamber and on the films, so the pressure was not increased further.The etch rate of the silicon oxide films monotonically increases with increasingdeposition pressure. The etch rate at 2 Pa was 515 nm/min. This increases to

81


16.5

17

17.5

18

18.5

19

19.5

20

20.5

150

200

250

300

350

400

450

500

550

0 50 100 150 200 250 300

Deposition rate

Etch rate

Dep

ositi

on ra

te (n

m/m

in)

Etch rate (nm

/min)

Deposition Temperature (oC)

Figure 3.28: Deposition rate and etch rate of SiOx films as a function of

deposition temperature see Table ??.

1100 nm/min at 6 Pa and to 1600 nm/min at 14 Pa. From this trend in depositionpressure vs etch rate it can be concluded that the silicon oxide films deposited athigher deposition pressure are less dense than films deposited at lower depositionpressure.

Figure 3.31 shows that with an increase in the deposition pressure the silicon oxidefilms become less compressively stressed. At 2 Pa deposition pressure the compres-sive stress is 214 MPa, which decreases by 50 % and reaches 111 MPa at 6 Pa. At10 Pa deposition pressure, the compressive stress reaches 25 MPa, and it attainsits lowest value of just 5 MPa at 14 Pa pressure. The low stress and high etch ratesilicon oxide that results at high deposition pressure would be very useful as anoptical layer or as a sacrificial layer. Table 3.8 summarizes the results of the siliconoxide thin film optimization process.

3.6 Summary

In this chapter based on the simulations in visible to MWIR wavelength range itwas established that silicon, air and silicon oxide can be choice of materials for thefabrication of DBRs and filters. Next, this chapter detailed on the silicon thin filmdeposition methods attempted in this research work. Through the optical, residualstress and stability test of deposited silicon thin films it was shown that ICPCVDdeposited silicon thin films are the best choice for the optical filters fabrication.An elaborate discussion on the optimization of amorphous silicon thin films waspresented. Optimization of the thin film properties was investigated as a functionof the variation in the deposition temperature, annealing temperature and ICPpower. It was concluded that change in the ICP power provides maximum control

82


-700

-600

-500

-400

-300

-200

-100

0 50 100 150 200 250 300S

tress

(MP

a)

Deposition Temperature (oC)

Figure 3.29: Residual stress in the SiOx films as a function of deposition

temperature (see Table 3.8).

13

14

15

16

17

18

19

20

21

400

600

800

1000

1200

1400

1600

1800

0 5 10 15

Deposition Rate

Etch Rate

Dep

ositi

on R

ate

(nm

/min

)

Etch R

ate (nm/m

in)

Deposition Pressure (Pa)

Figure 3.30: Deposition rate and etch rate of SiOx films as a function of

deposition pressure (see Table 3.8).

83


-300

-250

-200

-150

-100

-50

0

50

0 5 10 15

Stre

ss (M

Pa)

Deposition Pressure (Pa)

Figure 3.31: Residual stress in the SiOx films as a function of deposition

pressure (see Table 3.8).

over residual stress and optical properties of amorphous silicon thin films. Thischapter showed a strong correlation between decreasing ICP power and increasingrefractive index of the amorphous silicon films, regardless of the wavelength range.The extinction coefficient showed a negative correlation with ICP power for vis-ible wavelengths, and a positive correlation with ICP power for NIR and SWIRwavelengths. In particular, it has been shown that high quality films for opticalMEMS applications can be obtained despite the incorporation of a significant con-centration of hydrogen. The presence of hydrogen actually assists in the formationof stable and strong Si-H bonds, thus reducing the defect density in the films andimproving their optical quality. Finally, the chapter presented in-situ optimizationof residual stress in silicon oxide films by varying deposition pressure and depositiontemperature.

84


Table

3.8

:S

um

mar

yof

resu

lts

ofth

eop

tim

izat

ion

pro

cess

for

sili

con

oxid

e

film

s.

Sam

ple

No.

ICP

pow

erD

ep.

Dep

.SiH

4H

eN

2O

Ar

Dep

.E

tch

Str

ess

tem

per

ature

Pre

ssure

flow

rate

flow

rate

flow

rate

flow

rate

rate

rate

(W)

(◦C

)(P

a)(s

ccm

)(s

ccm

)(s

ccm

)(s

ccm

)nm

/min

nm

/min

MP

a

945

050

26.

512

370

120

2051

5-2

1410

450

130

26.

512

370

120

1734

0-2

4011

450

200

26.

512

370

120

1725

0-4

8112

450

275

26.

512

370

120

1716

7-5

8113

450

506

6.5

123

7012

018

1100

-111

1445

050

106.

512

370

120

1413

00-2

515

450

5014

6.5

123

7012

013

1600

-5

85

Chapter 4

Silicon-air-silicon DBRs for theSWIR and MWIR wavelengths

4.1 Introduction

In this chapter, the fabrication process and optical characterization of silicon-air-silicon based DBRs are the for the SWIR and MWIR wavelengths is presented. Itis shown that the given fabrication technology can be extended to fabricate DBRranging from ranging from 200 µm × 200 µm to 5000 µm × 5000 µm size. Thechapter concludes with discussion on optical and mechanical properties of silicon-silicon oxide-silicon based DBRs for the SWIR wavelength range.

4.1.1 Towards air-spaced tunable filters

A simple MEMS-based tunable Fabry Perot filter consists of two mirrors separatedby a variable air gap. Each mirror is generally a multi-layer structure consisting ofalternating high and low refractive index layers, each one-quarter of a wavelengthin optical thickness. It has already been discussed that for any given high indexmedium, the maximum mirror reflectivity is obtained if the low index mediumis air (i.e. an “air-spaced” mirror). A simple tunable Fabry Perot filter with 3-layer quarter wavelength silicon mirrors will have a structure such as that shown inFigure 4.1. Note that the requirement of the matching layer shown in Figure 4.1will depend on the particular substrate used.

The development of such a tunable filter requires a bottom-up approach, and this isthe process followed here. The first key step in this approach is the development ofthe bottom mirror. This chapter presents the design, fabrication process, and opti-cal characterization of bottom mirror DBR’s fabricated to operate in the short-waveinfrared and mid-wave infrared wavelength bands. Methods will also be examinedfor fabricating freely suspended DBR’s and filters for either post-assembly into afilter, or for fixed-filter applications.

86

CHAPTER 4. SILICON-AIR-SILICON DBRS FOR THE SWIR AND MWIRWAVELENGTHS

Silicon Fixed air cavity

Silicon

λ/2 tunable air cavity

Silicon Fixed air cavity

Silicon Matching Layer

Substrate

Top mirror (All layer λ/4 thick)

Bottom mirror (All layer λ/4 thick)

Figure 4.1: Optical arrangement of layers in a MEMS-based tuneable FabryPerot filter consisting of two air-spaced mirrors.

4.1.2 Filter structures for large area imaging

With the advent of large-area two-dimensional focal plane arrays (FPAs) from thenear infrared (NIR) to the long wave infrared (LWIR) wavelengths (0.7 µm to20 µm), it is now possible to image over a very large portion of the infrared spectrum.Among the reported FPA technologies, InGaAs based FPAs can be operated atroom temperature while covering the visible to short-wave infrared (SWIR, up to2.5 µm) wavelength range. These FPAs are now part of many commercial NIRand SWIR spectroscopic imaging systems [147, 148, 149, 150]. Although originallydeveloped for defence, security and aerospace applications, such imaging systemshave found new applications in crop germination studies, drug quality analysis, andfood security and quality analysis [151, 152, 153, 31, 154, 33]. The combinationof tunable optics with an FPA can provide new capabilities: wavelength selectiveFPAs or so-called adaptive FPAs. In traditional spectroscopic imaging systems, thisfunction is achieved by passing the incident spectrum through a series of complexoptical components, and then onto the FPA [155]. These traditional systems offerhigh resolution and a wide operating wavelength range. However, such systems areexpensive, bulky, fragile, and not readily field-portable. With recent advances inmicroelectromechanical systems (MEMS), miniature, highly-portable, robust, andlow-cost spectroscopic systems have become a reality [156, 157]. These systemsreplace the bulky and fragile optics components of traditional systems with MEMS-based miniature filters and mirrors. The integration of a MEMS-based Fabry-Perottunable filter with the detector can result in a very compact spectrometer, and anarray of such compact spectrometers can be used to provide individual on-pixelspectroscopy in an adaptive FPA.

Previous work has attempted monolithic integration of the detector with a MEMS-based filter to create multispectral detectors at SWIR wavelengths [27, 66]. Otherresearch groups have proposed resonant cavity enhanced detector structures bygrowing the detector on top of one of the filter mirrors for the mid-wave infrared

87


(MWIR) wavelength range [58, 158]. With monolithically integrated approaches,one is limited by the choice of materials compatible with both the FPA and theMEMS-based tunable filter fabrication. Furthermore, monolithic integration cannottake full advantage of the myriad of available processes to optimise the stress andoptical properties that are available exclusively to MEMS technologies, and themany optimisation technologies available exclusively for detector fabrication.

These limitations can be overcome by a hybrid approach, where the detector andMEMS-based optics are fabricated and optimized separately and, subsequently,bonded together with the help of a spacer/bonding pads. A conceptual diagramof such a device is shown in Figure 4.2. The MEMS-based tunable optical filterconsists of two distributed Bragg reflectors (DBRs) separated by a half-wavelengthoptical cavity of air. Each DBR is a stack of alternating high and low refractiveindex dielectric layers, with each layer thickness being a quarter wavelength. Theoptical cavity between the two DBRs determines the filter peak wavelength.

Detector or FPA

Incident Spectrum

DBR Tuneable Cavity

Substrate

Electrode Spacer

Figure 4.2: Working principle of proposed module consisting of a tunableFabry-Perot cavity hybrid bonded to a detector or imaging FPA.

4.2 DBR support structures

In order to successfully achieve a bottom mirror, a suspension mechanism needs tobe developed for creating the air gap between the two silicon layers. This sectiondescribes the design and fabrication process for the two MEMS structures thatwere assessed as candidate in order to develop functional silicon-air-silicon basedDBR’s: (1) In line with our previous microspectrometer designs, our initial devicestructures consisted of a suspended mirror supported by four support arms. (2) Anew design based on a mirror supported around the periphery was also examined asbeing potentially superior to the arm-based support structure. Initially, however,effort was focused on achieving the suspension of a single silicon layer, with theappropriate air-gap above a substrate, but without the bottom silicon layer of themirror.

88


Table 4.1: Silicon deposition recipe used for mirrors.

ICP power Pressure Dep. temperature SiH4 flow rate He flow rate(W) (Pa) (◦C) (sccm) (sccm)

26 4 300 5 95

4.2.1 Support structure with arm based support

Figure 4.3 shows the arms based DBR structure with support arms. Two types ofstructures were fabricated (1) with anti-stiction bumps and (2) without anti-stictionbumps. The support arms were 30 µm wide, and mirror diameters of 100 µm and200 µm were examined. A 15 µm diameter anti-stiction bump, made from thetop membrane thin film, was placed in the center of the mirror to prevent stictionbetween the top and bottom membranes during the release process. The thicknessof the anti-stiction bump was same as the thickness of the top membrane.

Top suspended

silicon membrane

Support Arms

(30 µm wide)

Bottom silicon

membrane

D

Substrate

Matching layer

Top suspended

silicon membrane

Support Arms

(30 µm wide)

Substrate

Anti‐stiction bump

(15 µm diameter)

Figure 4.3: Arm based DBR structure with an optional anti-stiction bumpat centre.

Mirrors were fabricated on 300 µm thick 〈100〉 oriented silicon substrates. Table 4.1shows the silicon deposition recipe used in fabricating these structures. The fabri-cation steps were (1) deposition of sacrificial material on substrate; (2) patterningsacrificial material into islands of appropriate shape and size; (3) deposition of topsilicon optical and support layer; (4) patterning of top silicon optical and supportlayer; and (5) etching of sacrificial layer material. The following fabrication processflow was used for the suspended membrane structures:

1. A 20% diluted PI2610 polyimide was spun, soft-baked at 100 ◦C for five min-utes and hard-baked at 300 ◦C for 20 minutes.

89


2. This 500 nm thick polyimide layer was patterned in an O2/CF4 plasma atroom temperature, to form the sacrificial layer islands. The etch recipe forpolyimide was: RF power of 26 W, pressure of 2.67 Pa, and O2/CF4 flow rateof 30/3 sccm.

3. A 141 nm thick ICPCVD silicon layer was deposited using the recipe shownin Table 4.1.

4. This silicon layer was patterned to form the arms and mirror structure using aCF4 based plasma recipe(RF power of 100 W, ICP power of 400 W, pressureof 4 Pa, and CF4 flow rate of 30 sccm).

5. The sacrificial layer was then etched in a March barrel asher in an oxygenplasma to release the 141 nm thick silicon membrane, thus yielding a finalsuspended structure.

Figure 4.4 shows 3-D optical surface profiles of the fabricated suspended structures,and Figure 4.5 shows line scans across these structures. As shown in Fig. 4.5 (a)the unannealed suspended membrane is compressively stressed and shows significantbowing, such that, the aircavity is several times greater than the targeted gap.Theedge to edge bowing is about 2000 nm, this reduces the effective finesse (FE) to0.5 [28, 43].In order to reduce the compressive stress, the fabricated structures wereannealed in a quartz tube furnace in a N2 environment at 310 ◦C for 45 minutes.Figure 4.5 (b) shows that the compressive stressed has been eliminated, and thatthe central 75 µm of the 100 µm diameter is flat within 10 nm, with such flatmirror effective finesse (FE) can be increased to 99. Maximum acceptable edgeto edge bowing is 30 nm which can give an effective finesse of 33. However, asshown in Figure 4.5 (c) the 200 µm diameter structures with anti-stiction bumps,even after annealing at 350 ◦C for one hour, do not obtain an acceptable level offlatness. Clearly, the presence of the bump significantly degrades the flatness of thestructure. These fabricated structures were stored in a laboratory ambient at roomtemperature for five days to test their environmental stability. It was found thatthese structures were collapsed during that period. The relatively weak arm-basedsupport structure, and contact between the edge of the circular membrane and thesubstrate, were identified as a possible causes of failure. These results promptedchanges to the suspension structure to develop a more mechanically robust design.

4.2.2 Support structure with full peripheral support

In order to ensure flatness of the suspended structures and provide environmentalstability, three changes were made to the DBR structures. First, instead of using anarm based support, a conformal support over the entire periphery of the structurewas provided. Second, 5 µm diameter holes were etched in the membrane to facili-tate removal of the sacrificial layer instead of relying on the etching from the edgesof circular membranes, and third, anti-stiction bumps were totally eliminated.

90


(a)

(b)

(c)

Figure 4.4: 3-D Optical surface profiles of fabricated circular suspended

structures: (a) 100 µm unannealed structure; (b) 100 µm structure annealed

at 310 ◦C for 45 minutes; and (c) 200 µm structure with anti-stiction bump

annealed at 350 ◦C for 60 minutes.91


(a) (b)

0

500

1000

1500

2000

2500

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16

Hei

ght (

nm)

Distance (mm)

0

100

200

300

400

500

600

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16

Hei

ght (

nm)

Distance (mm)

(c)

0

500

1000

1500

2000

2500

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Hei

ght (

nm)

Distance (mm)

Figure 4.5: Single line scan across the surface profiles of fabricated circular

suspended structures shown in Fig. 4.4: (a) 100 µm unannealed structure; (b)

100 µm structure annealed at 310 ◦C for 45 minutes; and (c) 200 µm structure

with anti-stiction bump annealed at 350 ◦C for 60 minutes.

The fabrication process was carried out on 300 µm thick 〈100〉 oriented siliconsubstrates. The fabrication steps used in this process identical to those for the arm-based design, with the exception of the mask layout used in the silicon patterningprocess. This mask allowed the support structure to extend around the entireperiphery of each polyimide island, and also allowed the formation of 5 µm diameterholes in the optical area. In the arm-based design, the etchant gasses had accessto the sacrificial material from the mirror periphery via the spaces between thearms. With the support structure now extending around the entire periphery, thesacrificial material is fully encapsulated and this access is no longer possible. Thus,the holes in the optical area are needed to allow the gasses to etch the polyimideand release the suspended mirror layer.

92


Figure 4.6(a) shows a 3-D optical profile of the suspended structure, and Figure4.6(b) shows a cross section of the suspended film. The sharp peaks in Figure 4.6(b)at either ends across the 270 µm diameter suspended structure are artifacts arisingfrom limitations of the optical measurement over sharp edges. In Figure 4.6(c),which is an expanded scale plot of Figure 4.6(b), it is noted that the suspended filmshows exceptional surface flatness, with a variation on the order of 6 nm across thestructure diameter. These fabricated structures were stored in laboratory ambientat room temperature for five days to test environmental stability, and negligiblechanges were observed in surface profile. Hence, this structure was selected for thefabrication of silicon-air-silicon mirrors.

(a)

7.4 IJm

5.6 IJm

(b) (c)

0

50

100

150

200

250

300

350

0.0 0.050 0.10 0.15 0.20 0.25 0.30 0.35 0.40

Hei

ght (

nm)

Distance (mm)

270

280

290

300

310

320

330

0.05 0.1 0.15 0.2 0.25

Hei

ght (

nm)

Distance (mm)

Figure 4.6: Optical surface profiles of fabricated 270 µm diameter circular

suspended structure, (a) 3-D surface profile, (b) extracted line scan across the

suspending membrane avoiding the etch holes, and (c) profile of fabricated

structure on an expanded vertical scale.

93


4.3 Si-air-Si based mirrors for SWIR wavelengths

With a functional support structure demonstrated, our this work moved on todemonstration of a functional mirror based on the Si-air-Si technology. This sectiondescribes the fabrication process and optical characterization of silicon-air-siliconbased DBRs. The fabrication process steps used in this section form the basis forfabrication of silicon-air-silicon DBRs.

4.3.1 Fabrication process

The fabricated mirrors were circular shaped with diameters of 270 µm, 320 µm,370 µm and 420 µm. Mirrors of each size were fabricated in two rows with each rowcontaining 8 mirrors. The center-to-center separation of each mirror was 750 µm.The mask contains four groups of 64 mirrors with a total 256 mirrors. It has been

270 µm

320 µm

370 µm

420 µm

Figure 4.7: The mask layout for circular mirrors ranging from 270 µm to

420 µm diameter in size .

shown in Chapter 3 that both 20 W and 26 W ICP power produces tensile stressedSi films with good optical properties. Further, it has been shown that between 20 Wand 26 W ICP power there is a sharp increase in the deposition rate. Consideringthe trade-offs involved between deposition rate, mechanical properties, and opticalperformance, the mirrors were fabricated with the deposition parameters for Sishown in Table 4.1.

In the fabrication of these mirrors, PI 2610 polyimide from HD microsystem wasused as the sacrificial layer to create the air cavity. Selection of an organic sacrificiallayer allowed us to use an O2 plasma to dry etch the sacrificial material, and thus

94


avoid the stiction issues associated with wet release of sacrificial layers [159]. Asper the manufacturer’s data sheet, PI 2610 cannot be spun and hard baked to athickness less than 1 µm. For fabricating an air spaced SWIR mirror operating in therange of 1600 nm – 2000 nm, the sacrificial layer thickness needs to be in the rangeof 400 nm – 500 nm. Hence PI 2610 was diluted by 50% in N-Methyl-2-pyrrolidone(NMP) solvent to attain the thickness required for forming the air cavity in SWIRmirrors. The polyimide was soft baked at a temperature of 100 ◦C for 5 min andhard baked at a temperature of 300 ◦C for 20 min. Since all our depositions werecarried out at 300 ◦C, the hard baking temperature for the polyimide was selectedas 300 ◦C.

The mirror fabrication flow diagrams are shown in Figure 4.8, and the process isdetailed below.

1. Initially a 142 nm thick silicon layer was deposited using the SENTECH SI500D ICPCVD system on a 280 µm thick sapphire substrate, as shown Figure4.8(a).

2. PI2610 polyimide was diluted by 50% in NMP, spun at 2300 rpm, soft bakedat 100 ◦C for 5 min, and then hard baked at a temperature of 300 ◦C for20 min. Dry etching of polyimide was used to pattern the sacrificial layer formirrors as shown in Figure 4.8(b). The sacrificial layer thickness was 400 nm.All dry etching steps were completed using an Oxford Instrument Plasmalab100 reactive ion etcher. The PI 2610 polyimide was etched using the etchingparameters in Table 4.2, at an etch rate of 200 nm min−1. Introducing a smallquantity of CF4, as shown in the recipe shown in Table 4.2, was found toreduces the roughness of the polyimide during etching.

3. A second 142 nm thick silicon layer was conformally deposited on top of thesacrificial layer, as shown Figure 4.8(c), using the same deposition process asthe first silicon layer.

4. As shown in Figure 4.8(d) the top Si layer was perforated with 5 µm diameteretch holes. The etch recipe for the Si is shown in Table 4.3. With theseparameters the etch rate of Si was 120 nm min−1.

5. The stress state of the silicon thin film on the polyimide was found to becompressive. Post–deposition annealing was chosen to convert this to a tensilestate. A series of annealing experiments were performed in a quartz tubefurnace in a N2 environment, keeping the annealing time fixed at 30 min.It was found that at 285 ◦C, the compressive stress is fully converted to atensile state. The sacrificial layer was ashed in an O2 plasma in a barrel asherwith 150 W RF power and a chamber pressure of 1 torr. The removal of thesacrificial layer leaves the top membrane suspended over the bottom layer,thus forming an air cavity of 400 nm thickness, as shown in Figure 4.8(e).

95


Table 4.2: Recipe for etching PI2610.

RF power Pressure Dep. temperature O2 flow rate CF4 flow rate(W) (Pa) (◦C) (sccm) (sccm)

26 2.67 26 30 3

Table 4.3: Silicon etch recipe used in fabricating mirrors.

ICP power RF power Pressure Temperature CF4flow rate(W) (W) (Pa) ◦C (sccm)

400 100 4 25 34

4.3.2 Optical characterisation of mirror array

A microscope image of a group of released mirrors is shown in Figure 4.9. Figure4.9(a) is an array of four, 270 µm diameter mirrors and Figure 4.9(b) is a close–upimage of a single mirror. A 3-D optical profile of the mirror array in Figure 4.9(a)is presented in Figure 4.10(a), and that corresponding to Figure 4.9(b) is shownin Figure 4.10(b). A line scan across the diagonal of a single mirror is shown inFigure 4.11, from which the maximum variation of the flatness across the 270 µmmirror was 20 nm. The extremely low curvature of the top membrane demonstratesthat this process can produce ultra-flat mirrors suitable for MEMS spectrometerapplications.

The optical performance of the fabricated mirrors was characterized using a spectral

(a) (b) (c)

(d) (e)

a b cc d

a b cc d

Figure 4.8: Fabrication process of SWIR mirrors. (a) Deposition of Si thin

film on sapphire substrate; (b) Spinning, hard-baking and patterning of PI

2610 polyimide ; (c) Deposition of the second Si thin film on top of polyimide;

(d) Patterning the etch holes through the top Si membrane; (e) Annealing and

dry release of mirror in an O2 plasma.

96


(a) (b)

Figure 4.9: Micrograph of the fabricated SWIR DBRs (a) A 2×2 array of

270 µm diameter DBRs (b) A magnified view of a single DBR.

(a) (b)

Figure 4.10: Optical surface profiles of the fabricated DBRs (a) Shows 3

D optical profile of a 2 × 2 array of 270 µm diameter DBRs, and (b) shows a

magnified image of a 3 D surface profile of the one of DBRs of the array shown

in (a).

and spatially resolved optical transmission measurement set-up developed in-house[160]. This system enables spatially localized measurement of calibrated spectraltransmission in the 1400 nm to 2600 nm spectral wavelength range. The lightbeam was focused to a spot size of 25 µm on the mirror surface. Figure 4.12shows the calibrated single point transmission through the mirror and the simulatedtransmission curve. The simulated transmission profile was obtained by the opticaltransfer matrix method discussed in Chapter 2, assuming an ideally flat mirrorprofile with no back-reflections from the back interface of the substrate.

Figure 4.12 shows the optical transmission measurement of this mirror, in compar-ison with that expected from the transmission matrix model. A minimum trans-mission of 0.064 was measured in the wavelength range of 1400 nm to 2300 nm.The transmission seems to be slightly less than that expected from the model. Weexpect that this small degradation is due to refractive index variations in the mir-ror silicon material. The extinction coefficient of silicon from the previous chapter

97


0.0

0.1

0.2

0.3

0.4

0.5

0 0.04 0.08 0.12 0.16 0.2 0.24 0.28 0.32

Hei

ght (µm

)

Position (mm)

Figure 4.11: Surface profile of the quarter wave mirror in Fig. 4.10, showing

the step height.

0

10

20

30

40

50

1400 1600 1800 2000 2200 2400

Tran

smis

sion

(%)

Wavelength (nm)

Figure 4.12: Measured (squares) and simulated (line) optical transmission

of a SWIR quarter wave mirror.

predicts a negligible extinction coefficient at these wavelengths. Hence, we expectnegligible absorption in the DBRs at these wavelengths.

However, with our current instrument setup, a reflection measurement in the SWIRwavelength range was not possible. Hence transmission measurements were used tocalculate the reflectance of the mirrors. Figure 4.13 shows the single point trans-mission through the mirror and the simulated transmission curve. The simulatedtransmission spectrum was obtained by the optical transfer matrix method [35],assuming an ideal flat mirror profile with no back reflections from the substrate.As shown in Figure 4.13, compared to the 5% simulated transmission of an idealMWIR mirror, the fabricated mirror shows a minimum transmission of 7%.

The optical transfer matrix model was also used to fit the measured transmisstiondata (“Measured Transmittance”) based on the known structure of the mirror. Thisoptical fit provided a better estimate of the actual thicknesses and optical materialparameters of the layers and, as a result, provides a better measurement of theactual reflectivity (“Calculated Reflectance”) of the mirror. In comparison to thesimulated 95% reflectivity, we attained 93% reflectivity in the measurements. The

98


mirrors exhibit greater than 85% reflectivity over the entire measurement range(1400 – 2300 nm). The reflectivity is close to 93% in the wavelength range 1620 -2020 nm. This indicates that this reflector is a good broadband reflector.

0

20

40

60

80

100

1200 1400 1600 1800 2000 2200 2400

Measured TransmittanceSimulated TransmittanceSimulated ReflectanceCalculated Reflectance

Tran

smitt

ance

/ R

efle

ctan

ce (%

)

Wavelength (nm)

Figure 4.13: SWIR mirror transmission and reflection. Measured Trans-

mittance: measured transmittance data; Simulated Transmittance: expected

transmittance based on a transmission matrix model using estimated thickness

of layers and initial silicon characterisation; Simulated Reflectance: expected

reflectance of mirror based on simulated transmission; Calculated Reflectance:

mirror reflectance based on measured transmission.

4.4 Si-air-Si mirrors for MWIR wavelengths

In this section, the fabrication process and optical characterization of silicon-air-silicon based DBRs operating at MWIR wavelengths is described. The fabricationprocess, optical surface profile, transmission and reflection characteristics are pre-sented.

4.4.1 Device layout

An essential consideration for fabricating large area DBRs is the robustness ofthe support structure under stress caused by the large area suspended membrane.The design used for fabricating large area DBRs is shown in Fig. 4.14. In thisdesign the suspended membrane has conformal support around the entire periphery,which provides the most robust support structure for a suspended membrane. Thefabrication and characterization of the DBRs with this design is presented in thispaper. The sacrificial material can be removed through the 5 µm diameter etch

99


holes perforated in the top suspended membrane. The adjacent etch holes are50 µm apart, thus, the average porosity of the second suspended silicon layer is lessthan 1%. With such a low porosity there is no impact on the refractive index andstress of the suspended silicon layer. In order to block stray light entering throughthe etch holes, metal pads with 8 µm diameter were formed directly under each etchhole. The presence of the metal pads reduces the effective optical area of the DBRby less than 1.5%, which is very minor and has no observed impact on the opticalperformance of the DBRs.

Underlying metal pad

Top suspended silicon membrane

Etch holes

SiOx (200 nm) thick

Silicon substrate

Figure 4.14: Structural design of large area DBRs with conformal support

around the entire periphery.


The mirrors were fabricated of various sizes, ranging from 200 µm × 200 µm to10 mm × 10 mm. The mirrors were either square or rectangular in shape. Thesilicon thin films were deposited using the deposition recipe shown in Table 4.1.

Figure 4.15 shows the fabrication flow of the Si-air-Si mirrors.

1. As shown in Figure 4.15(a) a silicon oxide layer was initially deposited on thesilicon substrate, which acts as the optical interference layer, The optimumthickness of this layer should be quarter-wave (optical), i.e. 865 nm, on topof which the first silicon layer of 350 nm thickness is deposited.

2. A chromium/gold metal bottom optical shield is then deposited and definedphotolithographically. This metal deposition is 100 nm thick and also depositsoptical-blocking metal pads underneath the subsequently etched release holes(Figure 4.15(b)).

3. Brewer science Prolift 100-16 was chosen as the sacrificial layer. As the thirdstep Apex-k1-100 adhesion promoter was spun at 4000 rpm and baked at150 ◦C for 30 seconds. Prolift100-16 was spun at 1800 rpm, soft-baked at100 ◦C for 2 minutes and then ramped to 250 ◦C for a total of 15 minutes for

100


(a) (b)

(c) (d)

(e) (f)

Figure 4.15: Fabrication process of MWIR mirrors. (a) Deposition of 350

nm Si thin film on silicon oxide coated substrate; (b) Deposition and etching

of metal optical shield; (c) Spinning, hard-baking and patterning of PL100–16

polyimide ; (d) Conformal deposition of the second Si thin film on polyimide;

(e) Patterning the etch holes through the top Si membrane; (f) Dry release of

mirror in an O2 plasma.

a full cure. The sacrificial layer was 1050 nm thick, and was wet patterned inAZ326 developer (Figure 4.15(c)).

4. After the removal of photoresist, a descum in oxygen plasma was performedto improve conformality and adhesion of the next silicon layer. On top of thesacrificial layer a second quarter wave thick silicon layer was deposited, whichconformally covers the sacrificial layer. The top silicon layer was perforatedwith 5 µm diameter etch holes using CF4 plasma (Figure 4.15(d) and (e)).

5. With the Prolift 100-16 sacrificial layer, the 350 nm thick silicon membraneswere tensile stressed as-deposited, and no post-deposition treatment was re-quired to tune the stress. The sacrificial layer was etched in an O2 plasma ina barrel asher with 160 W RF power and 133.3 Pa chamber pressure. Theremoval of the sacrificial layer leaves the top membrane suspended over thebottom layer thus forming an air cavity of 1050 nm (quarter-wave optical), asshown in Figure 4.15(f).

4.4.3 Optical characterisation of MWIR DBRs

The quality of the MWIR mirrors fabricated in this process was quantified throughoptical characterization of the mirrors. A microscope image of a fabricated mirroris shown in Figure 4.16. The 3-D optical profile of the 2 mm × 2 mm mirror ispresented in Figure 4.17, and a line scan across the same mirror is shown in Figure4.18. The sharp notches at the two ends of the optical profile are due to artifactarising from the limitation of the optical measurements over sharp edges. Note thatthe variation of the flatness variation across the 2 mm × 2 mm mirror was found tobe in the range of 20 nm – 30 nm. This profile is taken by avoiding the etch holes.Such a variation is only 2% – 3% of the air gap created and, as the transmissioncharacteristic will show, has negligible effect on the optical properties. Thus the

101


fabrication process demonstrates a high level of success in producing large-area,ultra-flat mirrors in the MWIR range.

Figure 4.16: Micrograph of fabricated 2 mm × 2 mm MWIR DBR.

Figure 4.17: 3-D optical surface profile of 2 mm × 2 mm MWIR DBR.

The optical transmission of a quarter-wave-mirror was investigated using a Spec-trum One FTIR spectrometer by Perkin Elmer. This system enables measurementof spectral transmission and absorption in the spectral range from 1280 nm to28500 nm with high accuracy. However with our current instrument setup a specu-lar reflection measurement in the MWIR wavelength range was not possible. Hencetransmission measurements were used to calculate the reflectance of the mirrors.Figure 4.19 shows the single point transmission through the mirror and the sim-ulated transmission curve. The simulated transmission spectrum was obtained bythe optical transfer matrix method [35], assuming an ideal flat mirror profile withno back reflections from the substrate. As shown in Figure 4.19, compared to 4.5%simulated transmission of an ideal MWIR mirror, the fabricated mirror shows aminimum transmission of 6%. Since silicon thin films have negligible extinctioncoefficient in the mid-wave infrared range. We expect negligible absoprtion in the

102


0

0.2

0.4

0.6

0.8

1

1.2

0 0.5 1 1.5 2 2.5 3

Hei

ght (

µm)

Distance (mm)

Figure 4.18: A line scan across the 3-D optical surface profile of a 2 mm ×2 mm mirror avoiding the etch holes.

DBR. Hence, in comparison to the theoretical 95.5% reflectivity we attained a mea-sured 94% reflectivity. The mirrors have more than 85% reflectivity for a 2720 nmwavelength range (3680 – 6400 nm). The reflectivity is close to 93% in the wave-length range 4100 - 4910 nm.

4.5 Silicon-silicon oxide based freely suspended

mirrors and fixed-filters

An issue to address for wavelengths shorter than 1.1 µm is that silicon substratesbecome heavily absorbing. As such, it may be necessary for the substrate supportingthe MEMS tunable filter to be etched away to create an opening in the opticallyactive area. Instead of using an array of individual on-pixel tunable filters, wepropose to fabricate an adaptive FPA that is based on a single large-area tunableoptical filter for the entire imaging FPA. This section also examines aspects of thefilter technology, including (1) removal of the substrate below the mirror to allowsilicon mounted filters to be hybridized and operate at visible and NIR wavelengthsand (2) fabrication of complete fixed-filter devices that are freely suspended forfixed-filter applications.

The requirements for integrating a MEMS-based tuneable optical filter to a large-area imaging FPA are:

1. The fabrication technology should be scalable to enable filters with dimensionsfrom a few hundred microns to several millimetres, which will allow them to beintegrated with either a single-element detector or with a large-area imagingFPA.

103


0

20

40

60

80

100

3000 4000 5000 6000 7000 8000

Measured TransmittanceSimulated TransmittanceSimulated ReflectanceCalculated Reflectance

Tra

nsm

ittan

ce /

Ref

lect

ance

(%)

Wavelength (nm)

Figure 4.19: MWIR mirror transmission and reflection. Measured Trans-

mittance: measured transmittance data; Simulated Transmittance: expected

transmittance based on a transmission matrix model using estimated thickness

of layers and initial silicon recipe characterisation; Simulated Reflectance: ex-

pected reflectance of mirror; Calculated Reflectance: mirror reflectance based

on measured transmission.

2. The optical layers must be extremely flat and free from any curvature thatwill otherwise degrade the peak transmission, the full width at half maximum(FWHM) of the filter, and the out-of-band rejection.

3. The filter and mirrors should have close to ideal optical performance in termsof transmittance and reflectance.

4. The large-area filter and mirrors must have a high degree of optical uniformityover the entire optically active area of the imaging FPA, in order to ensurethat all pixels in the FPA receive light that has been filtered in the samemanner.

5. The resonant modes of the suspended DBR mirror(s) must be mismatched tothe low-frequency vibrational excitation spectrum from the environment, toensure movement of the MEMS device is only due to deliberate actuation.

6. The filters need to be actuated by some well-established actuation method,such as electrostatic actuation.

In this section we will demonstrate a MEMS technology that meets all of the firstfive requirements, and we will present a clear pathway towards achieving the sixthrequirement.

104


4.5.1 Experimental Steps

In order to fabricate freely suspended DBRs, silicon has been used as the high re-fractive index material while silicon oxide (SiOx) has been used as the low refractiveindex material in the DBRs. All optical layers were deposited on a 300 µm 〈100〉 ori-ented silicon substrate using a Sentech SI500D inductively coupled plasma chemicalvapour deposition (ICPCVD) system. An Oxford Instruments Plasma 100 reactiveion etcher was used to perform low-temperature back-side etching of the substratewafer using SF6/O2 gases. This back-side etching process gives a high etch rate,vertical sidewalls, and a small degree of undercut. A Zygo Newview white lightoptical surface profiler was used to measure surface flatness of the suspended de-vices, an in-house built optical metrology system [160] was used to measure spatialdistribution of optical performance, and a Polytec MSA 500 Micromotion Analyzerwas used to investigate the vibrational behaviour.

4.5.2 Fabrication Process

This section describes the dielectric material growth conditions and device fabri-cation steps. Table 4.4 gives the deposition parameters for ICPCVD SiOx films,while Table 4.5 gives the deposition parameters for the Si films. In order to mea-sure stress in the Si and SiOx films, the thin films were deposited on 70 µm thick〈100〉 oriented Si substrates. Measuring the radius of curvature of the compositedeposited film and substrate films by surface profilometry, and applying Stoney’sformula [96, 98], thin film stress was calculated. The as-deposited Si thin films werefound to have 560 MPa of residual compressive stress, and the as-deposited SiO2

layer had 200 MPa of residual compressive stress.

Table 4.4: ICPCVD SiO2 deposition parameters.

ICP Pressure Temperature Flow rate (sccm)power SiH4 He N2O Ar(W) (Pa) (◦C)

450 2 130 6.5 123 70 120

Table 4.5: ICPCVD deposition parameters for Si thin films.

ICP power Pressure Temperature Flow rate (sccm)(W) (Pa) (◦C) SiH4 He

300 4 100 5 95

The Fabry-Perot filters were fabricated on 300 µm thick double-side polished 〈100〉Si substrates. A 200 nm thick SiOx layer was deposited on the front side as an etchstop layer for the low-temperature ICPRIE back-side silicon etching process (Figure4.20(a)). Figure 4.20(b) represents two different optical structures fabricated ontop of the 200 nm thick etch stop SiOx layer. The first structure was a Si-SiOx-Si

105


distributed Bragg reflecting (DBR) mirror and the second structure was a Si-SiOx-Si fixed-cavity Fabry-Perot optical filter. It is to be noted that the design thicknessof the optical layers was calculated for a SWIR wavelength of 2000 nm. The Silayers were targeted to have quarter-wave thickness of 145 nm. The SiOx layerswere targeted to have a quarter-wave thickness for the DBR of 360 nm, and a half-wave thickness of 714 nm for the Fabry-Perot filter cavity. Table 4.6 shows thetargeted and actual thicknesses for the optical layers.

(a) (b)

(c) (d)

(e)

Figure 4.20: Fabrication process flow of freely suspended Fabry-Perot op-

tical filters and optical mirrors: (a) deposition of SiOx etch stop layer (b)

deposition of the optical layers (c) patterning of the back-side of the handle

wafer with AZ 5214 resist (d) cryogenic back-side etching of handle wafer (e)

removal of SiOx etch stop layer in HF to leave the final suspended Fabry-Perot

optical filters and DBR mirrors.

As shown in Figure 4.20(c), AZ 5214 (positive) photoresist was patterned on theback-side of the Si substrate using a standard photolithography process. Subse-quently, to achieve suspended large-area structures, a low-temperature SF6/O2 IC-PRIE anisotropic etch with a high Si/SiOx selectivity was used to etch through theback-side of the Si substrate (Figure 4.20(d)). In the final step shown in Figure4.20(e), a selective HF etch was used to remove the 200 nm etch-stop SiOx layer

106


from the area directly underneath the final device structures.

Table 4.6: Targeted and actual thicknesses of the SWIR optical layers.

Layer DBR FilterTargeted (nm) Actual (nm) Targeted (nm) Actual (nm)

Si 145 136 145 150SiOx 360 387 714 683

Si 145 153 145 140

Devices with size varying from 500 µm × 500 µm to 5 mm × 5 mm were fab-ricated. One such fabricated chip is shown in Figure 4.21. This meets the firstrequirement of having a technology that can produce free-standing filters span-ning the size requirements from small single-element detectors to large-area FPAs.As shown in Figure 4.21, due to the compressive stress in the Si and SiOx, thefabricated and released structures exhibited significant surface distortions, whichnecessitated a post-fabrication flattening process. In Chapter 3 it was discussedthat a post-fabrication anneal of the ICPCVD deposited Si layer was an effectivemeans for tuning stress. The fabricated structures were annealed in a quartz-tubefurnace in a N2 environment. The critical anneal temperature to convert compres-sive films to tensile, was approximately 350 ◦C. However, due to the presence ofthe compressively stressed SiOx, both as an optical layer and an etch stop layer,the critical annealing temperature of the optical structures shifted to the range of400 – 420 ◦C. Additional studies have shown that in this temperature range theSiOx stress is not modified, and that annealing only modifies the stress in the Sithin film layers.

Figure 4.21: Fabricated SWIR devices before annealing showing significantdistortions in the optical layers due to high compressive stress. The coin is anAustralian one dollar piece of diameter 25 mm.

107


-40

-30

-20

-10

0

10

20

30

40

0 500 1000 1500 2000 2500

Hei

ght (

nm)

Distance (µm)

Available optical area

Figure 4.22: Optical surface profile of a 2 mm × 1 mm SWIR DBR after a

40 minute anneal at 400 ◦C in N2 atmosphere. The central 1.5 mm section is

flat to within 10 nm.

4.5.3 Optical characterization of SWIR devices

Optical surface profilometry of suspended devices

Figure 4.22 shows the measured optical surface profile of a 2 mm × 1 mm suspendedDBR. After annealing at 400 ◦C for one hour we see a total of 30 nm of bowing alongthe 2 mm length of the DBR. The DBR is flat to within 10 nm across the 1.5 mmcentral portion of the 2 mm length. The remainder of the bowing is concentratednear the anchor sections where the device attaches to the substrate. This highdegree of flatness is key to fabricating a tunable large area filter with two ultra-flatDBRs, which is essential to satisfying the second requirement of the process.

Figure 4.23 shows an optical surface profile of a suspended SWIR Fabry-Perot filterof size 2 mm × 1 mm. After annealing at 420 ◦C for one hour in a N2 environment,the filter shows a bowing of 220 nm along its length. This upward bowing at the endsof the filter may be associated with the “anchor effect” caused by a stress differentialbetween the SiOx and Si membrane layers at the anchor location. It is also likelythat at least part of this effect at the end is an artifact of the surface profilometer,induced by the sudden optical phase change at the edges. In the central section,the filter has a flatness variation of the order of 25 nm along 1.25 mm of its length.

Single point transmission spectra of fabricated SWIR mirrors and filters

In order to meet the third requirement of the process, that is to obtain close totheoretical optical performance, optical modelling of the mirror and filter was un-dertaken using the optical transfer matrix method [35], which assumed an ideal flat

108


50

100

150

200

250

-500 0 500 1000 1500 2000

Hei

ght (

nm)

Distance (µm)

Available optical area

Figure 4.23: Optical surface profile of a 2 mm × 1 mm Fabry-Perot filter

after a 1 hour anneal at 420 ◦C in a N2 atmosphere. The central 1.25 mm

section is flat to within 25 nm.

surface profile. The modelling results were compared to a single point measurementof the optical transmission through the structure using an in-house developed stan-dard bench top optical transmission measurement system. This system enables themeasurement of single point spectral transmission through the device under testat wavelengths from 600 nm to 2400 nm. The transmission was measured withthe light beam focused on the device under test, to a full-width at half-maximum(FWHM) spot size of 500 µm. Figure 4.24 shows the transmission spectrum of theFabry-Perot filter. The measured spectrum indicates that the peak of the filter is at1940 nm, which is close to the design target peak at 2000 nm. The FWHM spectrallinewidth of the filter was found to be 250 nm with a peak transmission of 94%,which matches very well with the simulation results. This fixed cavity filter has a750 nm free spectral range (FSR) and a finesse of 3.

Figure 4.25 shows the single point transmittance measurement through the DBRand model fitting to extract the reflectance. The modelled transmission profile(Model Tx) was obtained by the optical transfer matrix method assuming an idealflat mirror profile, using previously measured optical constants of ICPCVD de-posited silicon and silicon oxide films, and layer thicknesses measured during mirrorfabrication. This model predicts a best-case reflectance (Model Rx) of 90%.

The optical transfer matrix model was also used to fit the measured transmissiondata (Measured Tx) based on the known structure of the DBR. This fit providesa better estimate of the actual thicknesses and optical material parameters of thelayers and, as a result, provides a better estimate of the actual reflectivity (Esti-mated Rx) of the DBR. The measured transmission spectrum shows a peak at 1200nm and a plateau after 1500 nm. From 1560 to 2050 nm the transmission is in thevicinity of 10%. The measured transmission is in good agreement with the model.The estimated reflectance is also in good agreement with the modelled reflectance.

109


0

20

40

60

80

100

1200 1400 1600 1800 2000 2200 2400

Measured

Simulated

Tran

smitt

ance

(%)

Wavelength (nm)

Figure 4.24: Measured and simulated optical transmission spectra of the

SWIR filter shown in Fig. 4.23. Figure shows a close match between calculated

and simulated transmittance.

0

20

40

60

80

100

800 1000 1200 1400 1600 1800 2000 2200 2400

Model TxMeasured TxEstimated RxModel Rx

Tran

s/R

efl (

%)

Wavelength (nm)

Figure 4.25: Measured and modelled optical transmission and reflection

spectra of the suspended SWIR mirror shown in Fig. 4.22. Measured Tx:

measured transmittance data; Model Tx: mirror transmittance based on a fit

to the mirror structure using a transmission matrix model; Model Rx: ex-

pected reflectance of mirror; Estimated Rx: mirror reflectance based on fitted

transmission.

Interestingly the mirrors have almost 85% reflectance for a 700 nm range (1500-2200 nm). As such, this is a good broadband reflector for spectroscopic imagingsystems. A good agreement between measurement and model confirms the highoptical quality of the materials, and excellent flatness of the structure. As such, thethird requirement of fabricating a high quality optical device has been met.

110


01

23

4

01

23

405

1015

X (mm)Y (mm)

Min

imum

Tra

nsm

issi

on (%

)

0.0

5

10.0

Figure 4.26: Spatially resolved measurements of calibrated minimum trans-

mission of a 3 mm × 3 mm quarter-wave SWIR DBR.

Spatially mapped transmission profile of large-area quarter wave SWIRmirrors

To meet the fourth requirement, the large area filters and mirrors must have goodoptical spatial uniformity over the entire optically active area. The spatial uni-formity of a 3 mm × 3 mm quarter-wave SWIR mirror was investigated using anin-house developed optical metrology system [160]. This system enables the map-ping of calibrated transmission spectra across the full area of the mirror within thespectral band from 1400 nm to 2600 nm. Measurements were carried out across aphysical grid of 4.0 mm × 4.0 mm which included the 3 mm × 3 mm DBR. Figure4.26 shows the calibrated minimum transmission as a function of spatial positionacross the DBR surface. The light beam was focused down to a spot size of 25 µmon the mirror surface and the measurements were undertaken with a step size of50 µm in both the X and Y directions, while the monochromator was stepped inwavelength increments of 20 nm.

Histograms for the measurements on the mirror surface are presented in Figure 4.27.The minimum transmission can be characterized by µ ± σ = 0.100 ± 0.003, whichcorresponds to a 3 % variation across the mirror surface for minimum transmission.These optically homogeneous DBRs demonstrate that the fourth requirement forfabricating spatially uniform mirrors has been met.

4.5.4 Dynamic properties of SWIR DBR

To test the mechanical robustness and utility of the optical structures in dynamicoptical MEMS, and to assess immunity to spurious dynamic movements at lowfrequencies, the frequency response of the structure was measured for a 1 mm ×1 mm square DBR. The measurement setup was similar to that of Tsai et al. [161],

111


Figure 4.27: Spread of minimum spectral transmission from Figure 4.26.

in which a piezoelectric stack was used as an impact hammer to excite mechani-cal vibrations in the mirror. The resultant vibrations were measured using a laserdoppler vibrometer attached to an optical microscope. To eliminate air damp-ing, the sample was mounted in an evacuated chamber at a pressure of 0.1 Torr.Suspended membrane deflection was measured with reference to the substrate inorder to remove common-mode whole-sample movement. Figure 4.28 shows themeasured vibrational modes of the suspended mirror, which clearly indicates thepresence of six flexural resonant modes beginning with the fundamental vibrationalmode at the lowest frequency of 39 kHz. Such a high fundamental frequency sug-gests a predictable and fast dynamic response of the optical structure combinedwith immunity to low frequency stimuli from spurious external vibrations.The lowfrequency resonance at 19.5 kHz is a consequence of coupling from the electroniccircuitry. These measurements were corroborated with a finite element simulationof natural resonant frequencies of the suspended mirror using CoventorWare 2010MEMS design suite [162]. The mechanical properties of the silicon and silicon ox-ide thin films were measured and used as an input to the simulated model. Theresults on modal response are summarized in Table 4.7, which also compares thefinite element modelling results with the experimental observations. The modelindicates that in the 10 kHz to 100 kHz frequency range, five modes of vibrationare present, where the fundamental mode [1,1] occurs at approximately 40 kHz.Other modes and natural frequencies are also indicated in Table 4.7, and it is notedthat the modelled results and experimental measurements are in general agreementwithin approximately 5%. These variations can be attributed to the measurementaccuracy of the material properties. A high fundamental frequency value of 39 kHzsuggests a predictable and fast dynamic response of the optical structure combinedwith immunity to low frequency stimuli from spurious external vibrations.

112


Table 4.7: Resonant frequencies of 1 mm × 1 mm square shaped suspended

SWIR DBR.

Mode Finite Element Solution Experimental(kHz) (kHz)

(1,1) 40.91 39(1,2), (2,1) 58.63 56.5

(2,2) 73.2 68.7(3,1), (1,3) 83 80.3(3,2), (2,3) 94.56 93.2

Figure 4.28: Flexural modes in 1 mm × 1 mm suspended SWIR DBR

showing the fundamental and harmonic flexural modes.

4.6 Summary

Based on the results from the previous chapter, optimum silicon and silicon oxidedeposition recipes were selected for fabrication of SWIR and MWIR wavelengthrange DBRs. At first, two different DBR structures were discussed and merits offull peripheral support based structure for generating flat suspended membraneswas established. Next, fabrication process and optical characterization of silicon-air-silicon based DBRs for the SWIR and MWIR wavelengths was presented. Itwas shown that the given fabrication technology can be used to fabricate DBRranging from ranging from 200 µm × 200 µm to 5000 µm × 5000 µm size. Thevariability of sizes allows them to be used in conjunction with either single-elementphotodetectors or large-area focal plane arrays. The fabricated DBRs demonstrate a20-30 nm variation in flatness across several millimeters length. Single point spectralmeasurements on devices showed excellent agreement with simulated optical models.

113


The fabricated distributed Bragg reflectors had around 94% reflectivity comparedto the theoretical 95-96% reflectivity. The chapter concluded with discussion onoptical and mechanical properties of freely suspended silicon-silicon oxide-siliconbased DBRs for the SWIR wavelength range.

114

Chapter 5

Fabry-Perot filters for the SWIRand MWIR wavelength ranges

5.1 Introduction

This chapter presents the fabrication and characterization of silicon based Fabry-Perot filters for the SWIR and MWIR wavelength bands. This includes filterscontaining (1) single-layer silicon mirrors, (2) multi-layer solid mirrors with siliconas one material, and (3) silicon-air-silicon mirrors. Section 5.2 describes the designand limitations of the tether and beam based MEMS structures for fabrication oftunable silicon based Fabry-Perot filters. Section 5.4 provides details of the etchback technology employed in the fabrication of actuation structures and filters.Section 5.5 gives details of a notch based membrane suspension structure used inthe fixed cavity and tunable Fabry-Perot filters. Section 5.6 shows fabrication andactuation characterization of thin suspended structures utilizing the processes andstructures described in the earlier sections, and also gives the actuation characteri-zation of these membranes. Section 5.7 details fabrication, optical characterizationand actuation voltage vs displacement results of a tunable multi-spectral MWIRfilter. Section 5.8 describes the fabrication process and optical characterization ofa SWIR filter based on silicon-air-silicon mirrors.

5.2 Tether and beam based Fabry-Perot filters

5.2.1 Tether and beam based structures

Microelectronics research group at the University of Western Australia has in thepast used tether and beam based structures to successfully fabricate tunable Fabry-Perot filters. These structures offer a large actuation range for the filters [163], andin excess of the 1/3–gap actuation limit of traditional MEMS structures. Figure

115

CHAPTER 5. FABRY-PEROT FILTERS FOR THE SWIR AND MWIRWAVELENGTH RANGES

5.1 (a) shows one such design. This particular design uses angled silicon nitridetethers for stress-releif in the suspended mirrors, as the stress in Ge-SiO-Ge mirrorcaused severe deformation of actuators of previous prototype devices [28]. Figure5.1 (b) shows a straight tether based structure, which was used by Mao et al. [164]for fabrication of germanium based long wave infrared filters. In this structuregermanium was used as the mirror and beam material. It was identified that forthe germanium based filters the angled tethers were producing significant distortion.Hence authors opted for a straight tether based structure. Figure 5.1 (c) shows theschematic of the tether and beam based structure used in this thesis. The circulartop mirrors are 100 µm and 150 µm in diameter. The square top mirrors are of size100 µm × 100 µm and 150 µm × 150 µm. The width of beam and tether are 30 µmand 15 µm, respectively. The diameter of etch holes is 5 µm, which are used in therelease process to improve access for the etchant to the sacrificial material.

(a)

115

Figure 5.19: Lateral arrangement of the filters, listing the dimensions common to all filters and defining the parameters that vary between filters: tether angle, mirror diameter, aperture diameter and beam length. The width of the tethers was 10 μm on the mask, but was less on the fabricated devices because of lateral etching under the mask during the final silicon nitride etch.

20 μm

50 μm

Mirror diameter, Dm (80 μm to 220 μm)

Aperture diameter, DA (Dm - 10 μm)

Tether angle, θt

Less than10 μm

Beam length, L (Dm + 40 μm)

SEM of filter with Dm = 80 μm, θt = 50º

Figure 5.18: Cross-section showing the mirror profile, optical path and optical shield of a tuneable Fabry-Perot filter. The figure also defines the mirror and aperture diameter, which differ by a total of 10 μm in all mirrors. All fixed-fixed beams used to actuate the filters have a width of 20 μm.

Depicted cross-section

Compressive silicon nitride

Low-stress silicon nitride

Aperture diameter, Dm-10 μm

Mirror diameter, Dm

5 μm

20 μm

Optical path

Stray light blocked by bottom electrode

(b) (c)

(c) (d)

(e) (f) Fig. 4. Fabrication process flow of LWIR tunable Fabry-Perot filters

A scanning electron microscope (SEM) image with structural dimensions of the fabricated LWIR filter is shown

in figure 5. In this design, the length of actuator beams and the mirror diameter are the same. For comparison purposes, devices with two different mirror diameters (D = 150 μm and 200 μm) were fabricated. In order to characterize the dependence of actuation voltage on electrode separation, devices with 5.5 μm and 3 μm actuation gap length were fabricated by utilizing 0.5 μm and 3 μm thick lower polyimide layers, respectively.

Fig. 5. SEM image of a fabricated LWIR filter and its structure dimensions

3. Device characterization

Whereas ideal Fabry-Perot filters have smooth, flat, parallel mirrors, the mirrors in a real filter exhibit imperfections due to surface/interface roughness, mirror tilt and mirror curvature, which tend to broaden the bandwidth and reduce the transmittance of the filter. While roughness and tilt can be eliminated by carefully handling the fabrication process, mirror curvature, which is essentially induced by a stress gradient profile within the mirror, needs additional effort to eliminate. It has been demonstrated in our SWIR and MWIR filters that a PECVD SiNX layer as thin as 50 nm deposited underneath the Ge-SiO-Ge mirror can successfully be deployed as a stress compensator, and can be used to reduce the curvature of the mirror. Considering the highly absorbing nature of SiNX in the LWIR band, a much thinner SiNX layer must be used in order to ensure that no significant absorption is incurred, while also providing sufficient stress balancing. As such, all SiNX layers used must be highly stressed. In this work, SiNX with a predetermined compressive stress of 150 Mpa was used. In order to determine the optimal thickness of the SiNX compensation layer, four samples were prepared with different deposition times of 0, 14, 21 and 28 s, corresponding to layer thickness of 0, 4, 6 and 8 nm, respectively. Both curvature, which is defined as reciprocal of radius of curvature, and center-to-edge bowing of released filters were measured using an optical profilometer, and the results are summarized in figure 6(a) and (b), respectively.

D 30 μm

60 μm

20 μm

Etch Holes

30 µm

Etch Holes

30 µm

D

60 µm

15 µm

Bottom Electrode

Figure 5.1: Tether and beam based structures used by earlier works per-formed at microelectronics research group: (a) Angeled tether based filter forSWIR and MWIR wavelength range operation [28]; (b) straight tether basedfilter for LWIR wavelength range operation [164]; (c) straight tether basedstructure used in this thesis for fabricating Si-air-Si suspended mirror. Notethat the etch holes size is exaggerated for clarity.

116


5.2.2 Hex cell based structures

An air-spaced suspended mirror with two 150 µm wide membranes, separated by anarrow air gap, has a high probability that with such a narrow gap, the two mem-branes of the mirror may collapse on each other. In this work an experiment wasperformed to examine structures to provide better support. It is well known thata hexagonal structure provides good rigidity in the plane of the structure. There-fore, as an experiment hexagonal “ribbing” was provided to improve the rigidityof the suspended mirror and reduce the probability of any collapse due to mirrordistortion. To overcome the optical degradation induced by the etch holes and theribbing, a metal light-blocking shield on the substrate just below the etch holes andthe ribbing was deposited.

Figure 5.2 shows the hexagonal cells based structures for circular and square shapedmirrors, which were attempted to create a robust MEMS structure for the topsuspended mirrors. Each hexagonal cell is 50 µm in diameter with a 5 µm diameteretch hole in the center. Each hexagonal cell is connected to adjacent cells via a5 µm wide connecting channel.

Figure 5.2: Hexagonal cells based structures for circular and square shapedmirrors.

5.2.3 Fabrication process of silicon-air-silicon based filters

The center wavelength of the SWIR filter was chosen to be 2000 nm. The followingfabrication process was used to fabricate filters based on silicon-air-silicon DBRs:

1. As shown in Figure 5.3 (a) the first quarter-wave-thick (i.e. 141 nm) amor-phous silicon was deposited, using the recipe given in Table 5.1, on a 280 µmthick sapphire substrate .

2. A 40 nm-thick Aluminum metal pattern was deposited using liftoff (see Figure5.3 (b)). This step produces the thin bottom electrodes, and metal pads toblock the stray light through the etch holes. Aluminum has a very highabsorption coefficient in the SWIR wavelength range as compared to gold andchromium. Hence, it is a better choice for blocking stray light, as only a verythin layer is needed.

117


3. 50% diluted PI2611 polymide was spun at 1700 rpm, soft-baked at 130 ◦Cfor five minutes and hard-baked at 300 ◦C for 20 minutes. The 500 nm thickpolyimide was patterned in an O2/CF4 plasma at room temperature, whichserved as the sacrificial layer for the bottom silicon-air-silicon DBR (see Figure5.3 (c)). The etch recipe for polyimide is shown in Table 5.2.

Table 5.1: Silicon deposition recipe.

ICP Power Deposition Temp. Pressure SiH4 flow He flow

(W) (◦C) (mTorr) (sccm) (sccm)

26 300 30 5 95

4. A 141 nm thick second silicon thin film was then deposited as shown in Figure5.3 (d).

5. PI2610 polymide was spun at 4000 rpm, soft-baked at 130 ◦C for five minutesand hard-baked at 300 ◦C for 20 minutes. The 1200 nm thick polyimidewas patterned for the anchors in an O2/CF4 plasma at room temperature see(Figure 5.3 (e)). This layer serves as the sacrificial layer for the tunable cavitygap. The etch recipe for polyimide is shown in Table 5.2.

6. A third silicon thin film was deposited on top of this polyimide sacrificial layer(see Figure 5.3 (f)).

7. As shown in Figure 5.3 (g) 20% diluted PI2611 polymide was spun at 1700 rpm,soft-baked at 130 ◦C for five minutes and hard-baked at 300 ◦C for 20 minutes.The resulting 500 nm-thick polyimide was patterned in an O2/CF4 plasma atroom temperature, to serve as the sacrificial layer for the top silicon-air-siliconDBR. The etch recipe for polyimide is shown in Table 5.2

8. The fourth and last 141 nm silicon layer was then deposited (see Figure 5.3(h)).

9. A metal lift-off was performed to deposit 70 nm thick aluminum actuationelectrodes (see Figure 5.3 (i)) .

10. A final metal lift-off step was performed to deposit 100 nm thick aluminumfor the contact pads and power supply tracks.

Table 5.2: Recipe for etching PI2610.

RF power Pressure Dep. temperature O2 flow rate CF4 flow rate(W) (Pa) (◦C) (sccm) (sccm)

26 2.67 26 30 3

118


11. 5 µm diameter etch holes were perforated through the top three silicon andtop two polyimide layers (see Figure 5.3 (j)).

12. A Cr/Au (10 nm/60 nm) stray optical signal blocking layer was deposited onthe backside of the substrate (see Figure 5.3 (k)). Unlike Cr/Au, aluminumcan be etched by photo-resist developer, hence, choosing Cr/Au for backsideoptical shield, allowed further processing of the wafer without any need ofprotection of the backside metal layer. Of course, if this step is chosen as thelast step than aluminum can be used as the backside optical shield metal.

13. A final dry etching step was used to pattern the top DBR and beams.

14. The filters were annealed, pre-release, at 310 ◦C for 30 minutes in a N2 envi-ronment in a quartz tube furnace.

15. The fabricated filters were dry released in a March barrel asher in an O2

plasma at 150 W RF power at 1 Torr pressure (see Figure 5.3 (l)).

5.2.4 Optical Surface profilometry of filters

Tether and beam based filters in SWIR wavelength range

Figure 5.4 shows the optical surface profile of fabricated filters. Figure 5.4 (a)and (b) shows that in the center 60-80 µm diameter, the filters are flat to within20 nm. Each Hex-cell in the Figure 5.4 (c) is flat to within 10 nm. However,each freely suspended top DBR has a tilt of 140-250 nm. This high tilt is due togradient stress in the tethers. Furthermore, a line scan of surface profile, across anactuation-beam of a filter (Figure 5.5) depicts a “gutter” like profile. As a result ofthis actuation beam distortion, we expect large actuation voltages will be requiredto actuate these beams. This distortion most likely arises due to different stressescaused by differences in the thermal history between the third and fourth siliconlayers. Thus, it can be seen that tether and beam based structures are very muchaffected by gradient stress, resulting in a “guttering” effect on the beams, and wesee an obvious need to improve this actuation structure.

5.2.5 Tether and beam based structures in visible and NIRwavelength range

The tether and beam based design was used to fabricate suspended top DBRs for thevisible and NIR wavelength range on a silicon substrate. This test was performed toevaluate the suitability of this design for ultra-thin silicon film based MEMS filters.The following steps were used for fabrication of suspended top DBRs:

119


(a) (b)

(c) (d)

(e) (f)

(g) (h)

(i) (j)

(k) (l)

Figure 5.3: Fabrication process of SWIR tunable Filter. (a) Deposition of141 nm silicon thin film on sapphire substrate; (b) deposition and definitionof bottom electrode; (c) spinning, hard-baking and patterning of PI2611 poly-imide; (d) deposition of 141 nm silicon thin film for top membrane of bottommirror; (e) spinning, hard-baking and patterning of PI2610 polyimide; (f) de-position of 141 nm silicon thin film for bottom membrane of top mirror; (g)spinning, hard-baking and patterning of PI2611 polyimide; (h) deposition of141 nm silicon thin film for top membrane of top mirror; (i) deposition anddefinition of top electrode; (j) patterning top mirror and etch holes; (k) depo-sition of the optical metal shield on the back side of substrate to define theoptically active area; (l) pre-release annealing and dry release in an O2 plasma.120


(a)

0

500

1000

1500

2000

2500

3000

3500

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Hei

ght (

nm)

Distance (mm)

(b)

0

500

1000

1500

2000

2500

3000

3500

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

Hei

ght (

nm)

Distance (mm)

(c)

0

500

1000

1500

2000

2500

3000

3500

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

Hei

ght (

nm)

Distance (mm)

Figure 5.4: Optical surface profile and a line scan across the fabricatedsilicon-air-silicon DBR based Fabry-Perot filters; (a) Circular shaped filter;(b) Square shaped filter; (c) Hex-cell based filter.

1. As shown in Figure 5.6 (a) 20% diluted PI2611 polymide was spun at 1700 rpm,soft-baked at 130 ◦C for five minutes and hard-baked at 300 ◦C for 20 minutes.The 500 nm thick polyimide was patterned, for the anchors in an O2/CF4

plasma at room temperature. This layer also served as the sacrificial layer forthe tunable cavity gap. The etch recipe for polyimide is shown in Table 5.2.

121


0

500

1000

1500

2000

2500

3000

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04

Hei

ght (

nm)

Distance (mm)

Figure 5.5: A line scan across the width of a beam of fabricated silicon-air-silicon DBR based filters showing gutter-like profile.

2. Next, a one qarterwave thick silicon thin film (48 nm for visible and 70 nm forthe NIR wavelength range) was deposited on top of this polyimide sacrificiallayer (see Figure 5.6 (b)).

3. As shown in Figure 5.6 (c) PI 2610 polymide was 50% diluted by N-Methyl-2-pyrrolidone solvent to attain the thickness required for creating the air cavityin NIR and visible mirrors. In order to achieve quarter wavelength thicknessof the sacrificial layer, diluted PI 2610 was spun at 5600 rpm for the visiblemirrors and at 4500 rpm for the NIR mirrors, respectively. It was soft-bakedat 130 ◦C for 5 minutes, and then hard-baked at 300 ◦C for 20 minutes. Thispolyimide was patterned in an O2/CF4 plasma at room temperature, whichserved as the sacrificial layer for the top silicon-air-silicon DBR. The thicknessof polyimide layer for visible and NIR wavelength mirrors was 170 nm and250 nm, respectively.

4. The second quarter-wave thick silicon layer was then deposited (see Figure5.6 (d)).

5. 5 µm diameter etch holes were perforated through the top two silicon and toppolyimide layers. The top mirror, beam and tether were patterned by dryetching (see Figure 5.6 (e)).

6. The visible and NIR structures were pre-release annealed at 315 ◦C and 330 ◦C,respectively, for 30 minutes in N2 environment in a quartz tube furnace. Thefabricated structures were dry released in a barrel asher in an O2 plasma at150 W RF power at 1 Torr pressure (see Figure 5.6 (f)).

The optical surface profile and a line scan across it, for released NIR and visiblewavelength structures are shown in Figure 5.7. Figure 5.7 (a) and (b) show thatthese ultra-thin silicon film based structures could not survive the release process,

122


(a) (b)

(c) (d)

(e) (f)

Figure 5.6: Fabrication process of suspended top mirror for NIR and visiblewavelength range. (a) Spinning, hard-baking and patterning of 20% dilutedPI2611 polyimide; (b) deposition of quarterwave thick silicon thin film forbottom membrane of top mirror; (c) spinning, hard-baking and patterningof 50% diluted PI2610 polyimide; (d) deposition of second quarterwave thicksilicon thin film for top membrane of top mirror; (e) patterning top mirror andetch holes; (f) pre-release annealing and dry release in an O2 plasma.

and they had collapsed onto the substrate. At first, it was observed that for thesestructures the tethers and beams were not strong enough to support the suspendedsilicon–air–silicon DBRs. Hence, a new structure was prepared. In the new struc-ture 300 nm thick silicon film was used for tethers, beams and rim of the suspendedDBR, and 48 nm thick silicon films were used for the optically active areas of themirror, such structures with thick support and thin mirror area are reported in bulkmicromachined optical filters [26, 75]. Thus, the process was targeted to increasethe strength of the mirror support structure without changing the optical perfor-mance. An optical surface profile and a line scan through this more robust structureis shown in Figure 5.7 (c). It can be seen that although the tethers , beams, andrim around the mirror did not collapse during the release process, the top 48 nmthick silicon optical layer had still collapsed onto the suspended 48 nm thick bot-tom silicon layer. This failure of the tether-and-beam based structure for NIR andvisible wavelength was a clear indication of the need to modify the structure forfabricating ultra-thin silicon film based visible and NIR wavelength range tunablefilters.

123


(a)

0

500

1000

1500

2000

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

Hei

ght (

nm)

Distance (mm)

(b)

0

200

400

600

800

1000

0 0.1 0.2 0.3 0.4 0.5

Hei

ght (

nm)

Distance (nm)

(c)

0

200

400

600

800

1000

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

Hei

ght (

nm)

Distance (mm)

Figure 5.7: Collapsed suspended top mirror structures and line scan acrosstheir optical surface profile (a) NIR wavelength range structure (b) visiblewavelength range structure (c) thicker beam, tether and rim based structurefor visible wavelength range.

5.3 Stacked layer based cake structures

As discussed in earlier sections, the tether and beam based filter structure haveissues with tilt, flatness and integrity of devices. In order to reduce these problemswe developed a new structure, the “cake structure”, which has no support armsor tethers, but consists of a suspended membrane supported at the periphery and

124


sufficiently tensile to remain flat, such structure is reported in Reference [68] forfabricating bilayer air gap mirrors. On the same chip, three of these cake structureswere fabricated for comparison - the first circular, the second square, and the thirdcircular again but with hexagonal cells to improve support between the membranes.In these structures silicon layers and polyimide layers are stacked over each other.After dry release of sacrificial layers, the silicon layers provide conformal supportaround the periphery of the structures, thus forming the final filter structures.Figure 5.8 shows a conceptual cross sectional diagram of a released filter structure.

Figure 5.8: A conceptual cross section image of a circular cake structure.

Figure 5.9 shows the optical surface profile of fabricated cake like filters and a linescan across the surface profile, which shows flatness variation within 10-40 nm, andthere is no tilt in the top membrane. This is a very significant result because itprovided a clear pathway for developing improved structures for creating tunablefilters. A filter structure which is based on a conformal support around its peripherywill more likely be flatter than a tether and beam based structure.

All these fabricated filters were stored at room temperature in laboratory ambientfor two weeks, due to the unavailability of the optical measurement system forthis period. However, at the end of this period it was found that all these filtershad collapsed due to long exposure to air. Hence, any optical transmittance andactuation characterization could not be completed; yet, the inherent flatness of thecake structures paved the way for the notch based suspension structure which willbe discussed in Section 5.5.

5.4 The etch back process

To improve the structural rigidity and improve flatness, we attempted etch-back as amethod of fabricating a thick frame around a thin suspended membrane. The basicpremise behind this design was that it would be working with a single depositionof silicon for the top membrane rather than separate frame and optical depositions.With a single deposition, it was envisaged that stress gradients would be minimizedand a uniform tensile stress would be present in the top membrane and frame.

Etch back is a process whereby a thick silicon layer is selectively thinned down to therequired optical thickness in a pre-defined optically active area using a dry etching

125


(a)

0

500

1000

1500

2000

2500

0 0.05 0.1 0.15 0.2 0.25

Hei

ght (

nm)

Distance (mm)

(b)

0

500

1000

1500

2000

2500

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Hei

ght (

nm)

Distance (mm)

(c)

0

500

1000

1500

2000

2500

0 0.05 0.1 0.15 0.2 0.25

Hei

ght (

nm)

Distance (mm)

Figure 5.9: Optical surface profile and a line scan across the fabricatedsilicon-air-silicon cake structures; (a) Circular shaped cake structures; (b)Square shaped cake structures; (c) Hex-cell based cake structures.

process. Figure 5.10 shows the fabrication steps involved in this process. At first,a silicon film with the thickness required for the support structures is deposited ona sacrificial layer. Next, the optically active area is thinned down to the requiredoptical thickness by a dry etch process. In this way, the support structure forthe optical membrane can be made of thick silicon, providing a stronger mechanicalsupport for the suspended silicon membrane. Finally, the sacrificial layer is removed.

126


(a) (b)

(c) (d)

(e)

Figure 5.10: Etch back process steps (a) deposition and patterning of sac-rificial layer; (b) deposition of thick silicon layer; (c) etch back of the opticallyactive area; (d) formation of etch holes; (e) removal of sacrificial layer.

Table 5.3: Silicon etch recipe used in etch back.

ICP power RF power Pressure Temperature CF4flow rate(W) (W) (Pa) ◦C (sccm)

400 100 4 25 34

The dry etch recipe for the etch back must result in low surface roughness and highuniformity. In this thesis, a tetrafluoromethane (CF4) based etch recipe is selected,and the parameters of the recipe are shown in Table 5.3. The silicon etch rateusing this recipe is 120 nm min−1 and the recipe generates less than 2 nm surfaceroughness on an etched surface.

The effectiveness of this etch back based fabrication process was confirmed by fab-ricating a 200 nm thick, suspended circular membrane of 200 µm diameter and asuspended square membrane of 200 µm × 200 µm in area. Diluted PI2610 polyimidewas used as the sacrificial layer, which was spun and patterned to 500 nm thickness.After deposition of a 900 nm thick silicon film, the thick silicon layer was maskedand etched back to 200 nm thickness. After this step, 5 µm diameter etch holes wereperforated through the top membrane. The structures were pre-release annealed at285 ◦C for 30 minutes in a quartz tube furnace in a nitrogen environment. Finally,the sacrificial layer was removed in an oxygen plasma. Figure 5.11 (a) and (c) showoptical surface profiles of the approximately 200 µm wide and 200 nm thick sus-pended circular and square membranes, respectively. Line scans across the surfaceprofiles of these structures are shown in Figure 5.11 (b) and (d), respectively. It canbe seen from the surface profiles that the suspended top membranes are very flat.The optical surface profile shows a flatness variation of less than 10 nm. The 10 µmwide thick support is also visible around the perimeter of these figures. Thus, in

127


this experiment, only fabrication of membranes based on an etch back process wasexplored. Since these structure do not form an optical cavity, optical transmittancemeasurements were not undertaken.

(a) (b)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.0 0.050 0.10 0.15 0.20 0.25

Hei

ght (

µm)

Distance (mm)

(c) (d)

0

0.2

0.4

0.6

0.8

1

0.0 0.050 0.10 0.15 0.20 0.25 0.30 0.35

Hei

ght (

µm)

Distance (mm)

Figure 5.11: (a) Optical surface profile of 200 µm diameter circular struc-ture; (b) a line scan across surface profile of circular structure; (c) opticalsurface profile of 200 µm × 200 µm square structure; (d) a line scan acrosssurface profile of square structure.

5.5 Notch based suspension structure

Based on the findings from the work in the previous section, a new design for theMEMS filter structure was developed, which provides a novel contribution to thefield. A graphical layout of the suspension structure for the Fabry-Perot filter isshown in Figure 5.12, which is easily applicable to filters of size 200 µm × 200 µm,500 µm × 500 µm and 1000 µm × 1000 µm. With some modification to thesize of the electrodes and notches, these structures can be adapted for even largersize filters. The top and bottom mirror can consist of any combination of singlemembrane, multi-layer DBR, or air-spaced DBR. An optical area 10 µm smallerthan the mirror dimension is opened through the optical signal blocking metalshield deposited on the backside of the substrate. Four top electrodes and fourbottom electrodes provide symmetric actuation movement to the top mirror. Aspecific feature of the top membrane are the notches at the four corners of the top

128


mirror. These notches free the top membrane to actuate without bending-induceddistortion at the corners. The sacrificial layers are etched through the 5 µm × 5 µmetch holes. Unwanted optical signal through the etch holes is blocked with the helpof 8 µm × 8 µm metallic optical blocking pads.

Mirror size: D Size of opening in shield below mirror: D-20 μm

Top Electrode frame length:

D+100 μm

20 μm 25 μm

350 μm

100 μm Suspension

notch angle 45°

5 μm hole

125 μm

Bottom Electrode

Substrate Bottom Electrode Pad

Top Electrode Pad

Figure 5.12: Layout of the fabricated filters using the notch basedsuspension structure.

Tether and actuation beam based structures have been widely used by our group inthe past [163, 27] to fabricate microspectrometers, but they suffer from the followingdisadvantages:

1. The tethers can be very sensitive to out of plane stress or stress gradients,which will distort the suspended mirror.

2. Due to stress differences in the actuation beam materials, or due to differentthermal history of the materials, the actuation beams tends to attain a gutter-like shape. This distortion in the shape increases the stiffness of the actuationbeam and increases the actuation voltage.

3. The beam and tether based structures require a larger number of fabricationsteps, thus requiring additional time and resources.

4. Such structures cannot be used with large area (multi-mm × multi-mm) mir-rors, since the stress in the suspended mirror can actually pull off the posts

129


and beams from the substrate.

5. The number, shape and size of tethers and beams have to be modified tooptimize them for different materials and device areas.

In comparison to tether and beam based structures, the notch based structure offersthe following advantages:

1. The structure provides high immunity from the effects of out of plane stress.

2. The structure is simple to fabricate and requires fewer steps, thus saving timeand resources.

3. The structure can be expanded to many mm2 sized filters by simple scalingof the size of electrodes and notches.

4. This very specific arrangement of electrodes allows the filter to have a largertuning range than fixed-fixed beam actuators.

5. The suspension mechanism can be used with many different materials bysimply scaling the size of actuation electrodes and notches.

The following three sections of this chapter will demonstrate the applicability ofthis structure across a wide variety of suspended membranes and filters.

5.6 Ultra-thin silicon single membrane suspended

structures

This section describes the fabrication process, and optical and actuation character-ization of 70 nm thick suspended membranes. This thickness equals the quarterwavelength thickness of silicon films at 1000 nm design wavelength, which can beused for fabricating NIR wavelength range filters. These structures were fabricatedto demonstrate the efficacy of using an etch back process and the notch based sus-pension structure described above to open a path way for creating ultra-thin siliconfilm based tunable filters.


Figure 5.13 shows the 8-step fabrication process for a single membrane suspendedstructure.

1. At first a 100 nm thick gold bottom electrode was patterned on a sapphiresubstrate.

130


2. Prolift100-16 polyimide was used as the sacrificial layer which was spun at1950 rpm and soft-baked at 100 ◦C for 2 minutes and then hard-baked at250 ◦C for a total of 15 minutes. A 1.24 µm thick sacrificial layer was achievedwith this method.

3. On top of this sacrificial layer a 300 nm thick silicon layer was deposited. Therecipe for the silicon deposition is shown in Table 5.1 (see Figure 5.13 (c)).

4. On top of this silicon layer a 100 nm thick top electrode was deposited anddefined (see Figure 5.13 (d)).

5. The top silicon film was masked and selectively thinned down to 150 nm usingthe silicon etch recipe given in Table 5.3.

6. As shown in Figure 5.13 (f), etch holes were perforated through the top mem-brane.

7. The entire structure was pre-annealed at 265 ◦C for 15 minutes in a nitrogenenvironment. The sacrificial layer was then etched in a barrel asher in anoxygen plasma to freely suspend the 150 nm optical layer (see Figure 5.13(g)).

8. It was found that layers thinner than 150 nm were susceptible to collapse dueto the high oxygen plasma density in the barrel asher. Hence, a direct dryrelease of 70 nm thick silicon membranes in the barrel asher was not possible.In order to thin down the 150 nm thick silicon film to 70 nm, the released150 nm thick silicon membranes were dry etched an additional 80 nm in CF4

plasma in a reactive ion etcher (RIE). In the RIE the released membranes areexposed to CF4 for approximately 40 seconds. Such a short exposure does notresult in collapse of the suspended structures. This process also etches thesupport structure by 80 nm, but by careful selection of the initial thickness ofthe deposited layer, one can design the correct thickness for both the supportstructure and optical area of the suspended membrane (see Figure 5.13 (h)).Figure 5.14 shows a 70 nm thick 1 mm× 1 mm fabricated suspended structure.

5.6.2 Optical surface profile

Figure 5.15 (a) shows the optical surface profile of a released 1 mm × 1 mm sus-pended structure. The bottom and top pads are clearly visible in the profile. Figure5.15 (b) shows a line scan across the released membrane. Figure 5.15 (c) shows theline scan of the optically active central portion of the line scan of the suspendedmembrane shown in Figure 5.15 (b). The released membrane has flatness to within40 nm across its 1 mm length, and a flatness variation of only 15 nm in the 600 µmcentral portion. Note that this membrane is suspended over an air gap of 1.24 µmabove the substrate. Hence the flatness variation over its entire length is only 0.32%of the tunable air gap.

131


(a) (b)

(c) (d)

(e) (f)

(g) (h)

Figure 5.13: Fabrication process for single membrane suspended structure(a) deposition of bottom electrode on sapphire substrate; (b) spinning, hard-baking and patterning of Prolift100-16 polyimide; (c) deposition of thick siliconfilm for top membrane; (d) deposition and defining top electrodes; (e) etch backof optical area to 150 nm thick; (f) patterning of etch holes and notches; (g)pre-release annealing and dry release in O2 plasma; (h) etch back of additional80 nm in CF4 based etch recipe.

Figure 5.14: Fabricated 1 mm × 1 mm area 70 nm thick suspended mem-brane.

5.6.3 Actuation characteristics

The fabricated 1 mm × 1 mm area 70 nm thick suspended silicon membrane wasactuated by applying a DC voltage. Figure 5.16 shows the displacement of thetunable air-gap membrane as a function of applied DC voltage. At just 5 V themembrane shows a 260 nm reduction in cavity length, which reduces to 372 nm at

132


(a)

(b) (c)

-0.20

0.0

0.20

0.40

0.60

0.80

0.00 0.50 1.00 1.50 2.00

Hei

ght

(µm

)

Distance (mm)

-10.0

-5.00

0.00

5.00

10.0

15.0

0.50 0.60 0.70 0.80 0.90 1.0 1.1 1.2 1.3

Hei

ght (

nm)

Distance (mm)

Figure 5.15: (a) Optical surface profile of fabricated 1 mm × 1 mm area70 nm thick suspended membrane shown in Fig. 5.14; (b) a line scan across thesurface profile; and (c) a close up of the central 600 µm area of the membraneshown in (b).

10 V. The cavity gap continues to decrease with increasing applied voltage, and at10 V it reduces by 533 nm. At 25 V we see that initial cavity gap is reduced by694 nm. This is an actuation range of 56% of the starting cavity length, which isgreater than the theoretical 33% travel range for a simple fixed plate electrostaticactuator. Hence this actuation structure can achieve a wider spectral tuning rangefor MEMS based optical filters.

As shown in Figure 5.16 at low voltages, the thin membrane cavity gap decreasessharply with applied voltage, instead of following the standard MEMS electrostaticsactuation characteristic. In an ideal case we should only thin down the optical areaof the membrane to 70 nm and the remainder of the all support structure shouldremain at 300 nm thickness. However, in this single membrane structure, the entirestructure was thinned downed to 230 nm. Only silicon thin films under the metalelectrodes are not etched and remained as 300 nm. This thinning down of the entire

133


400.0

600.0

800.0

1000

1200

1400

0.00 5.00 10.00 15.00 20.00 25.00 30.00

Cav

ity le

ngth

(nm

)

Voltage (V)

Figure 5.16: DC actuation characteristic of fabricated 1 mm × 1 mm area70 nm thick suspended silicon membrane.

structure significantly reduces stiffness and hence spring constant of the membrane.In such cases it is expected that the membrane will actuate at low voltages. Ineffect, it may appear as a sort of mechanical amplification effect.

The solution to this problem is to mask entire support structures, when the opticalarea is being etched. It will be shown in subsequent sections that the protectionof support structure during the etch of optical area is successful in eliminating thisanomalous behavior of the DC actuation characteristic.

5.7 Multi-spectral MWIR tunable filter

5.7.1 Motivation

After demonstrating successful suspension and actuation of a single membrane, thenext step is to verify whether the actuation mechanism can be incorporated intoa tunable optical structure. In this case, a single-layer tunable mirror is used in aFabry-Perot filter. The MWIR wavelength range was chosen as a test vehicle, sincethe mirror layers are relatively thick.

134



Figure 5.17 shows the fabrication process flow for a single membrane tunable MWIRfilter. This filter has an asymmetric mirror configuration, i.e., the top mirror consistsof a quarter wavelength thick single layer of silicon and the bottom mirror consistsof a silicon-silicon oxide-silicon DBR. The center design wavelength of the filter waschosen as 4500 nm.

1. In the first step of the fabrication process, a 200 nm thick silicon oxide layerwas deposited on a 330 µm thick double side polished 〈100〉 oriented siliconsubstrate. This first layer operates as the optical interference layer.

2. Next, a silicon–silicon oxide–silicon based bottom mirror was deposited (Fig-ure 5.17 (a)). The recipes used for the silicon and silicon oxide depositionsare shown in Table 5.1 and Table 5.4, respectively. Each of the layers wastargeted to be a quarter wave thick; however the actual deposited thicknesseswere slightly different from the targeted thicknesses, as given in Table 5.5.

3. On top of the bottom mirror, a 100 nm thick gold bottom electrode wasdeposited and defined by photo lithography. As shown in Figure 5.17 (b),the same step also deposits metal optical blocking layers which prevents un-filtered light passing through the etch holes.

4. Prolift PL100-16 polyimide from Brewer Science was used for the sacrificiallayer. The PL100-16 was double spun at 1950 rpm, soft-baked at 100 ◦C for5 minutes and hard-baked by ramping the temperature up to 250 ◦C for atotal of 20 minutes (see Figure 5.17 (c)). This sacrificial layer was used tocreate the tunable air-cavity.

5. As shown in Figure 5.17 (d) another 575 nm thick silicon layer was depositedon top of the sacrificial layer.

6. A metal deposition and lift-off step was used to deposit and pattern a 100 nmthick gold top electrode.

7. Using reactive ion etching the top mirror and top membrane were perforatedwith etch holes and notches (see Figure 5.17 (e)).

8. A quarter-wave-thick anti-reflective coating based on silicon oxide was de-posited on the back side of the silicon substrate (Figure 5.17 (f)).

9. A metal layer to shield stray light was deposited (Figure 5.17 (g)) on thebackside of the substrate. The backside metal shield has an opening that is10 µm smaller than the mirror dimension under the top mirror of the filter.

10. Since deposited silicon thin film was slightly compressive, the filter structurewas annealed, pre-release, to convert the stress to a tensile state. The filterwas annealed at 245 ◦C for 5 minutes in a nitrogen environment. With this

135


short anneal the top-mirror was found to be tensile. The filter was releasedin a barrel asher at 150 W RF power at 1 Torr pressure in an oxygen plasma(see Figure 5.17 (h)).

Table 5.4: ICPCVD silicon oxide deposition parameters.

ICP Pressure Temperature Flow rate (sccm)power SiH4 He N2O Ar(W) (Pa) (◦C)

450 2 130 6.5 123 70 120

Table 5.5: Targeted vs actual thickness of the individual MWIR filter layers(from bottom to top).

Layer no. Material Substrate side Targeted thickness Actual thickness

(nm) (nm)

1 Silicon oxide Front 200 205

2 Silicon Front 320 310



5 PL100-16 Front 2250 2480


7 Silicon oxide Back 750 720

8 Au Back 80 75

5.7.3 Optical surface profile

Figure 5.18 shows a fabricated MWIR filter. Note that the etch holes and the fournotches at the four corners of the filter are clearly visible. The bottom electrode isvisible through the notches. Figure 5.19 (a) shows a surface profile of the fabricatedfilter and Figure 5.19 (b) shows a line scan across the optical surface profile of thefabricated filter. The profile was measured across the top electrodes of the filteravoiding the etch holes. Figure 5.19 (c) represents the line scan shown in Figure5.19 (b) on an expanded scale. The fabricated filter exhibits a 40 nm bow acrossan 800 µm length in the center part of the top mirror. This gives a maximum of2% variation in the air-gap cavity length between the top and bottom mirror.

5.7.4 Optical transmission characteristics

The optical transmission measurements of the fabricated filter were undertaken ina Spectrum One FTIR spectrometer from Perkin Elmer. Figure 5.20 shows optical

136


(a) (b)

(c) (d)

(e) (f)

(g) (h)

Figure 5.17: Fabrication process of MWIR tunable Filter. (a) Deposi-tion of silicon oxide–silicon–silicon oxide–silicon bottom mirror stack on sili-con substrate; (b) deposition and definition of bottom electrode; (c) spinning,hard-baking and patterning of Prolift100-16 polyimide; (d) deposition of sili-con film for top membrane mirror, deposition of top electrodes, and etch backof silicon membrane; (e) patterning of etch holes and notches; (f) depositionand definition of anti-reflective layer; (g) deposition of the optical metal shieldon the back side of substrate to define the optically active area; (h) pre-releaseannealing and dry release in an O2 plasma.

Figure 5.18: A micro-graph of fabricated MWIR filter.

transmission measurements taken at a single point on the fabricated 1 mm × 1 mm

137


(a)

(b) (c)

0.0

0.10

0.20

0.30

0.40

0.50

0.00 0.40 0.80 1.20 1.60

Hei

ght (

µm)

Distance (mm)

0.00

10.0

20.0

30.0

40.0

50.0

0.40 0.50 0.60 0.70 0.80 0.90 1.0 1.1 1.2

Hei

ght (

nm)

Distance (mm)

Figure 5.19: (a) Optical surface profile of fabricated 1 mm × 1 mm areaMWIR filter; (b) Line scan across the surface profile shown in (a); (c) Ex-panded scale surface profile of center 800 µm area of top mirror.

MWIR filter in comparison with simulated transmission spectra. The simulatedmodel takes into account the actual measured thicknesses of the deposited layers.It is noted that the measured optical results are very close to the simulated results.The fabricated filter shows a peak transmission of 66% and 410 nm FWHM ascompared to the theoretical 390 nm FWHM. The peak wavelength of the fabricatedfilter is centered at 4900 nm. There is a shift of 400 nm from the targeted centerwavelength of the filter which is attributed to a greater sacrificial layer thicknessthan the targeted value (see Table 5.5).

5.7.5 Wavelength tuning of fabricated filter

The fabricated filters were actuated by applying a DC voltage. Figure 5.21 showsthe displacement of the suspended silicon membrane mirror as a function of appliedDC voltage. The top mirror shows no sign of actuation below 10 V. The filter

138


0

20

40

60

80

100

3000 3500 4000 4500 5000 5500 6000

Tran

smitt

ance

(%)

Wavelength (nm)

Figure 5.20: Optical transmittance of 1 mm × 1 mm size fabricated MWIRfilter using the fabrication process outlined in Figure 5.17. Measured trans-mittance is shown as open squares against the simulated performance shownas a dashed line.

starts actuating after 10 V, and at 15 V it shows 100 nm of displacement frominitial cavity length. The cavity gap decreases by 220 nm at 17 V. At 21 V thefilter shows 510 nm of displacement from initial cavity length and, finally, it travels830 nm of physical distance at 25 V. Thus, the filter tunes to 33% of the initial airgap of 2480 nm. On application of further voltage the top suspended membranebroke, hence, more than 33% tuning range could not be demonstrated.

Figure 5.22 shows the measured and simulated wavelength tuning of the fabricatedfilter as a function of applied actuation voltage. With increasing actuation voltage,the central wavelength of the filter tunes towards shorter wavelengths, as the air-gap cavity length is reduced. Simultaneously, the peak transmittance of the filteralso increases slightly. This observation confirms that the thicker sacrificial layerhas shifted the operation of the filter to the lower reflectivity region of the bottommirror, and has resulted in a loss of peak transmittance for longer wavelengths.As the peak shifts back into the higher-reflectivity region of the mirror, the peak-transmittance increases. At 10 V the filter was centered at 4920 nm with 66% peaktransmittance. At 17 V the central wavelength shifts to 4830 nm with 67.4% peaktransmittance. At 19 V the filter central wavelength reaches 4610 nm and peaktransmittance increases to 70% and FWHM reduces to 380 nm. At 23 V the peakwavelength shifts to 4340 nm with 75% peak transmittance. Finally, at 25 V thefilter shows 75.22% peak transmittance, and 380 nm FWHM at a center wavelength

139


1600

1800

2000

2200

2400

2600

0 5 10 15 20 25 30

Cav

ity le

ngth

(nm

)

Voltage (V)

Figure 5.21: Displacement vs voltage characteristics of 1 mm × 1 mmMWIR filter.

of 4100 nm. In effect, the total achieved tuning range was 800 nm. As shown inFigure 5.22 the simulated model is a very close match to the measured results.

5.8 Fixed cavity SWIR Fabry-Perot filters with

silicon-air-silicon top mirror

5.8.1 Motivation

With a single-membrane structure demonstrated with actuation, we next increasethe device complexity by moving to a filter structure that contains a single air-spaced mirror. This section describes the fabrication process and optical character-ization of a fixed-cavity Fabry-Perot filter, based on a silicon-air-silicon top mirror,for the SWIR wavelength range. This filter again has an asymmetric mirror con-figuration i.e. the top mirror is made of silicon-air-silicon layers and the bottommirror is made of silicon-silicon oxide-silicon. A fixed-filter configuration was chosento simplify the fabrication process and demonstrate a proof of concept filter basedon a suspended silicon-air-silicon DBR.

140


0

20

40

60

80

100

3500 4000 4500 5000 5500

Tran

smitt

ance

(%)

Wavelength (nm)

0 V10 V

15 V17 V19 V21 V23 V25 V

Figure 5.22: Spectra showing wavelength tuning of 1 mm × 1 mm MWIRFabry-Perot filter with a suspended silicon membrane mirror. Measured trans-mittance is shown as open squares against the simulated performance shown asdashed lines. Applied actuation voltage is shown on top of each transmittancepeak.

5.8.2 Fabrication Process

Figure 5.23 shows the fabrication process flow for the SWIR filter. The centerdesign wavelength of the filter was chosen to be 2000 nm.

1. Initially, a 250 nm silicon oxide layer was deposited on a 330 µm thick doubleside polished 〈100〉 oriented silicon substrate. This layer functions as theoptical interference layer between the silicon substrate and the first depositedsilicon layer. On top of it, a silicon–silicon oxide–silicon based bottom mirrorwas deposited (Figure 5.23 (a)). The recipes used for the silicon and siliconoxide deposition are shown in Table 5.1 and Table 5.4, respectively. The actualdeposited thicknesses of the layers were slightly different from the targetedthicknesses, as shown in Table 5.6.

2. On top of the bottom mirror, a 100 nm thick gold bottom electrode wasdeposited and defined. As shown in Figure 5.23 (b) the same step also depositsmetal optical blocking areas which stop stray light from the etch holes.

3. Prolift PL100-16 polyimide from Brewer Science was used for the sacrificiallayers. PL100-16 was spun at 1950 rpm, soft-baked at 100 ◦C for 5 minutesand then hard-baked by ramping the temperature to 250 ◦C for a total of20 minutes (see Figure 5.23 (c)). This sacrificial layer was used to create theFabry-Perot air-cavity.

4. For the air gap based top mirror the silicon layers were chosen to be three

141


quarter wavelength thick. This provides two advantages: first, it strengthensthe support structures for the suspended top mirror; and, second it is morerobust against exposure in the oxygen plasma in the barrel asher during re-lease. As shown in Figure 5.23 (d) a second 435 nm thick silicon layer wasdeposited on top of the sacrificial layer.

5. PL100-16 was spun at 3650 rpm, soft-baked at 100 ◦C for 5 minutes and hard-baked at 250 ◦C by ramping the temperature for a total of 20 minutes (seeFigure 5.23 (e)). This layer was wet patterned in AZ326 developer to serveas the sacrificial layer for creating a fixed cavity quarter wave thick air-gapfor the top mirror. A short descum step was performed in oxygen plasma toimprove the adhesion of the next silicon layer.

6. On top of this layer the final three quarter wave thick silicon layer was de-posited.

7. A metal deposition and lift-off step were used to deposit and pattern a 100 nmthick gold top electrode (Figure 5.23 (f)). This layer was included to see theeffect of residual stress of the top metal layer on the flatness of top suspendedmirror.

8. Using reactive ion etching the top mirror and top membrane were perforatedwith etch holes and notches (Figure 5.23 (g)).

9. A quarter-wave thick anti-reflective coating based on silicon nitride was de-posited at the back side of the silicon substrate (Figure 5.23 (h)).

10. As shown in Figure 5.23 (i) a metal layer was deposited on the backside ofthe substrate to shield stray light. The backside metal shield has an opening10 µm smaller than the top mirror dimension under the top mirror.

11. Since silicon deposited on PL100-16 polyimide was compressive, the filter wasannealed prior to release. After the annealing at 265 ◦C for 15 minutes in anitrogen environment, the top-mirror was found to be tensile. The filter wasreleased in a barrel asher at 150 W RF power at 1 Torr pressure in an oxygenplasma (see Figure 5.23 (j)). Figure 5.24 shows a 500 µm × 500 µm SWIRfilter fabricated in this way.

5.8.3 Optical Characterization

Figure 5.25 (a) shows the surface profile of a 500 µm × 500 µm filter, and Figure5.25 (b) shows the surface profile of the top mirror on an expanded scale. Figure5.25 (c) shows a line scan across Figure 5.25 (a) through the notches of the filters.The small bumps in the profile are due to the presence of etch holes in the vicinityof the line scan. Figure 5.25 (d) shows a line scan of the surface profile of thetop mirror. The top membrane of the suspended mirror shows a 40 nm tilt across

142


Table 5.6: Targeted vs actual thickness of the SWIR filter layers (frombottom to top).

Layer no. Material Substrate side Targeted thickness Actual thickness

(nm) (nm)





5 PL100-16 Front 1000 1125


7 PL100-16 Front 500 560


9 Silicon Nitride Back 277 287

10 Au Back 100 105

500 µm, with no sign of any bowing. It was verified that this tilt was result of tiltedprofile of second sacrificial layer, hence this tilt can be reduced by proper patterningof the second sacrificial layer.

The optical transmission measurements of the fabricated filters were undertakenin a Spectrum One FTIR spectrometer from Perkin Elmer. Figure 5.26 shows thesingle point measured and simulated transmittance of the fabricated SWIR filter.The first and second order peaks are clearly visible for the measured and simulatedplots. For the first order peak, the measurement shows a peak transmission of70% and 90 nm FWHM at 2240 nm wavelength; the simulated spectrum predicts83% peak transmittance and 50 nm FWHM for this peak. The second order peakof the measured spectrum shows 53% peak transmission and 35 nm FWHM at1550 nm, as compared to a predicted 44% peak transmission and 30 nm FWHM.The relatively low peak transmittance, large FWHM and low out of band rejectionat long wavelengths of the measured first order peak of the SWIR filter are due tothe low reflectivity of the bottom mirror. Figure 5.27 shows measured and simulatedreflectivity of the bottom mirror of the SWIR filter. The measurement indicates 85%peak reflectivity for the bottom mirror as compared to 90% simulated reflectivity.The lower reflectivity of the bottom mirror can be improved in two ways: by usinga higher order bottom mirror made of silicon-silicon oxide, or by using a higherreflectivity bottom mirror. The first solution limits the free spectral range of thefilter, hence the second solution would be a better choice. In this case a silicon-air-silicon or a germanium-silicon oxide-germanium based bottom mirror are viablesolutions. These combinations would give a difference of approximately 2.5 betweenthe high and low refractive index materials, which can theoretically achieve closeto 95% peak reflectivity. A Higher reflectivity bottom mirror will also provide a

143


(a) (b)

(c) (d)

(e) (f)

(g) (h)

(i) (j)

Figure 5.23: Fabrication process of SWIR Fabry-Perot filter. (a) De-position of silicon oxide–silicon –silicon oxide–silicon bottom mirror stack onsilicon substrate; (b) deposition of bottom electrode; (c) spinning, and hard-baking of Prolift100-16 polyimide; (d) deposition of first three quarter wavethick silicon film for top mirror; (e) spinning, hard-baking and wet pattern-ing of Prolift100-16 polyimide for the top mirror air cavity; (f) deposition ofsecond three quarter wave thick silicon film for top mirror and deposition anddefinition of top electrodes; (g) patterning the etch holes and notches; (h) de-position of back-side anti-reflective layer; (i) deposition of the optical metalshield on the back side; (j) pre-release annealing and dry release in O2 plasma.

narrower FWHM for the filter.

144


Figure 5.24: A micro-graph of fabricated SWIR filter.

(a) (b)

(c) (d)

0.0

0.50

1.0

1.5

2.0

2.5

3.0

0.0 0.20 0.40 0.60 0.80 1.0

Hei

ght (

µm)

Distance (mm)

0.40

0.42

0.45

0.48

0.50

0.53

0.55

0.0 0.10 0.20 0.30 0.40 0.50 0.60

Hei

ght (

µm)

Distance (mm)

Figure 5.25: (a) Optical surface profile of fabricated 500 µm × 500 µmSWIR Fabry-Perot filter; (b) A close up of the top mirror; (c) line scan acrossthe surface profile; (d) line scan on top mirror on an expanded scale avoidingthe etch holes.

145


0

20

40

60

80

100

1400 1600 1800 2000 2200 2400 2600

Tran

smitt

ance

(%)

Wavelength (nm)

Figure 5.26: Experimental and simulated transmittance spectra of Fabry-Perot SWIR filter. The points represent measurements, while the dashed linerepresents model predictions.

0

20

40

60

80

100

1800 2000 2200 2400 2600 2800 3000 3200

Measured TransmittanceMeasured ReflectanceMeasured AbsorbanceSimulated Reflectance

Tran

s / R

efl /

Abs

or (%

)

Wavelength (nm)

Figure 5.27: Measured and simulated reflectivity of silicon-silicon oxide-silicon bottom mirror of SWIR filter shown in Figure 5.24. Also shown is thetransmittance and absorbance through the mirror.

5.9 Summary

In this chapter it was shown that the tether and beam based MEMS structures aremore susceptible to deformation of structure due to residual stress, hence there wasneed to improve the design of MEMS structure for fabrication of tunable siliconbased Fabry-Perot filters. A notch based suspension structure was proposed and

146


it was shown that this structure provide better immunity to stress induced defor-mation as compared to tether and beam based structure. Based on notch basedsuspension structure design three MEMS devices were fabricated. The first was a70 nm thick suspended membrane which actuated more than 50% of initial cavitygap. Second was a tunable multi-spectral MWIR filter. The multi-spectral MWIRfilter was actuated by 830 nm of physical distance at 25 V and 800 nm of wave-length range. Finally, fabrication process and optical characterization of a SWIRfilter based on silicon-air-silicon mirrors was presented. The measurement showed apeak transmission of 70% and 90 nm FWHM at 2240 nm wavelength; the simulatedspectrum predicted 83% peak transmittance and 50 nm FWHM for this peak. Thesecond order peak of the measured spectrum showed 53% peak transmission and35 nm FWHM at 1550 nm, as compared to a predicted 44% peak transmission and30 nm FWHM.

147

Chapter 6

Visible and NIR wavelengthsDBRs and filters

6.1 Introduction

A key objective of this project was to make in-roads towards extending the existingMEMS microspectrometer technology towards operation at visible and NIR (NIR:0.7 nm - 1.5 nm) wavelengths. The main issue identified in this transition was thatthe germanium based mirror technology of the existing technology prevents oper-ation at wavelengths below 1.5 µm due to excessive absorption in the germanium.While the use of silicon has already been proposed in this thesis as fulfilling the op-tical and structural requirements at these wavelengths, the very thin layers neededfor operation at these wavelengths will make fabrication exceedingly difficulty. Theobjective of the work presented in this chapter is to overcome these hurdles anddemonstrate true silicon based optical structures for tuneable Fabry-Perot filters inthe visible and NIR spectral bands.

The first step in demonstrating any Fabry-Perot filter technology is developmentof the mirrors. As such, we follow a step-by-step approach here, where we developthe mirror technology first, followed by demonstration of fixed (non-actuated) filterstuctures. While not forming part of the work presented in this thesis, future workwill pursue the next step of the process: to demonstrate tunable versions of silicon-air-silicon DBR based filters for the visible and NIR wavelength ranges.

This chapter describes distributed Bragg reflectors (DBRs) and Fabry-Perot filtersfabricated to operate in the near infrared and visible wavelength ranges. Section6.2 describes the fabrication process and optical results for silicon-silicon oxide-silicon and silicon-air-silicon based NIR and visible wavelength mirrors. Section6.3 presents fabrication and optical characterization of a fixed cavity silicon-siliconoxide-silicon based Fabry-Perot filter.

148

CHAPTER 6. VISIBLE AND NIR WAVELENGTHS DBRS AND FILTERS

6.2 Visible and NIR wavelengths DBRs

6.2.1 DBR architectures

This section presents the fabrication process of mirrors, and the subsequent sectiondetails their optical characterization. The fabrication processes of visible and NIRwavelengths DBRs are similar, differing only in the thickness of the deposited opticallayers. Two varieties of DBR’s were examined during this work, both based on athree-layer structure with silicon as a high-index medium (see Figure 6.1). Thefirst variety uses silicon oxide (SiOx) as the low index medium. The second varietycreates an air-gap for the low-index medium and attempts to yield the best possiblereflectivity of a silicon-based three-layer DBR structure. For the visible wavelengthrange DBR the central wavelength is 600 nm and for the NIR wavelength rangeDBR the central wavelength is 1000 nm. Each layer in the DBRs is a quarterwavelength thick, and Figure 6.1 indicates the thickness of each layer.

Figure 6.2 shows the modeled ideal optical response of these DBRs. As shown inFigure 6.2 (a), in the visible wavelength range a silicon-air-silicon DBR can achieve96% peak reflectivity, and a silicon-silicon oxide-silicon DBR can achieve 93% peakreflectivity. In addition to that, silicon-air-silicon based DBRs offer wider band-width than silicon-silicon oxide-silicon DBRs. It is to be noted that, for wavelengthsshorter than 500 nm, the reflectivity drops off sharply, and there is significant in-crease in the absorbance. This sharp increase in the absorbance is due to the highextinction coefficient of silicon thin films for wavelengths shorter than 500 nm. Asshown in Figure 6.2 (b), in the NIR wavelength range silicon-air-silicon DBRs canachieve 96% peak reflectivity, and silicon-silicon oxide-silicon DBR can achieve 92%peak reflectivity. It is to be noted that silicon-air-silicon DBRs maintain 90% ormore reflectivity for the entire simulated wavelength range (750-1400 nm). Onceagain, silicon-air-silicon DBR demonstrates a wider bandwidth and higher reflec-tivity as compared to that of silicon-silicon oxide-silicon DBRs. Although opti-cal performance of the silicon-air-silicon DBRs is superior to that of silicon-siliconoxide-silicon DBRs, however, from the perspective of ease of fabrication, siliconoxide based design offer significant benefits over silicon-air-silicon DBRs. Some ofthe advantages offered by silicon oxide based DBRs relate to its completely solidstructure, which renders them robust as well as requiring fewer fabrication steps.

6.2.2 Fabrication of silicon oxide cavity DBRs

The Si-SiOx-Si visible and NIR wavelengths DBRs were fabricated on 0.310 mmthick sapphire substrates. Based on the deposition rate, low tensile stress andgood optical constants, the silicon recipe chosen is shown in Table 6.1. For siliconoxide thin films the recipe and deposition parameters are given in Table 6.2. Thedeposition rate of SiOx for the given recipe was 0.33 nm/sec. The targeted thicknessof the silicon oxide layers and silicon layers was quarter-wave thick. The fabricated

149


(a) (b)

Si 38 nm (tH= o\4nH)

SiOx 105 nm (tL = o\4nL)

Si 38 nm (tH = o\4nH)

Sapphire Substrate


Air 150 nm (tL = o\4nL)


Sapphire Substrate

(c) (d)


SiOx 175 nm (tL = o\4nL)


Sapphire Substrate


Air 250 nm (tL = o\4nL)


Sapphire Substrate

Figure 6.1: Structure of visible and NIR wavelengths DBRs (a) visiblewavelength DBR with SiOx as the low index medium ; (b) visible wavelengthDBR with air as the low index medium; (c) NIR wavelength DBR with SiOx

as the low index medium ; (d) NIR wavelength DBR with air as the low indexmedium.

mirrors are a simple stack of quarter wave thick Si-SiOx-Si films deposited on asapphire substrate.

Table 6.1: Silicon deposition recipe.

ICP Power Deposition Temp. Pressure SiH4 flow He flow

(W) (◦C) (mTorr) (sccm) (sccm)

26 300 30 5 95

Table 6.2: Process parameters for deposition of silicon oxide films.

ICP Dep. Dep. SiH4 He N2O Arpower temperature pressure flow rate flow rate flow rate flow rate(W) (◦C) (Pa) (sccm) (sccm) (sccm) (sccm)

450 50 2 6.5 123 70 120

6.2.3 Fabrication of air-cavity DBRs

Unlike, fabrication of silicon oxide cavity based DBRs, the fabrication process ofair-cavity based DBRs posed significant challenges. One of the major challenge

150


(a)

0

20

40

60

80

100

350 400 450 500 550 600 650 700 750

Reflectance - Si-air-SiAbsorbance - Si-air-SiReflectance - Si-SiO

x-Si

Absorbance - Si-SiOx-Si

Ref

lect

ance

/ A

bsor

banc

e (%

)

Wavelength (nm)

(b)

65

70

75

80

85

90

95

100

700 800 900 1000 1100 1200 1300 1400 1500

Si-air-SiSi-SiO

x-Si

Ref

lect

ance

(%)

Wavelength (nm)

Figure 6.2: Optical modeling of NIR and visible wavelength DBRs shownin Figure 6.1 (a) Reflectance and absorbance of visible wavelength DBR withSiOxand air as the low index medium ; (b) Reflectance of NIR wavelengthDBR with SiOx and air as the low index medium.

is suspending ultra-thin membranes over a large area. In the visible and NIRwavelength range, the thicknesses of suspended layers are close to 40 nm and 70 nm,respectively, and the air cavity gap between two optical layer is less than 200 nm.Thus, two ultra-thin films suspended with such a narrow gap are highly susceptibleto collapse, and maintaining a uniform cavity gap is not a trivial task. It was

151


discussed in Section 4.2 that the support arm based structures were not successfulin fabricating SWIR wavelength range DBRs, and it was the conformal supportbased structure which proved to be the best structure for fabricating DBRs. Hence,visible and NIR based DBRs were fabricated using this structure.

In order to create a uniform cavity gap it was necessary that the silicon thin filmsshould be tensile enough to achieve the required level of flatness, and not so ten-sile that they rupture due to excessive stress. It was observed that as-depositedsilicon thin films used in visible and NIR DBRs had a residual compressive stress.Since the required air-gap between the two silicon layers is only 250 nm for the NIRwavelength mirrors and 170 nm for the visible wavelength mirrors, the presenceof any residual compressive stress causes the top membrane to collapse onto thebottom membrane, and any amount of post-release treatment could not recover thestructure. Hence, a series of pre-release annealing experiments were performed toascertain the optimum annealing temperature and time to attain sufficient tensilestress in the silicon thin films. As a result, it was concluded that for the visiblewavelength and NIR wavelength range the DBRs would undergo a pre-release tem-perature for a fixed interval of 30 minutes. It was observed that the pre-releaseannealing temperature for NIR wavelength mirrors and visible wavelength mirrorscan be chosen as 295 ◦C and 310 ◦C, respectively. Any increase in annealing time orannealing temperature was found to increase the tensile stress on membranes suchthat they were de-laminating at the edges (see Figure 6.3).

De‐lamination of

membrane at

the edges

Etch holes

Sacrificial layer

Figure 6.3: De-lamination of membrane at the edges during dry-releaseprocess due to excessive residual tensile stress.

It was also observed that these ultra-thin membranes, despite an appropriate levelof residual tensile stress, were collapsing during the dry-release step in the oxygenplasma in a barrel asher. A further investigation showed that the long exposureto oxygen plasma was causing these membranes to collapse. In addition, at 1 Torrchamber pressure the application of more than 160 W RF power was detrimentalto the release process and almost all DBRs collapsed with high RF power. Hence,the exact release time for these membranes was found by experiments at 160 W RFpower and 1 Torr chamber pressure . After release, these membranes were found

152


to be quite robust and didn’t demonstrate any failure due to mechanical vibrationsduring optical characterization. In conclusion, the successful fabrication of visibleand NIR DBRs was dependent on (i) selection of a robust structure, (ii) applyingthe optimum annealing conditions, and (iii) developing an appropriate dry releaseprocess in terms of power and time.

6.2.4 Final process for fabricating air cavity DBRs

Figure 6.5 shows the fabrication flow of NIR and visible wavelength air cavitymirrors. The mirrors were designed to be circular in shape with diameters of 270 µm,320 µm, 370 µm and 420 µm. Figure 6.4 shows the mask layout of the chip, withthe DBRs of different diameters grouped in boxes.

270 µm

320 µm

370 µm

420 µm

Figure 6.4: Mask Layout for the visible and NIR DBRs.

1. As shown in Figure 6.5(a), the first quarter wavelength thick silicon layer wasdeposited on a 280 µm thick sapphire substrate.

2. HD Microsystem PI 2610 polyimide was used as the mirror gap sacrificiallayer. PI 2610 was 50% diluted by N-Methyl-2-pyrrolidone solvent to attainthe thickness required for creating the air cavity in both NIR and visiblemirrors. In order to achieve quarter wavelength thickness of the sacrificiallayer, diluted PI 2610 was spun at 4500 rpm for the NIR mirrors and 5600 rpmfor the visible mirrors, respectively. The polyimide was soft-baked at 130 ◦C

153


for 5 minutes, and then hard-baked at 300 ◦C for 20 minutes. The polyimidethickness was 250 nm for the NIR and 170 nm for the visible wavelengthoperation. Using an O2 plasma etch recipe, the sacrificial layer was thenpatterned (see Figure 6.5 (b)).

3. Figure 6.5 (c) shows the next step, in which a second quarter wave thickSi layer was deposited on top of the sacrificial layer. This layer was perfo-rated with 4 µm diameter etch holes. These etch holes were used to etch thesacrificial layer at the final stage of the process.

4. The pre-release annealing ensured that the membranes were tensile enoughto be flat when suspended. The optimu 30 minutes pre-release annealingtemperature for NIR wavelength mirrors and visible wavelength mirrors wasdetermined to be 295 ◦C and 310 ◦C, respectively. A March barrel asher wasused to etch the sacrificial layer in an O2 plasma with 160 W RF power and1 torr pressure for 30 minutes to create the final air cavity shown in Figure6.5 (d).

Figure 6.6 shows the fabricated 270 µm diameter mirror for both the NIR andvisible wavelength ranges.

(a) (b) (c) (d)

Figure 6.5: Fabrication process of NIR and visible wavelength DBRs: (a)deposition of Si thin film on sapphire substrate; (b) spinning, hard-baking andpatterning of PI 2610 polyimide; (c) deposition of the second Si thin film ontop of polyimide and patterning of the etch holes through the top Si membrane;(d) pre-release annealing and dry release in an O2 plasma.

(a) (b)

Figure 6.6: 270 µm diameter fabricated DBRs: (a) visible wavelengthmirror; (b) NIR wavelength mirror.

154


Table 6.3: Summary results of fabricated silicon-silicon oxide-silicon DBRs.

NIR DBR Visible DBR

Wavelength Range (nm) 710–1000 500–700Thickness of Si-1 (nm) 60 41

Thickness of SiOx (nm) 132 100Thickness of Si-2 (nm) 62 36

Measured peak reflectivity (%) 89 88Simulated peak reflectivity (%) 92 93

Measured optical bandwidth (nm) 200 300

6.2.5 Optical characterization of silicon-silicon oxide-siliconDBRs

Visible and NIR silicon-silicon oxide-silicon DBRs were fabricated on sapphire sub-strates using the deposition conditions shown in Table 6.1 and Table 6.2, and theresulting thin film thicknesses are listed in Table 6.3. Figure 6.7 plots both the mea-sured and simulated transmission spectra of the DBRs. Simulation of the DBRswas undertaken using the optical matrix transmission method [35]. One of themost important things to note in Figure 6.7(a) and Figure 6.7(b) is the close to90% reflectance of the DBRs operating over most of the visible spectrum (500 –700 nm) and the NIR spectrum (700 – 1000 nm). This proves that despite the highabsorption of silicon at visible and NIR wavelengths, the impact of this increasedabsorption is minimized due to the very thin layers of silicon needed for quarter-wave DBRs. Compared to the theoretical 92% – 93% peak reflectivity, the ICPCVDsilicon based DBRs give 88% – 89% peak reflectivity. Note that these mirrors havemore than 80% reflectivity for a 200 nm wide optical band (500 – 700 nm) at visiblewavelengths, and for a 300 nm wide optical band in the NIR wavelength region.As such, these wide optical bandwidths are very suitable for applications such asspectroscopy. Table 6.3 shows the actual thicknesses of the DBR layers, measuredand simulated peak reflectivity, and measured optical bandwidth over which thereflectivity is greater than 80%.

6.2.6 Optical characterization of visible wavelength silicon-air-silicon DBRs

Figure 6.8 shows the single point reflectance, transmittance and absorbance mea-surements of a DBR in the 540–960 nm wavelength range, which covers a signif-icant part of visible wavelength range (540-700 nm). It is to be noted that thismeasurement was carried out in a spectrometer which can measure transmittanceand reflectance in the visible and NIR wavelength range (400 – 980 nm). Thesimulated reflectance profile (“Simulated Reflectance”) in Figure 6.8 was obtainedby the optical transfer matrix method assuming an ideally flat mirror profile usingknown optical constants of ICPCVD deposited silicon films. This simulation also

155


(a) (b)

0

20

40

60

80

100

650 700 750 800 850 900 950 1000 1050

Measured ReflectanceMeasured TransmittanceMeasured AbsorbanceSimulated Reflectance

Tra

ns /

Ref

l / A

bsor

(%)

Wavelength (nm)

0

20

40

60

80

100

400 450 500 550 600 650 700 750

Measured ReflectanceMeasured TransmittanceMeasured AbsorbanceSimulated Reflectance

Tran

s / R

efl /

Abs

or (%

)

Wavelength (nm)

Figure 6.7: Measured transmittance, reflectance and absorbance of fab-

ricated silicon-silicon oxide-silicon DBRs. Also shown is the simulated re-

flectance for each DBR: (a) NIR wavelength DBR; (b) visible wavelength DBR.

used layer thicknesses measured during mirror fabrication, and predicts a best-casereflectance of 92%. The measured reflectance (“Measured Reflectance”) had a max-imum value as 91.6% at 680 nm. The simulated and measured reflectance data arein close agreement with each other. It can also be seen from Figure 6.8 that thereflectance is less than 70% in the wavelength range of 540–570 nm due to high ab-sorption in the silicon thin films. From 650 nm to 810 nm the reflectivity is greaterthan 85%.

0

20

40

60

80

100

500 600 700 800 900 1000

Measured TransmittanceMeasured AbsorbanceMeasured ReflectanceSimulated Reflectance

Tran

s / R

efl /

Abs

(%)

Wavelength (nm)

Figure 6.8: Measured and simulated optical transmisttance, reflectance and

absorbance spectra of a silicon-air-silicon quarter wave mirror at visible wave-

lengths.

In order to confirm the spatial uniformity of the mirrors, a scanning spectroscopicmicroscope was used to produce a spatial map of the transmission spectra over the

156


540 nm to 960 nm wavelength range across a 5 mm × 3 mm spatial grid. The gridincluded an array of 28 DBRs with diameters ranging between 270 µm and 420 µm.Figure 6.9 shows the 2-D high resolution spatially resolved spectral transmissionmeasurement of the array of 28 mirrors demonstrating the variations in the min-imum transmission across a 5.0 mm × 3.0 mm physical grid at a wavelength of680 nm, which is the wavelength of minimum DBR transmittance. The 2-D spatialmap shows high homogeneity in optical transmission, across all the mirrors, indicat-ing the suitability of this fabrication process to produce high quality mirrors in thevisible/NIR wavelength range. Figure 6.10 shows a 3-D high resolution spatiallyresolved spectral transmission measurement of an array of 28 mirrors demonstrat-ing the variations in minimum transmission. From the results obtained by opticaltransmission spatial mapping of the DBRs, a transmission profile across a 320 µmdiameter DBR was extracted, and is shown in Figure 6.11. A very low variationin the minimum transmittance across a 320 µm DBR indicates that the air gapbetween the top layer and the bottom layer of the DBR is very uniform and, thus,the suspended top layer is very flat relative to the bottom layer of the DBR.

Table 6.4 shows the thicknesses of the layers, measured and simulated peak reflec-tivity and measured optical bandwidth where reflectivity is greater than 80%.

Table 6.4: Summary results of fabricated silicon-air-silicon DBRs.

NIR DBR Visible/NIR DBR

Wavelength Range (nm) 850–1300 540–960Thickness of Si-1 (nm) 60 41

Thickness of air cavity (nm) 250 170Thickness of Si-2 (nm) 62 36

Measured peak reflectivity (%) 92.7 91.6Simulated peak reflectivity (%) 95 92

Measured optical bandwidth (nm) 450 100/150

6.2.7 Optical characterization of NIR silicon-air-silicon DBRs

A scanning spectral microscope was built for the characterization of the NIR andvisible wavelength DBRs. This optical characterization system enables the measure-ment of calibrated transmission spectra with high spatial and spectral precision. Forthese transmission measurements, light was focused onto the DBRs to a spot size of20 µm and measurements were undertaken across a 5 mm × 3 mm grid. The areaof this scanned grid included the array of 25 DBRs. The NIR mirror measurementswere undertaken from 850 nm to 1350 nm, with mirror diameters ranging between370 µm and 420 µm.

Figure 6.12 shows the single point transmittance measurement through a NIR wave-length mirror and the model fitted in order to extract the reflectance. The simulatedtransmission profile was obtained by the optical transfer matrix method assumingan ideal flat mirror profile using known optical constants of ICPCVD deposited

157


Figure 6.9: High resolution, spatially resolved map of the calibrated mini-

mum transmission measuremnt on an array of quarter wave silicon-air-silicon

mirrors at a wavelengths of 680 nm.

Figure 6.10: High resolution, 3-D plot of a spatially resolved map of cali-

brated minimum transmission measurements on an array of silicon-air-silicon

DBRs at 680 nm wavelength.

158


0

10

20

30

40

50

60

-100 0 100 200 300 400 500 600 700Tr

ansm

ittan

ce (%

)

Distance (µm)

Figure 6.11: Transmission profile of 320 µm diameter visible wavelength

range one-quarter wave silicon-air-silicon mirror at a wavelength of 680 nm.

silicon films. This simulation also used layer thicknesses measured during mirrorfabrication, and predicts a best-case reflectance of 96%. The optical transfer matrixmodel was also used to fit the measured transmission data (“Fitted Tx” in Figure6.12) based on the known structure of the mirror. This fit provides a better estimateof the actual reflectivity (“Calculated Rx” in Figure 6.12) of the mirror. Comparedto the theoretically simulated minimum transmission of 4.3%, the fabricated DBRshave a minimum transmission of 7.7%. This corresponds to 92.3% estimated peakreflectance which is 3.7% lower than the simulated reflectance. From 850 nm to1210 nm the transmission is less than 10%, which corresponds to 90% reflectivityover a 360 nm wavelength range. The mirror has more than 80% reflectivity overthe entire measured wavelength range, which indicate that this silicon-air-siliconDBR qualifies as a good broadband reflector.

Figure 6.13 shows a 3-D high resolution spatially resolved spectral transmissionmeasurement of an array of 26 mirrors demonstrating the variations in the minimumtransmission across a 5.0 mm × 3.0 mm physical grid at a wavelength of 1000 nm.The 3-D spatial mapping shows high homogeneity of transmission across all themirrors, indicating the capability of this fabrication process to produce high qualitymirrors. Such spatial uniformity is key to spectroscopic imaging applications suchas adaptive focal plane arrays. Figure 6.14 shows a transmission profile across a370 µm diameter DBR. The flat transmission profile across the 370 µm mirror at1000 nm wavelength indicates that the top suspended layer is very flat relativeto the bottom layer of the DBR, thus producing a very uniform air gap in thesilicon-air-silicon mirror.

159


0

20

40

60

80

100

950 1070 1190 1310

Simulated TxMeasured TxFitted TxCalculated RxSimulated Rx

Tran

s / R

efl (

%)

Wavelength (nm)

Figure 6.12: Measured and simulated optical transmission spectra of the

quarter wave silicon-air-silicon NIR wavelength mirror. Measured Tx: mea-

sured transmittance data; Simulated Tx: expected transmittance based on

a transmission matrix model using estimated thickness of layers and initial

silicon recipe characterisation; Simulated Rx: expected reflectance of mirror

based on simulated transmission; Fitted Tx: mirror transmittance based on a

fit to the mirror structure using a transmission matrix model; Calculated Rx:

mirror reflectance based on fitted transmission.

0

1

2

3

01

23

45

60.0

0.2

0.4

0.6

0.8

1.0

X (mm)Y (mm)

Tran

smis

sion

0.0

0.2

0.4

0.6

0.8

1.0

Figure 6.13: High resolution, spatially resolved maps of the calibrated trans-

mission measuremnt of an array of one-quarter wave silicon-air-silicon NIR

mirrors.

160


0

20

40

60

80

100

-200 0 200 400 600 800 1000 1200Tr

ansm

ittan

ce (%

)

Distance (µm)

Figure 6.14: Transmission profile of 370 µm diameter NIR wavelength range

one-quarter wave silicon-air-silicon mirror at 1000 nm wavelength.

6.3 NIR Fabry-Perot filter using silicon-silicon oxide-

silicon based DBRs


This section describes the fabrication process and optical characterization of aFabry-Perot filter, incorporating silicon-silicon oxide-silicon based DBRs, designedto operate in the NIR wavelength range. Figure 6.15 shows the fabrication flow forthe NIR Fabry-Perot filter.

1. As shown in Figure 6.15 (a), a silicon-silicon oxide-silicon stack (bottom mir-ror), with each layer being a quarter wave thick, was deposited on a 430 µmthick double side polished BK7 glass substrate. This stack forms the bottommirror. The quarter wave thickness of the silicon and silicon oxide layers were74 nm and 173 nm, respectively.

2. On top of the bottom mirror stack a 100 nm thick gold bottom electrode waspatterned and lifted-off (Figure 6.15 (b)).

3. Prolift100-16 polyimide was then used as the sacrificial layer. It was spun at3600 rpm, soft-baked at 100 ◦C for 2 minutes and then hard baked at 250 ◦Cfor 20 minutes to fully cure it.

4. As shown in Figure 6.15 (d), a second silicon-silicon oxide-silicon stack (topmirror), with each layer being a quarter wave thick, was deposited on top ofthe sacrificial layer.

5. Finally, a 100 nm thick gold top electrode was patterned and deposited (Figure6.15 (e)).

161


(a) (b)

(c) (d)

(e) (f)

(g)

Figure 6.15: Fabrication process of NIR Filter. (a) Deposition of Silicon-Silicon Oxide-Silicon bottom mirror stack on BK7 glass substrate; (b) de-position of bottom electrode; (c) spinning, hard-baking and patterning ofProlift100-16 polyimide; (d) deposition of silicon-silicon oxide-silicon top mir-ror stack; (e) deposition of top electrode; (f) patterning of etch holes; (g)pre-release annealing and dry release in O2 plasma.

6. The top mirror was perforated with 5 µm× 5 µm etch holes using a CF4 basedetch recipe (Figure 6.15 (f)).

7. The fabricated mirrors were pre-annealed at 285 ◦C for 30 minutes in a quartztube furnace in a nitrogen environment. As a final step the sacrificial layerwas then etched in an O2 plasma in a March PMA 60 barrel asher at 160 WRF power and 1 torr chamber pressure, to yield the final suspended structure(see Figure 6.15 (g))

6.3.2 Optical characterization

Optical surface profilometry of NIR filter

The 3-D optical profile of the filter is presented in Figure 6.16. A line scan acrossthe filter shown in Figure 6.16, is shown in Figure 6.17 from which the variation ofthe flatness across the freely suspended 500 µm side-dimension filter was found tobe in the range of 10 nm. The line scan also shows a tilt of 30 nm in the top mirror.The filter shows a bowing of 150 nm along its length. This upward bowing at theends of the filter may is likely to be associated with the “anchor effect” caused by astress differential between the SiOx and Si membrane layers at the anchor location.

162


Figure 6.16: 3-D surface profile of 500 µm × 500 µm NIR Fabry-Perot

filter.

0.00

50.0

100

150

200

0 0.2 0.4 0.6 0.8 1 1.2 1.4

Hei

ght (

nm)

Distance (mm)

Figure 6.17: Line scan of the surface profile of 500 µm × 500 µm NIR filter

shown in Fig. 6.16.

163


0

10

20

30

40

50

60

1000 1050 1100 1150 1200 1250 1300 1350

Tran

smitt

ance

(%)

Wavelength (nm)

Figure 6.18: Single point optical transmission spectra of 500 µm × 500 µm

NIR Fabry-Perot filter with silicon-oxide-silicon DBRs. Measured data is

shown by markers and dashed line shows simulated transmittance.

Optical transmission characterization of NIR filter

Figure 6.18 shows the single point transmittance measurement through the NIRFabry-Perot filter measured in the wavelength range of 960–1500 nm, as well assimulated transmittance of the filter. The simulated transmittance profile was ob-tained by the optical transfer matrix method assuming an ideal flat top mirrorprofile using known optical constants of ICPCVD deposited silicon and silicon ox-ide films. The peak transmission of the filter at the center wavelength of 1140 nmis approximately 54 % with 40 nm FWHM. The fabricated filter shows an excel-lent out of band extinction. In order to check the optical transmittance uniformityof the entire optically active area of the fabricated filter, a calibrated transmissionmeasurement with high spatial and spectral precision was performed. For the trans-mission measurements, light was focused onto the filter down to a beam spot sizeof 20 µm, and measurements were made across a 2 mm × 2 mm grid. Figure 6.19shows the measurement results of the position of the peak wavelength. It can beseen that the optically active area shows excellent uniformity for a center transmis-sion wavelength of 1140 nm. This high optical transmission uniformity across theentire 500 µm × 500 µm area is key for imaging applications.

164


Figure 6.19: High resolution 3-D spatial map of the peak wavelength for

a 500 µm × 500 µm NIR Fabry-Perot filter with silicon-silicon oxide-silicon

DBRs.

6.4 Summary

This chapter presented the design, fabrication and optical characterization of silicon-silicon oxide-silicon and silicon-air-silicon based DBRs, for the visible to near in-frared wavelength range. It was demonstrated that silicon-silicon oxide-silicon basedDBRs give 88% – 89% peak reflectivity as compared to the theoretical 92% – 93%peak reflectivity. Next, successful fabrication of visible and NIR wavelength rangeDBR array consisting of DBRs ranging in diameter between 270 µm and 420 µmwas demonstrated. Calibrated optical measurements indicate that the measuredreflectivity is in close agreement with the theoretical reflectivity. For visible wave-length DBRs at wavelength of 680 nm the measured peak reflectivity was 92%.The DBRs demonstrated broadband reflectance having, 85% or higher reflectivityover a 160 nm wavelength range. The NIR DBRs showed a minimum transmis-sion of 7.7% as compared to the theoretically simulated minimum transmission of4.3%. This corresponded to 92.3% estimated peak reflectance which is 3.7% lowerthan the simulated reflectance. The DBRs had more than 80% reflectivity overthe entire measured wavelength range. The 2-D optical transmission spatial map

165


and optical transmission profiles of the DBRs demonstrated high uniformity acrossthe fabricated DBRs. Finally, the chapter concluded by detailing fabrication andoptical characterization of NIR wavelength range Fabry-Perot filter, incorporatingsilicon-silicon oxide-silicon based DBRs. The peak transmission of the filter at thecenter wavelength of 1140 nm was approximately 54 % with 40 nm FWHM.

166

Chapter 7

Conclusions and Future Work

The first section in this chapter presents conclusions from the research work pre-sented in this thesis. The future scope section identifies the technological challengesin fabricating multiple air-cavities, and provides future directions to transform thistechnology into a fully portable MEMS based micro-spectrometer.

7.1 Conclusions

This thesis has focused on creating a silicon based technology for the fabricationof micro-spectrometers. The thesis has discussed optimization of silicon and siliconoxide based ICPCVD materials, design, fabrication, and characterization of siliconbased distributed Bragg reflectors operating from the visible wavelength to MWIRwavelength range. The present work has demonstrated the possibility of using op-timized silicon thin films for fabrication of high reflectivity air-gap DBRs leading tohigh-performing silicon based micro-spectrometers. This work has also presented anotch based actuation structure to create large tuning range actuation structuresfor tunable Fabry-Perot filters. Thus the contents of this thesis represent a sub-stantial contribution to the subject of ICPCVD based materials, and silicon basedFabry-Perot filter technologies. The theoretical and experimental results have beenthoroughly analyzed, with good agreement between optical models and measuredoptical data. The work makes a significant contribution to the understanding ofthis subject in terms of materials and structures, and opens a pathway to extendthese techniques for large area imaging filters. Thus, the developed technologies arewell suited for creating high performance microspectrometers or wavelength-tunableadaptive focal plane arrays.

The key results and contributions of this thesis are summarized as follows:

1. In chapter 3 a systematic study and optimization of mechanical and opti-cal properties of ICPCVD based silicon was performed. It was found thathigher deposition temperature leads to higher quality films. Furthermore,

167

CHAPTER 7. CONCLUSIONS AND FUTURE WORK

at higher deposition temperatures, the decrease in the inductively coupledpower results in films with low tensile stress, higher refractive index and lowextinction coefficients. This work shows that hydrogen concentration alone isnot a sufficient parameter to determine film properties. We report that thehydrogen concentration and hydrogen-silicon bonding nature together play avital role in improving optical and mechanical quality of silicon thin films,such that even with high hydrogen concentrations, high quality films can beobtained, provided that the hydrogen is bonded with silicon in stable bondingconfigurations.

2. This work has presented the process of optimization of ICPCVD silicon oxide.It was observed that low compressive stress and a high wet etch rate of sili-con oxide can be obtained by depositing silicon oxide films at low depositiontemperature and high deposition pressure (see Section 3.5).

3. Section 4.3 and Section 4.4 presented silicon based surface micro-machined dis-tributed Bragg reflectors, for SWIR and MWIR wavelength, suited to MEMSspectrometer fabrication, with near theoretical reflectance properties. TheseDBRs consist of two silicon films separated by an air gap, with each layerbeing a quarter wave in optical thickness. A flatness variation in the order of10 nm was achieved across the DBRs. It was experimentally established thata conformal support structure around the entire periphery of the DBRs is thebest support structure to create flat DBRs.

4. This work has demonstrated that silicon-air-silicon based DBRs can be fabri-cated for the visible and near infrared-wavelength region. In spite of relativelyhigh absorbance of silicon in these wavelength ranges, the DBRs exhibitedgreater than 90% reflectivity. This work highlights the problem of precisecontrol of stress in silicon thin films for fabricating mirrors based on very-thinlayers of silicon (see Section 6.2).

5. This work has extensively used spatial optical mapping of the fabricated mir-rors to demonstrate their optical uniformity, and has also shown that anair-gap mirror based technology is highly scalable to very large areas. DBRsranging from 200 µm × 200 µm to 5000 µm × 5000 µm size with just 10-15 nm variation in flatness of the suspended membranes were fabricated andoptically characterized. The large area and high optical uniformity are thekey properties for employing them in imaging applications (see Section 4.3and Section 4.4).

6. The thesis has presented limitations of the beam-tether based support struc-ture in producing flat suspended mirrors (see Section 5.2). As a solution forthe limitation of this structure, a new notch based actuation structure hasbeen introduced. It was shown that with the notch based actuation structureit was possible to actuate a suspended membrane beyond 50% of the initialtunable air-gap. This planer actuation mechanism eliminates the need fortethers and beams, and is robust against out of plane stress on the suspended

168


mirrors. The new actuation mechanism can be applied across a wide array ofmirror materials and sizes (see Section 5.5).

7. Based on the optimized ICPCVD silicon recipe, an etch back process, andthe new actuation mechanism, a tunable multi-spectral MWIR filter was fab-ricated and optically characterized. As shown in Section 5.7 the filter showsnear ideal performance with close to 70% peak transmittance, 380 nm FWHMand 800 nm tuning range.

8. A air-gap mirror based SWIR filter was fabricated. The optical transmittancemeasurements indicated a first order peak with a peak transmission of 70%and 90 nm FWHM at a wavelength of 2240 nm. The simulated spectrum pre-dicts 83% peak transmittance and 50 nm FWHM for the first order peak. Thesecond order peak of the measured spectrum shows 53% peak transmissionand 35 nm FWHM at 1550 nm, as compared to a predicted 44% peak trans-mission and 30 nm FWHM. A relatively low transmittance, large FWHM andlow out of band rejection of the measured first order peak of the SWIR filteris due to the relatively low reflectivity of the bottom mirror (see Section 5.8).

7.2 Future work

It has been identified that by using the etch back process with a planer actuationstructure, a single membrane with less than 100 nm thickness can be suspended.However, air cavity based mirrors with less than 100 nm thick silicon films werefound to collapse during dry-release in an oxygen plasma. This raises new questionsto identify the proper choice of sacrificial layer and develop a high yield wet releaseprocess to fabricate ultra-thin silicon film based Fabry-Perot filters having multipleair cavities. A potential challenge in wet-release process is snapping down of mem-branes due to stiction forces while releasing sacrificial and drying the liquid betweenthe membranes. Hence it will require proper selection of sacrificial layers and rigidstructures to increase yield of wet-release process. Use of different sacrificial layersfor each air cavity may allow the use of a combination of dry and wet etch releaseprocesses to fabricate the multiple air-cavities.

Another solution for this issue is to minimize the dry-release time of the membranes,which can be achieved by substrate removal using a cryogenic dry-etch process andforming etch holes on both sides of the filter. The presence of etch holes on boththe top and bottom of the filter are expected to reduce the dry release time. Areliable wet-release process for MEMS filters can be developed as an alternative toa dry-release process. It is expected that such a process will eliminate the effect ofhigh temperatures associated with the dry-release process. Thus, any temperatureinduced deformation of MEMS device can be avoided. Developing a wet-releaseprocess for ultra-thin membrane based filters can be part of a future enhancementof the MEMS fabrication process.

The presented fabrication process and the actuation mechanism can be extended

169


to other materials, such as germanium, to create high finesse filters in the SWIR,MWIR and LWIR spectral ranges. This approach is highly recommended, especiallyfor LWIR tunable filters.

From chapter 4 to chapter 6 there has been extensive discussion and demonstra-tion of proof-of-concept silicon based air-gap mirrors and Fabry-Perot filters forspectroscopic systems. However, these filters require flip-chip bonding based inte-gration with a detector or a focal plane array in order to achieve miniaturization ofthe optics and actuation systems that will allow successful fabrication of portable,proof-of-concept microspectrometers or adaptive focal plane arrays.

In addition to the miniaturization of all system components, there are many otherconsiderations that need to be addressed in order to realize a practical microspec-trometer system, which includes signal-to-noise ratio of the measured data, thelong-term stability of the filter characteristics, the method of wavelength calibra-tion, and the speed of data acquisition. Since the complete development of a prac-tical microspectrometer system is beyond the scope of this thesis, these still remainas future work to be undertaken.

170

Bibliography

[1] B. Stenberg, R. Viscarra, A. Raphael, A. M. Mouazen, and J. Wetter-lind, “Visible and near infrared spectroscopy in soil science,” Advances inAgronomy, vol. 107, pp. 163–215, 2010.

[2] R. Kinoshita, B. N. Moebius-Clune, H. M. Es, W. D. Hively, and A. V. Bil-gilis, “Strategies for soil quality assessment using visible and near-infraredreflectance spectroscopy in a western kenya chronosequence,” Soil sciencesociety of america journal, vol. 76, no. 5, pp. 1776–1788, 2012.

[3] A. Ghulam, Z. L. Li, Q. Qin, H. Yimit, and J. Wang, “Estimating crop waterstress with etm+ nir and swir data,” Agricultural and Forest Meteorology,vol. 148, no. 11, pp. 1679–1695, 2008.

[4] S. Sankaran, J. M. Maja, S. Buchanon, and R. Ehsani, “Huanglongblng (cit-rus gntenlng) detection using visible,near lnfranred and thermal imaging tech-niques,” Sensors(Basel), vol. 13, pp. 2117–2130, 2013.

[5] J. Kumirska, M. Czerwicka, Z. Kaczyski, A. Bychowska, K. Brzozowski,J. Thming, and P. Stepnowski, “Application of spectroscopic methods forstructural analysis of chitin and chitosan,” Marine Drug, vol. 8, no. 5, p.15671636, 2010.

[6] G. G. Shepherd, Spectral imaging of the atmosphere. Academic Press, 2002,vol. 82.

[7] P. B. Fellgett, “On the ultimate sensitivity and practical performance of ra-diation detectors,” Journal of Optical Soceity of America, vol. 39, no. 11, pp.970–976, 1949.

[8] W. Kaye, C. Canon, and R. G. Devaney, “Modification of a beckman modeldu spectrophotometer for automatic operation at 210-2700 mu,” Journal ofOptical Society of America, vol. 41, no. 10, pp. 658–664, 1951.

[9] J. M. Montes, H. F. Utz, W. Schipprack, B. Kusterer, J. Muminovic, C. Paul,and A. E. Melchinger, “Near-infrared spectroscopy on combine harvestersto measure maize grain dry matter content and quality parameters,” PlantBreeding, vol. 125, no. 6, pp. 591–595, 2006.

171

BIBLIOGRAPHY

[10] “Fibre Optic Spectrometer Systems,” http://www.fibrephotonics.com/page/99/fibre-optic-spectrometer-systems.htm, 2015, [Online; accessed 11-May-2015].

[11] “Modular Spectrometers,” http://oceanoptics.com/product-category/modular-spectrometers/, 2015, [Online; accessed 11-May-2015].

[12] “Compact CCD Spectrometers,” https://www.thorlabs.com/newgrouppage9.cfm?objectgroup ID=3482, 2015, [Online; accessed 11-May-2015].

[13] “Mini-Spectrometers,” http://www.hamamatsu.com/us/en/4016.html, 2015,[Online; accessed 11-May-2015].

[14] K. K. M. B. D. Silva, “Optical coherence tomography: Technology enhance-ments and novel applications,” Ph.D. dissertation, School of EECE, UWA,Crawley, 2003.

[15] J. W. Goodman, Statistical Optics. John Wiley and Sons, Inc., 1985.

[16] “ANIR,” www.oceanoptics.eu/documents/ANIR.pdf, 2015, [Online; accessed11-May-2015].

[17] S. E. Alper and T. Akin, “A single-crystal silicon symmetrical anddecoupled MEMS gyroscope on an insulating substrate,” Journal ofMicroelectromechanical Systems, vol. 14, no. 4, 2005.

[18] S. T. Microelctronics, “Mems motion sensor:ultra-stable three-axis digital out-put gyroscope,” L3G4200D Datasheet, pp. 1–42, 2012.

[19] H. L. Kung, S. R. Bhalotra, J. D. Mansell, S. Member, Jr., S. Member,S. Member, D. A. B. Miller, and J. S. Harris, “Standing-wave transformspectrometer based on integrated MEMS mirror and thin-film photodetector,”in in Proc. of IEEE Lasers and Electro-Optics Society 2002 Annual Meeting,2002, pp. 10–14.

[20] S. Bhalotra, H. Kung, D. Knipp, H. Stiebig, and D. Miller, “All-silicon standing-wave microspectrometer with spectral resolution,” in OpticalMEMS, 2003 IEEE/LEOS International Conference on, Aug 2003, pp. 26–27.

[21] “MicroNIR Pro Spectrometer,” http://www.jdsu.com/en-us/optical-security-and-performance/products/a-z-product-list/Pages/micronir-spectrometer.aspx?rcode=micronir#.VVBa0 mqqko, 2015, [Online;accessed 11-May-2015].

[22] K. University, http://mems.ku.edu.tr/?page id=88, 2008, [Online; accessed02-Nov-2015].

[23] A. Tortschanoff, M. Lenzhofer, A. Frank, A. Kenda, T. Sandner, andH. Schenk, “Improved MEMS based FTIR spectrometer,” in InternationalSymposium on Optomechatronic Technologies, 2009. ISOT 2009., Sept 2009,pp. 116–121.

172

http://www.fibrephotonics.com/page/99/fibre-optic-spectrometer-systems.htm

http://www.fibrephotonics.com/page/99/fibre-optic-spectrometer-systems.htm

http://oceanoptics.com/product-category/modular-spectrometers/

http://oceanoptics.com/product-category/modular-spectrometers/

https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_ID=3482

https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_ID=3482

http://www.hamamatsu.com/us/en/4016.html

www.oceanoptics.eu/documents/ANIR.pdf

http://www.jdsu.com/en-us/optical-security-and-performance/products/a-z-product-list/Pages/micronir-spectrometer.aspx?rcode=micronir#.VVBa0_mqqko



http://mems.ku.edu.tr/?page_id=88

BIBLIOGRAPHY

[24] Y. Hongbin, Z. Guangya, C. F. Siong, L. Feiwen, W. Shouhua, and Z. Ming-sheng, “An electromagnetically driven lamellar grating based fourier trans-form microspectrometer,” Journal of Micromechanics and Microengineering,vol. 18, no. 5, pp. 1–6, 2008.

[25] A. Tortschanoff, M. Lenzhofer, A. Frank, A. Kenda, T. Sandner, andH. Schenk, “High resolution miniature FTIR spectrometer enabled by a largelinear travel MEMS pop-up mirror,” in Proc. of SPIE, vol. 7319, 2009, pp.73 190J1–8.

[26] M. Ebermann, K. Hiller, S. Kurth, and N. Neumann, “Design, operation andperformance of a Fabry-Perot-based MWIR microspectrometer,” in SENSORand Test Conference 2009, 2009, pp. 233–238.

[27] C. A. Musca, J. Antoszewski, K. J.Winchester, A. J. Keating, T. Nguyen,K. K. M. B. D. Silva, J. M. Dell, L. Faraone, P. Mitra, J. D. Beck, M. R.Skokan, and J. E. Robinson, “Monolithic integration of an infrared photondetector with a mems-based tunable filter,” IEEE Electron Device Letters,vol. 26, no. 12, pp. 888–890, 2005.

[28] J. S. Milne, “Micro-electromechanical technologies for next-generation spec-troscopic systems,” Ph.D. dissertation, School of Electrical, Electronics andComputer Engineering, 2009.

[29] J. Dell, A. Keating, J. S. Milne, J. Antoszewski, C. Musca, L. Faraone,D. Murphy, and O. Samardzic, “Micro-electromechanical systems-based mi-crospectrometers covering wavelengths from 1500 nm to 5000 nm,” in SPIEDefense and Security, Florida, USA, 2007.

[30] P. Hauff, An overview of Vis-NIR-SWIR field Spectroscopy as applied toprecious metals exploration. Spectral International Inc., 2001.

[31] C. Singh, D. Jayasa, J. Paliwala, and N. White, “Detection of insect-damagedwheat kernels using near-infrared hyperspectral imaging,” Journal of StoredProducts Research, vol. 45, pp. 151–158, 2008.

[32] E. Baldini, O. Facini, F. Nerozzi, F. Rossi, and A. Rotondi, “Leaf character-istics and optical properties of different woody species,” Trees, vol. 12, pp.73–81, 1997.

[33] R. D. Driver and K. Didona, “On-line high-speed NIR diffuse-reflectanceimaging spectroscopy in food quality monitoring,” Application notes,HeadwellPhotonics, pp. 1–8, 2008.

[34] W. H. Hayt Jr. and J. A. Buck, Engineering Electromagnetics. McGraw-HillHigher Education, 2006.

[35] H. A. Macleod, Thin-Film Optical Filters. New York: McGraw-Hill, 1989.

173

BIBLIOGRAPHY

[36] M. Born and E. Wolf, Principles of Optics, 7th ed. (Cambridge UniversityPress, Cambridge, 1999.

[37] G. Fowles, Introduction to Modern Optics, 1st ed. Dover Publications, NewYork, 1990.

[38] D. Griffiths, Introduction to Electrodynamics, 3rd ed. Benjamin Cummings,New York, 1999.

[39] E. Garmire, “Theory of quarter-wave-stack dielectric mirrors used in a thinFabry-Perot filter,” Applied Optics, vol. 2, pp. 5442–5449, 1992.

[40] D. I. Babic and S. W. Corzine, “Analytic expressions for the reflection delay,penetration depth, and absorptance of quarter-wave dielectric mirrors,” IEEEJournal of Quantum Electronics, vol. 28, pp. 514–524, 1992.

[41] J. M. Vaughan, The Fabry-Perot Interferometer History,Theory,Practice andApplications. Bristol: Hilger, 1989.

[42] G. J. Sloggett, “Fringe broadening in Fabry-Perot interferometers,” AppliedOptics, vol. 23, pp. 2427–2432, 1984.

[43] P. Atherton, N. K. Reay, T. R. Hicks, and J. Ring, “Tunable Fabry-PerotFilters,” Optical Engineering, vol. 20, pp. 806–814, 1981.

[44] G. Hernandez, Fabry-Perot Interferometers. Cambridge UniversityPress,London, 1988.

[45] C. Roychoudhuri and M. Hercher, “Stable multipass Fabry-Perot interferom-eter - design and analysis,” Applied Optics, vol. 16, pp. 2514–2520, 1977.

[46] K. J. Winchester and J. M. Dell, “Tunable Fabry-Perot cavities fabricatedfrom PECVD silicon nitride employing zinc sulphide as the sacrificial layer,”Journal of Micromechanics and Microengineering, vol. 11, no. 2001, pp. 589–594, 2001.

[47] D. Cristea, M. Kusko, C. Tibeica, R. Muller, E. Manea, and D. Syvridis,“Integrated tunable optical interferometer obtained using <111> orientedsilicon micromachining,” in International Semiconductor Conference, 2003.

[48] J. H. Correia, M. Bartek, and R. F.Wolffenbuttel, “High-selectivity single-chip spectrometer in silicon for operation at visible part of the spectrum,”IEEE Transaction on Electronics Devices, vol. 47, no. 3, pp. 553–559, 2000.

[49] R. F. Wolffenbuttel, “State-of-the-art in integrated optical microspectrome-ters,” IEEE Transaction on Instrumentation and Measurements, vol. 53, no. 1,pp. 197–202, 2004.

[50] M. Datta, M. W. Pruessner, D. P. Kelly, and R. Ghodssi, “Design of MEMS-tunable novel monolithic optical filters in InP with horizontal bragg mirrors,”Solid-State Electronics, vol. 48, pp. 1959–1963, 2004.

174

BIBLIOGRAPHY

[51] A. Hasse, S. Irmer, J. Daleiden, N. Dharmarasu, S. Hansmann, andH. Hillmer, “Wide continuous tuning range of 221 nm by InP/air-gap vertical-cavity filters,” Electronics Letters, vol. 42, no. 17, p. 974, 2006.

[52] M. Garrigues, J. Danglot, J.-L. Leclercq, and O. Parillaud, “Tunable high-finesse InP/Air moems filter,” IEEE Photonics Technology Letters, vol. 17,no. 7, pp. 1471–1473, 2005.

[53] J. Daleiden, S. Irmer, V. Rangelov, F. Ramer, A. Tarraf, C. Prott, M. Strass-ner, and H. Hillmer, “Record wavelength tuning of 127 nm for vertical cav-ity fabry-perot filter,” in IEEE/LEOS International Conference on OpticalMEMs, 2002.

[54] A. Spisser, R. Ledantec, C. Seassal, J. L. Leclercq, T. Benyattou, D. Rondi,R. Blondeau, G. Guillot, and P. Viktorovitch, “Highly selective and widelytunable 1.55 m InP/Air-gap micromachined FabryPerot filter for optical com-munications,” IEEE Photonics Technology Letters, vol. 10, no. 9, pp. 1259–1261, 1998.

[55] E. Ochoa, T. Nelson, R. Bedford, J. Ehret, L. Starman, M. Harvey, T. An-derson, and F. Ren, “Demonstration of hybrid AlxGa(1−x)Aspolysilicon mi-croelectromechanical tunable filter,” IEEE Photonics Technology Letters,,vol. 19, no. 6, pp. 381–383, 2007.

[56] C.-K. Kim, M.-L. Lee, and C.-H. Jun, “Electrothermally actuated fabryprottunable filterwith a high tuning efficiency,” IEEE Photonics TechnologyLetters, vol. 16, no. 4, pp. 1894–1896, 2004.

[57] T. Amano, T. Hino, F. Koyama, M. Arai, and A. Matsutani, “A thermallytunable GaAlAsGaAs micromachined optical filter with submillisecond tuningspeed,” IEEE Photonics Technology Letters, vol. 16, no. 6, pp. 1501–1503,2004.

[58] N. Quack, S. Blunier, J. Dual, F. Felder, M. Arnold, , and H. Zogg, “Mid-infrared tunable resonant cavity enhanced detectors,” Sensors, vol. 8, no. 9,pp. 5466–5478, 2008.

[59] C. F. R. Mateus, C.-H. Chang, L. Chrostowski, S. Yang, D. Sun, R. Pathak,Chang-Hasnain, and J. Connie, “Widely tunable torsional optical filter,”IEEE Photonics Technology Letters, vol. 14, no. 6, 2002.

[60] S. R. Mallinson and J. H. Jerman, “Miniature micromachined Fabry–Perotinterferometers in silicon,” Electronics Letters, vol. 23, pp. 1041–1043, 1987.

[61] S. Irmer, K. Alex, J. Daleiden, I. Kommallein, M. Oliveira, F. Rmer, A. Tar-raf, and H. Hillmer, “Surface micromachined optical low-cost all-air-gap fil-ters based on stress-optimized Si3N4layers,” Journal of Micromechanics andMicroengineering, vol. 15, no. 4, pp. 867–872, 2005.

175

BIBLIOGRAPHY

[62] M. Tuohiniemi and M. Blomberg, “Surface-micromachined silicon air-gap bragg reflector for thermal infrared,” Journal of Micromechanics andMicroengineering, vol. 21, no. 7, p. 075014, 2011.

[63] B. Saadany, M. Malak, F. Marty, Y. Mita, D. Khalil, and T. Bourouina,“Electrostatically-tuned optical filter based on silicon bragg reflectors,” inIEEE/LEOS international conference on optical MEMS and their applicationsconference, 2006, pp. 86–87.

[64] M. W. Pruessner, T. H. Stievater, and W. S. Rabinovich, “In-plane microelec-tromechanical resonator with integrated fabryperot cavity,” Applied PhysicsLetters, vol. 92, no. 8, p. 081101, 2008.

[65] A. Lipson and E. M. Yeatman, “A 1-D photonic band gap tunable opticalfilter in <100> silicon,” Journal of Micromechanical Systems, vol. 16, no. 3,pp. 521–527, 2007.

[66] A. J. Keating, K. K. M. B. D. Silva, J. M. Dell, C. A. Musca, and L. Faraone,“Optical characterization of fabryprot mems filters integrated on tunableshort-wave ir detectors,” IEEE Photonics Technology Letters, vol. 18, no. 9,pp. 1079–1081, 2006.

[67] T. J. Russin, M. Kerber, A. Russin, A. Wang, and R. Waters, “Fabrica-tion and analysis of a MEMS NIR Fabry-Perot interferometer,” Journal ofMicroelectromechanical Systems, vol. 21, no. 1, pp. 181–189, 2012.

[68] M.Tuohiniemi, M. Blomberg, A. Akujrvi, J. Antila, and H. Saari, “Op-tical transmission performance of a surface-micromachined Fabry-Perotinterferometer for thermal infrared,” Journal of Micromechanics andMicroengineering, vol. 22, no. 11, p. 115004, 2012.

[69] K. S. M. Noro, N. Kishi, H. Hara, and T. W. H. Iwaoka, “CO2/H2O gas sensorwith a MEMS Fabry-Perot Filter,” in SICE, CANADA, 2002.

[70] H.-K. Lee, K.-S. Kim, and E. Yoon, “A wide-range linearly tunable opticalfilter using lorentz force,” IEEE Photonics Technology Letters, vol. 16, no. 9,pp. 2087–2090, 2004.

[71] E. R. H. Halbritter, M. Aziz and P. Meissner, “Electrothermally tunable two-chip optical filter with very low-cost and simple concept,” Electronics Letters,vol. 38, no. 20, 2002.

[72] H. Z. D. Hohlfeld, M. Epmeier, “Tunable thermo-optic filter for WDM appli-cations,” in The Fifteenth IEEE International Conference on Micro ElectroMechanical Systems, 2002. IEEE, 2002, pp. 564 – 567.

[73] D. Hohlfeld and H. Zappe, “An all-dielectric tunable optical filter based onthe thermo-optic effect,” Journal of Optics A: Pure and Applied Optics, vol. 6,no. 6, pp. 504–511, 2004.

176

BIBLIOGRAPHY

[74] E. Pichonat-Gallois, V. Petrini, and M. d. Labachelerie, “Design and fabri-cation of thermal actuators used for a micro-optical bench: application to atunable fabry-perot filter,” Sensors and Actuators A: Physical, vol. 114, no.2-3, pp. 260–266, 2004.

[75] M. Meinig, M. Ebermann, N. N. an S. Kurth, K. Hiller, and T. Gessner,“Dual-band MEMS Fabry–Perot filter with two movable reflectors for mid-and long-wave infrared microspectrometers,” in 16th International Solid-StateSensors, Actuators and Microsystems Conference (TRANSDUCERS), 2011,pp. 2538 – 2541.

[76] P. A. Stupar, R. L. Borwick, J. F. DeNatale, P. H. Kobrin, and W. J. Gunning,“MEMS tunable Fabry-Perot filters with thick, two sided optical coatings,”in International Solid-State Sensors, Actuators and Microsystems Conference,2009, pp. 1357–1360.

[77] J. Jerman, D. Clift, and S. Mallinson, “A miniature fabry-perot intederometerwith a corrugated silicon diaphragm support,” in IEEE Solid-State Sensor andActuator Workshop, 1990, pp. 140–144.

[78] S. Li, S. Zhong, J. Xu, and Y. Wu, “A thermal tunable bulk-micromachinedoptical filter,” in 16th International Solid-State Sensors, Actuators andMicrosystems Conference (TRANSDUCERS), 2011, 2011, pp. 550 – 553.

[79] N. Kanbara, K. Suzuki, T. Watanabe, and H. Iwaoka, “Precisely tunableFabry-Perot filter for optical communications,” in IEEE/LEOS InternationalConference on Optical MEMs, 2002.

[80] M. Malak, N. Pavy, F. Marty, Y.-A. Peter, A. Liu, and T. Bourouina, “Stable,high-q Fabry-Perot resonators with long cavity based on curved, all-silicon,high reflectance mirrors,” in IEEE 24th International Conference on MicroElectro Mechanical Systems (MEMS), 2011. IEEE, 2011, pp. 720 – 723.

[81] R. P. Schneider and J. A. Lott, “InAlP/InAlGaP distributed bragg reflectorsfor visible vertical cavity surface-emitting lasers,” Applied Physics Letters,vol. 62, no. 22, pp. 2748–2750, 1993.

[82] N. Chitica, M. Strassner, A. Spisser, J. L. Leclercq, P. Viktorovitch, D. Rondi,E. Goutin, and e. al., “Tunable InP/Air gap Fabry-Perot filter for wave-length division multiplex fiber optical transmission,” in Eleventh InternationalConference on Indium Phosphide and Related Materials, 1999. IPRM., 1999,pp. 285–287.

[83] N. Nakada, M. Nakaji, H. Ishikawa, T. Egawa, M. Umeno, and T. Jimbo, “Im-proved characteristics of InGaN multiple-quantum-well light-emitting diodeby GaN/AlGaN distributed bragg reflector grown on sapphire,” AppliedPhysics Letters, vol. 76, no. 14, pp. 1804–1806, 2000.

177

BIBLIOGRAPHY

[84] S. Irmer, J. Daleiden, V. Rangelov, C. Prott, M. S. Friedhard Rmer, A. Tarraf,and H. Hillmer, “Ultralow biased widely continuously tunable FabryPerotfilter,” IEEE Photonics Technology Letters, vol. 15, no. 3, 2003.

[85] K. E. Petersen, “Silicon as a mechanical material,” Proceedings of the IEEE,vol. 70, no. 5, pp. 420–457, 1982.

[86] W. P. Eaton and J. H. Smith, “Micromachined pressure sensors: review andrecent developments,” Smart Materials and Structures, vol. 6, pp. 530–539,1997.

[87] G. T. A. Kovacs, N. I. Maluf, and K. E. Petersen, “Bulk micromachining ofsilicon,” in Proceedings of IEEE, vol. 86, 1998, pp. 1536–1551.

[88] M. Janai, D. Allred, D. Booth, and B. Seraphin, “Optical properties and struc-ture of amorphous silicon films prepared by CVD,” Solar Energy Materials 1,pp. 11–27, 1979.

[89] A. Pleschinger, J. Lutz, F. Kuchar, H. Noll, and M. Pippan, “Study of poly-crystalline and amorphous LPCVD silicon films by atomic forcemicroscopy,”Surface and Interface Analysis, vol. 25, pp. 529–532, 1997.

[90] W.-C. Hsiao, C.-P. Liu, and Y.-L. Wang, “Influence of RF Bias on hydro-genated amorphous silicon by high-density plasma chemical vapor deposi-tion,” Journal of The Electrochemical Society, vol. 154, no. 5, pp. 122–126,2007.

[91] Y. Qin, H. Yan, F. Li, L. Qiao, Q. Liu, and D. He, “The optoelectronicproperties of silicon films deposited by inductively coupled plasma CVD,”Applied Surface Science, vol. 257, no. 3, pp. 817–822, 2010.

[92] E. D. Palik, Handbook of Optical constants of Solids. San Diego, USA:Academic Press, 1998.

[93] I. Institute, http://www.ioffe.ru/SVA/NSM/Semicond/index.html, 2001,[Online; accessed 09-Jan-2015].

[94] H. Huang, Y. Liu, X. Hu, M. Bush, K. Winchester, C. Musca, J. Dell, andL. Faraone, “Mechanical design and finite element analysis of tunable Fabry-Perot MEMS structures for adaptive infrared detectors,” in Conference onOptoelectronic and Microelectronic Materials and Devices, 2004, 2004.

[95] H. Huang, K. Winchester, Y. Liu, X. Z. Hu, C. A. Musca, J. M. Dell,and L. Faraone, “Determination of mechanical properties of pecvd siliconnitride thin films for tunable mems fabryprot optical filters,” Journal ofMicromechanics and Microengineering, vol. 15, no. 3, pp. 608–614, 2005.

[96] G. G. Stoney, “The tension of metallic films deposited by electrolysis,”Proceedings of Royal Society, vol. 82, pp. 172–175, 1909.

178

http://www.ioffe.ru/SVA/NSM/Semicond/index.html

BIBLIOGRAPHY

[97] H. Guckel, D. Burnst, C. Rutigliano, E. Lovell, and B. Choi, “Diagnosticmicrostructures for the measurement of intrinsic strain in thin films,” JournalMicromechanics and Microengineering, vol. 2, pp. 86–95, 1992.

[98] X. Feng, Y. Huang, and A. J. Rosakis, “On the stoney formula for athin film/substrate system with non-uniform substrate thickness,” Journalof Applied Mechanics, vol. 74, pp. 1276–81, 2007.

[99] S. Bosch, J. Ferre-Borrull, and J. Sancho-Parramon, “A general-purpose soft-ware for optical characterization of thin films: specific features for microelec-tronic applications,” Solid-State Electronics, vol. 45, pp. 703–709, 2001.

[100] D. Basa and F. Smith, “Annealing and crystallization process in a hydro-genated amorphous Si-C alloy film,” Thin Solid Films, vol. 192, pp. 121–133,1990.

[101] C.-T. Lee, J.-H. Cheng, Y.-M. Chen, and T.-C. Tsai, “Investigation of silicon-germanium films deposited by plasma-enhanced chemical vapor depositionusing laser assistance,” ECS Transactions, vol. 3, pp. 197–206, 2006.

[102] R. A. Street, Hydrogenated amorphous silicon. Cambridge, 1991.

[103] E. Bustarret, M. Bensouda, M. Habrard, J. Bruyre, S. Poulin, and S. Gujrathi,“Configurational statistics in a− SixNyHz alloys: A quantitative bondinganalysis,” Physical Review B, vol. 38, no. 12, pp. 8171–8184, 1988.

[104] A. A. Langford, M. L. Fleet, B. P. Nelson, W. A. Lanford, and N. Maley, “In-frared absorption strength and hydrogen content of hydrogenated amorphoussilicon,” Physical Review B, vol. 45, pp. 13 367–13 377, Jun 1992.

[105] H. Shanks, C. J. Fang, L. Ley, M. Cardona, F. J. Demond, and S. Kalb-itzer, “Infrared spectrum and structure of hydrogenated amorphous silicon,”Physica Status Solidi B, vol. 100, no. 1, pp. 43–56, 1980.

[106] R. K. Nachimuthu, R. D. Jeffery, M. Martyniuk, R. C. Woodward, P. J.Metaxas, J. M. Dell1, and L. Faraone, “Investigation of cerium-substitutedeuropium iron garnets deposited by biased target ion beam deposition,” IEEETransaction on Magnetics, vol. 50, no. 12, pp. 1–7, 2014.

[107] D. A. Baldwin, M. Martyniuk, R. C. Woodward, C. C. Nunes, and R. D.Jeffery, “High resolution miniature FTIR spectrometer enabled by a largelinear travel MEMS pop-up mirror,” in Proc. 52nd Annu. Tech. Conf., Soc.Vac. Coaters, vol. 15, 2009, p. 1518.

[108] T. L. Hylton, B. Ciorneiu, D. Baldwin, O. Escorcia, J. Son, M. McClure,and G. Waters, “Thin film processing by biased target ion beam deposition,”IEEE Transaction on Magnetics, vol. 36, no. 5, p. 29662971, 2000.

179

BIBLIOGRAPHY

[109] J. Yun, S. Varlamov, J. Huang, K. Kim, and M. A. Green, “Effect of deposi-tion temperature on electron-beam evaporated polycrystalline silicon thin-filmand crystallized by diode laser,” Applied Physics Letters, vol. 104, no. 24, pp.1–5, 2014.

[110] T. Sontheimer, C. Becker, F. Ruske, C. Klimm, U. Bloeck, S. Gall, O. Kunz,T. Young, R. Egan, J. Hupkes, and B. Rech, “Challenges and opportunities ofelectron beam evaporation in the preparation of poly-si thin film solar cells,”in Photovoltaic Specialists Conference (PVSC), 2010 35th IEEE, June 2010,pp. 614–619.

[111] C. Becker, T. Sontheimer, S. Steffens, S. Scherf, and B. Rech, “Polycrystallinesilicon thin films by high-rate electronbeam evaporation for photovoltaic ap-plications influence of substrate texture and temperature,” Energy Procedia,vol. 10, no. 0, pp. 61 – 65, 2011.

[112] K. Bordo and H.-G. Rubhan, “Effect of deposition rate on the microstructureof electron beam evaporated nanocrystalline palladium thin films,” MaterialScience, vol. 18, no. 4, pp. 313 – 317, 2012.

[113] B. Amin-Ahmadi, H. Idrissi, M. Galceran, M. Colla, J. Raskin, T. Pardoen,S. Godet, and D. Schryvers, “Effect of deposition rate on the microstructureof electron beam evaporated nanocrystalline palladium thin films,” Thin SolidFilms, vol. 539, no. 0, pp. 145 – 150, 2013.

[114] W.-J. Liu, S. Chen, H.-Y. Cheng, J.-D. Lin, and S.-L. Fu, “Fabrication ofamorphous silicon films for arrayed waveguide grating application,” Surfaceand Coatings Technology, vol. 201, no. 15, pp. 6581–6584, 2007.

[115] A.Torkkeli, O. Rusanen, J. Saarilahti, J. Hietanen, H. Seppa, and H.Sipola,“Capacitive microphone with low-stress polysilicon membrane and high-stresspolysilicon backplate,” Sensors and Actuators, vol. 85, pp. 116–123, 2000.

[116] S. Chang and S. Sivoththaman, “Development of a low temperature MEMSprocess with a PECVD amorphous silicon structural layer,” Journal ofMicromechanics and Microengineering, vol. 16, no. 7, pp. 1307–1313, 2006.

[117] W. H. M. Lodder, “Powder formation in silane plasma,” Philips researchlaboratories, Tech. Rep., 1997.

[118] G. Oversluizena and W. H. M. Lodders, “Optimization of plasma-enhancedchemical vapor deposition of hydrogenated amorphous silicon,” Journal ofApplied Physics, vol. 83, no. 12, pp. 8002–8009, 1998.

[119] M. Netrvaloa, M. Fischer, J. Mullerova, M. Zeman, and P. Sutta, “Structureand optical properties of the hydrogen diluted a-Si:H thin films prepared bypecvd with different deposition temperatures,” in 8th International Conferenceon Advanced Semiconductor Devices and Microsystems, 2010.

180

BIBLIOGRAPHY

[120] G. Muller, “On the generation and annealing of dangeling bond defects inhydrogenated amorphous silicon,” Applied Physics A: Solids and Surfaces,vol. 45, pp. 41–51, 1988.

[121] F. Gaspari, Optoelectronics - Materials and Techniques. InTech, 2011.

[122] E. Fefer and Y. Shapira, “Direct determination of the band-gap states inhydrogenated amorphous silicon using surface photovoltage spectroscopy,”Applied Physics Letters, vol. 67, no. 3, pp. 371–373, 1995.

[123] S. Yamasaki, “Optical absorption edge of hydrogenated amorphous siliconstudied by photoacoustic spectroscopy,” Philosophical Magazine Part B,vol. 56, no. 1, pp. 79–97, 1987.

[124] A. Shyam, “Fabrication of high quality, low bandgap amorphous silicon andamorphous silicon germanium alloy solar cell by chemical annealing,” Ph.D.dissertation, Iowa state univeristy, School of Eleectrical engineering, 2011.

[125] D. Biegelsen, R. Street, C.Tsai, and . J Knights, “Hydrogen evolution anddefect creation in amorphous Si:H alloys,” Physical Review B, vol. 20, no. 12,pp. 4839–4846, 1979.

[126] R.Biswas, I. Kwon, and C. Soukouli, “Mechanism for the staebler-wronskieffect in a-Si:H,” Physical Review B, vol. 44, no. 7, pp. 3403–3406, 1991.

[127] F. J. Kampas and R. W. Griffith, “Hydrogen elimination during the glowdischarge deposition of a-si: H alloys,” Applied Physics Letters, vol. 39, no. 5,pp. 407–409, 1981.

[128] A. Matsuda and K. Tanaka, “Investigation of the growth kinetics of glowdis-charge hydrogenated amorphous silicon using a radical separation technique,”Journal of Applied Physics, vol. 60, no. 7, pp. 2351–2356, 1986.

[129] H.Kakinuma, S. Nishikawa, T. Watanabe, and K. Nihei, “Intrinsic stressand hydrogen bonding in glow-discharge amorphous silicon films,” Journalof Applied Physics, vol. 59, no. 9, p. 3110, 1986.

[130] M. Wu, W. Li, Y. Qin, J. Fu, and Y. Jiang, “Hydrogen bonding in hy-drogenated amorphous silicon thin films prepared at different precursor gastemperatures with undiluted silane,” Science China Technological Sciences,vol. 54, pp. 2310–2314, 2011.

[131] J. Gope, S. Kumar, S. D. A. Parashar, C. M. S. Rauthan, and P. C. Srivas-tava, “Effect of hydrogen content and bonding environment on mechanicalproperties of hydrogenated silicon films deposited by high-frequency pecvdprocess,” ISRN Nanomaterials, vol. 2012, pp. 1–9, 2012.

[132] H.-J. Lin and S.-H. Chen, “Effect of the hydrogen concentration on the growthmechanism of sputtered hydrogenated silicon thin films,” Optical MaterialExpress, vol. 3, no. 16, pp. 1215–1222, 2013.

181

BIBLIOGRAPHY

[133] L. Remache, A. Mahdjoub, E. Fourmond, J. Dupuis, and M. Lemiti, “Influ-ence of PECVD SiOx and SiNx:H films on optical and passivation propertiesof antireflective porous silicon coatings for silicon solar cells,” Physica StatusSolidi C, vol. 8, no. 6, pp. 1893–1897, 2011.

[134] F. A. Hassani, Y. Tsuchiya, and H. Mizuta, “In-plane resonant nano-electro-mechanical sensors: a comprehensive study on design, fabrication and char-acterization challenges,” Sensors (Basel), vol. 13, no. 7, pp. 9364–87, 2013.

[135] S. Keipert-Colberg, N. Barkmann, C. Streich, A. Schtt, D. Suwito, Pe-traSchfer, S. Mller, and D. Borchert, “Investigation of a PECVD silicon ox-ide/silicon nitride passivation system concerning process influences,” in 26thEuropean PV Solar Energy Conference and Exhibition, 2011, pp. 1–4.

[136] M. L. Green, M.-Y. Ho, B. Busch, G. D. Wilk, T. Sorsch, T. Conard, B. Brijs,W. Vandervorst, P. I. Risnen, D. Muller, M. Bude, and J. Grazul, “Nucleationand growth of atomic layer deposited HfO2 gate dielectric layers on chemicaloxide (SiOH) and thermal oxide (SiO2 or SiON) underlayers,” Journal ofApplied Physics, vol. 92, no. 12, pp. 7168–7174, 2002.

[137] M. Morita, T. Ohmi, E. Hasegawa, M. Kawakami, and K. Suma, “Controlfactor of native oxide growth on silicon in air or in ultra pure water,” AppliedPhysics Letters, vol. 55, no. 6, pp. 562–564, 1989.

[138] Y. Ono, M. Tabe, and H. Kageshima, “Scanning-tunneling-microscopy obser-vation of thermal oxide growth on Si <100> surfaces,” Physical Review B,vol. 48, pp. 14 291–14 300, 1993.

[139] K. Watanabe, T. Tanigaki, and S. Wakayama, “The Properties ofLPCVD SiO2 Film Deposited by SiH2Cl2 and N2O Mixtures,” Journal ofElectrochemical Society, vol. 128, no. 12, pp. 2630–2635, 1981.

[140] G. Lucovsky, M. J. Manitini, J. Srivastava, and E. A. Irene, “Low- temper-ature growth of silicon dioxide films: A study of chemical bonding by ellip-sometry and infrared spectroscopy,” Journal Vaccum Science and TechnologyB, vol. 5, no. 2, pp. 530–537, 1987.

[141] I. Pereyra and M. Alayo, “High quality low temperature DPECVD silicondioxide,” Journal of Non-Crystalline Solids, vol. 212, pp. 225–231, 1997.

[142] T. Chung, M. Kang, C. Chung, and Y. Kim, “Effects of process parameterson the properties of silicon oxide films using plasma enhanced chemical va-por deposition with tetramethoxysilane,” Current Applied Physics, vol. 9, p.598604, 2009.

[143] M. Moravej, G. S. Selwyn, R. F. Hicks, S. E. Babayan, J. Y. Jeong, A. Schutze,and V. Tu, “Deposition of silicon dioxide films with a non-equilibriumatmospheric-pressure plasma jet,” Plasma Sources Sci. Technology, vol. 10,pp. 573–578, 2001.

182

BIBLIOGRAPHY

[144] T. Chen, R. Ishihara, and K. Beenakker, “High quality SiO2 deposited at 80 Cby inductively coupled plasma enhanced CVD for flexible display application,”Electrochemical and Solid-State Letters, vol. 13, no. 8, pp. 89–91, 2010.

[145] A. Kshirsagar, S. Duttagupta, and S. Gangal, “Optimisation and fabrica-tion of low-stress, low-temperature silicon oxide cantilevers,” Micro and NanoLetters, vol. 6, no. 7, p. 476481, 2011.

[146] P. S. Waggoner, C. P.Tan, and H. G. Craighead, “Atomic layer depositedsilicon dioxide films on nanomechanical silicon nitride resonators,” Journal ofApplied Physics, vol. 107, no. 11, p. 114505, 2010.

[147] B. M. Onat, W. Huang, N. Masauna, M. Langea, M. H. Ettenberga, andC. Driesa, “Ultra low dark current InGaAs technology for focal plane arraysfor low-light level visible-shortwave infrared imaging,” in Proceedings of SPIE,vol. 6542, 2007, pp. 1–9.

[148] A. D. Hood, M. H. MacDougal, J. Manzo, D. Follman, and J. C. Geske,“Large-format InGaAs focal plane arrays for SWIR imaging,” in Proceedingsof SPIE, vol. 8353. SPIE, 2012, pp. 83 530A–83 537A.

[149] J. B. B. Theodore and R. Hoelter, “Extended short wavelength spectral re-sponse from InGaAs focal plane array,” in Proc. of SPIE, vol. 5074, 2003, pp.481–490.

[150] T. J. Martin, M. J. Cohen, J. C. Dries, and M. J. Lange, “InGaAs/InP focalplane arrays for visible light imaging,” in Proceedings of SPIE, vol. 5406,2004, pp. 38–45.

[151] E. N. Lewis, P. J. Treado, R. C. Reedor, C. Marcott, I. W. Levint, and G. M.Stay, “Fourier transform spectroscopic imaging using an infrared focal-planearray detector,” Analytical Chemistry, vol. 67, pp. 3377–3381, 1995.

[152] P. J. Treado, E. N. Lewis, and I. W. Levint, “Near-infrared acousto-optic fil-tered spectroscopic microscopy: A solid-state approach to chemical imaging,”Applied Spectroscopy, vol. 46, pp. 553–559, 1992.

[153] H. Ko, V. W. Smail, and D. L. Wetzel, “Reliability of InGaAs focal plane arrayimaging of wheat germination at early stages,” Journal of Cereal Science,vol. 48, pp. 394–400, 2008.

[154] R. C. Lyon, D. S. Lester, E. N. Lewis, E. Lee, L. X. Yu, and E. Jeffer-son, “Near-infrared spectral imaging for quality assurance of pharmaceuticalproducts: Analysis of tablets to assess powder blend homogeneity,” AAPSPharmSciTech, vol. 3, pp. 1–15, 2002.

[155] P. J. Treado, I. W. Levin, and E. N. Lewis, “Indium antimonide (InSb) focalplane array (FPA) detection for near-infrared imaging microscopy,” AppliedSpectroscopy, vol. 18, pp. 607–615, 1994.

183

BIBLIOGRAPHY

[156] R. A. Crocombe, “Miniature optical spectrometers: There’s plenty of roomat the bottom,” Spectroscopy, vol. 23, no. 38-56, 2008.

[157] D. Rossberg, “Silicon mircomachined ifrared sensors with tuable wavelengthslectivity for application in ifrared spectroscopy,” Sensors and Actuators, vol.46-47, no. 1995, pp. 413–416, 1995.

[158] F. Felder, M. Arnold, M. Rahim, C. Ebneter, and H. Zogg, “Tunable lead-chalcogenide on si resonant cavity enhanced midinfrared detector,” AppliedPhysics Letters, vol. 91, no. 10, p. 101102, 2007.

[159] N. Tas, T. Sonnenberg, H. Jansen, R. Legtenberg, and M. Elwenspoek,“Stiction in surface micromachining,” Journal of Micromechanics andMicroengineering, vol. 6, pp. 385–397, 1996.

[160] R. Rafiei, R. W. Basedow, K. K. M. B. D. Silva, J. T. Gurusamy, J. R. S.Castillo, D. K. Tripathi, J. M. Dell, and L. Faraone, “A versatile instrumen-tation system for MEMS-based device optical characterization,” Proceedingsof the SPIE, vol. 8923, pp. 8923–8932, 2013.

[161] H.-C. Tsai and W. Fang, “Determination of the poission ratio of thin filmsmaterial the resonant method,” Sensors and Actuators A, vol. 103, pp. 377–383, 2003.

[162] “CoventorWare,” http://www.coventor.com/mems-solutions/products/coventorware/, 2010, [Online; accessed 09-Jan-2015].

[163] J. S. Milne, J. M. Dell, A. J. Keating, and L. Faraone, “Extended tuning rangeFabry-Perot etalon with doubly-supported beam actuators,” in IEEE/LEOSInternational Conference on Optical MEMS, 2006.

[164] H. Mao, K. K. M. B. D. Silva, M. Martyniuk, J. Antoszewski, J. Bumgarner,and L. Faraone, “Towards longwave infrared tuneable filters for multispec-tral thermal imaging applications,” in Quantitative Infrared Thermography(QIRT) conference 2014, June 2014, pp. 1–6.

184

http://www.coventor.com/mems-solutions/products/coventorware/

http://www.coventor.com/mems-solutions/products/coventorware/

silicon-based optical mems for visible to mid-wave...

Documents