stanford photonics research centerstanford.edu/group/khuri-yakub/publications/02_cheng_01.pdf ·...

Stanford Photonics Research

Center

Stanford Photonics Research Center

now incorporating CNOM (Center for Novel Opto-electronic Materials)

Annual Report

2003

Stanford University

http://www.stanford.edu/group/SPRC

© 2003 Stanford University

http://www.stanford.edu/group/SPRC

Table of Contents

SPRC 2002 – 2003 Annual Report i

Table of Contents ....................................................................................................................................i

I. Executive Summary ..................................................................................................................1

II. Stanford Photonics Research Center - 2003 ...........................................................................5

III. SPRC Membership..................................................................................................................11

IV. SPRC/CNOM Characterization Facility...............................................................................13

V. Research Program:

A. Fundamental Science....................................................................................................................A-1

D.B. Jackrel, and J.S. Harris, “InGaAs and GaInNAs Photodiodes for Advanced LIGO (Laser Interferometer Gravitational Wave Observatory)”....................................................................A-3

Edo Waks, Eleni Diamanti and Yoshihisa Yamamoto, “Generation of photon number states” ...................................................................................................................................................A-6

H.B. Yuen, S.R. Bank, M.A. Wistey, V. Gambin, W. Ha, J.S. Harris Jr., A. Moto “Growth and characterization of MBE-grown GaNAsSb and GaInNAsSb for long-wavelength optoelectronic devices” ...................................................................................................A-10

B. Biophotonics ..................................................................................................................................B-1

G. Schuele, P. Huie, R. Brinkmann, R. Birngruber, A. Vankov, E. Vitkin, H. Fang, E. B. Hanlon, L. T. Perelman, D. Palanker, “Noninvasive monitoring of temperature during retinal laser treatments”........................................................................................................................B-3

E. Thrush, O. Levi, W. Ha, G. Carey, L. Cook, J. Deich, S.J. Smith, W.E. Moerner and J.S. Harris Jr., “Monolithically integrated fluorescence sensor”..........................................................B-6

K. Wang, N. Mehenti, H. Dai, H.A. Fishman, J.S. Harris, “Carbon Nanotubes as Microelectrodes for a Retinal Prosthesis” ............................................................................................B-9

X.J. Zhang, S. Zappe, C-C. Chen, O. Sahin, J. Harris, C. Quate, M. Scott and O. Solgaard, “Silicon microsurgery-force sensor based on diffractive optical MEMS encoders”............................................................................................................................................B-12

C. Nanophotonics ..............................................................................................................................C-1

D.P. Fromm, A. Sundaramurthy, K.B. Crozier, P.J. Schuck, G.S. Kino, C.F.Quate, W.E. Moerner, “Development of Electromagnetically Enhanced Au “Bowtie” Nanostuctures” .................C-3

Table of Contents

SPRC 2002 – 2003 Annual Report ii

J. Hwang, M. M. Fejer and W. E. Moerner, “Exploring Novel Methods of Interferometric Detection of Ultrasmall Phase Shifts”..................................................................................................C-6

Yang Jiao, Shanhui Fan, and David A. B. Miller, “Designing for beam propagation in periodic and nonperiodic photonic nanostructures: extended Hamiltonian method”...........................C-9

J. Matteo, D.Fromm, Y. Yuen, P.J. Shuck and L. Hesselink, “Spectral Analysis of Enhanced Transmission thru Single Nano-Apertures”.......................................................................C-12

Zheng Wang and Shanhui Fan, “Compact all-pass filters in photonic crystals as the building block for high capacity optical delay lines” .........................................................................C-15

Mehmet Fatih Yanik, Shanhui Fan, Marin Soljačić, J. D. Joannopoulos, “High-density Low-power All-optical Logic Gate in Photonic Crystals” .................................................................C-18

D. Telecom..........................................................................................................................................D-1

S. R. Bank, M. A. Wistey, H. B. Yuen, W. Ha, V. Gambin, K. Volz, and J. S. Harris, “1.5 Micron Lasers on Gallium Arsenide”...........................................................................................D-3

Hilmi Volkan Demir, Vijit A. Sabnis, Jun-Fei Zheng, Onur Fidaner, James S. Harris, Jr., and David A. B. Miller, “Novel scalable wavelength-converting crossbars” ......................................D-6

In-Sung Joe, Kyoungsik Yu, and Olav Solgaard, “Flexible, scalable 100Tbps Internet router with an optical switch fabric using optical tunable filters”........................................................D-9

A. Khalili, James S. Harris, “Side-Coupled Waveguide/Fiber Devices” ...........................................D-12

Jonathan R. Kurz, Jie Huang, Xiuping Xie, Martin M. Fejer, “Mode Multiplexing in Optical Frequency Mixers” ................................................................................................................D-16

Carsten Langrock, Rostislav V. Roussev, Jonathan R. Kurz, Martin M. Fejer, “Sum-frequency generation in a PPLN waveguide for efficient single-photon detection at communication wavelengths” ............................................................................................................D-19

Vijit A. Sabnis, Hilmi Volkan Demir, Jun-Fei Zheng, Onur Fidaner, James S. Harris, Jr., and David A. B. Miller, “Novel Optically-switched Electroabsorption Modulators for Wavelength Conversion” ...................................................................................................................D-22

R. Urata, L. Y. Nathawad, K. Ma, D. A. B. Miller, B. A. Wooley, and J. S. Harris, Jr., “Photonic A/D Conversion Using Low-Temperature-Grown GaAs MSM Switches Integrated with Si-CMOS”.................................................................................................................D-25

Table of Contents

SPRC 2002 – 2003 Annual Report iii

E. Optical Interconnects ................................................................................................................... E-1

Aparna Bhatnagar, Christof Debaes, Ray Chen, Noah Helman, Gordon Keeler, Salman Latif and David A. B. Miller, “Receiver-less Optical Clock Distribution using Short Pulses”................................................................................................................................................. E-3

C.H. Cheng, A.S. Ergun, and B.T. Khuri-Yakub, “Electrical through silicon wafer interconnects for high frequency photodetector arrays”....................................................................... E-7

Kai Ma, Ryohei Urata, David A. B. Miller and James S. Harris, Jr., “Low Temperature Growth of GaAs on Si Substrates for Ultra-fast Photoconductive Switches Used in a Hybrid CMOS/Photonic A/D Conversion System” ........................................................................... E-10

David Press, Rafael Aldaz, Michael Wiemer, James S. Harris and David A.B. Miller, “Mode-locking simulations for monolithically integrated Vertical Cavity Surface Emitting Laser” .................................................................................................................................. E-13

F. Optical MEMS / Optical Microsystems ...................................................................................... F-1

Ruslan Belikov, Xiang Li, Christophe Antoine-Snowden, Olav Solgaard, “Programmable Optical Wavelength Filters Based on Diffraction from MEMS Micromirror Arrays” ........................ F-3

U. Krishnamoorthy, I.W. Jung, P. Lu, Y.-A. Peter, E. Carr, R. L. Byer, O. Solgaard, “Development of segmented deformable mirrors for free space communications”............................. F-7

D. Lee, U. Krishnamoorthy, H. Ra, O. Solgaard , “Single-crystalline silicon micromirrors actuated by self-aligned vertical electrostatic combdrives with piston-motion and rotation capability” .......................................................................................................................................... F-10

Wonjoo Suh and Shanhui Fan , “Flat-top Reflection and All-pass Transmission Filter using Coupled Resonance in Photonic Crystal Slabs” ....................................................................... F-13

J. Wang, I. Jung, O. Solgaard, “Elastomer Spatial Light Modulators for Extreme Ultraviolet Lithography” .................................................................................................................... F-16

Thomas D. Wang, Michael J. Mandella, Ning Y. Chan, Christopher H. Contag, Gordon S. Kino, “MEMS Confocal Microscope with Dual Axes Architecture for In Vivo Molecular and Cellular Imaging”....................................................................................................... F-19

Kyoungsik Yu, Daesung Lee, Uma Krishnamoorthy, and Olav Solgaard, “Tunable bandwidth optical filter based on MEMS Gires-Tournois interferometer”........................................ F-23

Table of Contents

SPRC 2002 – 2003 Annual Report iv

G. Ultrafast and Nonlinear Optics ...................................................................................................G-1

Mathieu Charbonneau-Lefort, Bedros Afeyan and Martin M. Fejer, “Optical Parametric Amplification Using Chirped Quasi-Phase-Matching Gratings” .........................................................G-3

Rostislav Roussev, Arun Sridharan, Karel Urbanek, Robert Byer and Martin Fejer, “Parametric Amplification of 1.6 µm Signal in Annealed- and Reverse- Proton Exchanged Waveguide” .......................................................................................................................G-5

L. Scaccabarozzi, Z. Wang, X. Yu, S. Fan, M. M. Fejer, J. S. Harris, “Photonic Crystal AlGaAs Microcavities for non-linear Optical Applications” ...............................................................G-9

Andrew M. Schober, Mathieu Charbonneau-Lefort, Martin M. Fejer , “Parametric Oscillation of Ultrashort Pulses in Quasi-Phase-Matched Nonlinear Materials”...............................G-12

Xiuping Xie, Andrew M. Schober, Carsten V. Langrock, Martin M. Fejer, “Low Threshold and Near-transform-limited Pulse Generation with Cascaded Optical Parametric Generation in Reverse Proton Exchanged Lithium Niobate Waveguide” .......................G-15

H. Lasers ............................................................................................................................................H-1

A. E. Siegman, “Gain-Guided Optical Fibers for High-Power Fiber Lasers”......................................H-3

Arun Sridharan, “High Pulse Energy Yb:YAG MOPA and Non-Linear Frequency Conversion Module for Remote Sensing Applications” .....................................................................H-6

I. Optical Materials ............................................................................................................................ I-1

Junxian Fu, Seth bank, Mark Wistey, Homan Yuen, James S Harris, Jr , “Solid-Source Molecular Beam Epitaxy Growth of GaInNAsSb with Photoluminescence at 2.04µm”...................... I-3

M. Katz, R. Route, D. Hum, R. Roussev, K. Parameswaran, V. Kondilenko, G. Miller, M. Fejer, “Near-stoichiometric 1% Mg-doped LiNb03 and stoichiometric LiTa03 fabricated by vapor transport equilibration for freqeuncy conversion” ................................................................. I-5

P. S. Kuo, X. Yu, K. L. Vodopyanov, M. M. Fejer, J. S. Harris, D. Weyburne, D. Bliss, C. Yapp, K. O’Hearn, “Growth and Characterization of Thick-film Orientation-patterned GaAs” ................................................................................................................................................... I-8

Vincenzo Lordi, Homan Yuen, Seth Bank, Mark Wistey, James S. Harris, “Electroabsorption from GaInNAs and GaInNAsSb Multiple Quantum Wells”................................ I-11

Table of Contents

SPRC 2002 – 2003 Annual Report v

Ji-Won Son, Yin Yuen, Sergei S. Orlov, Bill Phillips, Ludwig Galambos, Vladimir Ya. Shur, and Lambertus Hesselink, “Direct e-beam domain engineering in LiNbO3 thin films grown by liquid phase epitaxy”.................................................................................................. I-14

Mark A. Wistey, Seth R. Bank, Homan B. Yuen, Lynford L. Goddard*, and James S. Harris“GaInNAs(Sb) Growth Improvements Leading to Record Long Wavelength Lasers”................................................................................................................................................. I-17

J. Sensing Applications ...................................................................................................................... J-1

Sameer R. Bhalotra, Helen L. Kung, Junxian Fu, Noah C. Helman, Ofer Levi, Jon Roth, Yang Jiao, Ryo Urata, David A. B. Miller, and James S. Harris, Jr., “Adaptive optical microsensors and spectra-selective imaging”........................................................................................ J-3

H. Chin, R. Chen, and D. A. B. Miller , “Linear electro-optic conversion of current pulses for a photonic-assisted electrical analog-to-digital converter”.............................................................. J-6

O. Levi, S. E. Bisson, T. J. Kulp, J. S. Harris, and M. M. Fejer, “Mid IR Cavity Ring Down spectroscopy using broadly tunable OP-GaAs light source”...................................................... J-9

Executive Summary

SPRC 2002 – 2003 Annual Report 1

I. Executive Summary

The third annual meeting of the Stanford Photonics Research Center (SPRC) will be held on September 15 – 17, 2003 at Stanford University. The meeting embraces a range of fundamental and applied sciences and engineering that fall under a broad definition of photonics. The range of activities considered at the meeting include nanoscale electronics and photonics, biophotonics, telecom, optical interconnects and optical MEMS, ultrafast optics, and advanced lasers and nonlinear materials. The three-day meeting includes contributions from invited speakers and from Stanford faculty and students.

During the past year three themes in photonics are evident as providing new opportunities for the field. First is the rapid growth in nanophotonics that includes new physical effects and devices at scales down to 10nm. The potential for novel telecommunication devices at the nanometer scale is being explored as is novel photonic materials with properties engineered to accomplish specific tasks related to the emission and control of light.

The research in photonic sensing especially as it relates to sensing in biology is also recognized as a new direction with significant potential. Noninvasive monitoring of biosystems down to the cellular level is now possible. The potential for biophotonics to provide opportunities in the commercial world is also appreciated. This year that potential will be explored by a BioPhotonics Industry Panel moderated by Professor Dennis Matthews, the director of the UC Davis Center for Biophotonics.

The third area of note is the rapid progress in ultrafast optics and in the fundamentally new approaches to the generation of ultrafast pulses of light as short as one optical cycle. Further, using a self-referencing frequency comb from modelocked lasers, with phase and amplitude control, it is now possible to synthesize optical frequencies across the visible and near infrared spectral regions. All-optical clocks are rapidly becoming a reality.

Are there developing trends in photonics that are visible over the horizon? There are certainly areas of research that hold promise in the long term. For example, the control of light in atomic media with the slowing of pulses of light, storage and then release represents a promising avenue of exploration. The integration of optical devices on silicon substrates represents a merging of electronics and optics. The development of optical guiding, filtering, modal conversion and control in photonic bandgap materials is an equally challenging and important area for research. For longer range vision, the exploration of the fundamentals of quantum mechanics with light, quantum communication, and perhaps quantum computing offer a new approach to information control and processing.

The growth of photonics and Stanford, in the broadest sense of the term, encompasses basic physics through device engineering. The field extends, in a strongly interdisciplinary way, beyond the boundaries of a single department or even a single school. The SPRC Annual Meeting is a reflection of the breadth and depth of photonics at Stanford. Further, the Annual Meeting provides the opportunity for faculty, students, invited speakers and participants to share and to contribute to the knowledge at the leading edge of photonics.

Executive Summary


The Third Annual SPRC Annual Meeting

The SPRC annual meeting is organized to facilitate interactions and communication among the attendees from industry, institutes and universities. As in the past, the meeting is organized to include presentations and discussions that range from fundamental research to device engineering and technology transfer to the commercial stream. The meeting highlights recent breakthroughs, advances in applications and technologies and ideas that are at the interface between those who conduct research and those who work to apply new approaches in the commercial world.

The rapid growth of photonics is reflected in the appointment of new faculty at Stanford and in the creation of new interdisciplinary research programs that bridge across well established disciplines. In turn, the recognition of the long range impact of photonics has attracted students into the field and has led to new courses and new approaches to learning. The faculty and students at Stanford benefit greatly from the interaction and support of the members of SPRC. The ongoing support has enabled us to create new courses and upgrade existing courses to meet the student demand. As is customary at the SPRC meeting, the more senior students in photonics will present talks as well as posters. The full range of the research in photonics is presented best by the students in the poster sessions. We urge you to take advantage of the poster sessions to meet with students and to learn firsthand about their projects and their progress toward an advanced degree.

The concept of a corporate affiliates meeting, with the goal of facilitating access to ideas and to research and students, was initiated at Stanford nearly fifty years ago. The success of the SPRC Annual Meeting is built on past experience and on the support of member companies. Edward W. (Ned) Barnholt, the CEO, President and Chair of the Board of Agilent Technologies, will present the opening Keynote address. Agilent is a founding corporate member of SPRC and a key company in photonics and opto-electronics technologies. The industry perspective of photonics is particularly timely as we head toward a growing economy with the potential for new opportunities and new markets.

We extend the theme of photonics in industry with a panel discussion on Biophotonics. Dennis Matthews, Director of the Center for Biophotonics, at UC Davis, will moderate the panel discussion. The panel discussion will be followed by a reception and dinner at the Stanford faculty club.

A tradition at the SPRC meetings is to have an opportunity to learn about other areas of science and research that extend beyond photonics. This year we are very fortunate to have Dr. Celine d’Orgeville, chief laser scientist of the Gemini Observatory, present the after dinner talk. The topic is the Adaptive Optics in modern astronomy and the role played by natural and laser guide stars. Modern astronomy has benefited greatly by the replacement of film by CCD detectors over the past twenty years. Equally important benefits lie in the future with the implementation of adaptive optics on large telescopes and arrays of telescopes.

The Annual meeting opens with a session on Fundamental Science followed by a session on Biophotonics. The first day’s sessions include a plenary talk followed by invited talks. We are indeed fortunate that speakers from locations across the country, and from Europe and Japan are contributing

Executive Summary


to the program. We have organized the meeting to feature new faculty members and students in each session. Following sessions on Nanophotonics, Telecom, Optical Interconnects and MEMS, we will introduce graduate students and the poster session with one-minute brief descriptions of their poster and research topic. The Poster Session on Tuesday afternoon is followed by dinner on your own – or, you may decide to continue your inquiry into research at Stanford by inviting a graduate student, or perhaps even two, to dinner!

The meeting continues on Wednesday with sessions on Ultrafast and Nonlinear Optics, Lasers, and Optical Materials. These are the more traditional areas of photonics but have continued to see new and often surprising developments. For late breaking developments, we have arranged for a post-deadline session on Wednesday afternoon. Please see one of the SPRC co-directors if you have a post-deadline contribution.

Following three days of discussion, contribution and meetings we invite you to an informal reception and BBQ in the Fairchild green late Wednesday afternoon. This is an opportunity to continue discussions with fellow attendees and to expand on issues and topics in a relaxed setting.

The SPRC Center continues to grow to reflect the growth of the field. The Center is fortunate to have the support of Associate Director, Nancy Christiansen and her staff. If you have a request for information regarding membership in SPRC, or if you have suggestions for future topics or for a special one-day symposium, please let us know.

Welcome to the third annual Affiliates meeting of the Stanford Photonics Research Center.

Robert L. Byer Martin M. Fejer David A. B. Miller

Co-Directors

Executive Summary


2003 Stanford Faculty, Students, Staff, Visitors, Member Companies


II. Stanford Photonics Research Center – 2003

Photonics Core Faculty and Senior Researchers

SPRC Co-Directors: Robert L. Byer Professor of Applied Physics www.stanford.edu/~rlbyer/ Solid state lasers, adaptive optics, nonlinear optics Martin M. Fejer Professor of Applied Physics www.stanford.edu/group/ginzton/faculty/fejer.html Nonlinear optics, nonlinear optical materials, applications in telecommunications David A.B. Miller Professor of Electrical Engineering, and by courtesy, of Applied Physics http://ee-www.stanford.edu/~dabm/ Optoelectronic physics, devices, and systems, optical switching and interconnects, optical sensing Photonics Faculty Mark L. Brongersma Assistant Professor of Materials Science and Engineering http://mse.stanford.edu/people/faculty/brongersma/brongersma.html Photonic nanoparticles and nanostructures Michel Digonnet Senior Research Engineer Fiber optics Abbas El Gamal Professor of Electrical Engineering http://www-isl.stanford.edu/people/abbas/ Digital imaging and wireless networks Shanhui Fan Assistant Professor of Electrical Engineering www.stanford.edu/group/fan Photonic crystals, computational electromagnetics Robert Feigelson Professor (Research) of Materials Science and Engineering, Emeritus http://mse.stanford.edu/faculty/feigelson.html Nonlinear optical materials

James S. Harris Professor of Engineering, Materials Science and, by courtesy, Applied Physics http://ee-www.stanford.edu/~harris/ Semiconductor optoelectronic materials, devices and applications Steve Harris Professor in the School of Engineering and Applied Physics http://www-ee.stanford.edu/~seharris/ Fundamentals of photonics and nonlinear optics Lambertus Hesselink Professor of Electrical Engineering and, by courtesy, of Applied Physics http://kaos.stanford.edu Nanophotonics and ultra-high density optical data storage Joseph M. Kahn (new faculty) Professor of Electrical Engineering - STAR Laboratory http://www-ee.stanford.edu/~jmk/ Optical fiber communications, free-space optical communications, associated devices and subsystems Mark Kasevich Professor of Physics and Applied Physics http://www.stanford.edu/dept/physics/people/ faculty/kasevich_mark.html High accuracy navigation and gravimetric sensors based on de Broglie wave interferometry; Future atom optics sensors which exploit the novel coherence properties of Bose-Einstin condensates Leonid Kazovsky Professor of Electrical Engineering http://ocrl.stanford.edu/ Optical telecommunications and network systems Thomas W. Kenny Associate Professor of Mechanical Engineering http://me.stanford.edu/faculty/facultydir/kenny.html Microsensors based on silicon micromachining B. (Pierre) T. Khuri-Yakub Professor (Research), http://piezo.stanford.edu/pierre/ acoustic sensors (temperature, film thickness, resist cure, ...), acoustic materials and devices

http://ee-www.stanford.edu/~dabm/

http://mse.stanford.edu/people/faculty/brongersma/

http://www-isl.stanford.edu/people/abbas/

http://mse.stanford.edu/faculty/feigelson.html

http://ee-www.stanford.edu/~harris/

http://www-ee.stanford.edu/~seharris/

http://kaos.stanford.edu

http://www-ee.stanford.edu/~jmk/

http://www.stanford.edu/dept/physics/people/faculty/kasevich_mark.html

http://ocrl.stanford.edu/

http://me.stanford.edu/faculty/facultydir/kenny.html

http://piezo.stanford.edu/pierre/

www.stanford.edu/~rlbyer/

http://www.stanford.edu/group/fejer/Faculty.htm

www.stanford.edu/group/fan



Gordon Kino Professor of Electrical Engineering, and, by courtesy, of Applied Physics, Emeritus www.stanford.edu/group/ginzton/faculty/kino.html Optical Fiber Sensors Michael McGehee Assistant Professor of Materials Science and Engineering http://mse.stanford.edu/people/faculty/mcgehee/ mc gehee.html Polymer optical materials and devices W. E. Moerner Professor of Chemistry http://www.stanford.edu/dept/chemistry/faculty/ moerner/ Photorefractive polymers Daniel Palanker Assistant Professor (Research) of Ophthalmology http://www.med.stanford.edu/school/eye/faculty/ biopalan.html Biomedical optics and electronics Calvin F. Quate Professor (Research) of Electrical Engineering Professor (Research) of Applied Physics (by courtesy) http://www.stanford.edu/group/quate_group/ Imaging and lithography applications of scanning probes. Krishna Saraswat Professor of Electrical Engineering http://nucleus.stanford.edu/ innovative materials, device structures, and process technology of silicon devices and integrated circuits Mark Schnitzer (new faculty) Assistant Professor of Applied Physics and Biological Sciences Biophotonics

John Shaw Professor (Research) of Applied Physics and Ginzton Laboratory, Emeritus www.stanford.edu/group/ginzton/faculty/shaw.html Optical Fiber Sensors Anthony E Siegman McMurtry Professor of Engineering Emeritus http://www-ee.stanford.edu/~siegman/ Research and consulting in lasers, optics and fiber optics, including technical and litigation consulting Olav Solgaard Assistant Professor of Electrical Engineering www.stanford.edu/~solgaard/ Optical micromechanical devices and applications Jelena Vuckovic (new faculty) Assistant Professor of Electrical Engineering http://ee.stanford.edu/~jela Photonic crystal-based optical and quantum optical devices and their integration; solid-state photonic quantum information systems; cavity quantum electrodynamics with quantum dots Brian A. Wandell Professor of Psychology, and by courtesy, of Electrical Engineering http://white.stanford.edu/wandell.html Image system engineering and visual neuroscience Yoshihisa Yamamoto Professor of Electrical Engineering and Applied Physics http://feynman.stanford.edu/people/yamamoto/ yamamoto.html Fundamental optoelectronic physics, structures, and devices, quantum computing

Adela Ben-Yakar Kenneth Crozier Moti Katz Uma Krishnamoorthy Ofer Levi Krishnan Parameswaran Yves-Alain Peter Tomas Plettner Roger Route Thomas Wang Stefan Frank Zappe Xiaolei Shi (GE)

Ulrich Wittrock (Univ. of Applied Sciences Münster) Jun-Fei Zheng (Intel)

SPRC Staff Nancy Christiansen Imelda White Sara Fahmy Special appreciation to: Gail Chun Creech Patricia Detton Vivian Drew Vera Haugh Eileen Varner

Photonics Core Faculty and Senior Researchers (Continued)

Research Staff, Post Doctoral and Visiting Scholars/Scientists

http://mse.stanford.edu/people/faculty/mcgehee/

http://www.stanford.edu/dept/chemistry/faculty/moerner/

http://www.med.stanford.edu/school/eye/faculty/biopalan.html

http://www.stanford.edu/group/quate_group/

http://ee.stanford.edu/~jela

http://white.stanford.edu/wandell.html

http://feynman.stanford.edu/people/yamamoto/yamamoto.html

http://www-ee.stanford.edu/~siegman/

www.stanford.edu/group/ginzton/faculty/shaw.html

www.stanford.edu/~solgaard/

http://nucleus.stanford.edu/

www.stanford.edu/group/ginzton/faculty/kino.html



Photonics Students and Recent Graduates

Rafael Aldaz Hatice Altug Christophe Antoine-Snowden Vlatko Balic Seth Bank Ruslan Belikov Sameer Bhalotra Aparna Bhatnagar Ueyn L. Block Danielle Braje Amber Bullington Emily Carr Peter Catrysse Mathieu Charbonneau-Lefort Ching-Hsiang Cheng Henry Chin Hoyeol Cho Sookyung Choi Alex Chow Chi On Chui Keith R. Cohn Laura Cook Hilmi Volkan Demir Hui Deng Eleni Diamanti Ali Ozer Ercan Onur Fidaner Dave Fromm Junxian Fu Vincent Gambin Martina Gerken Lynford Goddard Wonill Ha Noah Helman Yu-Li Hsueh Jie Huang David Hum Jaesuk Hwang David Jackrel Hemanth Jagannathan Xin Jiang Yang Jiao

In-Sung Joe Kerstin Johnsson David Julian Il Woong Jung Mark Kalman Georgios Kalogerakis Lance Kam Bianca Keeler Gordon Keeler Alireza Khalili Onur Kilic Sora Kim Sungchul Kim Seongsin Kim Doug King Helen Kung Paulina Kuo Jonathan Kurz Carsten Langrock Salman Latif Wah Tung Lau Daesung Lee Meredith Lee Daniel Levner Vincenzo Lordi Patrick P. Lu Kai Ma Kenneth Mai George Marcus Joseph Matteo Ammar Munir Nayfeh Caitlin O'Connell-Rodwell Ali Kemal Okyay Frank O'Mahony Aydogan Ozcan Jun Ren Jon Roth Rostislav Roussev Vijit A. Sabnis Khaled Salama Charles Santori

Shally Saraf Luigi Scaccabarozzi Johann M. Schleier-Smith Andrew Schober Georg Schuele Barden Shimbo Hocheol Shin Kapil Shrikhande Miroslav Shverdin Supriyo Sinha Ji-Won Son Arun Sridharan Wonjoo Suh Liying Sun Yuzuru Takashima Liang Tang Anuranjita Tewary Evan Thrush Ryohei Urata Edo Waks David R. Walker Ke Wang Jen-Shiang Wang Zheng Wang Jin Wang Michael Wiemer Katherine Willets Jeffrey Wisdom Mark Wistey Samuel Wong Kenneth Wong Xiuping Xie Hyunsoo Yang Mehmet Fatih Yanik Xiaobo Yin Kyoungsik Yu Xiaojun Yu Homan Yuen Yin Yuen John XJ Zhang Rashid Zia



SPRC Member Companies – 2002-2003

Founding Member Agilent Technologies, Inc. 3500 Deer Creek Road, Bldg 4A Palo Alto, CA 94304 http://www.agilent.com

Senior Members

GE Corporation Research and Development 1 Research Center Building K1 Niskayuna, NY 12309 http://www.ge.com

Texas Instruments 13560 North Central, MS 3735 Dallas, TX 75234 http://www.ti.com

Agilent Technologies is on the leading edge of nearly every major trend in communications and life sciences. From optical and wireless communications to disease and discovery research, Agilent delivers product and technology innovations that benefit millions of people around the world. Leading companies -- communications equipment manufacturers, Internet service providers, biopharmaceutical companies and more -- depend on Agilent's more than 20,000 test, measurement and monitoring devices, semiconductor products and chemical analysis tools to help drive the communications and life sciences revolutions that transform the modern world.

With more than 1,600 people in our Niskayuna facilities, virtually every scientific and technical discipline is represented. We're chemists, physicists, electrical and electronics engineers, microbiologists, metallurgists, information technologists and everything in between. People of different disciplines working on teams together to develop technologies in areas like electronic systems, manufacturing and business processes, physical metallurgy, and, polymer materials.

Texas Instruments Incorporated (TI) [NYSE: TXN] is a global semiconductor company and the world's leading designer and supplier of real-time signal processing solutions. The company's businesses also include sensors and controls, and educational and productivity solutions. Headquartered in Dallas, Texas, TI has more than 38,500 employees worldwide with corporate, sales and manufacturing facilities in more than 25 countries across Asia, Europe and the Americas. TI is the established market leader in DSP and analog technologies, the engines of the Internet age. TI's programmable digital signal processors (DSPs) and high-performance analog chips are fueling many innovative, high-growth applications, and TI's solutions are well suited for such fast-growing markets as digital wireless handsets, broadband access to and through the home and office, Internet audio players and other digital audio devices, high-resolution imaging and digital motor control.

http://www.agilent.com

http://www.ge.com

http://www.ti.com



SPRC Regular Members

Innovation Core SEI, Inc. 3235 Kifer Road, Suite 150

Santa Clara, CA 95051 www.sumitomo.com/sei_usa.htm

NTT Corporation

3-1, Morinosato-Wakamiya, Atsugi-shi, Kanagawa 243-0198 Japan www.ntt.co.jp/index_e.html

Northrop Grumman Corporation

600 Hicks Road MS L5200

Rolling Meadows, IL 60008-1098 www.northgrum.com/

Ricoh Company, Ltd.

16-1 Shinei-cho, Tsuzuki-ku Yokohama, 224-0035 Japan

www.ricoh.com

Sony Corporation

6-7-35 Kitashinagawa Shinagawa-ku, Tokyo 141-0001 Japan

www.sony.com

Suruga Seiki Co., Ltd. 549-1 Nanatsushinya, Shimizu-shi

Shizuoka 424-8566 Japan www.suruga-g.co.jp/

We would like to acknowledge Tera Beam for assistance to CNOM/SPRC this year:

SPRC Membership


III. SPRC Membership

The Stanford Photonics Research Center (SPRC) connects the photonics research and teaching at Stanford University with the photonics industry, locally, nationally, and internationally.

Membership in the Stanford Photonics Research Center (SPRC) is open to corporations interested in interacting with Stanford's photonics students, faculty, research and teaching. The Center provides a way for companies to be actively involved in Stanford's photonics activities. SPRC's goal is to support photonics for the mutual, sustained benefit of both Stanford's teaching and research and SPRC's corporate members. SPRC membership fees directly support research and teaching in photonics at Stanford. SPRC aims to connect Members to the fullest possible range Stanford photonics faculty, students, and projects at Stanford. Such a breadth of connection offers Members a broad perspective on current and emerging areas in photonics.

There are three levels of membership in SPRC:

Founding Members

Founding Members are involved in setting the strategic directions of SPRC, advise in general on technical directions, and participate on the SPRC Advisory Board. Founding Membership involves a multi-year commitment in membership dues at the Senior Member level (see below), together with a capital donation of $2M. Founding Members share all benefits of other levels of membership.

Senior Members

Senior Members participate on the SPRC Advisory Board. Senior members may send a visiting researcher to work for up to a full year at Stanford with a collaborating Stanford faculty member. Senior Members may participate as mentors in the Fellow/Mentor/Advisor (FMA) program and support Photonics research groups in accordance with SPRC Advisory Board and University policy. The fee for Senior Membership for AY 2004 (Sep 2003-Aug 2004) is $150,000.

Members

In addition to the Member benefits listed below for all Member companies, regular Membership offers visiting researchers from companies the opportunity to participate in research at Stanford, including research visits up to one month each year with a collaborating Stanford faculty member and research group. The annual fee for regular Membership is $50,000.

SPRC Membership


Benefits for all Members Benefits for all Members include • Access to Stanford’s photonics research activities and results, through review meetings, seminar

series, and other activities

• Access to students for recruiting

• Information on photonics seminars, recruiting, and other photonics events at Stanford

• Prompt alerting for Stanford invention disclosures in photonics and facilitated access to Stanford’s photonics intellectual property portfolio

• All Members are encouraged to have active involvement with Stanford’s photonics activities, including research collaborations, both formally and informally

• Members of the former Center for Novel Optoelectronic Materials (CNOM) are automatically now Members of SPRC at the appropriate level

SPRC Events With the growing interest in photonics, both within Stanford and in the external industrial community, the very successful Center for Novel Optoelectronic Materials (CNOM) broadened into the SPRC program. This program involves as broad a range of photonics faculty, students and research as possible, and aims to be a single point of contact for Stanford’s photonics work. In support of the Member benefits listed above, SPRC has initiated a program of activities for Member companies, including • The Annual Meeting in September, a major summary of current activities in photonics, and a

review of current Stanford photonics research

• Focused Workshops for our member companies (Fall and Spring)

• A student research review and Member company recruiting opportunity in February/March

• A database of student resumés for students near graduation, accessible to Member companies

• Ongoing enhancements to the Web site for Member companies, including student information, and electronic copies of research reports and theses.

Suggestions for other SPRC activities are welcome. Inquiries may be directed to Nancy Christiansen, Associate Director, email [email protected], phone (650) 723-4406, fax (650) 725-1822.


IV. SPRC / CNOM Characterization Facility

With prior DARPA/URI support, the CNOM Optical Materials Characterization Facility was established in 1992. This facility contains a variety of coherent sources and characterization tools to make possible the rapid measurement of the properties of optical materials and devices. The Characterization Facility is operated with funds derived from the Affiliates' Program, members of which have access to the Characterization Facilities for support of research on optical materials and devices.

The measurement capability of the CNOM Optical Materials Characterization Facility is summarized below. Contact Dr. Roger Route through the CNOM office/web page for detailed information about the characterization equipment and for access to the Facility.

CNOM Characterization Capabilities: 1) spatially- and temporally resolved spectroscopic absorption measurement,

2) Photoconductivity and photovoltaic currents at high intensities

3) scatter loss at 633 nm and 1064 nm (TMA Inst.),

4) waveguide refractive index profiles (Metricon),

5) variable angle spectroscopic reflectivity and ellipsometric measurements of thin films, waveguides and multilayers (SOPRA GESP) in the 250 - 1750 nm waveband

6) spectrophotometeric measurements with (Hitachi U4001) UV-VIS-NIR grating spectrophotometer and (Bio Rad) mid-IR and far-IR Fourier transform spectrophotometers

7) photorefractive gain, diffraction efficiency, and response rates.

CNOM Coherent Sources: 1) Spectra-Physics MOPO 730 Nd:YAG /BBO OPO system, ns pulses 1.84 – 0.21 µm

2) Coherent Mira femtosecond Ti:Sapphire laser system

3) Positive Light Spitfire Ti:S regenerative amplifier, with SHG and THG

4) Spectra-Physics Tsunami femto-second Ti-sapphire laser system,

5) Spectra-Physics OPAL femto-second OPO (1.3 - 2 µm),

6) Coherent Sabre Argon lasers

UV and Ultrafast Materials Characterization

The CNOM Optical Materials Characterization Facility has acquired an ultra-fast and UV materials characterization capability with a tunable Coherent Mira/Sabre Ti:Sapphire laser pumping a Positive Light Spitfire/Merlin regenerative amplifier with a frequency doubler/tripler option. Single Guassian mode pulses, either <130 fs or ~1 ps in duration, at a 1 KHz rep. rate are available from 950 to 233 nm, and sum-frequency generation is possible to generate wavelengths shorter than 200 nm. Stretched, flat-top pulses are also available from the harmonic package through the use of a longer set of doubling and tripling crystals. The high spatial and temporal quality output beam was used most recently to study UV degradation in nonlinear optical materials such as BBO.


Another addition to our ultra-fast capabilities was a Spectra-Physics OPAL-Tsunami system. Tsunami is a femtosecond actively mode-locked tunable Ti:sapphire laser producing <130 fs pulses at repetition rate of 80 MHz with average power > 1.5 W. OPAL is a synch-pumped OPO generating <130 fs pulses in a wavelength range covering 1.3 - 2 µm, with average power > 150 mW.

Spectroscopic Measurement of Absorption Loss

The spectroscopic absorption loss apparatus is based on a technique known as common-path photothermal interferometry, which uses the thermo-optic refractive index changes induced by absorbed optical power to monitor the absorptivity.. A pump beam at the wavelegnth at which the absorption is to be measured is focused coaxially with, and with a smaller waist than, a low-power probe beam. A phase shift is imposed on the central portion of the probe beam by the photothermal index change induced by the local temperature rise resulting from the absorbed pump power. A key step in the operation of the device is the use of a Fourier transforming lens to convert the localized phase shift into an intensity variation. This common path approach is much more robust than conventional methods based on Mach-Zehnder interferometry, allowing near shot-noise limited measurements of the induced phase in a simple, tabletop system. Sensitivities to absorption in the range of 10-6 cm-1 have been demonstrated. A variety of pump lasers have been used, including 1.06 µm and 532 nm Nd:YAG based systems and various Ar-ion laser lines, though the method is applicable with almost any convenient pump laser.

PDChopper

Pump

2nd

Pump

Crystal

Probe

PL

Dump

Fig. 1 Crossed-beam setup for low absorption spectroscopic loss measurements.

Modifications to the apparatus, using crossed pump and probe beams as illustrated in Fig. 1, allow spatially localized measurements for studying inhomogeneous bulk absorption, as well as surface and coating absorption effects. Through the use of two simultaneous pumps, we have characterized induced absoprtion effects such as gray-tracking in KTP and green-induced IR absorption (GRIIRA) in LiNbO3. The relatively rapid time response, faster than 100 ms, allows observation of transient absoprtion effects as well.

Photoconductivity and Photovoltaic Currents

Characterization of photorefractive transport properties, including photoconductivity and photovoltaic currents, at the high intensities characterisitc of nonlinear optical devices are difficult by conventional holographic methods. We have developed an apparatus, described in section E.3. "Characterization of


LiTaO3 by Photocurrents" that allows direct measurement of these currents in an experimentally convenient fashion. Our measurements with this apparatus has led to considerable insight into behavior of stoichiometric lithium niobate and tantalate.

Characterization of materials using spectroscopic ellipsometry

Spectroscopic and multiple angle of incidence ellipsometry is a valuable tool for the determination of the optical constants (n, k) of materials and the characterization of surface and interface morphologies. While optical transmittance and reflectance measurements provide information on bulk sample properties, ellipsometry has great sensitivity to the properties of the reflecting surface and interfaces in the case of multiple layers. When the optical constants of the materials studied are known, the technique can be used to characterize surface, interface roughnesses. On the other hand, the measurements can also be used to yield the optical constants of materials. The sample morphology must then be characterized by other techniques, since an accurate reduction of ellipsometric data to the physical quantities of interest requires knowledge of the constituent layer thicknesses.

A schematic of the experimental set-up is shown above in Fig. 2. The measurement consists in analyzing the change of polarization resulting from reflection off the surface studied. Wavelengths from 0.25-1.7µm and angles of incidence from 6-90o are accessible using our SOPRA instrument model GESP, which is of rotating polarizer type. The spectra consist of the ellipsometric angles (ψ,∆), measured typically at 200 points. Experimental data is fit using a linear regression with the purpose of minimizing an error function consisting of the difference

between calculated and measured values of

the quantities (ψ,∆). The ellipsometric angles are defined in terms of the ratio ρ = rp/rs = tgψei∆,

where rp, rs are the complex reflection coefficients for light polarized parallel and perpendicular to the plane of incidence. The reflection coefficients are related to the refractive index n and extinction coefficient k through Fresnel's equations. The optical constants are related to the microscopic properties of the material studied through the dielectric function which is a direct function of the material band structure. Anticipated imperfections, such as rough interfaces and surfaces are described as layers constituted of mixed materials

Fig. 2 Schematic layout of a spectroscopic ellipsometer.

Fundamental Science

SPRC 2002 – 2003 A-1 Annual Report

V. Research Program

A. Fundamental Science

D.B. Jackrel, and J.S. Harris, “InGaAs and GaInNAs Photodiodes for Advanced LIGO (Laser Interferometer Gravitational Wave Observatory)”................................................................ A-3

Edo Waks, Eleni Diamanti and Yoshihisa Yamamoto, “Generation of photon number states” ............................................................................................................................................... A-6

H.B. Yuen, S.R. Bank, M.A. Wistey, V. Gambin, W. Ha, J.S. Harris Jr., A. Moto “Growth and characterization of MBE-grown GaNAsSb and GaInNAsSb for long-wavelength optoelectronic devices” ............................................................................................... A-10

Fundamental Science


Fundamental Science


InGaAs and GaInNAs Photodiodes for Advanced LIGO (Laser Interferometer Gravitational Wave Observatory)

D.B. Jackrel, and J.S. Harris Department of Materials Science and Engineering

LIGO hopes to be the first instrument capable of detecting gravitational waves from astronomical sources, such as black holes and neutron stars. The LIGO installations are two interferometers, each with 4km long arms, located in Hanford, WA and Livingston, LA. In order to detect a gravitational event the interferometers must be sensitive enough to resolve a differential strain in the two arms of 10-18m. This amazing degree of sensitivity is largely dependent on the development of a 180W Nd:YAG laser source and a photodiode able to detect the 1064nm signal. There are several places within the instrument where photodiodes will be needed. The interferometer is operated near a dark fringe and the photodiode at the asymmetric port will detect roughly 30mW of optical power and needs to have a quantum efficiency over 90%. There will also be photodiodes used for laser intensity stabilization that need to be able to detect over 1 Watt of optical power and diodes used to control the auxiliary resonances within the instrument that require bandwidths of roughly 200MHz.

The devices utilize InGaAs or GaInNAs as the absorbing layer and are grown on standard n+ GaAs substrates using MBE (Molecular Beam Epitaxy). The devices are rear-illuminated PiN diodes, which are designed for high-power operation and a high degree of spatial uniformity. In a rear-illuminated device the absorbing layer is located within a micron of the heat sink. This provides roughly two orders of magnitude better heat dissipation than a front-illuminated device, in which the heat generated in the absorbing layer must diffuse through a relatively thick substrate (>100 microns) to reach the heat sink. Furthermore, the ring contact for a rear-illuminated device is much farther away from the intrinsic layer interface than in a conventional front-illuminated device and therefore there is a smaller variation in series resistance over the area of rear-illuminated devices. The range of device diameters reported in this study is between 0.5mm and 3mm.

Figure 1: Front vs. Rear-Illuminated photodiode schematics.

Fundamental Science


The intrinsic absorbing layers in the InGaAs devices contain roughly 25% indium in order to effectively absorb 1064nm light, and are thus highly lattice-mismatched from GaAs. Furthermore, in order to obtain high quantum efficiency the absorbing layer must be thick. In order to obtain over 90% quantum efficiency the layer must be nearly 2 microns thick. This thickness is well above the critical thickness for highly strained InGaAs. Thus, if one were to grow these thick layers directly on GaAs then misfit dislocations would form in order for the film to relax. Some of these dislocations would become threading dislocations that would short out the junction. In order to prevent the formation of dislocations within our intrinsic layer a graded buffer layer is used. This is a layer in which the composition of indium is linearly increased from zero percent at the substrate interface to 25% over two microns. Many dislocations form in this sacrificial layer but due to the slow linear grading none of these become threading and the device junction remains free from dislocations.1 However, even though there are no threading dislocations in the junction of these devices the surface of the graded buffer layer becomes rough due to the confined dislocations and creates rough interfaces in these devices as well as many point defects within the intrinsic absorbing layers. These poor materials properties often translate to poor electrical properties such as high dark current.

A potential solution to these issues is to move to a GaInNAs materials system. The addition of nitrogen into the material adds an extra degree of freedom, which allows one to simultaneously create a material with a band gap that is suitable to absorb 1064nm light and that is lattice matched to GaAs. Our group has an MBE system fitted with a plasma nitrogen source and using this modified MBE system we have been able to grow very thick, high quality, lattice matched GaInNAs films on GaAs. Below is a plot of dark current for two 0.5mm diameter devices with 2 micron thick intrinsic layers, one of InGaAs and one of GaInNAs. It can be seen from this plot that the dark current of the GaInNAs device is roughly 2 orders of magnitude lower than that of the InGaAs device. This is experimental evidence of the higher materials quality possible in the III-V-nitride material. Also it can be noted that neither device shows evidence of being close to breakdown at this bias level.

Figure 2: GaInNAs vs. InGaAs dark current, semilog plot.

To date we have fabricated InGaAs diodes that can detect over 300mW CW without saturating when biased at -10V. The response of the diodes below 300mW is also quite linear. These devices have a 3-

Fundamental Science


dB bandwidth of a few MHz, and an external quantum efficiency of 58%. Power levels of up to 450mW CW have been detected without showing any signs of degradation. The response above 300mW becomes non-linear (the efficiency drops to 45% at 450mW) due to a screening of the electric field within the devices. At bias levels less than -10V the saturation power levels are significantly reduced. In order to obtain higher saturation power levels future diode designs will be optimized to be biased at voltages beyond -20V.

Figure 3: External quantum efficiency vs. Optical power vs. Bias, InGaAs device.

In an effort to improve the quantum efficiency of the devices to 90% processing methods have been devised to thin the highly doped GaAs substrate. The rear-illuminated devices have the disadvantage that there is a significant amount of free-carrier absorption within the n+ substrate. Thus far substrates have been thinned to roughly 150 microns (from a standard 500 microns) on a few devices and have been tested at low power levels. Quantum efficiencies of close to 75% have been observed, even though low power densities were used to illuminate the devices. Future devices will have substrates thinned to below 100 microns with lower doping levels in order to push the quantum efficiencies of devices towards 90% without compromising heat dissipation or spatial uniformity.

References [1] Jackrel, D.B ., Harris, J.S., Ha, W., “High-Power High-Efficiency GaAs-Based Photodiodes for

the LIGO (Laser Interferometer Gravitational Wave Observatory) Project”, Stanford Photonics Research Center, Annual Report 2001, pages 24-26.

Fundamental Science


Generation of photon number states.

Edo Waks, Eleni Diamanti and Yoshihisa Yamamoto

Photon number states play an important role in quantum optics because they exhibit effects, which contradict the classical electromagnetic theory, such as anti-bunching and negativity of the Wigner function1. They also have important practical applications, for example in optical communications they achieve an optimal channel capacity2,3, while they can also improve the sensitivity of an interferometer to the Heisenberg limit4. Recently, there has been a tremendous effort to generate single photon states5 but, although theses sources are very promising, they suffer from substantial collection losses and the extension of their principle of operation to generate higher order photon number states remains a difficult problem. Up till now, the generation of a two photon number state has been demonstrated in the microwave regime, using Rydberg atoms in a high-Q cavity6.

In this work, we use the projection postulate of quantum mechanics to generate photon number states. This is achieved using the non-linear process of parametric down-conversion (PDC) in conjunction with a photon number detector, known as the Visible Light Photon Counter (VLPC). In non-collinear degenerate PDC, a bright pump field is injected into a material with a second-order non-linear dipole response and, with the appropriate phase-matching conditions, a pump photon splits into two photons of lower and equal energy that propagate in different directions, the signal and the idler. Since these photons always come in pairs, if a signal photon is detected, there must be an idler photon in the conjugate mode. Therefore, if a photon counter is placed in front of the signal arm and detects one photon, the corresponding idler arm is prepared in a state containing only one photon by the projection postulate of quantum mechanics. Clearly, the above scheme can be extended to the generation of higher order number states. If n signal photons are detected by a photon counter in a given pulse, then the idler pulse is projected onto a state iψ , which satisfies the condition: ii nn ψψ =ˆ , i.e. the idler mode is an eigenstate of the number operator n with eignevalue n . States which satisfy this special property are known as photon number states.

In order to implement the above idea, the photon counter must be able to determine the exact number of photons in the signal arm within the short time duration of the pump pulse. Conventional photon counters, such as avalanche photo-diodes (APD’s), cannot do that because they suffer from long dead time and large multiplication noise, and they can only distinguish the zero-photon case from the non-zero photon case in a pulse. On the contrary, the VLPC has been shown to have the capability to distinguish different photon number states with high quantum efficiency7,8. This is possible mainly because of two reasons. First, the VLPC has a relatively big active area (1mm in diameter) and when one photon is detected it forms a 5µm diameter dead spot, leaving the rest of the area available for subsequent detection events. Second, it has extremely low multiplication noise, which means that there are practically no fluctuations in the number of electrons emitted by the detector in response to a photo-detection event. Therefore, when two photons are simultaneously detected by the VLPC, it generates an electrical pulse, which is twice as large, and this pattern holds for higher photon numbers. The pulse height can then be used to determine the number of detected photons.

Fundamental Science


In our experiment, a 266nm pump source, generated from the fourth harmonic of a Q-switched Nd:YAG laser, is used. The pump pulses have a duration of 20ns, and a repetition rate of 45kHz. The fourth harmonic pumps a BBO crystal, set for non-collinear degenerate phase-matching, and the signal and idler waves both have a wavelength of 532nm and a divergence angle of 1 degree from the pump. Two VLPC detectors are used. Both are cooled down to 6-7K, which is the optimal operating temperature, and are appropriately protected from room temperature thermal radiation. The first VLPC (VLPC 1) is used as the triggering detector, which detects the number of photons generated in the signal arm on a given laser pulse. Its output is amplified to generate an electrical pulse, whose height is proportional to the number of emitted electrons, and subsequently the height of each pulse is discriminated by a Single Channel Analyzer (SCA). Figure 1a shows a pulse height histogram of VLPC 1. It features a series of peaks corresponding to different photon number states. The SCA can select pulse heights corresponding to one, two, three and four photon events, as is shown by the shaded areas in the figure. The second VLPC (VLPC 2) is placed in the idler arm, and is used to verify that the correct photon number state was generated. Its output is amplified and, whenever triggered from the SCA, the electrical pulse is integrated using a boxcar integrator. Figure 1b shows a pulse area distribution of VLPC 2 without any post-selection from VLPC 1. This distribution also features a series of peaks corresponding to different photon numbers, with the first peak representing a zero photon event. The photon number distribution, which is shown in the inset, is calculated by fitting each peak to a gaussian and normalizing the area of each gaussian by the total area of the peaks.

Both VLPC’s have imperfect detection efficiencies due to internal detection losses as well as external losses from the collection optics. The efficiency of VLPC 2 can and should be corrected for, because we are interested in how many photons were actually present in the idler arm, not how many have been detected. This can be done in a straightforward way by using a simple linear loss model, that takes into account the measured efficiency and dark counts of the VLPC. On the other hand, the efficiency of VLPC 1 plays a more subtle role. Detection losses can result in a higher order photon number state being misinterpreted by the detector as the correct photon number. Then, the probability distribution in the idler arm will no longer be an exact photon number, but a mixture of the desired number and higher order number states. This will cause a degradation of the fidelity, which is defined as the overlap between the desired and the actual generated state.

When the pumping intensity is low, however, the efficiency of the triggering detector (VLPC 1) does not play an important role because, in this case, there is a negligible probability of generation of higher order photon pairs, and therefore, when VLPC 1 detects n photons, it is true with very high probability that the same number of photons are present in both arms. Figure 2a shows the photon number distributions measured by VLPC 2, after correcting for detection efficiency and dark counts, when VLPC 1 post-selects a one, two, three and four photon event. For one, two and three photon post-selection, a nearly ideal photon number state is generated. For four photon post-selection, however, there contributions from three and five photon number states, which can be attributed to the smearing between the four photon peak and its nearest neighbours in Fig. 1a. This smearing is caused by build-up of multiplication noise, which puts a limit on the photon number resolution at higher numbers8.

Fundamental Science


As was previously discussed, increasing the pump power will result in degraded fidelity because of the imperfect detection efficiency of VLPC 1. On the other hand, higher pump powers will increase the probability that the correct number of photons was generated and hence the generation rate of the desired photon number state. Clearly, there is a trade-off between the state fidelity and the generation rate, which is apparent in Figure 2b, where the fidelity and the generation efficiency of the four photon number states are measured and plotted as a function of pump power. In this figure we also observe that, even with a pump repetition rate of only 45kHz, large enough for practical purposes generation rates can be achieved for reasonable fidelities.

In conclusion, we have generated photon number states using the projection postulate of quantum mechanics and conditional post-selection from a detector, the VLPC, which exhibits unique features. Further studies could include the generation of photon number states on demand, which would require for example a large parallel array of photon number generators, such as the one described here, with an optical switch, as well as the generation of Fock states, i.e. states containing n-photons with identical wavefunctions, which would require the use of a femtosecond pulse laser, so that the pulse duration is

Figure 1: (a) Pulse height histogram from VLPC 1 after an integrating amplifier. (b) Pulse area histogram of VLPC 2 with no post-selection from VLPC 1. The corresponding photon number distribution is shown in the inset.

Figure 2: (a) The photon number distributions, after correcting for efficiency and dark counts of VLPC 2, for reported numbers n = 1,2,3 and 4 by VLPC 1. (b) Generation rate (circles) and fidelity (squares) as a function of pump power.

Fundamental Science


on the order of the photon coherence length. This could open up the door for interesting new experiments in quantum information using photon number states.

References [1] D.Walls and G.Milburn, Quantum optics (Springer, New York, 1995). [2] Y.Yamamoto and H.Haus, Rev.Mod.Phys. 54, 1001 (1986). [3] C.Caves and P.Drummond, Rev. Mod. Phys. 66, 481 (1994). [4] M.Holland and K.Burnett, Phys.Rev.Lett. 86, 1502 (2001). [5] C.Santori et al., Phys.Rev.Lett. 86, 1502 (2001); J.Kim et al., Nature 407, 491 (1999); B.Lounis

and W.Moerner, Nature 407, 491 (2000); P.Michler et al., Science 290, 2282 (2000); E.Moreau et al., App.Phys.Lett. 79, 2865 (2001); A.Beveratos et al., Euro.Phys.J 18, 191 (2002); Z.Yuan et al., Science 295, 102 (2002).

[6] B.Varcoe et al., Nature 403, 743 (2002). [7] S.Takeuchi, J.Kim and Y.Yamamoto, App.Phys.Lett. 74, 1063 (1999); J.Kim, S.Takeuchi and

Y.Yamamoto, App.Phys.Lett. 74, 1902 (1999). [8] E.Waks, K.Inoue, W.Oliver, E.Diamanti and Y.Yamamoto, quant-ph/0308054 (2003).

Fundamental Science


Growth and characterization of MBE-grown GaNAsSb and GaInNAsSb for long-wavelength optoelectronic devices

H.B. Yuen, S.R. Bank, M.A. Wistey, V. Gambin, W. Ha, J.S. Harris Jr. Department of Electrical Engineering, Stanford University A. Moto Innovation Core SEI Inc.

Due to the dramatic increase in internet bandwidth demand, it is desirable to have low cost 1.3-1.6 µm lasers for optical metro area networks (MANs). Current technologies based on the InGaAsP/InP material system have several disadvantages including high cost and poor thermal performance. An alternative material system, GaInNAs/GaAs, has shown significant potential as an active region for 1.3 µm optoelectronic devices. GaInNAs can be grown on relatively inexpensive GaAs substrates and has superior thermal characteristics compared to InGaAsP. However, GaInNAs has not been able to optically emit at wavelengths longer than 1.4 µm due to material growth constraints such as relaxation and phase segregation.1 The addition of antimony (Sb) to GaInNAs, forming GaInNAsSb, has allowed for optical emission out to 1.6 µm and thus shows great promise for optoelectronic devices at 1.3 and 1.55 µm.

The addition of small amounts of nitrogen to GaAs was discovered to have the surprising effect of shrinking the bandgap while decreasing the lattice constant.2 Introducing nitrogen into InGaAs allows for optical emission at longer wavelengths while reducing epitaxial strain caused by indium when grown on GaAs substrates. The decrease in strain allows higher concentrations of indium to be incorporated, pushing the wavelength even longer. Unfortunately, there is a limit to the amount of nitrogen which can be incorporated due to the large miscibility gap between GaAs and GaN. GaNAs and its related alloys tend to phase segregate with significant nitrogen concentrations and thus results in inhomogeneous material.

The dilute nitride-arsenide alloys used in these experiments were grown by solid source molecular beam epitaxy (MBE) using a radio-frequency nitrogen plasma cell. The growth temperature of the substrate was kept low (~410-450ºC) in order to kinetically limit phase segregation and allow for larger amounts of nitrogen incorporation than thermodynamically allowed. These materials do not initially exhibit strong optical emission due to the non-radiative defects generated during the low-temperature, high-nitrogen growth. In order to improve photoluminescence (PL), the samples must undergo rapid thermal annealing (RTA) to remove defects from the active region and improve the overall crystal quality of the GaInNAs/GaAs quantum wells (QWs). However, during the RTA, nitrogen outdiffuses from the quantum well and blueshifts the optical emission. The usage of GaNAs barriers next to GaInNAs quantum wells reduces the blueshift by acting, to a certain degree, as a N reservoir.1

Antimony has been used in the past as a surfactant during MBE growth to improve the quality of semiconductors, including photoluminescence. Adding more In or N to GaInNAs with PL beyond 1.3 µm has typically decreased the PL intensity. However, it has been found that the introduction of Sb greatly increases optical emission intensity of GaInNAs alloys past 1.3 µm as seen in Figure 1. It was

Fundamental Science


also determined that Sb was incorporated in significant amounts to form GaInNAsSb. Sb has the same effect in GaAs as In and decreases the bandgap when incorporated.

0.75 0.80 0.85 0.90 0.95 1.00

10-3

10-2

10-1

100

1600 1500 1400 1300 1200

35% InGaInNAs

31% In GaInNAs

38% InSb: 7.2e-8 torrGaInNAsSb



Inte

nsity

(a.u

.)

Energy(eV)

Wavelength (nm)

Figure 1: Compared to high-In content GaInNAs PL samples, GaInNAsSb PL samples show a remarkable increase in intensity at longer wavelengths. Intensities are compared to best GaInNAs PL sample at 1.3 µm.

While adding Sb to GaInNAs has proven beneficial in obtaining longer wavelengths and improving material quality, adding Sb to GaNAs, the barrier material, was not extensively studied.3,4 In all previous studied of GaInNAsSb, Sb was also added to the barriers forming GaNAsSb. It was thought that since Sb had improved GaInNAs, it would do the same to GaNAs. However, after further examination, it was discovered GaNAs was preferred over GaNAsSb for the GaInNAsSb QW barriers.

With GaInNAsSb having such a large amount of compressive strain when grown on GaAs, it is important that the barriers provide tensile strain such that the active region may be strain compensated. This is important for growing good material and active regions of sufficient thickness for devices. It is also useful to note that in MBE, it is difficult to change growth rates, and thus compositions, during growth. In growing these samples and devices, it is the QW that dictates the composition of the barriers. Using x-ray diffraction (XRD), it was discovered that GaNAsSb barriers used for 1.55 µm were essentially lattice matched to GaAs. This provides no strain compensation to the QWs. GaNAs grown using 1.55 µm QW conditions does provide a sufficient amount of tensile strain such that the active region can be strain compensated. This can be seen in Figure 2.5

Other growth parameters such as the As overpressure (OP) and substrate temperature were adjusted to determine general properties of GaNAsSb. Using secondary ion mass spectrometry (SIMS), PL, and XRD, it was seen that as the As OP was increased, the incorporation of N remained constant but Sb decreased slightly. As the substrate temperature was increased, the N content again remained the same, but the amount of Sb dropped dramatically from 12% to 4% when raising the temperature by 150ºC. After obtaining the optimal growth conditions, PL was performed on GaNAsSb and GaNAs barrier material to determine optical material quality. It was discovered that although the signal from GaNAsSb was slightly higher than GaNAs, it was still 25 times worse than typical intensities obtained from QW material. With the trivial increase in optical material quality, lack of strain compensation, and possible band-offset issues discussed in literature, it was decided GaNAs was the preferred

Fundamental Science


barrier.5 Applying this new knowledge to device growth, world-record performance lasers were fabricated.6

Figure 2: 2θ/ω (004) scan of GaNAs and GaNAsSb QWs on GaAs. GaNAsSb barriers used for 1.55 µm GaInNAsSb QWs are essentially lattice matched to GaAs and provide no strain compensation. GaNAs has a significant amount of tensile strain and can strain compensate GaInNAsSb.

References [1] W. Ha, V. Gambin, M. Wistey, S.R. Bank, S. Kim, and J.S. Harris Jr., “Multiple-Quantum-Well

GaInNAs-GaNAs Ridge-Waveguide Laser Diodes Operating out to 1.4 µm,” IEEE Phot. Tech. Lett., 14, 5, 591-3, May 2002.

[2] M. Kondow, T. Kitatani, S. Nakatsuka, M.C. Larson, K. Nakahara, Y. Yazawa, M. Okai, and K. Uomi, “GaInNAs: A Novel Material for Long Wavelength Semiconductor Lasers,” IEEE J. Select. Topics Quantum Electron., 3, 719-30, June 1997.

[3] V. Gambin, W. Ha, M. Wistey, H.B. Yuen, SR. Bank, S. Kim, and J.S. Harris Jr., “GaInNAsSb for 1.3-1.6 µm Long Wavelength Lasers Grown by Molecular Beam Epitaxy,” IEEE J. Select. Topics Quantum Electron., 8, 4, 795, August 2002.

[4] K. Volz, V. Gambin, W. Ha, M. Wistey, H.B. Yuen, S.R. Bank, and J.S. Harris Jr., “The Role of Sb in the MBE Growth of (GaIn)(NAsSb),” J. Crys. Growth, 251, 360-366, April 2003.

[5] H.B. Yuen, S.R. Bank, M. Wistey, V. Gambin, W. Ha, J.S Harris Jr., and A. Moto, “Analysis of Material Properties of GaNAs(Sb) Grown by MBE,” 2003 Elec. Mat. Conf., Salt Lake City, UT, June 2003.

[6] S.R. Bank, M. Wistey, H. Yuen, L. Goddard, W. Ha, and J.S. Harris Jr., “A Low Threshold CW GaInNAsSb/GaAs Laser at 1.49 µm,” Electron. Lett., to be published.

Biophotonics

SPRC 2002 – 2003 B-1 Annual Report

Research Program

B. Biophotonics

G. Schuele, P. Huie, R. Brinkmann, R. Birngruber, A. Vankov, E. Vitkin, H. Fang, E. B. Hanlon, L. T. Perelman, D. Palanker, “Noninvasive monitoring of temperature during retinal laser treatments”........................................................................................................................B-3

E. Thrush, O. Levi, W. Ha, G. Carey, L. Cook, J. Deich, S.J. Smith, W.E. Moerner and J.S. Harris Jr., “Monolithically integrated fluorescence sensor”..........................................................B-6

K. Wang, N. Mehenti, H. Dai, H.A. Fishman, J.S. Harris, “Carbon Nanotubes as Microelectrodes for a Retinal Prosthesis” ............................................................................................B-9

X.J. Zhang, S. Zappe, C-C. Chen, O. Sahin, J. Harris, C. Quate, M. Scott and O. Solgaard, “Silicon microsurgery-force sensor based on diffractive optical MEMS encoders”............................................................................................................................................B-12

Biophotonics


Biophotonics


Noninvasive monitoring of temperature during retinal laser treatments

G. Schuele1, 2, P. Huie1, R. Brinkmann2, R. Birngruber2, A. Vankov1, E. Vitkin3, H. Fang3, E. B. Hanlon3, L. T. Perelman3, D. Palanker1, 1Department of Ophthalmology and Hansen Exp. Phys. Lab. (Stanford), 2 Medical Laser Center Lübeck (Lübeck), 3Biomedical Imaging and Spectroscopy Laboratory (Harvard)

Laser treatments of the retina are among the most common applications of lasers in medicine. The therapies range from established continuous wave (cw) photocoagulation to new ophthalmic laser applications like photo dynamic therapy (PDT), transpupillary thermo-therapy (TTT) and selective RPE treatment (SRT)[1]. Due to the lack of non-invasive means of measurement of retinal temperature, the laser induced heating in the retina can only be estimated theoretically. Since retinal pigmentation and blood perfusion strongly vary between patients, it is nearly impossible to accurately estimate the temperature rise during the laser treatment, and it would be highly desirable to measure the temperature increase in order to optimize the treatment and to limit the retinal damage.

In this work we introduce two novel approaches for noninvasive determination of the laser-induced temperature rise at the fundus.

Optoacoustic temperature measurement: Thermoelastic expansion of water induced by absorption of laser pulses in tissue, leads to emission of acoustic waves. It was shown that the maximum peak pressure is proportional to the thermal expansion coefficient β of water. For the strongest absorbing layer of the retina, the retinal pigmented epithelium (RPE), β increase nearly linearly in the temperature range from 20 to 60°C [2]. Therefore the amplitude of the optoacoustic stress wave can be used to determine the laser induced retinal temperature increase. To probe the temperature-related thermal expansion coefficient a short laser pulse must be applied to the retina to generate an optoacoustic signal. In the selective RPE treatment, where microsecond laser pulses are applied for retinal therapy, the treatment laser pulse itself can be used for generation of the optoacoustic signal and for monitoring of the laser-induced temperature increase. For this purpose a highly sensitive piezotransducer contact lens was developed, which is placed on the patient’ cornea prior to the retinal laser therapy, and the signals are amplified and recorded by a PC (fig. 1).

PC + data acquisitionPCI card with DSP

optoacoustic contact glass

laser

Figure 1: Setup for optoacoustic temperature measurements during patient treatment

Biophotonics


The typical clinical parameters for selective RPE treatment are 100 laser pulses at repetition rate of 500 Hz and 30 pulses at 100Hz, respectively (527nm, 1.7µs pulse duration, 160µm spot). Due to the heat accumulation during repetitive laser pulses, different retinal baseline temperatures build up at different therapeutic regimes. Figure 2 shows the results of the optoacoustic temperature measurements during the patient treatment. The experimental results are in good agreement with the heat diffusion-based calculations [2].

Temperature measurement by light scattering spectroscopy: Spectrum of the light scattered by a cell strongly depends on the size, refractive index and the shape of its cellular components (organelles). Thus, spectroscopy of the backscattered light can be used as a noninvasive tool for measurement of the sizes and optical density of cellular organelles [3]. We monitor the temperature-induced subcellular transformations in the human RPE cell culture by analyzing the spectrum of the backscattered light as a function of temperature. For calibration of the setup polystyrene beads of various sizes were solved in water. The samples were illuminated with a broad band light source and the spectrum of the backscattered light was measured with an optical multichannel spectrometer. The typical scattering spectra for the different bead sizes are shown in Fig. 3. Each scatter size has its unique spectral signature.

0 20 40 60 80 100

20

40

60

80

100

120

calculation measurement (upper) measurement (mean) measurement (lower)

tem

pera

ture

[°C]

pulse number [n]0 5 10 15 20 25 30 35

34

36

38

40

42

44

46

calculation measurement (upper) measurement (mean) measurement (unten)

tem

pera

ture

[°C]

pulse number [n]

A100 Hz 500 Hz

B

Figure 2: Optoacoustically determined laser-induced retinal temperatures for different pulse repetition rates

during selective RPE treatment. The heat diffusion temperature calculations with parameters from the literature are

in good agreement.

Biophotonics


Unpigmented human RPE cells (ATTC) were grown as a single layer on glass slides. The cells were placed on a temperature-controlled sample holder. During heating the back-scattering spectra were acquired (fig. 4A). From these spectra the size distribution of the scatterers has been deduced (fig.4B). In the preliminary measurements, the irreversible changes in the scattering spectra have been observed starting at 45°C.

• References

[1] Roider J, Brinkmann R, Wirbelauer C, Laqua H, Birngruber R: Subthreshold (retinal pigment epithelium) photocoagulation in macular diseases: a pilot study. British Journal of Ophthalmology 2000; 84: 40-7

[2] Schüle G, Hüttmann G, Framme C, Roider J, Brinkmann R.: Non-invasive optoacoustic temperature determination of the retinal pigment epithelium during laser irradiation of the eye, accepted and scheduled for publication in the Journal of Biomedical Optics, Jan/Feb 04

[3] Fang H, Ollero M, Vitkin E, Kimerer LM, Cipolloni PB, Zaman MM, Freedman SD, Bigio IJ, Itzkan I, Hanlon EB and Perelman LT, Noninvasive Sizing of Subcellular Organelles with Light Scattering Spectroscopy. IEEE J. Sel. Top. Quant. Elect. 2003;9,2

4 00 5 00 60 0 7 00 8 00 90 00 .0 0 0

0 .0 0 5

0 .0 1 0

0 .0 1 5

0 .0 2 0

0 .0 2 5

0 .0 3 0

scat

tere

d lig

ht in

tens

ity [a

.u.]

w a ve le n g th [n m ]

b e a d d ia m e te r 0 .5 4µm 0 .7 6µm 1 .0 5µm 2 .0 6µm 3 .1 7µm

p o lys ty re n e b e a d s in w a te r

Figure 3: Spectra of the light backscattered from the polystyrene beads in water.

0.30

0.35

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

0.80

450 500 550 600 650 700 750 800 850

λ (nm)

scat

tere

d lig

ht in

tens

ity [a

.u.]

surfa

ce d

ensi

ty [a

.u.]

scatter size [ m]µ

A

0.00

1.00

2.00

3.00

4.00

5.00

6.00

0 0.2 0.4 0.6 0.8 1 1.2

B

Figure 4: A: Scattering spectra of unpigmented RPE cells at various temperatures; B: Scatter size distribution deduced from the measured spectra.

Biophotonics


Monolithically integrated fluorescence sensor

E. Thrush, O. Levi, W. Ha, G. Carey, L. Cook, J. Deich, S.J. Smith, W.E. Moerner and J.S. Harris Jr.

Integrated biological analysis systems will find wide applicability in the areas of bio-warfare, clinical medicine and biological research. Medical diagnostics still rely upon labor-intensive, slow, and expensive laboratory techniques. Waiting for days for the results of a test for an infectious disease is a common reality of modern medical practice. Small portable micro-total analysis systems (µTAS, also referred to as biochips) based on micro-channels or micro-arrays promise to provide immediate point of care services that would facilitate detection of common diseases, bio-warfare agents, and early-stage cancer. Moreover, integrated biological analysis systems may enable parallel architectures for high throughput experimentation such as drug screening and give scientists a better understanding of complex biological processes like cellular transduction pathways, and protein interaction with potential drug targets. Finally, the development of extremely small analysis systems will improve and enable in vivo diagnostic tools such as imaging and real time drug delivery systems.

Fluorescence sensing remains one of the most widely used techniques in biotechnology. Unfortunately, traditional bio-fluorescence sensing systems use bulky and discrete elements, which are expensive and require large footprint and precise alignment. The advantages of integrated biological analysis systems are reduced when these systems rely upon large or fragile optical equipment. Integrated on-chip sensing architectures make portable and robust medical diagnostic equipment practical and promise to reduce the costs of research instrumentation.

Any practical fluorescence sensing system must contain three elements: excitation source, photodetector and spectral filters. Obtaining these elements in a simple and inexpensive way is a design challenge. Figure 1 illustrates the optoelectronic design for a semiconductor based monolithic sensor we have developed1. The excitation source is a vertical cavity surface emitting laser (VCSEL) epitaxial structure that includes two distributed Bragg reflectors (DBR) as mirrors and a quantum well gain region. The PIN photodetector is made by adding an intrinsic GaAs active layer underneath the VCSEL. The detector utilizes the bottom DBR of the VCSEL as an optical filter. The DBR is highly reflecting at the laser wavelength as designed and is relatively transparent to the Stokes shifted photons emitted from the dye. This technology is compatible with excitation wavelengths ranging from .6 – 1.1 µm. We have chosen to design for 770nm excitation because VCSEL technology is well developed in this spectral range and commercially available dyes (Li-Cor Inc.) are available.

Integrated semiconductor bio-sensors have been fabricated in a proximity architecture as shown in Fig. 2. Possible implementations of the sensor are illustrated in Fig. 3. The advantages of these architectures are simplicity of fabrication and large photon collection efficiency due to the large area detector. One disadvantage of the design is that the large area detector makes this design susceptible to laser background. Work continues to make other sensor architectures and explore design tradeoffs.

Biophotonics


Figure 1: Schematic of monolithic integration of Figure 2: SEM image of proximity sensor VCSEL, PIN photodetector, and optical filter

Figure 3: A) Proximity sensor detecting in a micro-array format. B) Proximity sensor detecting in a micro-channel format. To estimate the sensitivity of the sensors shown in Fig 3A, IR-800 Phosphoramidite dye (Li-Cor Inc.) was spun on a glass substrate in a PMMA film. The sensor was placed 500µm from the glass slide and the fluorescence signal was measured using 2mW excitation (773nm). The surface concentration of the dye on the glass substrate was determined by absorption using a spectrophotometer. Calibrated sample dilutions were used to estimate surface concentrations below the spectrophotometer’s detection limit. The sensor’s response is linear with the analyte’s surface concentration (Fig. 4), suggesting a lower detection limit of 105 molecules/µm2. This assumes that we can detect a 1 nA signal that is 1 part in 100 of the laser background (100nA).

The sensor sensitivity in a micro-channel format (Fig. 3B) was also determined. The experimental setup used a discrete lens (Geltech 370060, F=0.682 mm, N.A.=0.6) with a microfluidic chip while varying the distance to the sensor module. Flow channel width and depth are 100 and 50 µm respectively and were etched into bonded fused silica wafers (Corning 7980). The plano-convex lens is butted against the micro-fluidic channel and visual alignment of the lens to the microfluidics channel is achieved within a few microns. Following that, the sensor was visually aligned to the lens-channel assembly with micrometers. Dilute concentrations of fluorescent molecules (IR-800 Phosphoramidite in methanol) were flowed through the channel and detected with a laser excitation

Biophotonics


power of 2mW, see Fig. 5. The linear response of the fluorescence emission to the excitation power allows estimation of a lower bound for sensor sensitivity. Assuming that we can detect a 2nA response in a laser background level of about 200nA (1 part in 100), the sensor sensitivity is estimated to be approximately 1.0 µM.

Currently, the sensor sensitivity is limited by laser background and efforts are underway to reduce laser background and increase sensor sensitivity2. The large amounts of laser background result from placing the excitation source and photodetector in such close proximity. There are at least three sources of laser background: direct specular reflection from polished surfaces into the photodetector, spontaneous emission from the VCSEL side, and random scattering from various optical interfaces. Metal blocking layers have proved extremely effective in blocking the spontaneous emission from the VCSEL side. As a result, reflections from optical surfaces above the sensor are the dominant sources of laser background. Currently, the filter gives a filtration of about Optical Density (OD) 3.5. Future sensor designs will focus on increasing spectral filtration to greater than OD 5. In addition, optical simulations will be used to further optimize the system design to maximize the amount of spatial filtration.

Figure 4: Successful detection of fluorescent Figure 5: Successful detection of fluorescent molecules molecules in a microarray format. in a micro-channel format.

In conclusion, we have successfully fabricated a monolithically integrated fluorescence sensor. We believe that this is the first work to show a monolithic optoelectronic solution for fluorescence sensing. This study has found that placing the excitation source and photodetector in such close proximity limits the sensor sensitivity. The authors believe that greater sensitivity should be possible through the systematic reduction of laser background by increased spectral and spatial filtration.

References [1] E. Thrush, O. Levi, W. Ha, K. Wang, S. J. Smith, and J. S. Harris, Jr. Journal of

Chromatography A. In Press, 2003.

[2] E. Thrush, O. Levi, W. Ha, G. Carey, L. J. Cook, J. Deich,, S. J. Smith, W. E. Moerner, and J. S. Harris, Jr. Conference Proceedings of uTAS 2003. Lake Tahoe, CA, 2003.

Biophotonics


Carbon Nanotubes as Microelectrodes for a Retinal Prosthesis

K. Wang1, N. Mehenti2, H. Dai3, H.A. Fishman4, J.S. Harris5 1Applied Physics, 2Chemical Engineering, 3Chemistry, 4Ophthalmology, 5Electrical Engineering, Stanford University

About 50% of all blindness is caused by damage to the retina. Electrical stimulation has been shown to result in visual perception by Humayun and his co-workers1. Implantable retinal prosthetic devices consisting of microelectrode arrays are being built in attempts to restore vision2. The arrays transmit signals from an imaging system to the retina, and stimulate the remaining functional neurons (figure 1).

Current retinal prostheses use metals as planar electrodes. Common problems of metal planar electrodes include fouling, toxicity, antigenicity, and limited stimulation depth. We are seeking to develop a biocompatible, electrical-based retinal prosthetic device using carbon nanotubes (CNT) as penetrating electrodes. This device consists of an array of electrically conductive CNT towers projecting orthogonally from the surface of a silicon chip with pre-patterned microcircuitry (figure 2). Each tower consists of millions of self-aligned multi-walled carbon nanotubes. The entire device is covered with an insulating layer with only the tips of CNT towers exposed.

CNT tower insulating

layer

silicon chip

embedded

circuit

ganglion cells

rods and cones

bipolar cells

CNT array

Figure 1: Conceptual drawing by Jerry Lim of a retinal prosthetic device3. External electronics capture the image and transmit the image signal to implanted electronics on the retinal surface.

Figure 2: Design of a CNT microelectrode array for a retinal prosthesis. Left: CNT towers array connected to embedded circuit. Right: A CNT array implanted epiretinally4. Tower tips are in touch with retinal ganglion cells.

Biophotonics


Carbon nanotubes have intriguing electrical, mechanical and chemical properties for retinal prosthetic devices. They are extremely strong (with Young’s Modulus > 1Tpa, about five times stronger than steel)5, yet very flexible. Multi-walled carbon nanotubes are predominantly metallic. The three dimensional structure allows stimulation of retinal cells at different depths, which is key to generate vision.

A catalytic chemical vapor deposition (CVD) system is used for carbon nanotube tower synthesis6 (figure 3). Using traditional photolithography and an e-beam evaporator, 2nm of Fe was patterned as catalyst. Growth was done at 700°C, with 1200sccm ethylene. Ethylene decomposes and forms MWCNTs over the catalyst particles on sample. Vertical self-assembled nanotube towers form because of Van der Waals interactions between the individual tubes. Carbon nanotubes synthesized using this catalytic chemical vapor deposition method have high yield, good self-alignment, as well as high aspect ratio (figure 4). Towers have been grown at a variety of sizes with side-length ranging from 20µm to 250µm. Most towers are taller than 50µm after 20 minutes growth, and have an aspect ration of greater than 3. Some towers can reach the height of 150µm (comparable to the total thickness of human retina at the fovea).

furnace

C2H4 outlet

sample

quartz tube

Figure 3: Schematic drawing of a chemical vapor deposition system for multi-walled carbon nanotube synthesis.

A CB D

Figure 4. SEM pictures of the multi-walled carbon nanotube towers synthesized on plain silicon substrates with thermal CVD method. (A) An array of CNT towers grown on 38µm by 38µm catalyst patterns. (B) A 38µm diameter tower with a height of 150µm, indicating an aspect ratio of 4. (C) Edges and corners of a 250µm by 250µm CNT block, showing very good alignment of the nanotubes. (D) A CNT tower with underlying circuit.

Biophotonics


The biocompatibility of these CNT towers is tested by cell culture, as a first stage. Silicon substrate with as-grown towers were sterilized by 70% ethanol, coated with Poly(D-lysine) and laminin to improve cell adhesion, and then immersed in cell growth media. Rat (P7) retinal ganglion cells (RGC) were seeded onto the substrate and then kept in 37°C incubator. To examine cell survival, LIVE/DEAD dyes were loaded for scanning confocal microscope imaging (figure 5A). After cell growth was observed, the samples were fixed and dehydrated. Then SEM pictures were taken to study the cell growth on CNT towers in greater details (figure 5B and 5C).

In conclusion, we proposed a novel microelectrode array based on conductive multi-walled carbon nanotube towers. Uniform arrays of vertically aligned nanotube towers have been successfully synthesized using thermal CVD. Prototypes of the device have been fabricated. We have also shown that retinal ganglion cells can grow on these CNT towers without further modification. This device may hold significant potential as stimulating microelectrodes for retina prostheses applications. They may also act as recording electrodes to sense electrical and chemical activities in neural systems for fundamental neuroscience research.

References

[1] M. Humayun, et al., “Visual perception elicited by electrical stimulation of retina in blind humans”, Arch. Ophthalmology 114, 40 (1996)

[2] E. Margalit, et al., “Retinal prosthesis for the blind”, Survey of Ophthalmology 47, 335 (2002) [3] Original image from http://www.irp.jhu.edu/media/ [4] Diagram of the retina is originally from www.iit.edu/~npr/DrJennifer/ glossary.html [5] M. Treacy, et al., “Exceptionally high Young's modulus observed for individual carbon

nanotubes”, Nature 381 678 (1996) [6] S. Fan, et al., “Self-oriented regular arrays of carbon nanotubes and their field emission

properties”, Science 283, 512 (1999)

Figure 5: Retinal ganglion cell growth on CNT towers, after three days of culturing. (A) Scanning confocal microscope images of the culture with a LIVE/DEAD stain (green=alive, red=dead). Cells are all green, showing good viability. Cell bodies can be seen both ontop and near the CNT towers. Some neurites grow arround the towers, giving invisible towers an “outline”. (B) SEM picture of a ganglion cell growing on top of a CNT tower. (C) We observed that retinal ganglion cell neurites had a strong tendency to grow onto the side walls of CNT towers. This is an example.

A B C

http://www.irp.jhu.edu/media/

Biophotonics


Silicon microsurgery-force sensor based on diffractive optical MEMS encoders

X.J. Zhang, S. Zappe, C-C. Chen, O. Sahin, J. Harris, C. Quate, M. Scott* and O. Solgaard Dept. of Electrical Engineering, * Dept. of Developmental Biology, Stanford University

Localized and accurate microinjection of genetic material into biological model systems, such as Drosophila, will enable a variety of studies of developmental biology and genetics [1]. For such studies to be carried out in-vivo, the damage caused by the injection must be minimized. We have developed surface micromachined silicon-nitride injectors [2] with integrated force sensors for measurements of the injection force and needle-membrane interactions under various physiological conditions.

The force sensor is an optical encoder based on transmission phase gratings integrated with the injector. Precise displacement measurements using diffractive gratings is an established technology [3], and optical encoders have been developed for precise measurements of displacement and revolution angle for a variety of applications. However, the large size and expensive manufacture of conventional encoders make them unsuitable as integrated sensing devices. Recently there has been significant renewed interest in using diffractive micro-optical elements as displacement sensors in Atomic Force Microscopes (AFM) [4], MEMS capacitive ultrasonic transducers [5] and accelerometer with nano-g resolution [6]. For optical encoders, Sawada et al demonstrated a hybrid integrated encoder with a single grating on silicon [7]. Hane et al designed a dual-grating miniaturized displacement sensor using grating imaging [8]. These advancements in microfabricated diffractive grating optics enable integrated optical encoders for sensing and microscopy of embryos and single cells.

The force encoder consists of two identical constant-period transmission phase gratings that are vertically aligned in the static state [9], as shown in Figure 1. Phase gratings are advantageous in transmission because of their high optical throughput compared to amplitude gratings.

Biophotonics


When a force is applied to the injector (not shown), the upper index-grating is displaced with respect to the bottom grating. This changes the diffraction efficiency of the phase grating, and the relative position of the two gratings can be determined by the intensities in the diffraction orders. The diffraction characteristics of the dual transmission phase-grating can be analyzed by Frauenhofer diffraction theory. The trade-off between sensitivity and dynamic range is illustrated in Fig. 2. Encoders with larger grating period are used to have larger dynamic range, but low sensitivity (dotted line), while the opposite is true for encoders with finer pitch (solid line). For a given period, the sensitivity can be improved by increasing the number of grating periods that are illuminated, again at the cost of a reduced range (dash-dotted line). The periodicity of the encoder response can also be used to calibrate the relative displacement of the gratings.

Figure 2. Tuning of force encoder’s sensitivity and dynamic range by: I .Increasing half grating pitch period L=L1 for high dynamic range. II. Varying number of grating fingers N=(N2, N3) for given L for local high sensitivity enhancement within 2L injector displacement (L1=20L2, N2=4 N1,3).

(a) (b) (c)

Figure 3. SEM of (a) Injection force sensor. (b) Index and scale gratings with 20 um pitch and 2 um vertical gap. (c) Junction between the gratings and the supporting beams, with anti-sticking dimples.

Figure 3(a)-(c) shows scanning electron micrographs (SEMs) of the injector and force encoder. The sensor was illuminated by a HeNe laser (633nm/4mW) with spot sizes ranging from 60 to 160 µm. Figure 4 shows the measured power of the first diffraction mode as a function of absolute displacement of the injector. The grating displacement can be found from the known 20 µm period of the diffraction response. Using this calibration and a spring constant of 1.85 N/m, we find an injection

Biophotonics


force of 63 µN. The sensor as tested in Fig.4 (a) has large sensitivity around d=2L, but its output is ambiguous around the displacement where penetration takes place. The solution is provided by illuminating fewer periods of the force encoder. As shown in Figure 4(b), the same force sensor illuminated by a laser spot size of 60 µm (N=3), has an improved dynamic range (45% increase), at the cost of lower sensitivity (18%/um reduction). In this case, the diffraction is not ambiguous around the penetration displacement, so both the penetration force (48 µN) and the embryo membrane deformation (57

µm) at penetration can be determined. In a series of experiments, we found an average penetration force of 52.5±13.2% µN and an embryo deformation of 58±5.2% µm. The measurements are in reasonable agreement with the piezoresistive-scale calibration data, demonstrating that the microencoder force sensor has sufficient sensitivity and dynamic range for monitoring injection-force dynamics in Drosophila embryos.

In summary, we present an integrated dual diffractive micrograting based injection force sensor, with configurable sensitivity and sufficient dynamic range for monitoring injection force dynamics of injections into Drosophila embryos. The application of phase modulated optical encoder approaches is shown to provide enhanced quantitative understanding of embryo microinjection dynamics in terms of fundamental membrane mechanical properties. Current work includes further downscaling design for direct probing individual cellular interactions and dynamic operation of tunable-scale probe arrays for high-throughput instrument.

References [1] M.P. Scott, “Hox genes, arms and the man”, Nature Genetics 15: 117-118, 1997. [2] S. Zappe, X.J. Zhang, R.W. Bernstein, E.M. Furlong, M. Fish, M.P. Scott and O. Solgaard;

“Micromachined hollow needle with integrated pressure sensors for precise, calibrated injection into cells and embryos”, The Sixth Int. Sym. on Micro Total Analysis System (µTAS), Nara, Japan, Nov. 3-7, 2002.

[3] J. Guild, “Diffraction gratings as measuring scales”, Oxford University Press, 1960. [4] S.R. Manalis, S.C. Minne, A .Atalar and C.F. Quate; “Interdigital cantilevers for atomic force

microscopy”, Appl. Phys. Lett, 69 (25), Dec. 16, 1996 [5] N. Hall and F.L. Degertekin, “Integrated optical interferometric detection method for

micromachined capacitive acoustic transducers”, App. Phy. Lett., 80, 3859-3861, 2002

Biophotonics


[6] N. Loh, M.A. Schmidt and S.R. Manalis, “Sub-10 cm3 interferometric accelerometer with nano-g resolution”, J. Microelectromech. Sys., Vol.11, No.3, 182-187, 2002

[7] E. Higurashi, R. Sawada, “High-accuracy micro-encoder based on the higher-order diffracted light interference”, Int. Conf. on Opt. MEMS, Hawaii, Aug. 21-24, 2000.

[8] K, Hane, T. Endo, Y. Ito and M. Sasaki, “Si Micromachined Optical Encoder Based on Grating imaging”, The 11th Int. Conf. on Solid-State Sensors and Actuators, Munich, Germany, June 10-14, 2001.

[9] X.J. Zhang, S. Zappe, R.W. Bernstein, O. Sahin, C.-C. Chen, M. Fish, M.P. Scott and O. Solgaard; “Integrated Optical Diffractive Micrograting-based Injection Force Sensor”, The 12th Int. Conf. on Solid-State Sensors and Actuators, Boston, USA, June 9-12, 2003.

Nanophotonics

SPRC 2002 – 2003 C-1 Annual Report

Research Program

C. Nanophotonics

D.P. Fromm, A. Sundaramurthy, K.B. Crozier, P.J. Schuck, G.S. Kino, C.F.Quate, W.E. Moerner, “Development of Electromagnetically Enhanced Au “Bowtie” Nanostuctures” .................C-3

J. Hwang, M. M. Fejer and W. E. Moerner, “Exploring Novel Methods of Interferometric Detection of Ultrasmall Phase Shifts”..................................................................................................C-6

Yang Jiao, Shanhui Fan, and David A. B. Miller, “Designing for beam propagation in periodic and nonperiodic photonic nanostructures: extended Hamiltonian method”...........................C-9

J. Matteo, D.Fromm, Y. Yuen, P.J. Shuck and L. Hesselink, “Spectral Analysis of Enhanced Transmission thru Single Nano-Apertures”.......................................................................C-12

Zheng Wang and Shanhui Fan, “Compact all-pass filters in photonic crystals as the building block for high capacity optical delay lines” .........................................................................C-15

Mehmet Fatih Yanik, Shanhui Fan, Marin Soljačić, J. D. Joannopoulos, “High-density Low-power All-optical Logic Gate in Photonic Crystals” .................................................................C-18

Nanophotonics


Development of Electromagnetically Enhanced Au “Bowtie” Nanostuctures.

D.P. Fromm, A. Sundaramurthy, K.B. Crozier, P.J. Schuck, G.S. Kino, C.F.Quate, W.E. Moerner Department of Chemistry and E.L.Ginzton Laboratory, Stanford University

Optical electric fields (E-fields) are enhanced in close proximity to sharpened metal tips with a radius of curvature much smaller than the incident illumination wavelength, a so-called near-field optical effect. Near field optics are of interest because they offer a way to overcome the diffraction limit, and the E-field enhancement present in such structures can be exploited to produce an ultra-small, ultra-intense light source that could revolutionize optical lithography. Further, enhanced E-fields can be used to excite low-probability effects, such Raman scattering1 and enhanced fluorescence2.

To date, sharpened, pyramidal tips have been the focus of near field optics, mainly because atomic force microscope (AFM) tips can be made to have a very small (<10 nm) radius of curvature and are commercially available. Novotny et al3 have modeled the E-field enhancement for a pyramidal Au tip with a 5 nm radius of curvature and found a peak enhancement of E2 ~ 3000. To fully realize this large enhancement effect, it is important to use a tip made from a strongly conducting metal, such as Au or Ag. AFM tips are usually made from Si, and would need to be coated with a metal film to be effective near field probes. However, it is possible that thin films do not have sufficient free electron density present to achieve this full enhancement, and it would be preferable to construct a fully metallic tip. We have been concentrating our efforts on the examination of small metallic triangles and rods, which we can make with high-resolution electron-beam (e-beam) lithography. We can tune the size and shape of the triangles to modulate their enhancement properties. We are also investigating the enhanced field effects of “bowtie” junctions, arising from triangles arranged in an antiparallel fashion. “Bowties” with an overall length of 2.5 λ and tip spacing ~ λ /14 have been modeled and shown to have very large E-field effects for microwave experiments4.

Our investigation of these nanoscale structures is a combination of complementary simulations and experiments. We use Finite Difference Time Domain (FDTD) methods to study the strongly enhanced E-fields near the surface of a nanoparticle in the shape of a triangular prism, excited by light whose wavelength corresponds to the particle’s plasmon resonance. The FDTD software used was TEMPEST5. The goal of the simulation was to show that for a given size, shape and choice of material, the nanoparticle would resonate at a particular wavelength. The structure, modeled using TEMPEST, is shown in Figure 1A. The Au triangle is on quartz and is excited by a light source beneath it. Modeling the total E-field distribution 4nm away from the top surface of the triangle shows a peak enhancement of E2 = 1420 at a resonance of 605 nm with a bandwidth of ~ 100 nm FWHM. Figure 1B shows the change in field intensity at wavelength 605 nm for a 228nm X 200nm triangle.

Nanophotonics


Figure 1: (a) A triangular nanoparticle on quartz L = 228 nm; (b) Field enhancement near tip. The x and y axes represent the nodes in the simulation domain, each node being 4 nm wide. Approximate resonant wavelength is 605 nm and the enhancement in |E|2 is 1420 times |E|2incident.

We are using scattering experiments to measure the optical resonances of single, isolated triangles, rods, rod junctions, and “bowties.” Since scattering scales as E2 (linearly with intensity), the resonance of a nanostructure can be found by pumping the structure with a broadband source and measuring the scattered light intensity as a function of wavelength. Measuring the scattering of a single triangle or “bowtie” will allow us to find the intrinsic resonance for each individual structure, free of ensemble averaging. This ability will allow us to see how small heterogeneities in the size and shape of the structure impact the scattering resonance. Further, an isolated nanostructure is free from coupling to other tips and structures, so these resonance measurements will be that of only the triangle and can be directly compared with theoretical models and simulations. We use the light from a 75W W lamp that is collimated and brought into an oil-immersion darkfield condenser, with a large numerical aperture (N.A. = 1.4) and the center blocked to allow only high-angle light to excite the sample. A linear polarizer is used to control the polarization direction of the excitation light. A 40X- low N.A. (0.8) oil immersion microscope objective lens is used to collect scattered light, which is present at low angles. Directly transmitted (high N.A.) light is not collected by this objective. A 50 µm pinhole at the microscope image plane spatially filters our image to a diameter limited to ~ 1.25µm. The light is then focused onto either an avalanche photodiode (APD) to measure broadband scattering intensity or passed through a monochromator and imaged on a liquid nitrogen-cooled CCD to measure spectra. The sample is mounted on a closed-loop X-Y piezo scanner to accurately locate the desired particle for study. A 50 X 50 µm darkfield image of an array of isolated, single Au bowties and several of their spectra are shown below:

A B

Nanophotonics


Figure 2: (a) 50 X 50 µm darkfield image of an array of single Au “bowtie” junctions; (b) polarization-dependent scattering spectra of a bowties with a small gap (< 30 nm, bowtie 1,1), and a larger gap (40 nm, bowtie 3, 1).

The spectra shown above demonstrate a wide variation in peak position and widths, depending on the excitation polarization. Additionally, the bowties show a strong dependence on the gap width, where it appears that a different resonant mode exists when the particles are spaced by ~ 30 nm or so, but this mode disappears at distances farther than this. Single and double rods are also being examined, and the resonances of these particles also exhibit a strong sensitivity to excitation polarization and to the length of the gap between particles. We are currently investigating these coupling and polarization-dependent effects in more detail for both triangular and rod-like particles.

In conclusion, we have successfully manufactured nanoscopic Au and Ag triangles and “bowties” that are representative of the smallest structures made using e-beam lithography. We present simulations that predict a very large electric field enhancement for triangles/”bowties” of varying size, shape, and materials. We also have developed methods to measure the scattering from single particles. The results from these simulations will be compared with scattering experiments to determine the proper size, shape, and material to make a near field probe with a large E-field enhancement of an appropriate resonant frequency.

Acknowledgements:

D.F. would like to thank the NSF for a Graduate Research Fellowship. This work was supported by the Department of Energy and the National Science Foundation.

References [1] A.M. Michaels, J. Jiang, and L. Brus; J. Phys. Chem. B 104(2000) 11965. [2] H. F. Hamann, M. Kuno, A. Gallagher and D. J. Nesbitt; J. Chem. Phys. 114(2001) 8596. [3] L. Novotny, R.X. Bian, and X.S. Xie; Phys. Rev. Lett. 79(1997) 645. [4] R.D. Grober, R.J. Schoelkopf, and D.E. Prober; Appl. Phys. Lett. 70(1997) 1354. [5] TEMPEST 6.0, Electronics Research Laboratory, University of California at Berkeley.

Nanophotonics


Exploring Novel Methods of Interferometric Detection of Ultrasmall Phase Shifts

J. Hwang 1,2, M. M. Fejer2 and W. E. Moerner1

1Chemistry Department, Stanford University 2Applied Physics Department, Stanford University.

Current research in the area of single-molecule spectroscopy makes use of labeling with fluorescent molecules such as laser dyes or autofluorescent proteins. In spite of some advantages, fluorescence in general is plagued by a host of experimental issues, most notably the need to rigorously exclude from the probed volume all sources of unwanted fluorescence or Raman scattering.

An alternative detection method that we consider here is absorption spectroscopy, in which the absorption events are not detected by recording fluorescence. In this case, any spurious fluorescence from the sample or substrate are not important, and all the photons in the entire laser beam probing the sample can be used to sense the signal of interest. In fact, absorption detection was used in the first single-molecule spectroscopic experiments in 1989, which were performed at liquid-helium temperatures[1]. Here we concentrate on room temperature, which represents a considerable challenge for the following reason. The peak size of the absorption change for one molecule is ~104-105 larger at low temperatures than at room temperature. This means that the detection technique must work very hard to see an extremely small change in the probing laser beam on the order of 10-9 in relative size per molecule, without becoming overwhelmed with nonideal effects such as laser noise, vibrations and so on. This requires a technique that can operate at the quantum limit with strong rejection of interfering spurious effects.

The absorption of a photon by a molecule necessarily results in a phase shift of the probing beam. Instead of directly measuring the absorption, we choose to detect the phase shift. The advantage of measuring the phase shift instead of the absorption is that due to higher saturation intensity at the wavelength of peak phase shift than peak absorption, the measurement can be done with a high intensity probe without saturating the sample molecules. A higher intensity of probe laser yields better signal-to-noise ratio in the quantum limit. We choose to use a polarization Sagnac interferometer[2] to detect this phase shift. The common path nature of the polarization Sagnac interferometer provides high rejection of common mode nonidealities of the probe beam. Furthermore, the dark port output provided by our polarization Sagnac interferometer alleviates the restriction arising from optical damage of the photodiode and the single polarization along the loop enables polarization-sensitive detection. On the other hand, the common path nature of the Sagnac interferometer imposes strong constraints on the optical properties of the sample. Since the same sample is seen by both of the counter-propagating beams, the two phase shifts from the sample are cancelled at the dark port. For this reason, the sample must be placed asymmetrically inside the interferometer loop and the refractive index of the sample should be modulated on a time scale comparable to the transit time around the loop in order to create a phase difference between two beams. To achieve this for a molecular absorption, we have chosen to modulate the absorption by optical saturation, that is, by using a short, intense mode-locked laser pulse at a different wavelength to pump the molecule into its excited state in a repetitive fashion. This will cause the phase shift from the sample to disappear and reappear at an RF frequency equal to the laser pulse repetition rate. This high frequency phase shift signal can be

Nanophotonics


demodulated with the reference from the mode-locking RF source and the detection bandwidth can be reduced to improve sensitivity. This scheme of modulation of the sample with a separate pulsed laser is relatively noninvasive compared to other methods of modulating the sample we devised such as physically dithering the sample on a ultrasonic transducer. Furthermore it reduces the possibility of AM feedthrough; in particular, we expect negligible photo-thermal index modulation of the sample host matrix.

Figure 1: Layout of the setup. The red and green lines are the laser beam paths, and the dotted lines are electrical signal. PBS:polarizing beam splitter LP:linear polarizer HWP:half wave plate QWP: quarter wave plate SF:spatial filter DBS:dichroic beam splitter

As shown in Figure 1, the sample is placed asymmetrically in the interferometer loop and on a precision scanning stage in order to ultimately achieve microscopic scanning in case of solid state samples. Two lenses are used to focus the probe beam(presented as red) into the sample. With the help of a dichroic beam splitter, mode-locked pulsed laser light (presented as green) is sent into the interferometer path. Traveling through the same pair of lenses that were used to focus the probe beam, each pump pulse saturates the sample; then the pump pulse exits the interferometer loop at another dichroic beam splitter.

In the past, with an electro-optic modulator as a test sample, this interferometer was tested with a 10mW HeNe laser as the probe to show that 3x10-8 rad sensitivity can be achieved with 1Hz bandwidth[2]. However, this did not prove the feasibility of this scheme in the sense that the signal was not from molecules being modulated by mode-locked lasers. In order to obtain a proof-of-principle with a molecular sample, a less-sensitive detector and a relatively dense (~107 molecules in the focal volume) sample was used.

Contrary to our original speculation, molecules already known as good single-molecule fluorophores such as Cy3 and Cy5 did not work for the setup due to photobleaching. After a search for suitable sample molecules, we found that the saturable absorber molecule DASBTI in solution and terrylene in a p-terphenyl crystal work and give reasonable size of signals. In one experiment, a 2.3x10-5M solution of DASBTI in 1:1 ethanol and glycerol was contained between two coverslips with parafilm

Nanophotonics


as a spacer. This sample yielded 9.6x10-6 rad signal. With a focal spot measured to be 4µm and sample thickness of 127µm, the number of molecules in the focal volume is estimated to be on the order of 107 molecules. This number of molecules is expected to give roughly 10-4 rad phase shift from the bulk spectrum and a Kramers-Kronig calculation. The one order of magnitude discrepancy may arise from the error in estimating the number of molecules in the focal volume and also possibly from photobleachng of the saturable absorbers. A solid, crystalline sample of terrylene in p-terphenyl gave stable signals with less sign of photobleaching than the saturable absorbers.

In conclusion, we verified the feasibility of measuring small absorption signals from solid and liquid samples with a polarization Sagnac interferometer. In the future, we will complete a more precise calibration of the size of the signal and increase the sensitivity of the system by utilizing improved detectors and increased probe power.

References [1] Moerner, W. E. and Kador, L. Phys. Rev. Lett. (1989), 62, 2535-2538. [2] Beyersdorf, P. T.; Fejer, M. M.; Byer, R. L. J. Opt. Soc. Am. B (1999), 16, 1354-1358. [3] Hwang, J.;Fejer, M. M.;Moerner, W. E. Proc. SPIE (2003) 4962,110-120.

Nanophotonics


Designing for beam propagation in periodic and nonperiodic photonic nanostructures: extended Hamiltonian method

Yang Jiao, Shanhui Fan, and David A. B. Miller Department of Electrical Engineering, Stanford University

The possibility of radical dispersion in periodic photonic nanostructures has recently generated much interest. However, for applications such as group velocity dispersion compensation and frequency demultiplexing, periodic structures arguably do not offer enough design freedom to achieve the desired input/output dispersion relation over large frequency ranges. In such cases, non-periodic structures could perform much better than periodic structures1,2. One expects that introducing non-uniformities in 2-D and 3-D periodic structures will also allow better control of the device input/output dispersion relations over wider frequency ranges. However, there has been much less study of 2-D and 3-D non-uniform structures due to the difficulty in analyzing them. Hamiltonian optics offers one approach for nonuniform structures, at least for predicting beam paths3. In this work, we use Hamiltonian optics to design and analyze beam propagation in 2-D periodic structures with slowly varying non-uniformities. Furthermore, we extend the Hamiltonian optics method to analyze the width of a beam propagating in such structures; without such an understanding of beam width and its distortion in periodic and nonperiodic photonic nanostructures, practical design would be very limited. We validate our method with Finite Difference Time Domain (FDTD) simulations, and design a 2-D non-uniform structure that functions as a flat-top frequency demultiplexing device that steers beams in two different frequency ranges to two different locations.

Hamiltonian optics has traditionally been used to describe the optical ray path in a slowly varying dielectric medium4. Let X(τ) and K(τ) denote the optical beam path in space and the average wavevector along the beam path respectively, where τ is a parameter that varies continuously and monotonically along the beam path. Then the Hamiltonian equations governing the evolution of X(τ) and K(τ) are4:

0)(,, =∂∂

=∂∂

= kx,x

Kk

X HHddH

dd

ττ, (1)

where x and k are the position vector and the wavevector respectively. For a given x, the Hamiltonian H is the dispersion relation minus the frequency. For Eq. 1 to be valid, the nonuniforimlity in the structure needs to be slowly varying5. At each location in the structure, a local H can be obtained by Bloch wave analysis. With H defined, beam propagation in the structure can be analyzed using the Eq. 1, without the need for time consuming FDTD simulations.

We use the Hamiltonian equations to design a flat-top frequency demultiplexing device. The basic structure is a 200 by 200 square lattice of high index rods (n=2.54) in low index background (n=1.56), with gradual non-uniformity in the rod radius. Six beams equally spaced in frequency enter the structure from the top-left, and exits as two groups on bottom side (Fig. 1). The partial derivatives of H are needed in Eq. 1. We calculate H using the iterative eigenmode solver MIT Photonic-Bands package (MPB)6.

Nanophotonics


0 100 2000

50

100

150

200

y, la

ttice

con

stan

ts

0 100 200

0.285

x, lattice constantsx, lattice constants

Figure 1. Frequency demultiplexing structure. a) Beam path (Eq. 1) for six frequencies (ωa/c) from 0.28 to 0.285; b) Beam width profile for ωa/c=0.285 (profile of the fence) and ωa/c=0.28 (profile of the 3 dotted lines).

To steer the beams to the desired exit locations, we start with an initial guess for the rod radius distribution ρ(x), ρ(x)=a(x2+y2). The distribution is altered in an optimization algorithm employing Eq. 1, which sequentially sets the beam path of each frequency to the desired locations5. The algorithm takes 10 minutes on a Pentium 4 computer. The FDTD simulation of the beam propagation for a specific ρ(x) takes many days on the same machine. Optimizing ρ(x) by running an FDTD simulation multiple times would be prohibitively expensive computationally.

We extend the Hamiltonian optics method with equations for the beam width and the radius of curvature in periodic structures with slowly varying non-uniformities. To the best of our knowledge, beam width equations have not previously been proposed for periodic nanostructures. We follow the method developed by Poli et. al. for plasmas7. Assume a monochromatic field has the form A(x)exp(-φ(x)+is(x)). If we expand the complex phase s(x)+iφ(x) up to the 2nd order around the beam path, the 2nd derivatives of s(x) and φ(x) will give the beam width Wα(x), and the radius of curvature Rα(x):

22

2

2

2 2,α

ααα

αα φαφω

α WcRss

=≡∂∂

=≡∂∂ (2)

where α is the space coordinate x or y. Substituting the assumed form of the field into Maxwell's equation, equations for the 2nd derivatives of s(x) and φ(x) along the beam path are obtained7:

2 2 2 2 2ds H H H H Hs s s sd x x x k x k k k k k

αβαγ βγ αγ βδ αγ βδ

α β β γ α γ γ δ γ δ

φ φτ

∂ ∂ ∂ ∂ ∂= − − − − +

∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ (3)

αγβδδγγβ

βγαδδγγα

αβ φφτ

φ⎟⎟⎠

⎞⎜⎜⎝

⎛

∂∂∂

+∂∂

∂−⎟

⎟⎠

⎞⎜⎜⎝

⎛

∂∂∂

+∂∂

∂−= s

kkH

kxHs

kkH

kxH

dd 2222

For homogenous media, Eq. 3 reduces to beam width equations for a freely propagating Gaussian beam. For the periodic medium with non-uniformity in our example, Eq. 3 can be solved numerically. We evaluated the 2nd derivatives of H numerically using H calculated from MPB. Fig. 1b shows the beam width calculated for two beams, and verifies that the two chosen output locations are far enough apart. Both beams have a flat phase front at the input and would diverge in vacuum. But, as shown in Fig. 1b, Eq. 3 predicts the beams narrow at some points because the curvature in ρ(x) acts like a lens.

Fig. a) Fig. b)

Nanophotonics


x, lattice constants

Figure 2. Comparison of the extended Hamiltonian method with FDTD simulation. The center dotted line is the beam path, the dotted lines on each side show the beam width as calculated using Eq. 3. The underlying shading shows the intensity of the e-field from FDTD.

We verify the validity of the proposed extended Hamiltonian method with the FDTD simulation. The structure is a 200 by 200 square lattice of high index rods (n=2.54) in low index background (n=1.56), with a quadratic rod radius distribution ρ(x), ρ(x)=a(x2+y2). The nonuniformity in the lattice causes a beam entering from the upper left corner to turn 90 degrees. The beam path and beam width calculated with Eq. 1 and Eq 3 are plotted over the FDTD simulation results in Fig. 2. 2-D FDTD method with perfectly matched layer boundary conditions8 is used in the simulation. The two methods are in fairly good agreement. The transverse beam profile is diverging from a Gaussian due to the large nonuniformity of this structure. This is not modeled by Eq. 3 because the beam width calculation only uses the 2nd order derivative of the nonuniformity along the beam path.

We have demonstrated the use of the Hamiltonian optics to design a flat-top frequency demultiplexing device. We have introduced a set of equations for tracking the beam width in periodic nanostructures with nonuniformities, and confirmed the validity of the equations with FDTD simulations. We believe this extension will supplement the Hamiltonian equations to enable more quantitative design and analysis of useful devices such as frequency demultiplexers, beam steering devices, and wavelength dependent (e.g., gas) sensors, without resorting to FDTD simulations.

YJ gratefully acknowledges the support of the NSF Graduate Research Fellowship and the Reed-Hodgson Stanford Graduate Fellowship.

References [1] N. Matuschek, et. al., J. of Quantum Electronics 35, 129-137 (1999) [2] M. Gerken, D.A.B. Miller, Applied Optics 42, 1330-1345 (2003) [3] P. Russell, J. Lightwave Tech. 17, 1982-1988, (1999) [4] J.A. Arnaud, Beam and Fiber Optics, NY, Academic Press, (1976) [5] Y. Jiao, S.H. Fan, D.A.B. Miller, IEEE LEOS 2003 Annual Meeting, Tucson, California (2003). [6] S. Johnson and J. D. Joannopoulos, Optics Exp. 8, 173-190 (2001). [7] E. Poli, et. al., Physics. of Plasmas 6, 5-11, (1999) [8] J. P. Berenger, J. Comput. Phys., vol. 114, p. 185, (1994)

0 20 40 60 80 100 1200

20

40

60

80

100

120

y, la

ttice

con

stan

ts

Nanophotonics


Spectral Analysis of Enhanced Transmission thru Single Nano-Apertures

J. Matteo, D.Fromm, Y. Yuen, P.J. Schuck , L. Hesselink, and W.E. Moerner EE/SSRL Laboratory, Stanford University

The diffraction limit represents a fundamental limit in the spatial resolution for far-field optical intensity. Early attempts at overcoming this limit with sub-wavelength apertures was greatly hampered by their low transmission which falls off rapidly with aperture size (~d/wavelength4)1. Arrays of nano-apertures have recently been shown to exhibit enhanced transmission due to surface plasmon resonances. This method has the practical limitation that the entire array must be illuminated. Although the relative power transmission may be high, due to the widely expanded beam, the total power transmitted by any given aperture is actually quite low. Shi et. al have recently reported a C-shaped aperture design that is able to overcome the low power throughput of the single metallic apertures without the practical disadvantages, and low efficiency of the plasmon enhanced aperture arrays2. This single aperture is able to achieve spot sizes on the order of one-tenth of a wavelength with normalized power throughputs 1000 times greater than a square aperture which produces a similar spot size. Figure 1 shows a comparison between both spot size and power throughput for a C-aperture and a one-tenth wavelength square aperture.

Figure 1: E-Field intensity distribution 50nm away from a).100nm x 100nm square aperture b). C-aperture. The C-aperture produces similar sized spot with a 1000 times transmission enhancement.

Looking at this point from a spectral point of view, the C-aperture performs well because, despite its small size, it is resonant at a longer wavelength than a conventional aperture. Therefore, since the spot size is determined by the physical size of the aperture, the greater the ratio between the resonant wavelength and the aperture size, the greater the resolution. Figure 2 shows the calculated transmission spectra of a C-aperture compared to that of a square aperture of the same area, and a rectangular aperture of the same area.

Nanophotonics


Figure 2: FDTD calculated spectra for a a square aperture(128nm x 128nm), C-aperture1(50nm feature size), and vertical rectangular aperture (50nm x 270nm). (same area)

Fabricating C-apertures for use in the near infrared and visible ranges requires the ability to make features on such small scales the only feasible fabrication procedure is focused ion beam milling. Using a dual beam FEI machine we were able to fabricate C-apertures which are resonant from 550nm thru 650nm. As the wavelengths become shorter, the fabrication becomes more difficult. Nonetheless, it appears that apertures can be made which operate throughout the visible range.

Experimental Results

Using a white light setup and confocal detection, we’ve measured the transmission spectra of individual C-apertures, and compared their performance to that of different shaped apertures.

Figure 3: Transmission spectra for a square aperture(128nm x 128nm), C-aperture1(50nm feature size), and vertical rectangular aperture (50nm x 270nm). (same area)

Note that the C-aperture has a stronger resonance than the square of the same area, and that it occurs at a much longer wavelength. The vertical rectangular aperture transmits even more light, however, it has a spot size that is considerably larger than the other two. It is worth noting that the 50nm square aperture which should produce the same spot size as the C, did not even transmit enough light to exceed the noise level of the detector.

We also showed the scalability of this technology through tuning the resonance by scaling the dimensions of the aperture accordingly.

Nanophotonics


Figure 4: Transmission spectra for C-apertures ranging from 40nm-55nm in their minimum feature size.

The C-aperture resonance is tuned over 100nm simply by scaling the dimensions in steps of roughly 10%. Using this technique it is conceivably possible to design an enhanced transmission aperture for any wavelength spanning the visible regime. Finally, we showed that the resonance peak can further be shifted by changing the index of the adjacent media. This provides us with potentially even greater resolution, much like a solid immersion lens.

Figure 5: Transmission spectra for C-apertures as the index of the adjacent medium changes from air(n=1), to water(n=1.33), to oil (n=1.5).

References: [1] H. A. Bethe; Physical Review. 66 (1944) 163. [2] X. Shi, and L. Hesselink; Japanese Journal of Appl. Phys. 41 (2002) 1632-5.

Nanophotonics


Compact all-pass filters in photonic crystals as the building block for high capacity optical delay lines

Zheng Wang1 and Shanhui Fan2 1 Department of Applied Physics, Stanford University, Stanford, CA 94305-4090 2 Department of Electrical Engineering, Stanford University, Stanford, CA 94305

Optical all-pass filters are resonant devices that produce strong on-resonance phase variation with respect to the wavelength, while maintaining unity reflectance or transmission over the entire resonance frequency bandwidth. Such filter can generate large group delay and thus dispersion with no reflection and interference between the stages, and has been used in optical delay lines where multiple stages of all-pass filters are cascaded1, as shown in Figure 1a. Importantly, the capacity of optical delay line based upon multiple stages of all-pass filters is intrinsically limited by the physical dimensions of each stage. The quest for high capacity delay line provides a strong motivation for our implementation of all-pass filter concepts in photonic crystal resonators (PCR), since a PCR possesses the smallest footprint achievable using transparent dielectric materials.

For a general optical delay line structure illustrated in Figure 1a, its capacity is defined as the total number of pulses that the structure can accommodate without significant spatial overlap between different pulses. Given a spectral bandwidth δω, the shortest pulse has a Fourier-transform limited temporal width τ of approximately 2π/δω. Hence, the total number of pulses that a delay line can accommodate is determined by the total phase shift φ across the entire structure L over the signal bandwidth:

∫∫ =⋅===δωδω π

φπω

ωπδω

τ 222dd

dkdL

vL

vLN

gg

where vg is the group velocity at the carrier frequency. Since each optical resonance within the bandwidth generates a 2π phase shift, the capacity of an optical buffer is in fact limited by the number of resonance that one could fit within the physical length of the delay line as N=L/l, where l is the size of one resonance mode along the waveguide. Therefore, it is of critical importance to develop all-pass filter structures based on photonic crystal resonators, as such resonator possesses the smallest modal volume for a given index contrast.

In this paper, we specifically study a first-order optical all pass filter structure consisting of a two-dimensional photonic crystal waveguide and two side-coupled cavities, as the fundamental building block for optical delay lines. The all-pass characteristics is realized by enforcing an accidental degeneracy between two standing-wave cavity modes with opposite mirror symmetry in a resonator system2. The photonic crystal structure, as shown in Figure 1b, comprises of a square lattice of high-index dielectric rods with radius 0.2a and dielectric constant of 11.56, where a is the lattice constant. The waveguide is created by removing a single row of dielectric rods in the lattice, and the two cavities are created by reducing the radius of two defect rods. Each cavity supports a localized singly degenerate monopole mode for the TM modes, with the electric field component parallel to the

Nanophotonics


cavity cavity cavity

waveguide

(b)

(a)

Figure 1.A multi-stage all-pass filter consisting of a waveguide and side-coupled resonators. (b) Schematic of a first order all-pass filter in a two dimensional photonic crystal. The two single mode cavities are marked by red dots (ε=4.2, r=0.05a). The dielectric constants and the radii of the cylinders represented by the green rods (ε=4.2, r=0.05a) and the black circles (ε=4.2, r=0.05a) are chosen to equalize the direct and the indirect coupling.

cylinders3. The two cavity modes are evanescently coupled to each other. They are also coupled to the waveguide through the photonic crystal.

Using the temporal coupled mode theory4, the degeneracy of the resonances can be maintained2, when the cavity spacing is (n+1/4) times the propagating mode wavelength and the evanescent coupling through tunneling in the photonic crystal equals the indirect coupling through the waveguide. The frequency splitting effect of the mutual coupling in this case is exactly cancelled off by indirect coupling and the degeneracy is restored. Under this condition, the compound cavity modes can be decomposed as a forward decaying mode F and a backward decaying mode B, where the forward decaying mode F couples only with the forward traveling guided mode. Thus the forward traveling wave is completely de-coupled from the backward traveling wave and complete transmission is achieved both on resonance and off resonance. The structure behaves as an ideal first-order all-pass filter, where the transmission is given by:

T = − jj(ω − ω0) −

1τ

j(ω − ω0) +1τ

where ω0 is the resonance frequency; τ is the decay rate of the cavity mode due to the coupling to the waveguide,

In the photonic crystal structure as shown in Figure 1b, we accomplish the degeneracy condition, by modifying the dielectric constant and the size of four specific rods in the photonic crystal (black circles and green dots in Figure 1b), and by choosing the size and the dielectric constant of the defect rods in such a way that the cavity spacing is equal to 1.25λ. To demonstrate the all-pass filter characteristics numerically, we perform a two-dimensional finite-difference time-domain (FDTD) simulation. In the FDTD cell terminated by perfected matched boundary conditions, the system is excited with a temporal Gaussian source located on the left end of the waveguide. As can be seen in the simulated spectra exhibited in Figure 2, the structure indeed generates a strong group delay at 0.372 (c/a) with a quality factor Q=1135 and a maximum delay at 3764 (a/c), and the transmission exceeds 99.9% throughout the entire frequency bandwidth.

The dynamic behavior of the all-pass filter can be exhibited by analyzing the electric field patterns as the pulse propagates through the structure. A temporal Gaussian pulse is excited at the left end of the waveguide, travels along the waveguide and excites the cavity mode (Figure 3a). The pulse has a center frequency at 0.372 (c/a), which coincides with the resonant frequency of the cavities, and a temporal width of 2.69 (a/c), which is much shorter than the cavity lifetime. After the initial pulse has propagated through the resonator, the energy in the resonator continues to slowly decay into the

Nanophotonics


waveguide structure. Most importantly, we note that the decay occurs only in the forward direction (Figure 3.b). Such a uni-directional decay indicates that only the forward propagation cavity mode is excited, in consistency with the coupled mode theory presented above for the all-pass filter response.

The additive nature of the group delay allows us to cascade multiple resonances along the waveguide to obtain stronger delay. Consequently, additional cavity pairs downstream can have optimized resonance frequency and quality factors to synthesize a flattop delay spectrum1. For the carrier wavelength of 1550nm, an estimation of the maximum capacity of such optical delay line would be 192 bit/mm, where adjacent stages, separated by 9 lattice constant a, are placed on different side of the waveguide for good modal isolation.

As final remarks, we note that the scaling relations, as derived here for optical delay lines, also apply in other applications of cascaded all-pass filter structures, such as the syntheses of arbitrary infinite impulse response filters5, and on-chip soliton generation6. Thus, the implementation of photonic crystal micro-resonators should result in enhanced performance. Finally, it has been demonstrated that two-dimensional results can in principle be realized in realistic three dimensional photonic crystal structures7. Thus, high capacity optical delay line, therefore, can be created based on this all pass filter design which maintains a very low insertion loss and small footprint.

This work is supported in part by NSF under contract ECS-0200445. The authors acknowledge stimulating discussions with D. A. B. Miller.

References [1] G. Lenz, et al., IEEE J. Quant. Electron. 37, 525 (2001). [2] S. Fan, et al., Phys. Rev. Lett. 80, 960 (1998). [3] P.R. Villeneuve, et al., Phys. Rev. B, Condens Matter (USA) 54, 7837 (1996). [4] C. Manolatou, et al., IEEE J. Quant. Electron. 35, 1322 (1999). [5] C.K. Madsen, J. Lightwave Technol. 18, 860 (2000). [6] J.E. Heebner, et al., Phys. Rev. E, Stat. Nonlinear Soft Matter Phys. (USA) 65, 036619/1 (2002). [7] M.L. Povinelli, et al., Phys. Rev, B, Condens, Matter Mater. Phys. (USA) 64, 075313/1 (2001).

0.365 0.370 0.375 0.3800.0

0.5

1.0

Normalized Frequency (2πc/a)

Tran

smis

sion

0

1000

2000

3000

4000

5000

6000

Delay(a/c)

Figure 2. Transmission and group delay spectrum for the photonic crystal structure shown in Figure 1(b), as calculated by FDTD simulations.

Figure 3. Electrical field patterns as a pulse propagates through the photonic crystal structure shown in Fig. 1(b). Red and blue represent large positive or negative fields, respectively. The black circles indicate the position of the rods.

(a)

(b)

Nanophotonics


High-density Low-power All-optical Logic Gate in Photonic Crystals

Mehmet Fatih Yanik Department of Applied Physics, Stanford University, Stanford, California 94304

Shanhui Fan Department of Electrical Engineering, Stanford University, Stanford, California 94304

Marin Soljačić, J. D. Joannopoulos Department of Physics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139

Low-Power all-optical devices in photonic crystals could enable dense and large-scale integration of optical information processing components [1-5]. We demonstrate high-contrast all-optical switching in a nonlinear photonic crystal cross-waveguide geometry with instantaneous Kerr nonlinearity, in which the transmission of a signal is reversibly switched on and off by a low power control input. The signal and control inputs/outputs are spatially separated using the cross-waveguide geometry [6] shown in Figure 1.

Figure 1. Electric field distributions in a photonic crystal cross-waveguide switch (a) control input is present, signal output is high; (b) control input is absent, signal output is low. Red and blue represent large positive or negative electric fields, respectively. The same color scale is used for both panels. The black circles indicate the positions of the dielectric rods in the photonic crystal.

In the absence of the control, the signal exhibits bistable transmission behavior (blue curve in Figure 2a). Presence of the control input modulates the bistability transition threshold of the signal (red and green curves in Figure 2a), yielding high contrast transitions in the signal output. The structure thus behaves as an optical transistor.

Nanophotonics


Figure 2. (a) Input versus output power for the signal in waveguide X, calculated by keeping output power in waveguide Y (and hence the energy in cavity mode Y) at constant levels. Blue, red and green curves correspond to control output powers appropriate for various times in the switching process as shown in part (b). (b) The input and output power level for the signal and the control as a function of time. The lines are from theory calculations, and the open circles and triangles are from FDTD simulations. The labels A, B, and B’ indicate the control output power levels that were used to calculate the bistability curves of part (a).

The switching dynamics, as revealed by both coupled mode theory [7] and finite-difference time-domain simulations [8], exhibits collective behavior that can be exploited to generate high contrast and robust logic levels. Our theoretical model perfectly accounts for the entire switching dynamics as shown in Figure 2b.

Our geometry accomplishes both spatial and spectral separation between the signal and the control in the nonlinear regime as proven by our analytical model, and verified with FDTD simulations. Figure 3 shows the field spectrum at the output ports taken during the entire switching process, indicating complete signal and control isolation.

Figure 3. The output power spectra for the signal and control respectively, through the entire switching process shown in Figure 2a.

The device occupies a small footprint of a few micrometers square, and requires only a few milliWatts at 10Gbit/s switching rate using Kerr nonlinearity in AlGaAs below half the electronic band-gap [9,10]

Nanophotonics


for a three-dimensional structure operating at 1.55µm, with the optical mode confined in the third dimension to a width about half a wavelength, and cavity quality factor of Q=5000 achievable in Photonic Crystal Slabs [11]. The switching process is also robust to small fluctuations in the structure and power levels.

References [1] E. Centeno and D. Felbacq, Phys. Rev. B 62, R7683 (2000). [2] S. F. Mingaleev and Y. S. Kivshar, J. Opt. Soc. Am. B 19, 2241 (2002). [3] M. Soljacic, M. Ibanescu, S. G. Johnson, Y. Fink and J. D. Joannopoulos, Phys. Rev. E 66, 55601

(R) (2002). [4] M. Soljacic, C. Luo, J. D. Joannopoulos and S. Fan, Opt. Lett. 28, 637 (2003). [5] M. F. Yanik, S. Fan and M. Soljacic, Applied Physics Letters (submitted). [6] S. G. Johnson, C. Manolatou, S. Fan, P. R. Villeneuve, J. D. Joannopoulos and H. A. Haus, Opt.

Lett. 23, 1855 (1998). [7] H. A. Haus, Waves and Fields in Optoelectronics (Prentice-Hall, New Jersey, 1984). [8] A. Taflove and S. C. Hagness, Computational Electrodynamics (Artech House, Norwood MA,

2000). [9] M. N. Islam, C. E. Soccolich, R. E. Slusher, A. F. J. Levi, W. S. Hobson and M. G. Young, J.

Appl. Phys. 71, 1927 (1992). [10] A. Villeneuve, C. C. Yang, G. I. Stegeman, C. Lin and H. Lin, J. Appl. Phys. 62, 2465 (1993). [11] K. Srinivasan and O. Painter, Opt. Express 10, 670 (2002).

Telecom

SPRC 2002 – 2003 D-1 Annual Report

Research Program

D. Telecom

S. R. Bank, M. A. Wistey, H. B. Yuen, W. Ha, V. Gambin, K. Volz, and J. S. Harris, “1.5 Micron Lasers on Gallium Arsenide”...........................................................................................D-3

Hilmi Volkan Demir, Vijit A. Sabnis, Jun-Fei Zheng, Onur Fidaner, James S. Harris, Jr., and David A. B. Miller, “Novel scalable wavelength-converting crossbars” ......................................D-6

In-Sung Joe, Kyoungsik Yu, and Olav Solgaard, “Flexible, scalable 100Tbps Internet router with an optical switch fabric using optical tunable filters”........................................................D-9

A. Khalili, James S. Harris, “Side-Coupled Waveguide/Fiber Devices” ...........................................D-12

Jonathan R. Kurz, Jie Huang, Xiuping Xie, Martin M. Fejer, “Mode Multiplexing in Optical Frequency Mixers” ................................................................................................................D-16

Carsten Langrock, Rostislav V. Roussev, Jonathan R. Kurz, Martin M. Fejer, “Sum-frequency generation in a PPLN waveguide for efficient single-photon detection at communication wavelengths” ............................................................................................................D-19

Vijit A. Sabnis, Hilmi Volkan Demir, Jun-Fei Zheng, Onur Fidaner, James S. Harris, Jr., and David A. B. Miller, “Novel Optically-switched Electroabsorption Modulators for Wavelength Conversion” ...................................................................................................................D-22

R. Urata, L. Y. Nathawad, K. Ma, D. A. B. Miller, B. A. Wooley, and J. S. Harris, Jr., “Photonic A/D Conversion Using Low-Temperature-Grown GaAs MSM Switches Integrated with Si-CMOS”.................................................................................................................D-25

Telecom


Telecom


1.5 Micron Lasers on Gallium Arsenide

S. R. Bank, M. A. Wistey, H. B. Yuen, W. Ha, V. Gambin, K. Volz, and J. S. Harris Solid State and Photonics Laboratory, Stanford University

1. Introduction

Optical fiber is replacing copper wire in high speed computer networks - they can carry many times more data than comparable electrical cables and are immune to crosstalk and radio frequency interference. Wavelengths of light from 1.45-1.6 µm can travel tremendous distances—hundreds of kilometers—through fiber without significant loss, however, generating light at these wavelengths using conventional semiconductor lasers has been difficult. We have developed lasers using a new material, GaInNAs, which can generate wavelengths ideal for optical fiber communication on local area networks (LANs) and metro-area networks. Initially developed as a solution for applications at the fiber dispersion minimum ~1.3 µm, this GaAs-based material system has matured into a viable commercial product. These lasers utilize a smaller forward voltage and reduced bandgap, so they are also promising candidates for CMOS-compatible chip-to-chip interconnects. By adding antimony (Sb) to the alloy, we have demonstrated high performance edge-emitting lasers at 1.5 µm. These results demonstrate that GaAs is capable of covering the full range of fiber wavelengths with high performance.

There are two distinct classes of devices that are needed. The first is high speed (~10 Gbps) low power (~1 mW) devices, in the 1.45-1.6 µm region, for use as transmitters. These devices must be directly modulated and temperature insensitive, in terms of both wavelength and threshold current. The second class is continuous wave (CW) high-powered pump lasers for Raman amplification to replace/improve erbium-doped fiber amplifiers (EDFAs). Raman is a non-linear process and requires high power lasers (>300 mW) at slightly shorter wavelengths (1.40-151 µm) for efficient amplification. GaInNAsSb lasers on GaAs offer advantages over InP-based devices for both applications. GaAs-based lasers feature superior high-temperature performance as well as inherently higher gain allowing substantially higher output powers. The GaInNAsSb lasers produced in this work are grown by solid source molecular beam epitaxy (MBE), using an rf plasma to crack (inert) nitrogen gas into reactive nitrogen atoms.

2. Reaching Long Wavelengths

GaInNAs must be grown at low temperatures to avoid phase segregation which causes point defects including vacancies, interstitials, and arsenic antisites to permeate the material. Consequently, all samples require post-growth rapid thermal anneal at 700-820°C for 1-3 minutes, depending on the sample. Annealing removes these defects improving luminescence efficiency by an order of magnitude, but has the undesirable effect of shifting the emission to shorter wavelengths (we refer to this as the “blue shift” although the light is still in the infrared and not actually blue). The blue shift is attributed to two principal effects. First, nitrogen atoms change their nearest neighbors upon anneal. During the initial growth, nitrogen preferentially surrounds itself with gallium owing to the relative strengths of In-N bonds compared to Ga-N. In the bulk, however, the larger indium atoms induce

Telecom


strong compressive strain on the lattice. During anneal, the nitrogen atoms, which are small, tend to move toward the large indium atoms to minimize local strain. The second cause of blue shift on anneal is that some of the nitrogen diffuses out of the GaInNAs quantum wells into the surrounding GaAs. Our group was the first to use nitrogen-rich GaNAs barriers to minimize this effect. GaNAs has a smaller lattice constant than GaInNAs or GaAs, so it can also be used to compensate some or all of the compressive strain from the GaInNAsSb quantum wells. This ancillary advantage allows the growth of multiple quantum wells which improves gain in lasers. Using these improvements, we have demonstrated several lasers operating at long wavelengths. Our group was the first to demonstrate a long-wavelength VCSEL based on GaAs (Coldren 2000) which was both room temperature and continuous-wave. More recently, we have demonstrated 1.5 µm edge-emitting lasers with high output powers that show great promise for use in Raman amplifiers.

3. Antimony Incorporation Allows Longer Wavelengths and Superior Crystal Quality

We have recently discovered that adding antimony (Sb) to the alloy produces even more remarkable results. It had been generally accepted that antimony acted as a surfactant during MBE growth of GaAs-based materials - antimony is a very large atom, so it was believed to “ride” the surface during growth, rather than being incorporated into the crystal itself. Although antimony does act as a surfactant by increasing crystal quality and the rate of indium and nitrogen incorporation, we have also found it also incorporates into the crystal. This decreases the bandgap of the material even further, extending emission beyond that possible with antimony-free GaInNAs. Without antimony, it is extremely difficult to produce wavelengths longer than 1.30 µm: increasing the indium concentration even modestly to 35% degrades material quality, and so very little light is emitted (Figure 1, bottom). But small amounts of antimony restore crystal quality, and also increase the amount of indium possible, thereby extending the wavelength of emission (longer wavelength peaks in Figure 1).

Figure 1: Photoluminescence spectra of several potential active layers; higher intensity indicates superior luminescence efficiency. The introduction of antimony improves material quality by approximately three orders (purple to orange curve). Additionally, with more indium and antimony high material quality is maintained throughout the telecom bands.

0.75 0.80 0.85 0.90 0.95 1.00

10-3

10-2

10-1

100

1600 1500 1400 1300 1200

35% InGaInNAs

31% In GaInNAs




Inte

nsity

(a.u

.)

Energy(eV)

Wavelength (nm)

Telecom


Based on these results, edge-emitting lasers containing similar active regions were grown and fabricated. Figure 2 shows the CW spectrum and light versus current (L-I) curve from edge-emitting lasers, which are by far the highest performance GaAs-based devices ever reported in this wavelength range. These devices lased at 1.498 µm and exhibited CW threshold current densities of 1.06 kA/cm2 with maximum CW output power of 140 mW (both facets). These results are for as-cleaved devices (without facet coatings) mounted epitaxial-side up on a copper heatsink. Additionally the devices are substantially less temperature sensitive than InP devices owing to their much improved characteristic temperature, T0. The CW T0 of these devices is 101 K, compared with ~50-80 K for typical of InP lasers. Moreover, these lasers have a relatively low diode turn-on voltage ~1.5 V. Through improvements in electrical properties of the device, it should be possible to reduce the turn-on to <1 V allowing these lasers to be driven directly with CMOS. We are currently developing electrically-pumped VCSELs in this material system, and hope to have results in the near future.

Figure 2: (a) CW L-I and (b) spectra of an as-cleaved 20 µm x 2450 µm edge-emitting device.

4. Conclusion

We have developed edge-emitting lasers, grown on relatively inexpensive GaAs substrates, which emit light at longer wavelengths than other GaAs-based lasers. These lasers show the viability of high-power GaInNAsSb lasers suitable for Raman amplifier pumping near 1.5 µm. These results have been keyed by the addition of antimony to GaInNAs which allows greater indium and nitrogen incorporation to reduce the bandgap, improves crystal quality, and itself incorporates to further reduce the bandgap. The results to date are stepping stones toward the ultimate goals of low-cost, CW VCSELs and high power edge-emitting lasers. We believe these lasers will lay the foundation for low-cost, high-speed metro area optical networks and will enable low-voltage interconnect applications due to the reduced bandgap.

1495 1500

0.0

0.2

0.4

0.6

0.8

1.0

Out

put P

ower

(A.U

.)

Wavelength (nm)0 250 500 750 1000 1250 1500

0

20

40

60

80

100

120

140

CW

Out

put P

ower

(mW

)

Current (mA)

20 µm x 2450 µm

(a) (b)

1495 1500

0.0

0.2

0.4

0.6

0.8

1.0

Out

put P

ower

(A.U

.)

Wavelength (nm)0 250 500 750 1000 1250 1500

0

20

40

60

80

100

120

140

CW

Out

put P

ower

(mW

)

Current (mA)

20 µm x 2450 µm

(a) (b)

Telecom


Novel scalable wavelength-converting crossbars

Hilmi Volkan Demir, Vijit A. Sabnis, Jun-Fei Zheng, Onur Fidaner, James S. Harris, Jr., and David A. B. Miller Department of Electrical Engineering, Stanford University

All-optical wavelength conversion has been extensively investigated, and different wavelength converters have been demonstrated with promising results.1 These wavelength converters were, however, not suitable for conveniently scaling them into two-dimensional arrays to realize chip-scale reconfigurable multi-channel wavelength conversion systems. For instance, a wavelength-converting optical crossbar switch that allows for arbitrary mapping of the input wavelengths to the output wavelengths could not be made. To address this problem, we introduce a novel, scalable, wavelength-converting, optical crossbar switch that consists of a two-dimensional array of optically-controlled optoelectronic switches placed at its cross nodes.

Figure 1: A plan view microscope picture of a 2x2 wavelength-converting crossbar

Figure 1 shows a plan view of a fabricated 2x2 wavelength-converting crossbar with a total of four optoelectronic switch elements. Each switch in the crossbar includes an InGaAsP quantum well waveguide modulator and an InGaAs surface-normal photodiode as a part of its optoelectronic circuit on the same InP susbtrate.2 The photodiodes and modulators of the switches are monolithically integrated into a compact circuit area (few hundreds of microns in length and width), using a selective area regrowth technique.3 The function of these switches is to transfer optical signals from their input data streams incident on their photodiodes onto the continuous-wave beams coupled into their

cw beam 1

cw beam 2

@λΟ1

@λΟ2

output 1@λΟ1

output 2@λΟ2

input 1@λΙ1 input 2@λΙ2

λΙ1 → λΟ2

λΙ2 → λΟ1

1mm

0.6mm

-+Von+Von

-+Von+Von

-+Voff+Voff

-+Voff+Voff

Telecom


modulator waveguides when their photodiodes and modulators are properly reverse-biased. The photodiodes will not extract photocurrent and the modulators will be transparent if they are slightly forward biased. Each of these switches located at the cross nodes can, therefore, be independently enabled or disabled electrically to reconfigure the crossbar as desired by simply changing their biasing accordingly. The individual optically-controlled optoelectronic switches provide unconstrained, full C-band wavelength conversion, because the InGaAs photodiodes detect over the entire telecommunication wavelength range and the InGaAsP quantum well modulators are designed to modulate optical transmission over the C-band with appropriate reverse biasing. Our preliminary experimental results demonstrate proof-of-concept wavelength conversion up to 2.5 Gb/s NRZ and RZ operation with >10dB RF extinction ratios. 4

The 2x2 array depicted in Figure 1 includes two side-by-side waveguides, each incorporating two switches. For crossbar operation, with the use of a diffractive spot generator or fiber splitters, the optical input 1 at λI1 is optically distributed to the two leftmost photodiodes, each connecting to a different waveguide, and similarly, the optical input 2 at λI2 is distributed to the rightmost photodiodes. Two continuous-wave beams at λO1 and λO2 are also separately coupled into the two waveguides. By electrically enabling the bottom left wavelength converter and the top right one, and disabling the rest of them, the crossbar is configured to transfer optical information from input 1 to output 2 and from input 2 to output 1, with selected wavelength mapping of λI1 λO2 and λI2 λO1. Similarly, the complimentary mapping of λI1 λO1 and λI2 λO2 is possible through electrically enabling the top left wavelength converter and the bottom right one, and disabling the rest.

Here, we report proof-of-principle, full C-band wavelength conversion results from a 2x2 array with four switch elements. Figures 2Ai through 2Aiii show three eye diagrams taken from one of the switch elements, A, at the output wavelengths of (i) 1530.0 nm, (ii) 1550.0 nm, and (iii) 1565.0 nm, respectively, at 1.25Gb/s using an optically-preamplified receiver. Similarly, Figures 2Bi through 2Biii, Figures 2Ci through 2Ciii, and Figures 2Di through 2Diii present the other sets of three different output wavelengths for the other switch elements B, C and D. In all cases, the input wavelength was 1551.7 nm and the input optical power was <8mW. While testing one of the wavelength converters, the other one along the same waveguide was disabled through slightly forward-biasing its photodiode and modulator to remove its background quantum well absorption and the potential crosstalk between the two input channels. In an nxn crossbar operation, since each input channel will need to be distributed over n photodiodes, each input channel will suffer 1/n power loss. Despite this optical distribution loss, the mW-level input optical power requirement of the individual wavelength converters makes large-size crossbars feasible. For very large n, however, 1/n optical input loss will limit the array size, n.

In summary, we present a novel, 2x2, low-power, compact, electrically-configurable, wavelength-converting crossbar. Our experimental results demonstrate full C-band wavelength coverage. The preliminary characterization leads us to anticipate that it is possible to operate large-size wavelength-converting crossbars at 10 Gb/s and beyond, by reducing the on-chip resistor and device capacitances of the individual switches appropriately, optimizing their modulator quantum wells, and integrating them more tightly.

Telecom


Figure 2: Eye diagrams of the four switch elements (A, B, C, and D) of an 2x2 array at three exemplary wavelengths (i) 1530.0 nm, (ii) 1550.0 nm and (iii) 1565.0 nm covering the entire C-band.

Acknowledgements:

We would like to acknowledge Intel Corporation for funding our research, OEPIC Inc. for providing epitaxial wafer growth service, and Melles Griot for supplying waveguide alignment system. HVD also acknowledges Intel for supporting his research assistantship.

References: [1] S. J. B. Yoo, IEEE J. Light. Tech., 14 (6), pp. 955-966, (1996). [2] H.V. Demir, V. A. Sabnis, O. Fidaner, S. Latif, J. S. Harris, Jr., D. A. B. Miller, J-F. Zheng, N. Li,

T-C. Wu, and Y-M. Houng, Conference on Lasers and Electro-optics (LEOS), Tuscon, AZ, 2003 (paper WU1).

[3] V. A. Sabnis, H.V. Demir, O. Fidaner, J.S. Harris, Jr., D. A. B. Miller, J-F. Zheng, N. Li, T-C. Wu, and Y-M. Houng, Integrated Photonics Research Meeting (IPR), OSA Technical Digest, pp. 12-14, Washington, DC, 2003 (paper IMB3).

[4] V. A. Sabnis, H.V. Demir, J-F. Zheng, O. Fidaner, J.S. Harris, Jr., and D. A. B. Miller, Technical Report, Stanford Photonics Research Center Annual Meeting, Stanford, CA, 2003.

(Ai) (Aii) (Aiii)

(Bi) (Bii) (Biii)

(Ci) (Cii) (Ciii)

(Di) (Dii) (Diii)

Telecom


Flexible, scalable 100Tbps Internet router with an optical switch fabric using optical tunable filters

In-Sung Joe, Kyoungsik Yu, and Olav Solgaard Department of Electrical Engineering, Stanford University

As Internet traffic approximately doubles every year [1], we need fast routers. Larger routers are desirable at point-of-presence nodes (POPs), because the required number of interconnections within a POP decreases. The Stanford optical router (OR) project investigates how to incorporate advanced optics into routers to improve capacity with less power consumption. All-optical switches are not feasible, because routers require per-packet address lookup and large buffers to resolve congestion [2]. In addition, efficient, large capacity, and fast accessible optical memories or optical buffers are not yet available. Therefore, scalable 100Tbps Internet router designs are proposed with linecards implemented in electronics, and a switch fabric utilizing optics [3]. The Stanford OR project also adopted a two-stage, load-balanced switch design [4]. This two-stage switch is scalable, and it provides 100% throughput. Most importantly, it does not require a central scheduler, which simplifies the switch fabric by allowing the traditionally frequently scheduled and reconfigured switch to be replaced by two identical switches that follow a fixed switching sequence [3].

Linecard 1

Linecard 2

Linecard 3

Linecard 4

Linecard 5

Linecard 6

Linecard 7

Linecard 8

Linecard 1 3

Linecard 14

Linecard 15

Linecard 16

λ 1 λ 2 λ 3 λ 4

4X4 Star

Coupler

4X4 Star

Coupler

4X4 Star

Coupler

Linecard 1

Linecard 2

Linecard 3

Linecard 4

Linecard 5

Linecard 6

Linecard 7

Linecard 8

Linecard 1 3

Linecard 14

Linecard 15

Linecard 16

Wavelength Legend λ 2 + λ 4+

4X4

WGR

Telecom


Figure 1. Schematic of the optical switch fabric. (16X16 case) Linecard 15,16 covers failed linecards 13, 14.

Figure 1 shows the schematic of the switch fabric. It is composed of tunable lasers, tunable filters, waveguide grating routers (WGR), and star couplers to perform fixed, round-robin switching. Each linecard has two tunable lasers with different wavelength ranges; a primary laser that covers one free spectral range (FSR) of the WGR (from λ1 to λN), and a secondary laser that covers the next FSR of the WGR (From λ1+ to λN+).

Input linecards are sending data packets using different wavelengths, and the WGRs rout these packets to different output ports according to their wavelength. Star couplers, connected to WGRs, are broadcasting collected data packets carried by different wavelengths to output linecards, and tunable filters at output linecards are selecting data packets. Since each input linecard and output linecard has a connection, we can achieve a full connection between input linecards and output linecards by tuning the lasers and filters m times, where m is the number of linecards present. If the input linecards and output linecards are physically the same, a two-stage switch can be implemented with a single switch by repeating a full round robin twice. This switch architecture enables optical packet switching because tunable lasers with less than 45ns tuning time [5], as well as tunable filters using heterodyne detection with a tunable laser as a local oscillator have been demonstrated [6].

This switch fabric is scalable by adding or removing linecards, and it fully functions as long as the half of the linecards in each group are working. It can also be scaled down by removing all the linecards from the same input port of each WGR. Therefore, if the number of input ports of WGRs and star couplers is N, the switch fabric can support from N to N2 linecards. For example, two linecards connected to the first and the second input port of each WGR is sufficient for a fully functioning switch fabric. In addition, when all N2 linecards are present as shown at Fig. 1, and linecards 13 and 14 have failed, linecards 15 and 16 can cover by transmitting a half of the incoming data to a primary tunable laser (covers wavelength from λ1 to λN), and the other half to a secondary tunable laser (covers wavelength from λ1+ to λN+) to allow the switch fabric to fully operate without linecard 13 and 14. Secondary tunable filters on linecards 15 and 16 allow these to cover for the failed linecards 13 and 14. Secondary tunable lasers and filters are functioning only when linecards are missing. This provides flexibility and prevents the shut down the whole switch in case of linecard failures.

To demonstrate the switch fabric design, we are building an experimental setup in our lab. We will show the switching function of an optical tunable filter using heterodyne detection. Figure 2 shows the schematic of the experimental setup.

Freq info

RF Mixer

DFB λ1

DFB λ2

MZ Mod

MZ Mod

Coupler

LO Polarization Controller

Coupler

Photodetector

ESA FG

LPF

Oscilloscope

Telecom


Figure 2. Schematic of the heterodyne detection experimental setup

Distributed feedback (DFB) lasers in butterfly packages are used as input signals, and they are modulated by Mach-Zehnder (MZ) external modulators. An external cavity tunable laser is used as a local oscillator (LO), but this will be replaced by a fast tuning sampled grating distributed Bragg reflector (SG-DBR) laser in the near future. One of the input signals is selected by tuning the wavelength of the LO. The difference in wavelength between the selected DFB and the LO is detected by the electrical spectrum analyzer (ESA), and the RF function generator (FG) provides this frequency difference to the RF mixer. The RF mixer combines the generated intermediate frequency (IF) with the output of the photodetector. A RF low pass filter (LPF) is used to remove unwanted high frequency components, and the signal carried by the selected DFB laser is shown on the oscilloscope.

Once the concept is demonstrated, we will replace the external cavity laser with a fast tuning SG-DBR laser with tuning time of less than 50ns, and a RF phase locked loop (PLL) and a voltage controlled oscillator (VCO) will be added for better RF performance. Total tuning time of the tunable filter will be dominated by the tuning time of the LO, which is fast enough for packets switching in the Stanford 100Tbps Internet router switch fabric.

References [1] A. M. Odlyzko, “Internet traffic growth: Sources and implications,” in ITCOM, Proc. SPIE, to

appear (2003) [2] S. Chuang, I. Keslassy, and N. Mckeown, “Optics in Routers: Overview,” in C2S2 annual review,

poster session (2003) [3] I. Keslassy, Shang-Tse Chuang, Kyoungsik Yu, David Miller, Mark Horowitz, Olav Solgaard, and

Nick Mckeown, “Scaling Internet routers using optics,” in SIGCOMM, proc. ACM, to appear (2003)

[4] C.-S. Chang, D.-S. Lee and Y.-S. Jou, “Load balanced Birkhoff-von Nuemann switches, Part I: one-stage buffering,” Computer Communications, Vol. 25, pp.611-622, 2002

[5] J.E. Simsarian, A. Bhardwaj, J. Gripp, K. Sherman, Su Yikai, C. Webb, Zhang Liming, M. Zirngibl, “Fast switching characteristics of a widely tunable laser transmitter,” Photonics Technology Letters, IEEE, Volume: 15 Issue: 8, pp.1038-1040, Aug. 2003

[6] K.Y. Eng, M.A. Santoro, T.L. Koch, J. Stone, W.W. Snell, “Star-coupler-based optical cross-connect switch experiments with tunable receivers,” Selected Areas in Communications, IEEE Journal on , Volume: 8 Issue: 6, pp.1026-1031, Aug 1990

Telecom


Side-Coupled Waveguide/Fiber Devices

A. Khalili, James S. Harris. Department of Electrical Engineering, Stanford University

Side-coupled devices can be used in a telecommunication link for a variety of different applications, such as emission, modulation and detection of the optical signal. As depicted in figure 1, a simple transmit/receive WDM network can be realized once the individual functionalities are implemented. The major benefits associated with this approach are the compactness, simplicity, possibility of integration and ease of packaging. The network in figure 1 consists of side-coupled lasers, modulators and WDM notch filter/detectors that together with the AWG multiplexer, conceptualize a simple data transmission system.

Figure 1. Conceptual side-coupled waveguide-based data-transmission system.

Fig. 2a shows the structural layout of an arbitrary element (e.g. the modulator) in figure 1. The device operates when the light in the waveguide reaches the vicinity of the coupling region, where the evanescent fields’ overlap causes the electromagnetic energy at the phase-matched wavelength –which can be designed to be highly narrowband- to couple to the top Anti-Resonant Reflective Optical Waveguide (ARROW)[1]. An ARROW is a partly lossy waveguide that can reach a very low effective propagation index in a high-index medium like semiconductors and it is usually realized in a VCSEL-like resonant cavity architecture; except that the cavity is resonant at an oblique angle rather than normal incidence.

An alternative approach to the fabrication of side-coupled devices is to use fibers instead of planar waveguides as the main propagation media, as depicted in figure 2b. All aforementioned functionalities can be realized here in a similar manner. These in-line fiber devices, in which light is evanescently coupled between the fiber and the electrooptic waveguide, are superior to the conventional fiber-coupled devices, which involve the interruption of the fiber and the insertion of the device; mainly since the latter approach comes with several drawbacks including high insertion-losses

Telecom


and high alignment/packaging costs. Although liquid crystals[2], electrooptic polymers[3] and Lithium Niobate[4] substrates have been previously used for this purpose, GaAs and other semiconductor-based substrates (e.g. in an ARROW architecture) seem advantageous, for the possibility of integration with other electrooptic devices and electronic circuits.

Figure 2 a)A coupled waveguide-ARROW structure, b)Architecture of a typical in-line fiber-coupled ARROW.

The modulator[1] made previously at this group using this method (figure 2b) showed a transmission change from 10% to 40% with a 9V swing (the low breakdown voltage of the device imposed the limit on the tuning range) in a more or less linear fashion, which is especially attractive for analog applications. The modulation on/off ratio can be improved significantly by increasing the interaction length. The useful linewidth of the device is around 1nm and it can be made smaller by increasing the cavity thickness. The device size was about 1mm*5um*0.4um, which translates to an RC bandwidth of about 10GHz (lower in practice due to parasitics). Although this modulator was designed to operate at 850nm, it can be easily redesigned for 1550nm, using GaInNAs or InP as active materials. In general, adjusting the center wavelength for the modulator or the detector is possible by precise control over the ARROW core thickness. Although such high precision requires state of the art monitoring techniques, it’s also possible to tune the detectors by adjusting the reverse bias voltage or by tuning the temperature.

Design and fabrication of an ARROW laser is the main course of research at the present time. This composite-cavity laser design would greatly reduce the alignment and packaging complexity of fiber-coupled semiconductor laser systems. As opposed to conventional butt-coupled edge-emitter/fiber packages, this novel device uses the evanescent light to efficiently couple light between the fiber and the semiconductor waveguide. Here, an ARROW is fabricated on an MBE-grown GaAs wafer and is designed to resonate with the optical mode of the fiber at the operating wavelength. The coupling mechanism provides an inherent wavelength filter that would allow single-mode operation of the laser, without any Bragg grating. As depicted in figure 3, part of the laser cavity and one of the laser mirrors are located in the fiber, while the active region and the other mirror are realized in the semiconductor.

Telecom


Figure 3. Schematics of an in-line fiber ARROW laser

Figure 4 shows the electric field profile of the ARROW mode, which is phase-matched to the mode of an 830nm single-mode fiber. The effective index of this mode is 1.45, which is close to that of the fiber.

Figure 4.The field profile of the ARROW designed for the in-line laser at 830nm. Horizontal lines indicate the borders of high/low-index layers of the DBRs and the cavity. This design has 5 top and 30 bottom DBR pairs. The y-distribution of the electric fieid is shown on the right. An SEM picture of a grown sample is also shown.

Figure 5 depicts the process flow for fabricating these ARROWs, following the initial MBE growth of the aforementioned epitaxial structure. The fabrication and testing of these devices are underway.

Unpumped Region

This evanescent tail penetrates inside the fiber and would cause the coupling

Telecom


Figure 5. Fabrication process flow for in-line ARROW lasers

References

[1] E. Mao, PhD thesis, Stanford University (2000) [2] Z.K. Ioannidis, Appl. Opt. 30, 328 (1991) [3] G. Fawcett, Electron. Lett. 28, 985 (1992) [4] S. Creaney, IEEE photonics technol. Lett. 8, 355 (1996)

Etch Larger Mesas Oxidation / Etch ARROW Mesas

Deposit contact/Interconnect Layer Deposit Si3N4 Passivation Layer

Telecom


Mode Multiplexing in Optical Frequency Mixers

Jonathan R. Kurz, Jie Huang, Xiuping Xie, Martin M. Fejer E. L. Ginzton Laboratory, Stanford University.

Optical-frequency (OF) mixers based on periodically-poled lithium niobate (PPLN) waveguides have demonstrated many of the functions needed for future optical networks including wavelength conversion [1], dispersion compensation by spectral inversion [2], and optical time-division-de-multiplexing [3]. Mode multiplexing will extend the functionality of OF mixers even further, providing new solutions to longstanding problems such as spatial separation of the nonlinear mixer output from the residual input. The recent development of high-contrast mode converters and sorters enables bi-directional wavelength conversion and spectral inversion without offset. We demonstrate a proof-of-principle odd/even mode wavelength converter that suppresses the residual input waves by 30 dB without sacrificing mixing efficiency.

In a standard OF mixer all the interacting waves propagate in the (lowest order) 00-mode, and the wavelength-converted output, E3, can be separated from the pump and residual input (E2, E1) by spectral filtering (top of Fig. 1). In bi-directional wavelength conversion or spectral inversion without offset, however, the input and output wavelengths are identical, making them impossible to distinguish, let alone spatially separate. A Type-II phasematching scheme combined with a polarizing beam splitter can solve this problem when two polarizations are available (which is not the case for highly efficiency proton exchanged waveguides), although this approach is generally less efficient and flexible than quasi-phasematching (QPM). An alternative (bottom of Fig. 1) is to use higher order waveguide modes to distinguish and spatially separate the interacting waves. If the mixer output E3 is produced in a 10-mode, it can be separated using integrated optics structures. A year ago we showed that odd and even waveguide modes can be mixed efficiently using angled and staggered QPM gratings [4]. Here we demonstrate that asymmetric Y-junctions can be used as high contrast mode launchers and filters, enabling an odd/even mode wavelength converter.

The mode-converting properties of asymmetric Y-junctions have been known for decades [5,6] and have recently been used in high-contrast switches. Launching into the narrow port of an asymmetric Y-junction converts a 00-mode into a 10-mode, while launching into the wider port leaves the mode unchanged (Fig. 2). Viewed in the opposite propagation direction, the junction acts as a 00/10-mode sorter. This sorting behavior occurs when the branches join gradually enough that their modes evolve adiabatically, remaining local normal modes throughout the structure. We have fabricated asymmetric Y-junctions that sort TM00 and TM10 modes with over 30 dB of contrast. The smooth refractive index profiles formed by annealed proton exchange seem well suited for creating adiabatic designs. Calculations using local-normal mode theory and BPM simulations have guided these designs.

One challenge faced in characterizing asymmetric Y-junction designs is measuring the mode content (TM00 versus TM10) of a waveguide with a resolution better than 0.1%. Simply imaging the modal interference at the waveguide output may not provide such precise information without some knowledge of the mode shapes or their relative phase. We have developed a sensitive technique for measuring waveguide mode content at the first-harmonic (FH) wavelength using quasi-phasematched

Telecom


nonlinear mixing between the modes, which we label FH00 and FH10. With the proper waveguide design, we can produce three second-harmonic generation / sum-frequency generation (SHG/SFG) products, (FH00)2, (FH10)2, and (FH00*FH10), with nearly the same QPM period. Furthermore, we can arrange so that output the second-harmonic modes are different in each process, (SH00, SH20, SH10, respectively) making them easily distinguishable. Measuring the relative power produced by the three mixing processes yields an accurate measurement of the ratio FH00/FH10.

Using an asymmetric Y-junction design with 30-dB of contrast, we fabricated the odd/even mode OF mixer shown at the bottom of Fig. 1. In the experiment shown in Fig. 3, a signal at 1557.0 nm was launched into a TM10 mode (through the odd input port), producing a wavelength-shifted output in a TM00 mode at 1555.5 nm by cascaded χ2 mixing. In this proof-of-principle device, the conversion efficiency was only –18.8 dB because of shortened QPM gratings; standard devices often operate with 0-dB conversion efficiency. Measurements of the two output ports confirm that the residual signal and the mixed output have been spatially separated to the expected degree. The even output port (black trace) contains most of the converted signal, while the odd output port (gray trace) contains most of the residual input signal. Due to the relatively low conversion efficiency, the contrast between mixed signal and residual input at the even port is only –12.5 dB; with 0-dB conversion efficiency, however, this contrast would be greater than 30 dB. This experiment contains another variation on the standard cascaded OF mixer device. Instead of producing the SH needed for difference-frequency mixing by SHG of a 1.5-µm pump, this device uses two 1.5-µm pumps and SFG. This change allows shifting of the pump wavelengths to the outer edges of the conversion band where they can be spectrally filtered; typically 30-dB of filtering would not be sufficient to prevent pump cross-talk. With this odd/even mode mixing configuration, and a more efficient device, it will be possible to demonstrate spectral inversion without offset.

Figure 1: Ways of distinguishing and spatially separating the interacting waves in a standard waveguide QPM device (top) and by using higher-order waveguide modes (bottom).

Telecom


Figure 2: Mode sorting behavior in an asymmetric Y-junction. The modes of the larger / smaller input branches, which have higher / lower effective indices, evolve into the first / second modes of the combined junction.

Figure 3: Measured power vs. wavelength for an odd/even mode DFG device. The odd port (gray) contains most of the residual signal (1557.0 nm) while the even port (black) contains 12.5dB more mixer output (1555.8 nm) than residual signal. The pump transmission shows the full device contrast to be greater than 30 dB.

References [1] J. Yamawaku, H. Takara, T. Ohara, K. Sato, A. Takada, T. Morioka, O. Tadanaga, H. Miyazawa,

and M. Asobe, Electron. Lett. 39, 1144 (2003). [2] M. H. Chou, I. Brener, G. Lenz, R. Scotti, E. E. Chaban, J. Shmulovich, D. Philen, S. Kosinski, K.

R. Paramewaran, and M. M. Fejer, Opt. Lett. 26, 1285 (2001). [3] S. Kawanishi, M. H. Chou, K. Fujiura, M. M. Fejer, and T. Morioka, Electron. Lett. 36, 1568

(2000). [4] J. R. Kurz, X. P. Xie, and M. M. Fejer, Opt. Lett. 27, 1445 (2002). [5] H. Yajima, Appl. Phys. Lett., 22, 647 (1973). [6] W. K. Burns and A. Fenner Milton, J. Quant. Elec., QE-11, 32 (1975).

Telecom


Sum-frequency generation in a PPLN waveguide for efficient single-photon detection at communication wavelengths

Carsten Langrock, Rostislav V. Roussev, Jonathan R. Kurz, Martin M. Fejer Edward L. Ginzton Laboratory, Stanford University

Single-photon detection at wavelengths in the fiber-optic communications band1,2 is important for quantum optics applications, such as quantum cryptography. Current detection technologies operating at these wavelengths cannot deliver the performance needed to implement, for example, quantum key distribution as required by the BB84 quantum coding scheme3, over distances longer than a few tens of kilometers due to limitations imposed by high dark counts and low detection efficiencies. Two of the more prominent technologies currently used are InGaAs/InP avalanche photodiodes (APDs) and solid-state photomultipliers. Both suffer from low detection efficiencies4 (< 16% at 1.3 µm and < 7% at 1.55 µm) and high dark counts (> 104-105/s). In contrast, single-photon detection in the near-infrared (NIR) (600-800 nm) can be performed very efficiently using silicon APDs. Single-photon counting modules (SPCMs) with detection efficiencies in the range 50-70% with dark counts below 25/s are commercially available. Improved detection has been reported with quantum efficiencies in excess of 75% at 700 nm 5. One can take advantage of these detectors if efficient conversion from 1.55 µm to the NIR is available.

Single-photon counting via frequency up-conversion in bulk PPLN has been recently demonstrated6. This detection scheme relied on the ability to efficiently convert photons from the C-band to 630 nm in a ring cavity, where they were detected by an SPCM. The overall detection efficiency presented was 55% at 20 W of circulating pump power. We report here a single-pass non-resonant waveguide SFG device with comparable detection efficiency, and convenient connectivity to fiber-based systems by pigtailing a fiber to the input of the chip.

Frequency up-conversion of weak signals is possible by sum-frequency generation (SFG). SFG in periodically poled LiNbO3 (PPLN) waveguides has been used to demonstrate low-power all-optical gates7 and optical sampling systems8.

It has been shown that nonlinear interactions in PPLN can be implemented very efficiently using quasi-phasematching (QPM) by means of periodic sign reversal of the nonlinear coefficient9. For

Figure 1. Experimental Setup

Telecom


efficient conversion in nonlinear optical frequency mixing, high field intensities and long interaction lengths are necessary, both of which can be achieved simultaneously in a guided-wave structure10,11. The largest normalized second harmonic generation efficiency reported to date has been demonstrated in reverse-proton-exchanged (RPE) PPLN waveguides12.

Our experimental setup for wavelength conversion is shown in Fig.1. An external cavity tunable diode laser (ECDL) at 1551 nm followed by an erbium-doped fiber amplifier (EDFA) is used as the pump source. The signal at 1340 nm is provided by a second ECDL. Note that the same device will work equally well with the roles of pump and signal reversed, which we will demonstrate in a future experiment. A dichroic mirror and a prism at the output of the chip are used to separate the SFG, the signal, and the pump. A Newport 1830C power meter was used to measure the pump power after the chip and to calibrate the two optical spectrum analyzers, which were used to detect the residual signal and the SFG.

The wavelength conversion was performed in an RPE PPLN waveguide. The normalized internal SFG efficiency was ηnor= 330+/-10%/W-cm2. The poling duty-cycle was 37+/-2%, reducing the efficiency by 10-20% compared to the maximum achievable QPM efficiency. Signal depletion ηNL exceeding 99% was observed at 88mW of pump power coupled into the waveguide (Fig. 2). The overall internal quantum efficiency (QE) of our device is approximately 94% and is currently only limited by the propagation losses at the signal

and sum wavelength mentioned above. Taking the free-space to waveguide coupling efficiency of the signal into account as well as Fresnel reflections off of the uncoated end facets of the device, we arrive at an external QE of 63%. The reduction in QE is mainly caused by Fresnel reflections for the signal at the input and the sum-frequency at the output of the chip, which each amount to 13%. Anti-reflection coating the end facets will increase this number significantly. Free-space to waveguide mode-matching losses of 0.5 dB have been achieved, which could be further reduced by improvements in the waveguide design and fabrication. Prior to these improvements, the overall detection efficiency of our device is currently 41%, assuming a silicon APD with a QE of 75%. Eliminating the Fresnel reflections off of the end facets will increase the detection efficiency up to 53%. Refinements in the waveguide design and processing will result in an increased coupling efficiency as well as lower

Figure 2. Experimental and theoretical results obtained by numerical integration of coupled-mode equations

Telecom


propagation losses (~0.1 dB/cm), which will lead to overall detection efficiencies exceeding 60%. Optimizing the poling duty-cycle will reduce the pump power requirements by up to 20%.

References [1] “Single-Photon Counter for long-haul Telecom”, Anders Karlsson, Mohamed Bourennane,

Gregoire Ribordy, Hugo Zbinden, Juergen Brendel, John Rarity, and Paul Tapster; IEEE Circuits and Devices Magazine, November 1999, No. 6, Vol. 15, pp. 34-40

[2] “Single-photon counters in the telecom wavelength region of 1550 nm for quantum information processing”, Mohamed Bourennane, Anders Karlsson, Juan Pena Ciscar, Markus Mathes; Journal of Modern Optics, 2001, No. 13, Vol. 48, pp. 1983-1995

[3] “Quantum Cryptography: Public Key Distribution and Coin Tossing”, C.H. Bennett and G. Brassard; IEEE International Conference on Computers Systems and Signal Processing, December 1984, pp. 175-179

[4] “Performance of InGaAs/InP avalanche photodiodes as gated-mode photon counters”, G. Ribordy, J.-D. Gautier, H. Zbinden, and N. Gisin; Applied Optics, 1998, No. 12, Vol. 37, pp. 2272-2277

[5] “Absolute efficiency and time-response measurement of single-photon detectors”, P. G. Kwiat, A. M. Steinberg, R. Y. Chiao, P. H. Eberhard, and M. D. Petroff, Applied Optics, April 1994, No. 10, Vol. 33, pp. 1844-1853

[6] “Efficient single-photon counting at 1.55 µm via frequency upconversion”, M. A. Albota and F. N. C. Wong, CLEO/QUELS 2003

[7] “Low-power all-optical gate based on sum frequency mixing in APE waveguides in PPLN”, K. R. Parameswaran, M. Fujimura, M. H. Chou, and M. M. Fejer; IEEE Photon. Technol. Lett., June 2000, Vol. 12, pp. 656-659

[8] “High sensitivity waveform measurement with optical sampling using quasi-phasematched mixing in LiNbO3 waveguide”, S. Kawanishi, T. Yamamoto, M. Nakazawa, and M. M. Fejer; Electronics Letters, June 2001, Vol. 37, pp. 842-844

[9] “Second Harmonic Generation of Green Light in a Periodically-Poled Planar Lithium Niobate Waveguide”, E. J. Lim, M. M. Fejer, and R. L. Byer; Electronics Letters, February1989, Vol. 25, pp. 174-175

[10] “Blue Light Generation by Frequency Doubling in a Periodically Poled Lithium Niobate Channel Waveguide”, E. J. Lim, M. M. Fejer, R. L. Byer, and W. J. Kozlovsky; Electronics Letters, May 1989, Vol. 25, pp. 731-732

[11] “Infrared Radiation Generated by Quasi-Phase-Matched Difference Frequency Mixing in a Periodically-Poled Lithium Niobate Waveguide”, E. J. Lim, M. L. Bortz, and M. M. Fejer; Applied Physics Letters, October 1991, Vol. 59, pp. 2040-2041

[12] “Highly efficient second-harmonic generation in buried waveguides formed by annealed and reverse proton exchange in periodically poled lithium niobate”, K. R. Parameswaran, R. K. Route, J. R. Kurz, R. V. Roussev, M. M. Fejer, and M. Fujimura; Optics Letters, February 2002, Vol. 27, pp. 179-181

Telecom


Novel Optically-switched Electroabsorption Modulators for Wavelength Conversion

Vijit A. Sabnis, Hilmi Volkan Demir, Jun-Fei Zheng, Onur Fidaner, James S. Harris, Jr., and David A. B. Miller Department of Electrical Engineering, Stanford University

All-optical wavelength conversion technologies have recently been extensively investigated to improve the flexibility of wavelength-division-multiplexed optical networks [1]. High-performance optical and optoelectronic techniques that avoid the high packaging costs and performance constraints of traditional optical-electronic-optical conversion have been reported [2]. These techniques, however, offer limited wavelength conversion capabilities or cannot be conveniently scaled for multichannel operation. In this letter, we introduce a novel, dual-diode, optically-switched electroabsorption modulator that offers bi-directional, unconstrained wavelength conversion over the entire C-band (i.e., 1530-1565 nm). Our device’s compact size, surface-normal optical control, electrical reconfigurability, and low optical and electrical power requirements allow for single chip implementation of two-dimensional arrays of multichannel wavelength converters [3]. Fig. 1 shows a top view microscope picture of a fabricated switch [4] that includes (i) a surface-illuminated, InGaAs-InP, p-i-n diode, photodetector (PD), (ii) a multiple-quantum-well, InGaAsP-InP, p-i-n diode, waveguide electroabsorption modulator (EAM), and (iii) a TaN thin-film resistor (R). The switch is monolithically integrated onto an InP substrate in a 300 µm by 300 µm area [5]. The switch inputs consist of a high-speed optical data stream at λ1 incident on the PD and a continuous wave beam at λ2 incident on the input facet of the EAM. The switch output is the output beam from the EAM. The switch transfers optical data from the incident PD beam to the output beam of the EAM. The PD is capable of broadband photodetection for wavelengths smaller than 1650 nm. The EAM is designed to efficiently modulate wavelengths over the C-band by appropriately presetting its DC bias point. Hence, the switch is capable of performing unconstrained, bi-directional wavelength conversion over the entire C-band. A circuit schematic of the device is presented in Fig. 2. To operate the switch, the PD is biased to allow for high-speed, linear photocurrent extraction and the EAM is biased to strongly absorb at the wavelength of the continuous wave input beam, λ2. Switching occurs when an optical data stream is incident on the PD, generating a photocurrent, which is then discharged through the on-chip resistor. The photocurrent creates a voltage drop across the resistor, which in turn reduces the applied bias across the modulator. Through the quantum confined Stark effect, this reduction in voltage results in the initially absorbed EAM input to be strongly transmitted through the waveguide. Hence, data is transferred from the PD input beam at λ1 to the EAM output at λ2. The switch can be electrically reconfigured by setting the PD bias appropriately. By slightly forward biasing the PD, photocurrent from the incident data stream can not be extracted, disabling the switching capability. Along with the compact size of the switch, such a capability enables the realistic fabrication of arrays of these switches for creating a multichannel wavelength-converting crossbar switch.

Telecom


Using an Agilent Lightwave Component Analyzer, a switch comprising a 2.65 µm wide EAM and a 650 Ω resistor was measured to have a 3-dB optical bandwidth of ~ 1 GHz. Fig. 3 illustrates 1.25 Gb/s eye diagrams for this switch, demonstrating wavelength conversion from 1551.7 nm (PD input) to three exemplary wavelengths (EAM output) across the C-band. An optically preamplified receiver and digital communications analyzer were used to record the eye diagrams. The average absorbed PD input power was 5.6 mW, fixed for all cases, and resulted in > 10 dB RF extinction ratios. There are no constraints on the choice of input wavelengths within standard telecommunication bands because of the wide wavelength photodetection capabilities of the PD. This allows our switch to perform truly arbitrary wavelength conversion over the entire C-band. Fig. 4 shows return-to-zero (RZ) and non-return-to-zero (NRZ), 2.5 Gb/s eye diagrams with > 10 dB RF extinction ratios for a switch, comprising a 2.1 µm wide EAM and a 340 Ω resistor, that was measured to have a 3-dB optical bandwidth of ~ 2 GHz. The average absorbed PD input power was < 8 mW for both the RZ and NRZ cases. The pattern effect, observed in these eye diagrams, illustrates that the optical bandwidth of this device is slightly insufficient for ideal 2.5 Gb/s operation. We expect optimized versions of these devices, exhibiting increased switching bandwidths with smaller required optical switching powers, will result in cleaner eye diagrams. Our theoretical model suggests that switches operating at 10 Gb/s with > 10 dB extinction ratios are practically feasible, and our current work focuses on fabricating such devices. In conclusion, we have demonstrated a compact, low power, optically-switched electroabsorption modulator performing unconstrained wavelength conversion up to 2.5 Gb/s with extinction ratios exceeding 10 dB. Optimization of the multiple quantum well design and a corresponding progressive reduction in the size of the on-chip resistor and PD and EAM capacitances will allow for high-speed switching at 10 Gb/s and beyond. Furthermore, our switch architecture inherently offers two-dimensional scalability, allowing the possibility of creating a multichannel wavelength-converting crossbar switch.

Figure 1: Top view microscope picture of a fabricated switch.

Telecom


Figure 2: Simplified circuit diagram of the switch.

Figure 3: 1.25 Gb/s eye diagrams demonstrating full C-band wavelegnth conversion.

Figure 4: Return-to-zero and non-return-to-zero eye diagrams at 2.5 Gb/s.

References [1] S. J. B. Yoo, Journal of Lightwave Technology 14, 955 (1996). [2] K. E. Stubkjaer, IEEE Journal of Quantum Electronics 6, 1428 (2000). [3] H. V. Demir, V. A. Sabnis, J.-F. Zheng, O. Fidaner, J. S. Harris, Jr., and D. A. B. Miller, see 2003

Stanford Photonics Research Center Annual Report. [4] H. V. Demir, V. A. Sabnis, J.-F. Zheng, O. Fidaner, S. Latif, N. Li, T.-C. Wu, Y.-M. Houng, J. S.

Harris, Jr., D. A. B. Miller, submitted to Electronics Letters. [5] V. A. Sabnis, H. V. Demir, O. Fidaner, J. S. Harris, Jr., D. A. B. Miller, J.-F. Zheng, N.

Li, T.-C. Wu, and Y.-M. Houng, Integrated Photonics Research (OSA Technical Digest, Washington DC, 2003), pp. 12-14.

RZ

NRZ

RZ

NRZ

1551.7 nminput

1530 nmOutput

VEAM = 8 V

1547.5 nmOutput

VEAM = 10 V

1565 nmOutput

VEAM = 10 V

1551.7 nminput

1551.7 nminput

1530 nmOutput

VEAM = 8 V

1530 nmOutput

VEAM = 8 V

1547.5 nmOutput

VEAM = 10 V

1547.5 nmOutput

1547.5 nmOutput

VEAM = 10 V

1565 nmOutput

VEAM = 10 V

1565 nmOutput

1565 nmOutput

VEAM = 10 V

Telecom


Photonic A/D Conversion Using Low-Temperature-Grown GaAs MSM Switches Integrated with Si-CMOS

R. Urata, L. Y. Nathawad, K. Ma, D. A. B. Miller, B. A. Wooley, and J. S. Harris, Jr. Ginzton Laboratory, Electrical Engineering Department

With rapidly increasing signal bandwidths along with the predominance of digital technologies and techniques, the need for higher speed analog-to-digital (A/D) conversion arises in order to interface between the analog and digital domains. Applications for A/D conversion at tens of giga-samples/s (GSa/s) exist in the areas of high-speed test and measurement, optical communications, and wireless communications and radar. Despite the demand for higher speed A/D conversion, improvement of conventional, electrical A/D converters has been limited by phase error in the clock (timing jitter and skew) [1]. As a solution, the idea of combining the low-timing-jitter and high-speed advantages of photonics with electrical A/D converters in a hybrid system has produced a number of photonic A/D conversion systems. Our proposed photonic A/D conversion system utilizes a sample-and-hold scheme with low-temperature-(LT) grown GaAs metal-semiconductor-metal (MSM) switches. The switch is attached to a high-speed electrical transmission line, with the other end attached to a hold capacitor. As the input electrical signal propagates down the line, a mode-locked laser pulse optically triggers the switch, sampling the signal onto the hold capacitor. An electrical A/D converter running at the laser repetition rate then digitizes this held signal. By time-interleaving a number of these channels, with an appropriately phase aligned optical clock for each channel, the aggregate sampling rate of the system is increased. The short carrier lifetime of LT-grown GaAs leads to picosecond order switch turn-off times. Thus, the large bandwidth of the optically triggered electrical switches allows direct utilization of the low-timing-jitter, high-bandwidth nature of the mode-locked laser optical pulses to provide the desired broadband input sample-and-hold. In this paper, we demonstrate a two-channel test chip, with CMOS A/D converters fabricated in a 0.24 µm CMOS technology for 1 GSa/s, 4-bit operation per channel. LT GaAs MSM switches were integrated with the CMOS circuits using a flip-chip bonding technique. This prototype system demonstrates a peak signal-to-noise-plus-distortion-ratio (SNDR) of ~23 dB [3.5 effective-number-of-bits (ENOB)] and a spurious-free-dynamic-range (SFDR) of ~29 dB over a 40 GHz input bandwidth. For a 25 µs measurement period, an rms timing jitter of < 80 fs is obtained.

The CMOS A/D converters were designed and fabricated in a National Semiconductor 0.24 µm CMOS process. Critical design constraints were the need for a low input capacitance due to the limited responsivity of the MSM switches, a differential input with high common-mode rejection to match the differential configuration sample-and-hold [2], low area in order to fit as many channels as possible in a given area, and low power consumption. Fig. 1 shows a diagram of a single channel. A front-end differential buffer amplifier provides a low input capacitance and drives a 4-bit flash converter. The fabricated prototype chip consists of two A/D converter channels. In order to reduce parasitics at the buffer input nodes, a flip-chip bonding process was used to integrate the differential configuration MSM switches with the CMOS A/D converters.

The prototype chip was tested using an 80 MHz titanium/sapphire mode-locked laser to trigger the MSMs, and a low-timing-jitter electrical source (Agilent E8244A) for the input signal. Fig. 2

Telecom


Figure 6: Diagram of a single channel. Differential configuration MSMs allow feedthrough cancellation. Input/hold capacitance (CH1, CH2) is ~30 fF.

Figure 2: Measurement setup for two-channel test chip.

illustrates the test setup. The laser beam is split in two to generate an optical clock for each channel, with a translation stage used for proper phase alignment. Although the sampling rate per channel is limited to the repetition rate of the laser, the 80 MHz fundamental is multiplied up to 640 MHz with a frequency multiplier and used to clock the A/D converters to confirm CMOS circuit functionality at this higher repetition rate. The 4-bit output of each channel is fed into a logic analyzer (hp 16702A), with every 8th point taken to give the final output. Input sinusoids from 3 MHz up to 40 GHz were sampled using ~50 pJ optical pulses for a measurement time window of 25 µs (limited by logic analyzer memory). Although aliasing occurs for input frequencies beyond the Nyquist limit, linearity of the system can be characterized by looking at the output signal spectrum within the Nyquist band. Frequency-dependent loss of the test setup and test chip was compensated for by increasing the amplitude of the input signal.

Results for testing of a single channel are shown in Fig. 3, with peak SNDR and SFDR plotted as a function of input frequency. For the entire 40 GHz input band, SNDR is ~25 dB, corresponding to 3.8 ENOB, and SFDR is > 30 dB. The SNDR is fairly constant as input frequency increases, indicating that jitter from the input source and the laser is not limiting the performance of the system. By taking the output signal, subtracting quantization noise and the largest harmonic distortion contribution, and

Telecom


attributing all other noise to timing jitter, an rms value of σjitter~80 fs is obtained for the system. Most of the distortion in the system can be attributed to buffer amplifier non-linearity. For higher A/D converter clock speeds, the performance worsened due to inadequate settling time for the buffer amplifier. Results for two-channel testing are also shown in Fig. 3. With the exception of the low frequency data point (3 MHz), the peak SNDR is ~23 dB, corresponding to 3.5 ENOB, and SFDR is ~29 dB. In comparison to single-channel results, degradation in performance is caused in part by channel mismatch errors, present in all time-interleaved systems [3]. Channel-to-channel variations in gain and offset of the conversion transfer function, along with phase error due to skew between channels, can increase distortion. Overall, performance is not limited by timing jitter, demonstrating the advantage of utilizing the mode-locked laser optical clock.

Figure 3: Results of single-channel (solid line) and two-channel (dashed line) testing. Peak SNDR (circles) and SFDR (triangles) are plotted for various input frequencies.

References

[1] R. H. Walden, “Performance trends for analog-to-digital converters,” IEEE Commun. Mag., vol. 37, pp. 96-101, Feb. 1999.

[2] R. Urata, R. Takahashi, V. A. Sabnis, D. A. B. Miller, and J. S. Harris, Jr., “Ultrafast differential sample and hold using low-temperature-grown GaAs MSM for photonic A/D conversion,” IEEE Photon. Technol. Lett., vol. 13, pp. 717-719, July 2001.

[3] W. C. Black, Jr. and D. A. Hodges, “Time interleaved converter arrays,” IEEE J. Solid-State Circuits, vol. SC-15, pp. 1022-1029, Dec. 1980.

Telecom


Optical Interconnects

SPRC 2002 – 2003 E-1 Annual Report

Research Program

E. Optical Interconnects

Aparna Bhatnagar, Christof Debaes, Ray Chen, Noah Helman, Gordon Keeler, Salman Latif and David A. B. Miller, “Receiver-less Optical Clock Distribution using Short Pulses”................................................................................................................................................. E-3

C.H. Cheng, A.S. Ergun, and B.T. Khuri-Yakub, “Electrical through silicon wafer interconnects for high frequency photodetector arrays”....................................................................... E-7

Kai Ma, Ryohei Urata, David A. B. Miller and James S. Harris, Jr., “Low Temperature Growth of GaAs on Si Substrates for Ultra-fast Photoconductive Switches Used in a Hybrid CMOS/Photonic A/D Conversion System” ........................................................................... E-10

David Press, Rafael Aldaz, Michael Wiemer, James S. Harris and David A.B. Miller, “Mode-locking simulations for monolithically integrated Vertical Cavity Surface Emitting Laser” .................................................................................................................................. E-13



Receiver-less Optical Clock Distribution using Short Pulses

Aparna Bhatnagar, Christof Debaes, Ray Chen, Noah Helman, Gordon Keeler, Salman Latif, and David A. B. Miller Ginzton Laboratory, Stanford University

Clocks are used in computing systems to maintain synchronicity between unequal logic paths. Data only proceeds in the system at clock edges, demanding an ever-increasing clock frequency to maximize the computational bandwidth. As microprocessor clock frequencies rise, the skew and jitter budgets of clocks become proportionally smaller. At 10GHz, less than 10ps of jitter or skew is allowable. The bandwidth of wires used for distributing clocks does not keep pace. A sharp edge at the input of such a wire has a much slower rise time at the destination, resulting in higher noise susceptibility and greater variation in the delay. Each buffer, repeater, or amplifier required to avoid the limitations of electrical wires adds jitter and consumes power.

Optics offers a low jitter clock source with 10GHz and higher repetition rates in the form of the mode-locked laser. Since the light in a laser cavity effectively makes several round trips prior to emission, the Q of a mode-locked laser is very high, making the generated pulse stream a very stable clock source. Fundamentally, a mode-locked laser producing sub-picosecond pulses with gigahertz repetition rates can have jitter on the order of a few hundred femtoseconds or less [1,2,3]. To benefit from the fast rising edges and low timing jitter of these pulses we believe it is best to introduce as little circuitry between the photodetector and the clocked node as possible. The receiver-less ideal is to drive the input capacitance of a clocked element directly with the photocurrent from the detector, without any intervening circuitry by directly creating large voltage swings over the photodetectors. This eliminates the power, jitter and latency of the clock receiver or distribution network, thereby addressing key clocking challenges.

Our experiments show the injection of a noise-immune sharp optical clock onto CMOS chips using the receiver-less approach. To create a 50% duty cycle clock from the modelocked pulse stream, we split the beam, delay one arm by half a bit period and focus the two beams on a pair of detectors connected in series. Figure 1 shows the use of this technique for driving actual digital circuitry on a chip, where we clocked a pseudo-random bit sequence generator (PRBS) consisting of 4 D-flip-flops and an XOR gate. 100fs pulses from an 80MHz Ti:Sapphire laser were used.



H-tree

k=2

k=3

Optical insertion

A close-up of the eye-diagram on the oscilloscope is also shown in Fig. 1 when 160µW is shone on each detector. The histogram on the falling edge of a flip-flop output shows an rms-jitter of 6ps. The jitter of the clock at the input of the flip-flop is likely to be significantly less. The measured jitter includes also the jitter introduced by the source follower output driver that switches large currents. The chip was fabricated using a commercial 0.5µm Ultra-thin Silicon-on-Sapphire (UTSi) CMOS process and integrated with GaAs/Al0.3Ga0.7As multiple quantum-well (MQW) p-i-n detectors via flip-chip bonding. Thus, we demonstrated a stable, functioning optically clocked digital circuit, and measured an upper bound on the jitter of the output data.

A realistic clock distribution network must distribute a noise-free clock to over a hundred thousand points with the same precision and efficiency as we have demonstrated for a single point. To understand the relative merits of the receiver-less scheme for such applications, we model a symmetric clock distribution network consisting of inverter buffers each of which drives four inverters in the next level of a balanced H-tree as shown in Fig. 2. We require that each inverter drive four times its input capacitance. Such a fan-out-of-four (FO-4) configuration generally results in a close-to-minimum delay [4]. As a consequence, at every level, the driving inverter is slightly larger than that at the next level since it has to drive not only four inverters but also the intervening wire capacitances.

Figure 2: Schematic of modeled clock distribution (left). Table 1: Model parameters (right).

The top of the distribution tree is usually an active time alignment circuit such as a PLL or DLL (delay- locked loop) which has a limited driving capability. Therefore we force the input capacitance of the top of the distribution tree to be equal to the inverter input capacitance at the leaf nodes. Thus, we add additional FO-4 inverter gain stages at the top of the H-tree for extra capacitance drive. This clock distribution network, which is taken as a starting point, is compared to an optical equivalent where the top of the network has been replaced by an optical clock distribution. For a clock tree with n H-tree branches, we define the optical insertion level (parameter k) as the level above which the distribution is done optically as shown in Fig. 2. The electrical power consumption, the delay of the clock distribution network and the required optical power, were modeled for various optical insertion

Symbol Typical value Leaf node inverter capacitance input capacitance

Cin 30fF

Wire capacitance/unit length CW 430fF/mm Resistance of global wires RW 20Ω/mm Fan-out of 4 delay tF04 90pS Supply voltage Vdd 1.8V Clock frequency f 1Ghz Chip total dimension L 20mm Total capacitance of the latches

Cclock 9nF

Detector transition time ttrans 10pS Direct injection detector capacitance

Crec 30fF

DC-Responsivity R 0.5 A/W TIA-Receiver latency tTIA 300 pS TIA-Receiver Analog Power PTIA 500 µW



levels. The modeling was done for both receiver-less optical injection as well as a more traditional approach with TIA receivers. The parameters used in the model are summarized in Table 1. The FO-4 delay, the wire resistance and the wire capacitance characteristic of a typical 0.18 µm technology were used [4,5,6].

Figure 3 plots the results of our modeling. The graph on the left shows the ratio between the electrical power consumption of a distribution network with different optical insertion levels and the conventional FO-4 electrical distribution tree. We present this ratio for various different optical detector capacitances. It is clear that by introducing optics in the distribution tree only a small amount of power can be gained except with very low detector capacitances in the receiver-less case. The graph on the right in Fig. 3 compares the total distribution delay between a conventional clock distribution tree, and the optical equivalents using either direct injection or trans-impedance receivers, or receiver-less with additional buffers placed such that the load on the receiver-less diodes is always four times the capacitance Cin. One sees that, in contrast with the power gain, the introduction of optics does make a large difference in the delay. For example, with an optical insertion at level 7 of this 9-level distribution tree, we would reduce the delay from 1.7ns to only 190ps in the buffered receiver-less case, corresponding to a 90% reduction.

Figure 3: Power ratio (left) and Delay comparison (right) between electrical clock distribution tree and an optical equivalent for different levels of insertion.

In conclusion, we experimentally demonstrated low-jitter receiver-less injection of clocks that can drive a PRBS on a chip. In conclusion to the modeling of the electrical power consumption, the delay, and the required optical power, we found that choosing optical clock distribution will likely not create significant savings in the power consumption. There can however be a large decrease in the total delay through the distribution tree, making the optical solution potentially superior in terms of skew and jitter performance.

References [1] L. Krainer , R. Paschotta, S. Lecomte, M. Moser, K.J. Weingarten, U. Keller, “Compact

Nd:YVO4 Lasers with Pulse Repetition Rates up to 160 GHz”, IEEE JSTQE, Vol 38, No.10, pp. 1331 – 1338, Oct 2002



[2] A.M. Braun, V.B. Khalfin, M.H. Kwakernaak, W.F Reichert, L.A. DiMarco, Z.A. Shellenbarger, C.M DePriest, T. Yilmaz, P.J. Jr. Delfyett, J.H.Abeles, “Universality of Mode-Locked Jitter Performance”, IEEE PTL Vol 14, No. 8, pp. 1058 –1060, August 2002

[3] Leaf A. Jiang, M.E. Grein, H.A Haus, E.P. Ippen, “Noise of Mode-Locked Semiconductor Lasers”, IEEE JSTQE, Vol 7, No 2, pp. 159-167, March 2001

[4] R. Ho, K.W. Mai, M. A. Horowitz, “The future of wires”, Proc. of the IEEE, vol. 89, no. 4, pp. 490 –504, April 2001.

[5] P. J. Restle, C. A. Carter, J. P. Eckhardt, B. L. Krauter, B. D. McCredie, K. A. Jenkins, A. J. Weger, A. V. Mule, “The clock distribution of the Power4 microprocessor”, IEEE ISSCC, 2002 Digest of Technical Papers, Vol. 1, pp. 144-145, Feb 2002

[6] F.E, Anderson, J.S. Wells, E. Z. Berta,, “The core clock system on the next-generation ltanium microprocessor”, Proceedings of the ISSCC conference, Vol. 2, pp 110-112, Feb 2002



Electrical through silicon wafer interconnects for high frequency photodetector arrays.

C.H. Cheng, A.S. Ergun, and B.T. Khuri-Yakub

In optoelectronic applications, it is advantageous to have electronic circuitry as near the detector/emitter as possible. However, integrating both the optoelectronic devices with electronics on the same wafer often leads to a compromise between the performance of either or both systems. An excellent solution to this problem is to construct the optimum optoelectronic devices and electronics on separate wafers, provide a through wafer interconnect with minimum attenuation on the optoelectronic wafer, then flip-chip bond the two wafers. In this fashion, the optoelectronic wafer can be fully populated, as in applications of infra-red (IR) focal plane arrays or three-dimensional ultrasound imaging [1-5]. Finally, since the optoelectronic and electronic wafers can be fabricated in different facilities, the overall yield of the manufacturing process is enhanced.

For low loss and high frequency applications, we propose to use a through wafer interconnect shown in Fig. 1. The interconnect presents a parallel capacitance and a series resistance to the input impedance of the photodetector. Thus, for operation that is not limited by the interconnect, both the capacitance and resistance have to be very small. We present a technology where the parasitic capacitance is reduced to 0.05 pF in a wafer that is 400 µm thick with a resistivity of 1000 Ω-cm and with a via that has a diameter of 20 µm.

PN JunctionThrough-Wafer

Interconnect

PN JunctionBackside Pad

Ohmic Contact

Solder PadGround Strip Line

SiliconSubstrate

PolysiliconSealed

PIN Photodetector

Image Processing ChipSolder Bump

Figure 1: Schematic of the through-wafer interconnect.

Two-sided deep reactive ion etching (RIE) is used to make through-wafer holes with an aspect ration of 20:1. In previous work [3], we made vias with MIS junctions that had a parasitic capacitance of 0.28 pF. However, in this work [1][2] we report vias with PN junction diodes that have a measured parasitic capacitance of 0.05 pF at a reverse bias higher than 10 V. The parasitic capacitance can be further reduced by reducing the thickness of the wafer, reducing the diameter of the via, and increasing the resistivity of the silicon wafer.



The process flow is shown in Fig. 2. We start with a 400 µm thick double-sided polished n-type <100> Si wafer which is thermally oxidized to 2 µm thick. Both sides are then patterned with 20 µm diameter openings for each interconnect (Fig. 2a). The through-wafer deep etch is done by etching half way from both sides of the wafer (Fig. 2b). The wafer is then heavily doped with boron to build the pn junction diode inside the holes (Fig. 2c). The interconnect holes are then filled with polysilicon (Fig. 2d). The polysilicon on both sides is then etched back and stopped by the oxide (Fig. 2e). It is ready to be etched for the front and backside oxide opening (Fig. 2f). The wafer is then doped with boron which makes up the pn junctions for the front and backside pads. The oxide is etched and the backside is patterned with photoresist for phosphorous ion implantation for ohmic contact (Fig. 2g). At the very end, the backside metal pads for flip-chip bonding are formed by lift-off (Fig. 2h). The SEM pictures show the cross section of a finished interconnect (Fig. 3). The wafer is now ready for flip-chip bonding to a signal processing wafer or PCB.

a) Oxidation 2µm & Via Open

b) Deep Etch

c) Via Diffusion

d) Poly Seal

e) Etch Back

f) Oxide Open & Diffusion

g) Oxide Remove & GND

h) Metalization

400 µm

20 µm

20 µm

Figure 2: Fabrication process for through wafer via. Figure 3: SEM pictures through a via.

A testing device with a through-wafer interconnect, connected with both frontside and backside pads and a ground to the substrate, was employed to measure the C-V characteristic at 1 MHz frequency. A reversed DC bias was applied to drive both pn-junction pads into the depletion region. The total capacitance is the capacitance of the through-wafer interconnect plus both pads. The series resistance for each via was 900 Ω, which was more than our prediction of 434 Ω. This is due to the non-uniform doping profile throughout the via-hole. The leakage current coming from both pads and the via was 7nA.

A C-V characteristic at 1 MHz frequency was tested. By applying a reverse bias for more than 10 volts, the capacitance decreased to 0.055 pF because of depletion into the substrate. This experimental result was very close to what we expected from the simulation (Fig. 4). The C-V measurements revealed that electrical through-wafer interconnects with parasitic capacitance of 0.05pF have been



achieved. This result makes its applications to low-capacitance devices possible, such as silicon high frequency photodetectors.

Figure 4: Comparisons of experimental capacitance-voltage relationships (CV) of a reversed biased pn junction through-wafer interconnect with theoretical simulation.

Because the series resistance of the through-wafer interconnects is not yet optimized for current fabrication processes, it remains as a problem for future development. This series resistance can be substantially reduced by employing doped polysilicon or metal to seal the via-holes, which is the subject of future work.

Acknowledgement: This work is sponsored by the Office of Naval Research.

References:

[1] C. H. Cheng, A. S. Ergun, B. T. Khuri-Yakub, "Electrical Through Silicon Wafer Interconnects for High Frequency Photodetector Arrays," Photonic Devices and Systems Packaging Symposium PhoPack 2002, Stanford, California, 7/14-16/2002, pp. 54-7

[2] C. H. Cheng, A. S. Ergun, B. T. Khuri-Yakub, "Electrical Through Wafer Interconnects with 0.05 Pico Farads Parasitic Capacitance on 400 µm Thick Silicon Substrate," Solid-State Sensor, Actuator, and Microsystems Workshop 2002, Hilton Head Island, South Carolina, 6/2-6/2002, pp. 157-160

[3] C. H. Cheng, A. S. Ergun, B. T. Khuri-Yakub, "Electrical Through-Wafer Interconnects with Sub-PicoFarad Parasitic Capacitance," Proceedings of MEMS Conference 2001, Berkeley, California, 8/24-26/2001, pp. 18-21

[4] C. H. Cheng, E. M. Chow, A. S. Ergun, and B. T. Khuri-Yakub, "An Efficient Electrical Addressing Method Using Through-Wafer Vias for Two-Dimensional Ultrasonic Arrays," 2000 IEEE International Ultrasonics Symposium, San Juan, Puerto Rico, 10/22-25/2000, pp. 1179-82.

[5] S. Calmes, C. H. Cheng, F. L. Degertekin, X. C. Jin, and B. T. Khuri-Yakub, "Highly Integrated 2-D Capacitive Micromachined Ultrasound Transducers," presented at the 1999 IEEE International Ultrasonics Symposium, Lake Tahoe, Nevada, Octobor 17-20, 1999; in Ultrasonics Symposium Proceedings, pp. 1163-6.



Low Temperature Growth of GaAs on Si Substrates for Ultra-fast Photoconductive Switches Used in a Hybrid CMOS/Photonic A/D Conversion System

Kai Ma, Ryohei Urata, David A. B. Miller and James S. Harris, Jr. Solid States and Photonics Laboratory, Stanford University

Modern communications and high-speed instrumentation require much higher speed analog-to-digital converters (ADC) with bandwidth up to several tens of GHz. Electronic ADCs are powerful in signal processing, but low input bandwidth and timing jitter problem limit their performance at high speeds. Since photonic devices have high speed and good timing accuracy advantages, a hybrid system could poterntailly combine the advanteages of both technologies. Based on this idea, we proposed a CMOS/photonic A/D conversion system utilizing a sample-and-hold scheme with low temperature GaAs (LT-GaAs) metal-semiconductor-metal (MSM) photoconductive switches [1, 2]. Figure 1 shows the configuration of our system. When the input analog signal propagates down the transmission line, the LT-GaAs switches would be triggered on by a fs-order mode locked laser pulse, sampling the corresponding voltage onto a series of hold capacitors along the transmission line. Sampled data would then be digitized at a much slower rate using CMOS ADCs attached to the sample and hold circuit. We use a time-interleaved architecture with multiple channels to achieve ultra-fast aggregate sampling rate.

We chose LT-GaAs material for the switches. LT-GaAs is usually grown at 200°C to 350°C substrate temperature, followed by a post-growth anneal to improve carrier mobility. Due to the low growth temperature, a large density of excess As incorporates into the material and forms As-related deep-level defects. These defects result in ps carrier lifetimes. LT-GaAs has a good combination of short carrier lifetime and high carrier mobility, as well as high dark resistivity and high breakdown field, making it ideal for ultra-fast switch application [3].

To integrate the GaAs switches with Si chips, one approach is the flip chip bonding hybrid technique. Using this technique, we have demonstrated a two-channel prototype ADC with ~3.5 effective bits of resolution for an input bandwidth up to 40GHz and estimated total jitter smaller than 80fs [4].

To minimize the input parasitics, monolithic integration is explored. One alternative is to finish the final level of metallization of the Si IC after the growth of the GaAs devices. But this approach creates significant fabrication perturbation. In our application, the material quality requirements for the switches are more forgiving and the low growth temperature is compatible with completely fabricated

electrical input signal

LT GaAs switch

CMOS ADC

hold capacitor

electrical transmission line fs-order optical pulse

Figure 1. Schematic configuration of the hybrid CMOS/photonic A/D conversion system.



CMOS circuits. Furthermore, by using low temperature process, the thermal stress caused by the 60% difference in thermal expansion coefficients between Si and GaAs is greatly reduced. Therefore, we are investigating the direct growth of LT-GaAs on finished Si CMOS chips. In our approach, the critical step is cleaning the Si surface at temperatures low enough not to affect the metal interconnects and the reliability of low-k dielectrics in Si circuits [5]. This report shows preliminary study on LT-GaAs growth on Si substrates.

The LT-GaAs layers were grown on nominal (100) and vicinal (100) Si substrates oriented 4° off (100) toward [011]. Details of the wafer cleaning and film growth are published in [6]. A final HF dip was done immediately before loading the wafers into the loadlock chamber to achieve hydrogen passivation. During the heat cleaning in the growth chamber, the Si surfaces exhibited a streaky 2 × 1 reconstruction RHEED pattern at temperatures as low as ~450 °C. Samples in this report were baked at ~550 °C and then cooled down to growth temperature (~250 °C). Following the LT-GaAs growth, the samples were annealed in-situ at ~600°C for 10 minutes under an As overpressure. Laser annealing could be used to avoid this high temperature anneal when actually growing on circuits. Upon annealing, the RHEED patterns showed 2 × 4 reconstruction, indicating a high quality single domain GaAs layer with a smooth surface.

For 0.5 µm thick LT-GaAs grown on a nominal (100) Si substrate, the root-mean-square (rms) roughness measured by AFM was ~1.0 nm. The same thickness LT-GaAs layer on a vicinal (100) substrate showed ~0.5 nm rms roughness. The peak-to-peak variation across the surface was ~6 nm for the film grown on the nominal substrate and ~3 nm for the mis-oriented substrate. This roughness is approximately an order of magnitude smaller than the previously reported best data [7]. For both samples, the XRD GaAs (002) superstructure peak is not broadened more than the (004) fundamental peak, indicating that the APD density is below the XRD detection limit [8].

Interdigitated MSM switches were fabricated by depositing titanium/gold on the LT-GaAs epi-layers through a lift-off process. The switch was placed in the middle of a coplanar waveguide transmission line structure for high-speed characterization (figure 2). During measurement, the switch was dc-biased and optically triggered by a titanium/sapphire mode-locked laser with ~150 fs FWHM pulse width and ~850 nm center wavelength. The movement of photo-generated carriers causes an electrical transient to propagate down the transmission line, characterizing the responsivity and speed of the switch. A time-resolved electro-optic (EO) sampling technique was used to measure the transient, with a lithium tantalate (LiTaO3) crystal placed on top of the transmission line [9].

A typical output signal is shown in figure 3. Signal FWHM was ~2ps for both samples. Dependence of the response on applied bias and optical pulse energy was similar to that seen for devices made from LT-GaAs grown on GaAs substrates. Figure 4 shows that the responsivity of switches is comparable for LT-GaAs on Si versus LT-GaAs on GaAs, although presumably, a much higher density of dislocations exist in the material grown on Si. By optimizing the material quality, LT-GaAs grown directly on Si presents a promising alternative to LT-GaAs on GaAs material.

In conclusion, we have grown GaAs directly on silicon substrates by molecular beam epitaxy (MBE) at low substrate temperatures. Electro-optic sampling measurements showed that the performance of LT-GaAs switches on a Si substrate is comparable to switches on a GaAs substrate.



References [1] R. Urata, R. Takahashi, V. A. Sabnis, D. A. B. Miller, and J. S. Harris, Jr., IEEE Photo. Technol.

Lett., 13, 717 (2001). [2] R. Urata, L. Y. Nathawad, K. Ma, R. Takahashi, D. A. B. Miller, B. A. Wooley, and J. S. Harris,

Jr., in Proceeding of IEEE LEOS Annual Meeting, Vol. 2, Glasgow, Scotland, 2002, pp. 809-810. [3] F. W. Smith, H. Q. Lee, V. Diadiuk, M. A. Hollis, A. R. Calawa, S. Gupta, M. Frankel, D. R.

Dykaar, G. Mourou, and T. Hsiang, Appl. Phys. Lett. 54, 890 (1989) [4] L. Y. Nathawad, Stanford University Ph.D oral. [5] Private communication with Professor Krishna Saraswat at Stanford University. [6] K. Ma, R. Urata, D. A. B. Miller and J. S. Harris, Jr., Mat. Res. Soc. Symp. Proc. Vol. 768, G3.14

(2003) [7] C. Kadow, S. B. Fleischer, J. P. Ibbetson, J. E. Bowers, and A. C. Gossard, Appl. Phys. Lett., 75,

2575 (1999) [8] S. F. Fang, K. Aomi, S. Iyer, H. Morkoc, H. Zabel, C. Choi, and N. Otsuka, J. Appl. Phys., 68, R31

(1990). [9] J. A. Valdmanis, and G. Mourou, IEEE J. Quantum Electron. QE-22, 69 (1986).

Figure 2. SEM picture of MSM switch characterization structure.

signal line

ground line

ground line

switch

0

0 .5

1

1 .5

2

2 .5

3

0 10 20 30 4 0

time (ps)

sign

al (a

.u.)

~2ps

Figure 3. Switch response measured by EO sampling

0 5 10 15 20 25

0

10

20

30

40

50

LT-GaAs on Si

phot

ocur

rent

(µA)

optical pump power (mW)

0.8µm spacing 1µm spacing 5µm spacing

0 5 10 15 20 25

0

10

20

30

40

50

2µm finger width0.8µm finger spacing

at 10V bias

phot

ocur

rent

(µA

)

Optical pump power (mW)

LT-GaAs on Si LT-GaAs on GaAs

(a) (b)

Figure 4. (a) Switch responsivity comparison between LT-GaAs on Si sample and LT-GaAs on GaAs sample. (b) Finger spacing effect on switch responsivity.



Mode-locking simulations for monolithically integrated Vertical Cavity Surface Emitting Laser

David Press, Rafael Aldaz, Michael Wiemer, James S. Harris and David A.B. Miller Department of Electrical Engineering, Stanford University.

Mode-locked lasers are promising sources for clock distribution, WDM and OTDM systems as well as other applications. Among these lasers, VCSEL structures are very interesting given their low cost and easy of mass production, high quality circular beam, and potential high repetition rates with low jitter. In this paper we will present our approach to simulating the mode-locking characteristics of this novel device.

One widely used mode-locking model is Haus’ master equation1, which tracks the complex field envelope A at a single point inside the cavity:

( ) ( ) AtTqt

glgAAjt

jDtTAT

Tg

R ⎟⎟⎠

⎞⎜⎜⎝

⎛−

∂∂

Ω+−+⎟⎟

⎠

⎞⎜⎜⎝

⎛+

∂∂

−=∂∂ ,, 2

2

22

2

2

δ (1)

Effects such as gain g, saturable loss q, unsaturable loss l, dispersion D, self-phase modulation δ, and gain bandwidth Ωg are treated as perturbations to the pulse occurring over one round trip time TR. The split-step fourier-transform method2 can be used, which handles the linear effects of a round-trip in one half-step in the frequency domain and non-linear effects in another half-step in the time domain.

This technique has been used to accurately model mode-locking in solid-state lasers3, where the gain is saturated to a constant level by the average cavity power. In a semiconductor laser however, the gain can recover significantly between pulses. In our model, rate equations are used to describe the carrier concentrations in the saturable gain and loss regions of the semiconductor mode-locked laser:

( )ag

ag

ag

agag

eLJ

areahANN

dtdN

,

2,

,

,,

*2

+Γ

−−=να

τ (2)

The subscripts g and a indicate different variables applicable to the gain and absorber regions. N is the carrier concentration, τ the recovery time, hv the photon energy, area the effective area of the gaussian mode, J the injected current (zero for the absorber), e the fundamental electric charge, Γ the confinement factor, and L the total length of gain or absorber region. α represents the material absorption coefficient, which was semi-empirically fit to a logarithmic α(N) dependence using Kuznetsov’s parameters4. The time dependence of the gain and loss can then be calculated as.

( ) ( ) ( )( ) agag LtNtqtg ,,, αΓ= (3)

Similar rate equations have been used to model edge-emitting semiconductor mode-locked lasers5,6, but these models tracked only the photon density and could not account for phase effects such as dispersion. Traveling-wave equations were used rather than the Haus master equation because the gain and loss are spread throughout the cavity, not lumped at one end as in a VCSEL.



Figure 1 below shows the electric field magnitude, as well as cavity gain and loss, for a typical pulse. The pulse repetition period is around 48ps, and the figures show only a small portion of this total time. Before the pulse arrives, the loss has recovered to its unsaturated value because the absorber recovery time is much less than the repetition period. The gain is still recovering at a rate proportional to the injected current. When the pulse arrives, the loss saturates more quickly than the gain, allowing a brief window of net gain to occur over the center of the pulse. Once the pulse has passed, the loss quickly recovers to attenuate the pulse’s tail. The net gain window in figure 1 is much shorter and shaped differently than it would be if the gain were assumed to be at a constant level (as in solid-state models).

Figure 1: Pulse envelope (left) and cavity gain and loss (right)

As the injected current is increased, the non-constant gain profile begins to play an even more important role. At very high current levels, the gain can actually recover beyond the loss before the pulse arrives. This leads to an unstable pulse train, and can even support two pulses within the cavity.

A number of simulations were run with a cavity length set to give 21GHz repetition rate, and a beam radius of 10 um. A dispersion-free cavity was assumed, although dispersion could be added later if appropriate. An absorber recovery time of 3 ps should be achievable with a reverse-biased quantum well absorber. With proper heat-sinking, the 42 um diameter mesa could handle currents up to 300 mA. A confinement factor of 2 was used, corresponding to the gain and absorber wells being placed at the peak of the cavity’s standing wave pattern.

Figure 2 shows a comparison of average output power from the mode-locking simulations with power from the CW model described by Kuznetsov4. For each simulation, the CW laser’s cavity loss was set to match the average loss in the mode-locked laser. The two show good agreement, although the mode-locking simulator predicts slightly lower powers because its material absorption model is somewhat more conservative. The predicted full-width half-max pulse duration is also shown in figure 2, with values ranging from nearly 1 ps to 140 fs. The pulse was unstable for currents below 25 mA and above 500 mA.



Figure 2: Simulated optical output power (left) and pulse duration (right).

In the future, cavities with appropriate dispersion parameters will be analyzed to determine expected pulse durations, output powers and threshold currents. Also, different arrangements and structures for the gain and absorber will be studied to find the best possible configuration for the device.

References: [1] H. Haus, J. Fujimoto, E. Ippen. “Structures for additive pulse mode locking”, J. Opt. Soc. Am. B.,

Vol. 8, No. 10, pp 2068-2076. [2] Agrawal, G. 1989. Nonlinear Fiber Optics. San Diego, CA: Academic Press. [3] F. Kartner, J. Aus der Au, U. Keller. “Mode-Locking with Slow and Fast Saturable Absorbers –

What’s the difference?” IEEE J. of Sel Topics in Quan Elec. Vol. 4, No. 2, pp 159-168. [4] M. Kuznetsov, F. Hakimi, R. Sprague, A. Mooradian, “Design and Characterization of High-

Power (>0.5-W CW) Diode-Pumped Vertical-External-Cavity Surface-Emitting Semiconductor Lasers with Circular TEM00 Beams”, J. of Sel Topics in Quan Elec., Vol. 5, No. 3, May 1999.

[5] J Yu, D. Bimberg. “Suppression of self-pulsation for tens of gigahertz optical pulses from passively mode-locked semiconductor lasers”, Appl. Phys. Lett., Vol. 67, No. 22, pp 3245-3247.

[6] W. Yang, A. Gopinath, “Study of passive mode locking of semiconductor lasers using time-domain modeling”. Appl. Phys. Lett. Vol. 63, No 20, pp 2717-2719.

Optical MEMS / Optical Microsystems

SPRC 2002 – 2003 F-1 Annual Report

Research Program

F. Optical MEMS / Optical Microsystems

Ruslan Belikov, Xiang Li, Christophe Antoine-Snowden, Olav Solgaard, “Programmable Optical Wavelength Filters Based on Diffraction from MEMS Micromirror Arrays” ........................ F-3

U. Krishnamoorthy, I.W. Jung, P. Lu, Y.-A. Peter, E. Carr, R. L. Byer, O. Solgaard, “Development of segmented deformable mirrors for free space communications”............................. F-7

D. Lee, U. Krishnamoorthy, H. Ra, O. Solgaard , “Single-crystalline silicon micromirrors actuated by self-aligned vertical electrostatic combdrives with piston-motion and rotation capability” .......................................................................................................................................... F-10

Wonjoo Suh and Shanhui Fan , “Flat-top Reflection and All-pass Transmission Filter using Coupled Resonance in Photonic Crystal Slabs” ....................................................................... F-13

J. Wang, I. Jung, O. Solgaard, “Elastomer Spatial Light Modulators for Extreme Ultraviolet Lithography” .................................................................................................................... F-16

Thomas D. Wang, Michael J. Mandella, Ning Y. Chan, Christopher H. Contag, Gordon S. Kino, “MEMS Confocal Microscope with Dual Axes Architecture for In Vivo Molecular and Cellular Imaging”....................................................................................................... F-19

Kyoungsik Yu, Daesung Lee, Uma Krishnamoorthy, and Olav Solgaard, “Tunable bandwidth optical filter based on MEMS Gires-Tournois interferometer”........................................ F-23



Programmable Optical Wavelength Filters Based on Diffraction from MEMS Micromirror Arrays

Ruslan Belikov, Xiang Li, Christophe Antoine-Snowden, Olav Solgaard Edward L. Ginzton Laboratory, Stanford University, Stanford, CA 94305-4085

Introduction

Diffractive optics can be used to generate optical transfer functions for use in correlation spectroscopy, femtosecond pulse shaping, and other filter applications [1]. When implemented with MEMS technology [2], such diffractive optics can be configured as programmable optical filters capable of synthesizing arbitrary spectral reflectance profiles. An analytical theory has been developed for specifying the diffractive surface needed to generate an arbitrary complex-valued optical filter [3]. It enables fast calculations of the required diffractive surfaces. The theory has been experimentally verified using the commercial Texas Instruments DMDTM array [4] as the programmable diffractive element. It can also be used with many other MEMS devices that create diffractive surfaces, polychromator included.

The theory relating the desired complex-valued spectral filter H(f) [3] to the diffractive surface can be understood by considering the impulse response h(t), where H(f)=FTh(t). To see how the DMD generates arbitrary impulse responses, consider the setup in Fig. 1, where the input is a collimated beam Ein(t), normal to the +120 DMD micromirror tilt state.

Figure 1. Synthesis of arbitrary impulse response h[n] by the DMD array. Here, the n-th row has no mirrors turned to +120, while (n+1)-th row has 2, so that h[n] = 0 and h[n+1] = 2/N.

The output is the diffracted beam Eout(t), collected in the far field and is given by the convolution ( ) ( )thtEtE inout ⊗= )( (1) Let ( ) ( )ttEin δ= which will yield the impulse response at the output. A single mirror turned to +120 on the nth row will lead to a delta function output with a relative time delay of

Part of DMD array

n-th row

(n-1)-st row

(n+1)-st row Input Ein(t) = δ(t)

Output Eout(t) = h(t)

h[n-2] h[n-1]

h[n+1]

h[n+2]

τ

h[n]

τ/2

+120 state

-120 state

Optical axis direction

fs7.16sin2⋅== n

cdnn θτ

(2)



where d=17µm is the mirror pitch, and θ = 12Ο is the tilt angle. The full impulse response is a superposition of all the individual mirror responses:

where h[n] is the fraction of the illuminated mirrors turned to +120 on the n-th row (modulated by the input beam distribution). Here h[n] is arbitrary within the constraints 1][0 ≤≤ nh , but it can be shown that any arbitrary complex-valued filter can be synthesized to a multiplicative constant under some bandwidth and resolution constraints that depend on N and τ [5]. Furthermore, note that the 1/N factor in (2) implies a power loss factor on the order of N on average, implying a trade-off between filter complexity and power efficiency. This power loss is inherent to all diffractive filters and is not particular to the DMD implementation described here.

Spectral Filters

The spectral filter H(f) corresponding to h(t) will be periodic because h(t) is discrete. Each period constitutes a spectral range of size

θλ 2sin

21d

FSR =∆= (4)

Furthermore, since h(t) is real, H(f) will exhibit hermitian symmetry. However, half of each FSR will still be arbitrary, and any N-tap FIR filter can be designed on either half, subject to some excess power loss [3]. The spectral resolution of such a filter is FSR/N.

Figure 2. Synthesized spectral power reflectances generated by the DMD array

( )τδ ntnhN

thN

n−= ∑

=1][1)( (3)



Experimental measurements of some synthesized filters are shown in Fig. 2, along with the theoretical curves. The theoretical curves show ( ) 2λH , with FSR as a fitting parameter, which yielded half-FSR of 775nm-861nm, in good agreement with Eq. (4). The peak at 775nm corresponds to the left edge of FSR, and is always present.

Figure 2a shows single band-pass peaks in various locations, and 2b shows that relative intensities of two peaks can be varied. A 30x30 sub-array of mirrors centered on the laser beam was used in these measurements. The broadening of the measured peaks is due to the large acceptance angle at the photodetector. (The artifacts around 787nm in 2b are likely triggering errors.) In Figure 2c, we demonstrate spectral phase control by generating two adjacent peaks in and out of phase. A 20x20 sub-array was used, increasing the available acceptance angle and hence eliminating peak broadening, but reducing total power.

Femtosecond Pulse Shaping

Femtosecond pulse shaping systems are used to generate ultra-fast optical bursts with user-specified time waveforms and are needed for many applications [7]. Spectral filters are the enabling element of such pulse shapers, and our method of filtering is particularly well-suited for this application because we can control both spectral amplitude and phase. For a full discussion of advantages and disadvantages compared to traditional systems, see [ 6] and [8].

To synthesize an arbitrary desired pulseshape, one simply needs to create an impulse response that is congruent to the desired pulseshape (see (1)). Fig. 3 shows experimental measurements of pulseshapes created by the DMD array, along with the theoretical curves. Two sample waveforms are shown: 1,1…1 and 1…1,0…0,1…1 (period=50), on the left and right, respectively. Both are windowed by the nearly top-hat, 2.7mm diameter transverse beam profile, which illuminates 224 rows.

(a) -- theory (b) -- experiment

Figure 3(a) Two examples of pulseshapes and their corresponding power spectra. The input pulse and its spectrum are superimposed in grey. (b) Same measured pulse shapes with the input pulse superimposed.

The optimal resolution of the DMD array is just τ = 16.7fs (see (2)), provided the input pulsewidth is not longer. The resolution in Fig. 3 is limited by the 100fs input pulse shown in grey. The pulseshape duration in Fig. 3 is roughly 3ps and is limited by the input beam size. The maximal duration is roughly 7ps, which corresponds to most of the array being illuminated.



References M. B. Sinclair, M. A. Butler, A. J. Ricco, and S. D. Senturia, “Synthetic spectra: a tool for correlation

spectroscopy,” Applied Optics 36, 3342-3348 (1997). G. B. Hocker, D. Youngner, M. Butler, M. Sinclair, T. Plowman, E. Deutsch, A. Volpicelli, S.

Senturia, A. J. Ricco, “The polychromator: a programmable MEMS diffraction grating for synthetic spectra,” Solid State Sensor and Actuator Workshop, Hilton Head Island, SC, June 4-8, 2000, pp. 89-91.

R. Belikov, O. Solgaard, "Optical Wavelength Filtering by Diffraction from a Surface Relief”, Opt. Lett., vol 28, No. 6, pp. 447-449, March 2003.

L. A. Yoder, W. M. Duncan, E. M. Koontz, J. So, T. A. Bartlett, B. L. Lee, B. D. Sawyers, D. A. Powell, P. Rancuret, “DLPTM Technology: Applications in Optical Networking,” Proceedings of the SPIE, 2001, vol. 4457, pp. 54-61.

R. Belikov, X. Li, O. Solgaard, “Programmable Optical Wavelength Filter Based on Diffraction from a 2-D MEMS Micromirror Array,” Proceedings of CLEO, Baltimore, MD, June 1-6, 2003.

R. Belikov, C. Antoine-Snowden, O. Solgaard, “Femtosecond Direct Space-toTime Pulse Shaping with MEMS Micromirror Arrays,” Optical MEMS, Waikoloa, HI, August 2003.

A. M. Weiner, “Femtosecond Pulse Shaping Using Spatial Light Modulators,” Review of Scientific Instruments, Vol 71, No. 5, May 2000.

D. E. Leaird, A. M. Weiner, "Femtosecond Direct Space-to-Time Pulse Shaping”, IEEE JQE, vol 37, No. 4, pp. 494-504, April, 2001.



Development of segmented deformable mirrors for free space communications

U. Krishnamoorthy, I.W. Jung, P. Lu, Y.-A. Peter, E. Carr, R. L. Byer, O. Solgaard

E.L. Ginzton Laboratory, Stanford University, CA.

Optical free-space communication is of interest to a variety of civilian and military projects because it enables high-speed links between earth, satellites, and planes (Fig. 1). Optical signals are, however, compromised by atmospheric turbulence. The wavefront of light is distorted due to non-homogeneity of the refractive index of air. Long-distance free-space optical links therefore require dynamic correction of these time varying wavefront aberrations. Wavefront correction is achieved by reflecting the emitted optical beam from the receiver back to the transmitter, where its wavefront is measured. The wavefront at the receiver is then optimized by predistorting the emitted laser beam using a deformable mirror. The feedback loop, including the deformable mirror, must have a fast response time, in the range of 10µs, to effectively compensate for the varying conditions of the atmosphere.

We are developing segmented, deformable mirrors for this application. We are collaborating with other academic institutions, industry, and Lawrence Livermore National Laboratory under DARPA’s Coherent Communications, Imaging, and Targeting (CCIT) initiative. The aim of this project is to develop secure, multi-gigabit/s, free-space communication links and aberration-free three-dimensional imaging at ranges larger than 1,000 km.

MEMS-Deformable Mirrors (DM) that fit these applications include microfabricated membrane mirrors2, continuous face sheet mirrors backed by individual actuating elements3 and segmented mirrors4. Segmented deformable mirrors offer several advantages including smaller size and lower cost over conventional adaptive optics mirrors1. Although they suffer some losses from diffraction effects relative to continuous membrane mirrors, the segmented deformable mirrors are capable of much higher speeds. We are focusing on the design of scalable microfabricated mirror arrays for these applications to maximize the speed advantage, to simplify fabrication and minimize cost.

The specifications for the CCIT deformable mirror are as follows: 1) 100 kHz response, 2) 0.75 µm of stroke (piston motion), which is λ/2, 3) λ/50 surface quality, 4) scaleable to 1 million segmented actuators/degrees of freedom, and 6) a limit of 100V for the drive voltage. We first designed mirror arrays driven by piston-style electrostatic actuators. We developed mechanical and electrostatic

Figure 1: Free space communication between earth,

200µ

Figure 2: 4x4 segmented mirror array fabricated in the MUMPS process. Mirrors are 200x200µm.



models that met the specifications and fabricated the mirrors in a commercial process called the Multi-User MEMS Process (MUMPs®). We fabricated 4x4 arrays to test the electro-mechanical theory of our actuators (Fig. 2). We developed post-processing technologies to assemble the final deformable mirror arrays (Fig. 3). These included a gold thermo-compression-based flip-chip bonding process, an isotropic Xenon difluoride-based substrate release process. We successfully tested the static and dynamic responses of these structures. The mirror arrays were characterized using a ZYGO white-light interferometer. We measured mirror radius of curvatures as large as 5cm and surface roughness around 40nm. Deflections as high as 1 micron were measured for applied voltages < 200 V.

Although the mirror arrays met most of the design specifications, the optical surface quality of the mirrors needed improvement. Hence we developed an independent process to replace polysilicon mirrors with single crystal silicon mirrors. We successfully demonstrated 5x5 mirror arrays in this process that met all our design specifications. Deflection characteristics are shown in Fig. 4. Mirror radius of curvature >1m and surface roughness <30nm were measured (Fig. 5).

Next, to demonstrate the scalability of our mirror arrays, we fabricated 32x32 silicon mirror arrays in the same process (Fig. 6). While the fabrication process scaled well, the XeF2-based substrate release step did not. Weak sections of oxide membrane in the structure cracked during the release etch leading

Figure 4: Deflection characteristics of single crystal silicon mirrors in 5x5 array

Figure 3: Post-processing includes gold-thermocompression flip-chip bonding of the mirror arrays to the actuators followed by a substrate release and an HF-based sacrificial oxide layer removal.

or

Walls between pixels eliminate cross talk

Flip chip Bond & Release

Mirr (Bulk Si)

bond padActuator

(polysilicon) Gold

Figure 6: Photograph of 32x32 silicon mirror array fabricated using an SOI-based process. Size of each individual pixels is ~ 200x200µm.

Figure 5: Surface profile of 5x5 array of silicon mirrors measured on a white light interferometer (WYKO). Measured Ra=30nm rms, 1m radius of curvature.



to loss of structures. Additionally the thermo-compression assembly step for the larger 32x32 array chips requires a larger force for adequate bonding. Hence we are currently in the process of improving our bonding procedure. We are also developing a release process that stiffens the oxide membrane by patterning the substrate prior to release. We expect the new single crystal silicon mirrors to meet the scaling (1000x1000), speed (10µs), and optical surface quality specifications (rms of less than 30nm) required for multi-gigabit/s communications and near diffraction-limited imaging at ranges larger than 1,000 km.

To increase the functionality of these segmented deformable mirrors, we designed new tip-tilt mirror arrays that have the same capabilities as their piston-actuated predecessors but can also be tilted by large mechanical angles (-10 < θ <+10) in two-dimensions. While the tip-tilt motion provides beam-steering capability for better targeting, the piston motion provides λ/2 phase correction at 1.55µm. These devices will demonstrate fast response times of 10µs (piston) and 100µs (tilt) and low actuation voltages (< 200V). Additionally, these mirror elements have a high fill factor (> 94%) while maintaining a high quality mirror surface.

Our mirror designs use vertical comb drive actuators to maximize force and simplify mechanical and electrical isolation between electrodes. We have completed the modeling and design of these structures to meet the above specifications. We have already developed a new bottom-up scheme to fabricate these mirrors without the need for flip-chip bond-based post-processing assembly. Currently, we are in the process of fabricating these devices.

References: [1] R. K. Tyson, "Adaptive Optics Engineering Handbook." New York, NY: Marcel Dekker, Inc.,

2000, pp. 339. [2] G. Vdovin, S. Middelhoek, P. M. Sarro, “Technology and applications of micromachined silicon

adaptive mirrors,” Opt. Eng., vol. 36, no. 5, pp 1382-1390, 1997. [3] T. G. Bifano, J. Perreault , R. Krishnamoorthy Mali, M. Horenstein, “Microelectromechanical

Deformable Mirrors,” IEEE Journal of Selected Topics in Quantum Electronics, vol. 5, no. 1, pp 83-89, 1999.

[4] W. D. Cowan, M. K. Lee, B. M. Welsh, V. M. Bright, M. C. Roggemann, “Surface Micromachined Segmented Mirrors for Adaptive Optics,” IEEE Journal of Selected Topics in Quantum Electronics, vol. 5, no. 1, pp 90-101, 1999.

SiSOISiO2

Vertical comb-teeth

Inner torsional spring

Ground layer

Bottom View

Outer torsional spring

SiSOISiO2

SiSOISiO2

Vertical comb-teeth

Inner torsional spring

Ground layer

Bottom View

Outer torsional spring

Vertical comb-teeth

Torsional springs

Mirror base

Top View

Electrode 2

Electrode 3

Electrode 4

Ground

Oxide Layer

Electrode 1

Vertical comb-teeth

Torsional springs

Mirror base

Top View

Electrode 2

Electrode 3

Electrode 4

Ground

Oxide Layer

Vertical comb-teeth

Torsional springs

Mirror base

Top View

Electrode 2

Electrode 3

Electrode 4

Ground

Oxide Layer

Electrode 1

Figure 7: Two-dimensional tip/tilt mirror design layout. These devices are driven by vertical comb drive actuators and are capable of vertical piston motion >1 mm and a tw-dimensional tilt of +/- 10 degrees.



Single-crystalline silicon micromirrors actuated by self-aligned vertical electrostatic combdrives with piston-motion and rotation capability

D. Lee, U. Krishnamoorthy, H. Ra, O. Solgaard

Scanning micromirrors enable a variety of optical systems including displays, confocal microscopes, optical coherence tomographs, and optical fiber switches. Among the advantages of MEMS technology over more traditional fabrication methods for these applications are higher speed, lower mass, and the potential for lower cost through parallel processing. The optical apertures that can readily be realized using MEMS technology are not sufficient for many important applications, however. A promising solution is to use arrays of micromirrors that can be controlled with interferometric precision to form phased arrays [1]. Controlling the relative phase of the light being reflected from different scanning micromirrors in the array requires two degrees of freedom of actuation of each mirror. The mirrors must both be able to perform the scanning function and must be able to move vertically to adjust the phase of the reflected light during the scan. Such micromirrors have been demonstrated in surface micromachining [2] and with double-sided Deep-Reactive Ion Etching (DRIE) [3].

We present the design, fabrication process, and experimental results of dual-mode micromirrors for phased array operation based on front-side-only processing of Silicon-on-Insulator (SOI) material. The micromirrors are actuated by vertical combdrives that have been shown to enable high-resolution, high-speed scanning micromirrors [4]. In the process described here, the vertical combdrives are self-aligned, resulting in high yield and stable operation even with relatively small electrode gaps [5,6]. Two layers of device silicon are used to allow the required two degrees of actuation, as well as two-sided operation for maximum scan angle. Two oxide layers between the device layers provide electrode isolation and etch stops for thickness control.

The fabrication process for the dual-mode, SOI micromirrors with vertical combdrives is illustrated in Fig. 1. The process starts with coarse patterning of the bottom combteeth (Mask 1) in SOI wafers with 50 µm device-layer thickness (Fig. 1a). Next, 0.4 µm of thermal oxide is grown on an unpatterned silicon wafer. The unpatterned wafer is bonded by thermal fusion bonding to the patterned SOI wafer. This step is followed by grinding and polishing of the unpatterned wafer to create a second device layer of about 30 µm thickness (Fig. 1b). Next, an alignment mask (Mask 1A), consisting of two rectangular windows, is used to etch through the upper silicon layer and expose the alignment marks in the lower silicon layer for subsequent mask alignment steps.

Two mask layers are then deposited and patterned to avoid resist spinning on high-aspect ratio structures. The first (Mask 2) of the two masks is required to provide electrical connection to the lower combteeth. First, a low-stress nitride film (1.0 µm) is deposited and patterned to define contact areas for the lower comb (Fig. 1c). Then a low-temperature oxide (LTO) film (0.4 µm) is deposited and this LTO and the previously patterned nitride film is patterned to define the self-aligned upper and lower combteeth (Mask 3) (Fig. 1d). The upper silicon layer is then etched to define the upper combs. The LTO mask, as well as exposed oxide between the device layers, on the sidewalls of the coarse lower combteeth, and on the backside of movable structures defined in the upper silicon layer, are



removed in a timed vapor HF etch (Fig. 1e). The aligned lower combteeth and their contact holes are then etched by DRIE using the previously patterned nitride layer as the etch mask (Fig. 1f). Finally, the remaining nitride film and exposed parts of the middle oxide layer are removed. Figure 2a shows an SEM of the self-aligned combteeth, and Fig. 2b shows the top view of a finished device.

Figure 1: Process flow: (a) DRIE of coarse lower combteeth (Mask 1). (b) Thermal oxidation, fusion bonding, grinding and polishing to create the upper device layer. (c) Definition of contact areas for the lower combteeth (Mask 2). (d) LTO deposition and patterning (Mask 3). (e) DRIE of the upper device layer followed by vapor HF etching. (f) DRIE of the lower comb defined by Mask 3 and the contact holes defined by Mask 2.

Figure 2: (a) SEM picture of aligned combteeth after the process is completed. The gap between adjacent gaps is 6 µm. (b) SEM picture of the top view of a double-sided micromirror. (c) SEM picture of aligned combteeth after the upper fixed combteeth are removed. The gap between adjacent combs is 6 µm.

If the upper fixed combteeth must be removed, they will be etched using the DRIE step that creates contact holes to the lower combs (Mask 2). Figure 2c shows an SEM of the self-aligned combteeth after removal of the upper fixed combteeth. We observe a reduction of the width of the lower fixed combteeth and some roughness of the sidewalls. The reason is undercutting of the intermediate oxide layer during the process that limits the exact pattern transfer from the upper silicon to the lower layer. The alignment of the combteeth is not affected, however. In both variations of the process (with and without removal of the upper fixed combteeth) we easily achieved combdrives with 4 µm gaps.

The multiple isolated electrodes that can be created in the two isolated device layers enable a variety of modes of operation. The two device layers we use in our process allow us to create 5-electrode

(a) (b) (c)

Movable combteeth

Fixed upper combteeth

Fixed lower combteeth

Oxide layer

A

Contact areas to the lower combteeth

B

Micromirror Vertical combdrive Movable

combteeth

Fixed lower combteeth

(a) (b)

(d)

(c)

(e) (f)

Masking nitride filmSilicon layers Masking LTO filmThermal oxide



devices as shown in Fig. 3. Among the extra capabilities enabled by the 5-electrode devices are lateral combdrives and capacitive position sensors for closed loop control.

In most of our devices, we create a 3-electrode device from the 5-electrode device of Fig. 3 by connecting the upper fixed combteeth to ground. With electrode 5 (the movable comb) also connected to ground, bi-directional rotation is achieved by applying a driving voltage on electrode 1 or 2. Pure piston motion is achieved by applying the same voltage on both electrodes 1 and 2. Using this actuation scheme, we have achieved two-sided, static optical deflection from –9.2 degree to +9.2 degree with 155 V actuation voltage. The static vertical deflection ranged from 0 µm to –7.8 µm with 110 V actuation voltage.

Rotation and piston motion is achieved by applying voltages to electrodes 1 and 2 only, but extra electrodes (Fig. 3c) add flexibility and increase the range of combined rotation and piston motion. Electrode 3 or 4 can be used to adjust the rotation of mirrors brought to a specific vertical position by electrode 1 and 2.

Figure 3: Multi electrodes device: (a) Schematic cross-section along AB (see Fig. 2b) with no actuation. (b) Pure-piston motion is achieved by applying the same voltage to electrode 1 and 2. (c) Electrodes 3 and 4 add flexibility in the adjustments of tilt and piston motion as shown.

References: [1] K. Li, U. Krishnamoorthy, J.P. Heritage, O. Solgaard, “Coherent micromirror arrays”, Optics

Letters, vol. 27, no. 5, Mar. 2002, pp. 366-368. [2] U. Krishnamoorthy, K. Li, K. Yu, D. Lee, J.P. Heritage, O. Solgaard, “Dual-mode Micromirrors

for Optical Phased Array Applications”, Transducers 2001, Munich, Germany, Jun. 2001. [3] V. Milanovic, S. Kwon, L. P. Lee, “Monolithic Vertical Combdrive Actuators for Adaptive

Optics”, 2002 IEEE/LEOS International Conference on Optical MEMS, Lugano, Switzerland, Aug. 2002, pp. 57-58.

[4] R.A. Conant, J. Nee, K.Y. Lau, R.S. Muller, “A Fast Flat Scanning Micromirror”, Solid-State Sensor and Actuator Workshop, Hilton Head, SC, USA, Jun. 2002, pp. 6-9.

[5] U. Krishnamoorthy, O. Solgaard, Self-Aligned Vertical Comb-drive Actuators for Optical Scanning Micromirrors, 2001 IEEE/LEOS International Conference on Optical MEMS, Okinawa, Japan, Sept. 2001.

[6] U. Krishnamoorthy, D. Lee, O. Solgaard, “Self-Aligned vertical electrostatic combdrives for micromirror actuation”, Journal of Microelectromechanical Systems, vol. 12, no. 4, Aug. 2003,

pp. 458-464.

(a) (b) (c)

Electrode 1 Electrode 2 Electrode 3 Electrode 4 Electrode 5

Electrode 1 Electrode 2

Electrode 3 Electrode 4Electrode 5

Electrode 1 Electrode 2

Electrode 3 Electrode 4Electrode 5



Flat-top Reflection and All-pass Transmission Filter using Coupled Resonance in Photonic Crystal Slabs

Wonjoo Suh and Shanhui Fan Department of Electrical Engineering, Stanford University, Stanford, CA 94305 [email protected]

In communication systems, an all-pass transmission filter, which generates significant delay at resonance, while maintaining 100% transmission, is useful for applications such as optical delay or dispersion compensation. Also, a narrow-band flat-top reflection filter is needed to achieve the wavelength selectivity in wavelength division multiplexing systems. Here, we introduce a novel mechanically tunable photonic crystal structure consisting of coupled photonic crystal slabs, which can generate both flat-top reflection and all-pass transmission responses in a single device.

The structure, as shown in Figure 1, consists of two photonic crystal slabs. Each slab is constructed by introducing a periodic array of air holes into a high-index guiding layer. Unlike all previously reported all-pass reflection filters based on Gires-Tournois interferometers, our structure generates all-pass transmission spectrum, which significantly simplifies signal extraction and optical alignment.

Figure 1. Schematic of a mechanically tunable photonic crystal filter consisting of two photonic crystal slabs.

Figure 2. Transmission spectrum through a single photonic crystal slab for normally incident light. The open circles are FDTD simulation results and the solid line is analytic theory.

Guided resonance, which is responsible for the spectral function of our device, is a class of optical mode that is strongly confined by the dielectric slab, and yet can couple into radiation modes due to the phase matching mechanism provided by the periodic index contrast 00. As light is normally incident upon the slab, the wave can either pass through the slab directly, or indirectly by exciting the resonance. The interference of these two pathways determines the spectral response of light through the photonic crystal slab. In the particular case where the transmission through the direct pathway is unity, the interference is forced to be destructive due to energy conservation. Therefore the reflection from the crystal exhibits a Lorentzian line shape with a 100% reflectivity at the resonant frequency 0. Such a Lorentzian line shape can be seen in Figure 2 for the transmission through a single slab 0, with a dielectric constant of 11.4 (which is appropriate for GaAs at 1.55µm), a thickness of 1.05a, and a radius of 0.1a for the air holes (a is the lattice constant).



We are able to create various filter functions when we have two of these slabs to form a coupled resonator system. Let’s first consider an all-pass transmission filter for optical delay applications. To accomplish an all-pass characteristic with no intensity variation over the resonance bandwidth, it is necessary to use at least two resonances since a single resonance can only generate optical delay with a strong variation of the transmitted intensity as a function of frequency. For simplicity, we consider a system with two identical slabs, as shown in Figure 1. Each slab supports a single resonance within the bandwidth of interest. Since there is mirror symmetry parallel to the slab, the resonant modes of the coupled system can be decomposed into either even or odd mode with respect to the mirror plane. Since the two modes have different symmetry, the decaying amplitudes to the forward and backward direction acquire opposite phase and interfere destructively. Therefore, complete transmission over the bandwidth of interest becomes possible, provided that the even and odd modes possess the same resonant frequency and the same width 0. To achieve this, we place two slabs close to each other, such that the modes in the two slabs can couple through an evanescent tunneling pathway, in addition to the free space propagation of light between the slabs.

For physical realization, we consider the two slab structure each of which has been studied in Figure 2, with a resonance frequency at 0.694 (c/a). In a finite-difference time-domain (FDTD) simulation, the line shape of even and odd modes can be obtained by Fourier-transforming the temporal decay of the resonance amplitude. By choosing the displacement between the slabs to be 0.4a, the resonant line

Figure 3. (Left) Spectral response for the two slabstructure (a) Resonance amplitudes of the even mode (dotted) and the odd mode (solid). (b) Transmission spectrum (c) Group delay.

Figure 4. (Right) Transmission spectra through the two slab structure as we vary the distance betweenthe slabs to be (a) 0.5a, (b) 1.1a. The solid line represents the theory and the open circles correspond to FDTD simulations.



shapes of the even and odd modes are almost completely overlapping (Figure 3(a)). For such a structure, the transmission spectrum indeed shows near 100% transmission over the entire bandwidth both on and off resonance (Figure 3(b)); and yet a large resonant delay is generated in the vicinity of the resonant frequency (Figure 3(c)). Moreover, the spectral response function of the two-slab structure can be tuned by mechanically varying the distance between the slabs. As we increase the distance between the slabs, the evanescent coupling becomes negligible and it is no longer possible to generate all-pass transmission. Rather, significant reflection occurs in the vicinity of the resonance. (Figure 4(a)) In particular, by choosing h=1.1a, one could generate a flat-top reflection spectrum as demonstrated by FDTD simulations in Figure 4(b). Thus, with mechanical tuning, a guided resonance device can generate two types of filter responses that are useful for optical communication systems.

This work was partially supported by the US Army Research Laboratories under Contract No. DAAD17-02-C-0101, and by the National Science Foundation (NSF) grant ECS-0200445. The computational time was provided by the NSF NRAC program. We acknowledge Mehmet Fatih Yanik for developing the software code used in this work.

References M. Kanskar, P. Paddon, V. Pacradouni, R. Morin, A. Busch, J. F. Young, S. R. Johnson, J. Mackenzie,

and T. Tiedje, Appl. Phys. Lett. 70, 1438 (1997). S. Fan and J.D. Joannopoulos, Phys. Rev. B, 65, art. No. 235112 (2002). R. Magnusson and S.S. Wang, Appl. Phys. Lett. 61, 1022 (1992). S. Fan, W. Suh and J. D. Joannopoulos, J. Opt. Soc. Am. A 20, 569 (2003). S. Fan, P. R. Villeneuve, J. D. Joannopoulos and H. A. Haus, Phys. Rev. Lett., 80, 960 (1998).



Elastomer Spatial Light Modulators for Extreme Ultraviolet Lithography

J. Wang, I. Jung, O. Solgaard

Extreme ultraviolet (EUV) lithography is one of the most promising technologies for fabricating electronic devices with critical dimensions (CDs) smaller than 50nm. Due to the short wavelength (13nm), it is hard to make a defect-free mask and protect the masks with pellicles. This makes the cost of masks extremely high. As a consequence, alternatives to the traditional mask technology are demanded. A promising approach is to use spatial light modulators (SLM) to replace masks in EUV Lithography1. The schematic architecture of a maskless EUVL system is shown in Fig. 1. The image pattern on the wafer is modulated by a two-dimensional micromirror array, in which the image can be controlled electronically. The light from a pixel that is phase-shifted by π radians with respect to its surroundings will diffract outside the numerical aperture of the imaging optics and will appear dark in the image pattern.

2-D SLM

CondenserOpticsEUV

Source

Imaging Optics

Wafer

Mechanical Scan

2-D SLM

CondenserOpticsEUV

Source

Imaging Optics

Wafer

Mechanical Scan

π/2 phase shiftπ/2 phase shiftπ/2 phase shift

Electrode

Elastomer

Reflective Multilayer

Silicon Substrate

Electrode

N 81 layersR~70%

4 nm2.8 nm

MolybdenumSilicon

MolybdenumSilicon

MolybdenumSilicon

Interconnect

0.2µm

1µm

N

0.1 ~ 2µm

1µm0.3µm

Electrode

Elastomer

Reflective Multilayer

Silicon Substrate

Electrode

N 81 layersR~70%

4 nm2.8 nm

MolybdenumSilicon

MolybdenumSilicon

MolybdenumSilicon 4 nm

2.8 nm

MolybdenumSilicon

MolybdenumSilicon

MolybdenumSilicon 4 nm

2.8 nm

MolybdenumSilicon

MolybdenumSilicon

MolybdenumSilicon

Interconnect

0.2µm

1µm1µm

N

0.1 ~ 2µm0.1 ~ 2µm

1µm1µm0.3µm0.3µm

Figure 1. The architecture of a maskless EUV lithography system. A collimated EUV beam is modulated by a two-dimensional micromirror array, and then focused on the wafer. When all the pillars are in the same plane, incident light is specularly reflected. When some of the pillars are depressed, the incident light is diffracted.

The design and fabrication of SLMs for EUV lithography pose several technological challenges. The optical surfaces must be very smooth and compatible with Mo/Si multilayer technology to achieve high reflectivity. Large arrays are required so the fabrication process must allow direct integration with electronics for multiplexing. Finally, the individual pixels must be as small as possible to minimize the need for de-magnifying optics in the EUV. To meet these challenges we have developed an SLM-fabrication process as shown in Fig. 2. Our process allows the formation of very small pixels (limited by the electrode size) with polished surfaces in a process with few high-temperature steps so it can be vertically integrated with electronics, thus meeting the requirements of EUV maskless lithography. Meanwhile, because only one common top electrode is used, the wire connection and multiplexing are simplified. It should also be pointed out here that before filling the cavity defined by the nitride shell, the process does not contain any elastomer materials. This not only circumvents material incompatibilities and contamination, but also allows testing of different elastomers without changing the process.



1. Deposit and pattern bottom electrodes on insulation layer

2. Deposit, pattern, and polish the sacrificial layer

Silicon Substrate

Silicon Substrate

Silicon Substrate

3. Deposit and pattern the nitride shell and the top electrode

Silicon Substrate

4. Deposit and pattern the Mo/Si multilayer

1. Deposit and pattern bottom electrodes on insulation layer

2. Deposit, pattern, and polish the sacrificial layer

Silicon SubstrateSilicon Substrate



3. Deposit and pattern the nitride shell and the top electrode

Silicon SubstrateSilicon SubstrateSilicon Substrate

4. Deposit and pattern the Mo/Si multilayer

Mo/Si Multilayer

Oxide

Nitride

Poly1

Poly2 (Top Electrode)

Elastomer

Silicon Substrate

Silicon Substrate

5. Release the sacrificial layer to form a cavity

6. Inject the elastomer into the cavity by capillary forces

Mo/Si Multilayer

Oxide

Nitride

Poly1


Elastomer Mo/Si Multilayer

Oxide

Nitride

Poly1


Elastomer



5. Release the sacrificial layer to form a cavity

6. Inject the elastomer into the cavity by capillary forces

Figure 2. Process flow of elastomer spatial light modulators One critical step of the process is the elastomer replacement. The injection test is performed with materials with different viscosities and the results are listed in Table 1. The injection time, as expected, is highly related to the viscosity of the material, but the results show that injection in shallow channels is possible even with relatively high viscosity elastomers. The surfaces of a SLM before and after the elastomer injection were measured under a white-light interferometer (Fig. 3). The result shows that the surface profile is not adversely affected by the injection.

Table 1. Injection time of different elastomers

Material Viscosity (mPa-Sec) Injection time Dow Corning Sylgard 527 450 ~3 minutes Dow Corning Sylgard 182 5000 ~27 minutes Masterbond Mastersil 773 60-70 <1 minutes

The operation principle requires isolated piston motion of each pixel, or localized deformation, in the elastomer SLMs. Fig. 4 shows the localized deformation when one pixel of a 4 by 4 SLM is actuated with 160 V DC. The surface profiles at 40V to 200V DC with an interval of 40V are shown in Fig. 5. Along with a main lobe, two side lobes with significant opposite deflection are evident. The formation of the side lobes is caused by the large ratio of the deformation to the thickness of the elastomer (~6% when the voltage is 200 volts). The area of the deformed region (~100µm by 100µm) is substantially larger than the pixel size (20µm by 20µm), but the result shows that the restoring force provided by the elastomer allows localized deformations of the membrane. To reduce the area of deformation, the thickness, and therefore the strength, of the membrane must be reduced. Our theoretical analysis indicates that, in a practical EUV SLM with pixel sizes of 1 by 1 µm or less, the Mo/Si layer, as well as the nitride shell and the top electrode, must be pixilated to reduce the strength of the top membrane.



0 200 400 600 800 1000

-1500

-1000

-500

0

500

1000

1500

2000

2500

Position (µm)

Def

orm

atio

n (n

m)

(a) before injection(b) after injection

(a) (b) (c)

Figure 3. Surface profiles of elastomer SLM before (a) and after elastomer injection (b). 1-D profiles from (a)

and (b) are shown in (c). The membrane is not adversely affected by the injection.

Figure 4. The surface profile of an elastomer SLM when one pixel is actuated with 160 volts.

0 50 100 150 200 250 300-120

-100

-80

-60

-40

-20

0

20

40

Position (µm)

Def

orm

atio

n (n

m)

40v80v120v160v200v

Figure 5. The surface profiles at a cross-section of an elastomer SLM when one pixel is actuated with 40 volts

to 200volts.

References: [1] N. Choksi, D. S. Pickard, M. McCord, R. F. W. Pease, Y. Shroff, Y. Chen, W. Oldham, and D.

Markle; J. Vac. Sci. Technol. B 17(6) (1999) 3047

300µm

Localized Deformation

Membrane

α γβ

α γβ α γβ



MEMS Confocal Microscope with Dual Axes Architecture for In Vivo Molecular and Cellular Imaging

Thomas D. Wang,1 Michael J. Mandella,4 Ning Y. Chan,4 Christopher H. Contag,2 Gordon S. Kino3 Department of Medicine,1 Department of Pediatrics,2 Department of Elect Engineering,3 Stanford University, Optical Biopsy Technologies, Inc4

Abstract

We present a MEMS confocal microscope that has dual axes architecture and post-objective scanning that is well-suited for collection of fluorescence and reflectance images in vivo. This design uses two low numerical aperture lenses to achieve high axial resolution and long working distance (WD). The scanning mirror located distal to the lenses to produce arc-surface images over a large field of view (FOV). With fiber optic coupling, this microscope can be scaled down to millimeter dimensions with standard optical micro-electro-mechanical systems (MEMS) fabrication techniques.

Background

Confocal microscopy is a powerful tool for imaging in biological tissue, and can be used in vivo if the lenses and scanning mechanisms can be made sufficiently small. This method performs optical sectioning in a highly scattering media with sub-cellular resolution, high signal-to-noise ratio (SNR), and good contrast.1 At present, in vivo confocal imaging is performed on exposed tissue surfaces such as that of skin.2,3 Internal access to tissue is limited by the large physical dimensions of conventional microscopes objectives. In the conventional single axis configuration, the same lens is used for illumination and image collection, and the optics cannot be reduced to millimeter scale easily. Also, the scanning mechanism is located proximal to the objective, thus multiple optical elements are required to correct for aberrations and achieve high performance. Furthermore, the use of standard objectives in vivo is rather cumbersome.

Previously, a laser scanning confocal microscope with single axis configuration has been miniaturized with MEMS technology.4,5 This instrument uses a single mode optical fiber as the pinhole, and the optical path is folded into a zig-zag pattern by a silicon spacer coated with reflective surfaces, resulting in a scanhead size of 1.2 x 2.5 x 6.5 mm. The axial resolution of 18 µm was not adequate for in vivo use, and the binary lens objective was too dispersive to collect fluorescence. We have redesigned the optics to improve the axial resolution to 2 µm using a dual axes architecture that has separate objectives for illumination and collection.6 In addition, with reflectance images, we have shown the ability of the dual axes configuration to differentiate against noise from scattering over a depth of 1 mm using illumination at 1345 nm.7 Furthermore, we have demonstrated fluorescence (non-coherent) images collected in horizontal cross-sections parallel to tissue surface using bi-axial post-objective scanning.8



Dual Axes Architecture

The dual axes uses separate low NA objectives for illumination (IO) and collection (CO) at angle θ, as shown in Fig. 1.9,10 The NA of each lens is determined by the angles αi and αc. The point spread function of either objective (gray ovals) has a long axial but narrow transverse dimension.

The overall axial resolution of the system is represented by the length of the overlapping region (black oval). The length of this overlap depends primarily on the transverse rather than the axial dimension of the beams at the focus. The illumination beam is incident to the tissue at angle θ, and the collection originating from the overlapping focal volume at the intersection of the two beams, is collected off-axis. As a result, light scattered along the illumination path (checkered region) outside of the focal volume is unlikely to arrive at the collection objective with the proper angle necessary for detection because the correct combination of scattering events required occurs with low probability. On the other hand, a standard confocal microscope objective has a single axis and requires a much higher NA to achieve an equivalent axial resolution. As a result, the light collected emerges from within the same large angle cone (dashed lines) as that traversed by the illumination, thus resulting in increased noise from scattering.

αc

αi

θ

I

C

.

i

θ

IO

CO

..

Tissue

α

WD

z

x

αc

αi

θ

I

C

..

i

θ

IO

CO

..

Tissue

α

WDWD

zz

xx

Fig. 1 Dual Axes Architecture.

MEMS Design

The MEMS design of the dual axes architecture uses similar parallel elements and zig-zag optical path to that of the first MEMS single axis prototype, as shown in Fig. 2a. The illumination exits the first fiber and reflects off a folding mirror, which is integrated into the walls of the silicon spacer, and then the rays are focused by an elliptical mirror on the lens plate to the uniaxial scan mirror, then directed through the index matching prism, and finally passing through the interface window, which is placed in contact with the tissue. The return path from the tissue is symmetric to the illumination path, thereby passing back through the interface window, water, prism, scan mirror, a second elliptical mirror, and a second folding mirror and finally focused onto the end of the collection fiber. The proper ray trace of this optical design was confirmed with ASAP ver 7.1 (Breault Research Corp), an optical modeling software.



The MEMS fabrication will be consist of four parts: 1) an integrated window with a prism shape made of Teflon or plastic, 2) a bulk-micromachined precision silicon spacer having reflective walls, 3) a silicon substrate containing the micromachined biaxial scan mirrors, and 4) a plastic molded lens plate containing elliptical focusing mirrors. The 4 parts are designed to be bonded together in the same fashion as that of the first MEMS prototype. The illumination and collection fibers are attached to V-grooves that are precision etched into the dish cavity. The micromirror provides accurate scanning in two dimensions at relatively high speeds. The optical surface has low roughness and curvature under both static and dynamic operation. The component assembly is shown in Fig. 2b. The optical fibers for illumination and collection are anchored within the inner housing and aligned with the rest of the optics by adjusting its position on an alignment plate. The inner housing slides into the outer housing of the scan head, which protects the optics and provides a seal from bodily fluids. The interface window is epoxied to the front face of the outer housing.

Fig. 2 a) Exploded view of zig-zag optical path, and b) Assembly of MEMS housing.

The dual axes design demonstrates several advantages over that of the single axis for purposes of miniaturization with MEMS. First, sub-cellular resolution can be achieved in the axial dimension. Second, the low NA lenses used are relatively less sensitive to aberrations and easy to fabricate. Third, a long WD is created which allows for a miniature mirror to be placed on the focused beam side of the objective to provide a large FOV. Fourth, light scattered along the illumination path outside of the focal volume is less likely to be collected, thus enhancing detection sensitivity and dynamic range. Finally, this method of imaging can be used for collection of non-coherent light such as fluorescence.

References

[1] J. Pawley, ed., Handbook Biological Confocal Microscopy, 3rd ed., Plenum, New York (1996).



[2] Langley RG, Rajadhyaksha M, Dwyer PJ, Sober AJ, Flotte TJ, Anderson RR. J Am Acad Dermatol 2001;45(3):365-76.

[3] Busam KJ, Hester K, Charles C, Sachs DL, Antonescu CR, Gonzalez S, Halpern AC. Arch Dermatol 2001;137(7):923-9.

[4] D. L. Dickensheets, G. S. Kino. J. of Microelectromechanical Systems 7, 38-47 (1998). [5] D. L. Dickensheets, G. S. Kino. Optics Letters 21, 764-66. (1996) [6] T. D. Wang, M. J. Mandella, C. H. Contag, G. S. Kino. Optics Letters 28, 141-44 (2003). [7] T. D. Wang, C. H. Contag, M. J. Mandella, N.Y. Chan, G. S. Kino. Optics Letters 2003;28, in

press. [8] Wang TD, Mandella MJ, Contag CH, Chan NY, Kino GS. Submitted, Journal Biomedical

Optics. [9] R. H. Webb, F. Rogomentich. Applied Optics 38, 4870-75 (1999). [10] S, Lindek, E. H. Stelzer. J. Opt. Soc. Am. 13, 479-82 (1996).



Tunable bandwidth optical filter based on MEMS Gires-Tournois interferometer

Kyoungsik Yu, Daesung Lee, Uma Krishnamoorthy, and Olav Solgaard Department of Electrical Engineering, Stanford University, Stanford, CA, USA

High-capacity, long-haul optical dense wavelength division multiplexed (DWDM) communication systems require narrow bandwidth optical filtering at the transmitter and/or receiver to optimize spectral efficiency and to improve signal quality by rejecting out-of-band noise1, 2. As the nonlinear penalties become more severe due to narrower WDM channel spacing, higher bit rate, and increasing transmission distance, the center wavelength, passband shape, and the bandwidth of passband of the optical filters must be tightly controlled. Recent experiment also shows that narrow bandpass filtering can mitigate waveform distortion due to higher-order polarization modal dispersion (PMD)3. For practical purposes, the peak wavelength of the narrow bandpass optical filter is often slightly detuned from the carrier wavelength of the optical signal to obtain the vestigial side band filtering of return-to-zero (RZ) modulation1, 4. Such systems require very precise tuning of the center wavelength on the order of sub nanometer. In addition to the tuning of the center wavelength, it is desirable to have the ability to tune the optical bandwidth of the optical filter. The optimal bandwidth of the optical filter is approximately equal to the bit rate1,2,3,4, but the exact value depends on many parameters including electrical filter bandwidth, PMD, and filter detuning.

Accurate real-time control of the center wavelength and bandwidth is therefore an attractive property of optical bandpass filters for pre- and post-filtering. In this paper, we demonstrate an optical filter, in which the center wavelength and 3-dB bandwidth can be tuned using MEMS technology. The proposed device has a periodic characteristic in the spectral domain with a free spectral range (FSR) matching the channel spacing of DWDM system. This spectral periodicity makes this type of device suitable for colorless narrow bandpass filters. When combined with wavelength multiplexers/demultiplexers with the channel spacing equal to the FSR of the device, for example, this optical filter can be used in any ports to tune the center wavelength and the bandwidth of individual WDM channel. Continuous tuning of the center wavelength and the FSR for these devices have been demonstrated earlier5. In this paper, the focus is mainly on variable optical bandwidth. Compared to filters implemented in planar lightwave circuit6, the devices reported here offer improved insertion loss, FSR, and speed of tuning.

A schematic diagram of the device structure and the experiment setup is shown in Fig. 1. By replacing the back reflection mirror plane in a Gires-Tournois (GT) interferometer5 with MEMS-tunable micromirrors and modulating the phases of emergent light beams, the spectral characteristic of the filter can be tuned. The amounts of phase-shift, given by the positions of the micromirrors, are selected to optimize the filter performance by generating the desired spectral bandwidth and, in general, the shape of passband. Less than 1 micron of translational motion on one axis is required to modulate the phase of the optical beams. The FSR of the device is controllable by changing the distance between beam splitter and micromirrors in Fig. 1. A one-dimensional array of six micromirrors was fabricated using deep-reactive-ion-etching of silicon-on-insulator material7. The 300micron by 500micron micromirrors were electrostatically actuated by vertical combdrives for simple up-and-down piston motion. An array of six micromirrors imaged by an interferometric optical



profilometer is shown in the inset of Fig. 1. The dynamic characteristic of the micromirrors was measured by stroboscopic optical profilometry. The magnitude response in Fig. 2 shows a natural frequency of ~15kHz. Additional mirrors can be used for more degrees of freedom in the tuning of the filters, although that will lead to more difficult alignment and packaging. The first surface of the beam splitter in the interferometer has a reflectance and transmission of 0.31 and 0.62 respectively. The second surface was wedged and anti-reflection coated to prevent multiple reflections within the beam splitter.

For center wavelength tuning, all micromirrors in the array have to be moved by the same amount as described in 5. However, for variable bandwidth and shaping of passband in general, we need differential movement of the individual micromirrors with respect to the other mirrors in the array. Although there are many possible mirror configurations that can change the optical 3-dB bandwidth of the filter, one of the simplest ways is to move only the first mirror while maintaining the positions of other mirrors. Fig. 3 shows insertion loss curves for different input voltages to the first micromirror in the array. Good agreement between the experimental result (Fig. 3(a)) and the simulation based on Gaussian beam theory (Fig. 3(b)) was obtained. The change of 3-dB bandwidth for various input voltages to the first micromirror is summarized in Fig. 4. The 3-dB bandwidth of the optical filter was continuously controllable from 0.288nm to 0.508nm while maintaining the FSR of 0.8nm (100GHz). The required mirror deflection to obtain this tuning range was approximately a quarter of the wavelength. The micromirror deflection curve as a function of the input voltage is also shown in Fig. 4. To demonstrate the variable bandwidth as well as the center frequency tunability, we measured the insertion loss when the optical filter is connected with the standard arrayed waveguide grating multiplexer with 100GHz channel spacing. We were able to maintain the center frequency of the optical filter while changing the 3-dB optical bandwidth from 0.28nm to 0.42nm as shown in Fig. 5.

In conclusion, we have reported on a bandwidth-variable optical filter based on a tunable Gires-Tournois interferometer using MEMS actuators. When the FSR is 100GHz (0.80nm), the optical bandwidth of the device can be continuously tuned from 36GHz (0.288nm) to 63.5GHz (0.508nm). Fast tuning speed and wide optical bandwidth tuning range make this type of device suitable for signal-conditioning pre- and post-filters for high-bit-rate (>10Gb/s) DWDM optical communication systems.

Figure 1: Schematic diagram of the tunable bandwidth optical filter. Inset: Micromirror array imaged by WYKO optical profilometer.

Figure 2: The measured frequency response of the micromirror is shown with the best-fit theoretical curve.



Figure 3: Measured (a) and simulated (b) insertion loss as a function of input wavelength for different mirror configurations.

Figure 4: Optical 3-dB bandwidth and mirror deflection with respect to the applied voltage are shown.

Figure 5: Insertion loss as a function of wavelength when combined with AWG wavelength demultiplexer.

References [1] S. Bigo, “Improving spectral efficiency by ultra-narrow optical filtering to achieve multi-terabit/s

capacities”, Proc. OFC 2002, Paper WX3, Mar., 2002. [2] P.J. Winzer, M. Pfennigbauer, M.M. Strasser, and W.R. Leeb, “Optimum filter bandwidths for

optically preamplified NRZ receivers,” Journal of Lightwave Technology, vol. 19, no. 9, pp. 1263-1273, Sep., 2001.

[3] L. Moller, et. al., “Higher order PMD distortion mitigation based on optical narrow bandwidth signal filtering,” Photonics Technology Letters, vol. 14, no. 4, pp. 558-560, Apr., 2002.

[4] H. Kim and A.H. Gnauck, “10Gbit/s 177km transmission over conventional single mode fibre using a vestigial side-band modulation format,” Electronics Letters, vol. 37, no. 25, pp. 1533-1534, Dec., 2001.

[5] K. Yu and O. Solgaard, “MEMS optical wavelength deinterleaver with continuously variable channel spacing and center wavelength,” Photonics Technology Letters, vol. 15, no. 3, pp. 425-427, Mar., 2003.

[6] E. Pawlowski, et. al., “Variable bandwidth and tunable centre frequency filter using transversal-form programmable optical fitler,” Electronics Letters, vol. 32, no. 2, pp. 113-114, Jan., 1996.

[7] D. Lee, et. al., “High-resolution, high-speed microscanner in single-crystalline silicon actuated by self-aligned dual-mode vertical electrostatic combdrive with capability for phased array operation,” Transducers ’03, Paper 2E46.P, June, 2003.

Ultrafast and Nonlinear Optics

SPRC 2002 – 2003 G-1 Annual Report

Research Program

G. Ultrafast and Nonlinear Optics

Mathieu Charbonneau-Lefort, Bedros Afeyan and Martin M. Fejer, “Optical Parametric Amplification Using Chirped Quasi-Phase-Matching Gratings” .........................................................G-3

Rostislav Roussev, Arun Sridharan, Karel Urbanek, Robert Byer and Martin Fejer, “Parametric Amplification of 1.6 µm Signal in Annealed- and Reverse- Proton Exchanged Waveguide” .......................................................................................................................G-5

L. Scaccabarozzi, Z. Wang, X. Yu, S. Fan, M. M. Fejer, J. S. Harris, “Photonic Crystal AlGaAs Microcavities for non-linear Optical Applications” ...............................................................G-9

Andrew M. Schober, Mathieu Charbonneau-Lefort, Martin M. Fejer , “Parametric Oscillation of Ultrashort Pulses in Quasi-Phase-Matched Nonlinear Materials”...............................G-12

Xiuping Xie, Andrew M. Schober, Carsten V. Langrock, Martin M. Fejer, “Low Threshold and Near-transform-limited Pulse Generation with Cascaded Optical Parametric Generation in Reverse Proton Exchanged Lithium Niobate Waveguide” .......................G-15



Optical Parametric Amplification Using Chirped Quasi-Phase-Matching Gratings

Mathieu Charbonneau-Lefort and Martin M. Fejer E. L. Ginzton Laboratory, Stanford University

Bedros Afeyan

Polymath Research Inc., Pleasanton, CA

Quasi-phase-matching (QPM) offers significant technological advantages over birefringent phase-matching techniques. By periodically alternating the sign of the nonlinear susceptibility of the material, it became in principle possible to phase-match nonlinear interactions at any wavelength. Recently, the idea of engineering optically nonlinear materials was pushed a step further, when it was realized that appropriately chirped QPM grating could be used to simultaneously generate and shape output light pulses. The generation of chirped pulses using chirped QPM gratings was demonstrated experimentally [1] and the possibilities offered by second-harmonic generation (SHG) and difference-frequency generation (DFG) using aperiodic gratings were investigated theoretically [2].

This work focuses on optical parametric amplification (OPA) using chirped QPM gratings. OPA is analytically more complex than SHG or DFG because both the signal and idler waves are allowed to grow. Moreover, we aim to understand the interaction of short pulses, in which case the group velocity mismatch between the three waves must be taken into account.

This paper presents a first stage in our investigation of OPA with short pulses and aperiodic QPM gratings. We make some simplifying assumptions to make the analysis tractable. First, we neglect the depletion of the pump wave. Second, we assume that the pump pulse duration is much longer than that of the signal pulse, so that the pump can be considered to be constant. On the other hand, we do account for group velocity mismatch (GVM) between the signal and idler pulses. We consider linearly-chirped QPM gratings, that is, we assume that the difference between the wavevectors involved changes linearly with position around the perfect phase-matching period. These assumptions lead to the following two coupled linear partial differential equations desribing the space-time evolution of the signal and idler pulses, with a quadratic phase mismatch:

where vs and vi are the signal and idler group velocities, γ is the gain coefficient, κ is the chirp rate and xpm and xo are the positions of the perfect-phase-matching point and crystal input facet, respectively. The mathematical formulation of this problem is identical to that of parametric amplification of a laser in an inhomogeneous plasma, and its solution is well-known in the field of plasma physics [3-4].

We first solve for the response to an impulse excitation, or Green’s function of the problem. The final pulse shapes can then be recovered by integration of the impulse response over the initial pulse shape. The main steps of the calculation are the following. First, we take a Laplace transform in time; the time derivatives are replaced by the transform variable. We then eliminate the signal wave to obtain a

[ ]

[ ]⎭⎬⎫

⎩⎨⎧ −−−=⎟⎟

⎠

⎞⎜⎜⎝

⎛∂∂

+∂∂

⎭⎬⎫

⎩⎨⎧ −−−=⎟⎟

⎠

⎞⎜⎜⎝

⎛∂∂

+∂∂

20

2*

20

2*

)()(2

exp1

)()(2

exp1

pmpmsii

pmpmiss

xxxxiaiatvx

xxxxiaiatvx

κγ

κγ

(1)



second-order ODE in space for the idler wave. By a proper changes of variables, it can be cast in the form of a standard equation of mathematical physics whose solutions are parabolic cylinder functions. Therefore at this stage we have obtained the solution in the Laplace domain, and all we are left with is the inversion of the transform. This is best carried out using the integral representations for the parabolic cylinder functions. We obtain the following expression for the envelope of the idler wave:

( ) [ ] dzzz

zzUiUiai

isi11)(expexpsinh

/1

0 21

22 κγ

⎟⎠⎞

⎜⎝⎛ −

−Φ⎟⎟⎠

⎞⎜⎜⎝

⎛κπγ

πδν−γ

= ∫ , (2)

where 4/2/ π+=Φ iXU is the phase, )1/()( α−−α= XTUs and )1/()( α−α−= TXUi are positions inside the pulse relative to the wave fronts, xX κ= and tvT s κ= are the normalized space and time variables and α = vi /vs is the ratio of the group velocities. The integral can be evaluated in different limits. First, in the case of a uniform QPM grating, the integral can be written exactly in terms of modified Bessel functions [5]. When the chirp is non-zero, the integral has to be evaluated approximately using, for instance, the method of steepest descents. Two distinct regimes exist depending on the magnitude of the chirp, and they lead to completely different behaviors.

In the small-chirp regime, the interaction is phase-matched throughout the grating. This condition is given by Lvvvv isis /)(2 γ+<<κ . Fig. 1a compares the approximate analytical solution obtained after asymptotic evaluation of the integral appearing in Eq. (2) with the numerical solution of the coupled equations (1). The signal and idler impulse responses are approximately Gaussian. The maximum amplification factor is exp(g), with the gain given by )arcsin()2/(1)2/( 22 RRRg βΓ+β−Γβ= ,

where )/(2 isis vvvv +=β is a measure of the GVM, γκ= 4/LR is the chirp-to-gain ratio and Lγ=Γ is

the gain in absence of QPM chirp or GVM. The center of the pulse moves at velocity )/(2 isis vvvv + , and the duration of the pulse is γδν /2|| L , where is vv /1/1 −=δν is the GVM parameter.

As mentioned earlier, the signal and idler pulse shapes can be obtained by convolving the impulse response (or Green’s function) with the envelope of the input signal pulse (which play the role of initial conditions in the temporal evolution). In the frequency domain, it is useful to think of the amplification as described by a transfer function, which can be obtained by taking the Fourier transform of the impulse response. In the small chirp limit, the transfer function is approximately gaussian, with an amplification factor equal to exp(g), and a bandwidth (1/e full width) equal to

L/2|)|/4( γδν .

In the large chirp regime, i.e. Lvvvv isis /)(2 γ+>>κ , a particular wavelength is phase-matched only at a precise location inside the crystal. As such, different frequency components experience growth at different positions along the crystal. These different frequencies are amplified by roughly the same amount but they travel at different speeds due to GVM. Therefore, we expect the amplified pulse to be more or less flat and to exhibit a frequency chirp. This is exactly what is found by asymptotic evaluation of the integral in Eq. (2) in the large-chirp regime. The maximum amplification factor is

)/exp( 2 κπγ , and the frequency chirp is linear across the pulse, with a proportionality constant equal to )2/( 2δνκ . Since the amplitude of the pulse is flat, it can be approximated by a rectangular pulse with a quadratic phase. Fig. 1b compares this approximation to the numerical solution.



A simple transfer function in the large-chirp regime can be obtained using the square pulse approximation. It is proportional to ∫

+Ω

−Ωκπγ

X

Xdyiy )exp()/exp( 22 , where κδωδν=Ω /|| is related

to the frequency detuning and 2/)2/( κ= LX depends on the magnitude of the chirp. This transfer function becomes a square passband when the chirp is large enough, i.e. 4>>κL . The gain bandwidth is proportional to the chirp and is equal to ||2/ δνκL .

Figure 1: Comparison between the approximate analytical expression obtained from the asymptotic evaluation of the integral in Eq. (2), on one hand, and numerical solution of Eq. (1), on the other hand. These curves represent the pulse amplitude as a function of position at various times. a) Small-chirp regime, with 1=ΓL , sv/1=δν , 2/1=κ . b) Large-chirp regime, with 1=ΓL , sv/1=δν , 2=κ .

This work is a first step towards understanding the possibilities offered by carefully engineered QPM OPA devices. We have calculated how the chirp of the grating decreases the gain, increases the amplification bandwidth and introduces frequency chirp. It is important to allow for the simultaneous presence of a pulsed pump [6] and a QPM grating. This requires the analytic techniques introduced in [6] for treating short pump pulses to be generalized to include a space-time WKB description. The method presented in [6] can easily depict the overlap between three pulses that walk off each other at different velocities, which limit the amplification process, but wihout a QPM chirp. Including the QPM chirp in the calculation will be our next step. Beyond pulsed OPA’s themselves, we are also interested in synchronously-pumped optical parametric oscillators (SPOPO’s). In such devices, the pump and signal are both pulsed and the challenge is therefore allow for more degrees of freedom to design a desired light pulse by engineering the QPM grating appropriately. A SPOPO is being developed in our group [7] that will allow us to provide further experimental impetus for and validation of our theoretical investigations.

References [1] M. Arbore, O. Marco and M. M. Fejer., Opt. Lett., Vol. 22, 865 (1997). [2] G. Imeshev, M.A. Arbore, M. M. Fejer, A. Galvanauskas, M. Fermann and D. Harter, JOSA B,

Vol 17, 304 (2000). [3] M. N. Rosenbluth, R. B. White, and C. S. Liu, Phys. Rev. Lett., Vol. 31, 1190 (1973). [4] F. W. Chambers, Appendix A4, Ph.D. dissertation, Dep’t. of Physics, MIT, 1975. [5] D. L. Bobroff and H. A.Haus, J. Appl. Phys., 38, 390, 1967. [6] B. Afeyan and M. M. Fejer, Proceedings of NLO, 2002. [7] A. Schober, M. Charbonneau-Lefort and M. M. Fejer, poster presented at SPRC 2003.



Parametric Amplification of 1.6 µm Signal in Annealed- and Reverse- Proton Exchanged Waveguides

Rostislav Roussev, Arun Sridharan, Karel Urbanek, Robert Byer and Martin Fejer Edward Ginzton Laboratory, Stanford University, Stanford, CA94305 [email protected]

Quasi-phase-matched (QPM) optical parametric amplifiers (OPA) and oscillators (OPO) in bulk periodically poled lithium niobate (PPLN) have been used extensively for generation of mid-IR radiation. Difference frequency mixing in a waveguide allows high conversion efficiency at low and intermediate pump power and can provide energy-efficient optical parametric pre-amplification for high-power pulse generation in the 1.6 µm band for remote sensing applications. In addition, pulsed and CW tunable narrow-line radiation for spectroscopic or other applications in the vicinity of 3 µm can be generated efficiently by mixing a moderately strong readily available pump at 1.064 or 0.98 µm with standard telecommunication lasers in the 1.55 or 1.6 µm band. Recently, PPLN waveguide difference-frequency mixing between 813 nm pump and signal around 1060 nm was reported [1], generating mid-IR idler around 3000 nm.

We report experimental results on optical parametric amplification in annealed proton exchanged (APE) and reverse-proton exchanged [2] (RPE) waveguides in periodically-poled lithium niobate using a 1.064-µm pulsed Nd:YAG laser as the pump and a tunable L-band external cavity diode laser as the signal.

For efficient OPA, it is essential that the losses in the waveguide be minimized while maintaining adequate conversion efficiency for net gain, especially for the case of operation in the saturated regime where the maximum output power of the signal is limited by the losses at the pump and signal wavelengths. Another important factor is the coupling of each of the interacting wavelengths into its appropriate propagation mode. Usually, coupling into the fundamental mode at each wavelength is achieved via a mode filter (single mode waveguide) and an adiabatic taper [3] is used to convert the input mode into the fundamental mode of the tightly confining wavelength conversion section of the waveguide without appreciable coupling to higher-order guided modes.

Due to the depth asymmetry of APE waveguides, designing a single mode filter for both pump and signal presents a challenge when the two input wavelengths are widely different. Using numerical modeling [4], we determined a narrow range of fabrication parameters for single mode operation at both wavelengths. The proton exchange depth is limited to about 0.85 µm (1.3 µm after soft anneal[4]) reducing the confinement of the 3µm idler and the conversion efficiency. With PE depths in the range 0.75-0.85 µm, the maximum mode-filter width for single mode pump is 2-2.5 µm. The mode size of the signal in such mode filters is fairly large (>15µm) and a very long taper is necessary for adiabatic conversion to the tightly confined (~3µm in depth) mode of the mixing region.

Considering the above limitations and using modeling [4] to determine the approximate QPM period, we fabricated APE PPLN waveguides with widths between 10 to 18 µm by proton exchange in benzoic acid at 161 °C for 22.5 hours to a PE depth of 0.83 µm (soft-annealed depth of 1.27 µm) and



annealing at 332.8 °C for 21.4 hours. The propagation losses of the signal in the mode filters were estimated using the Fabry-Perot (FP) fringe contrast method [5] by testing single mode waveguides without tapers. The losses measured for APE and RPE waveguides are summarized in Table 1. The total device length was 52 mm, with 46 mm of periodic poling. The input of each waveguide consisted of a 2 mm long mode filter and a 2 mm linear taper, while the output consisted of a 2 mm long linear taper only. Note the significant difference in measured losses between the single mode and the tapered waveguides, showing that the taper design was not adiabatic. Phase-matching was obtained for a period of 23.9 µm for signal wavelengths ranging from 1596 to 1617 nm at waveguide widths between 12 and 18 µm. Waveguides narrower than 12µm did not support a guided mode at the 3.1 µm idler. Using 200 ns input pump pulses with peak power of 100±40 W, a maximum gain of 10 dB was obtained at 1617 nm at waveguide widths of 16-18 µm. The amplified signal at 10dB gain was 6.5 mW outside and 7.3 mW inside the chip. With over 25 W pump inside the waveguide, the gain is in the unsaturated regime. Assuming perfect QPM, the normalized conversion efficiency necessary to produce a 10-fold increase of the signal power is η=0.6 %/Wcm2 to be compared to our numerical estimates of 3-6%/Wcm2. Since only the fundamental mode is amplified, one factor reducing the observed efficiency is the scattering of the signal power between multiple waveguide modes in the sub-optimal taper.

WG type APE RPE QPM period 23.5 23.7 23.9 25.2 25.28 25.36 MF width 1.5 1.75 2 2 2.5 3 MF loss dB/cm

0.16 0.10 0.07 0.13 0.13 0.13

WG ‘loss’ dB/cm

0.8-1.2 0.4-1 0.35-0.75

0.5-0.65 0.24-0.3 0.21-0.3

Psout/Ps

in - 0.12 0.36 0.37 0.32 0.33 Pp

out/Ppin - 0.12 0.26 0.16 0.23 0.18

Reverse proton exchanged waveguides are characterized by a refractive index profile that is much more symmetric in depth compared to APE waveguides, allowing easier design of a single-mode mode filter for both inputs. Better confinement for the idler and alignment of the modes at all wavelengths in the interaction region help increase the conversion efficiency typically by a factor of two or more compared to APE waveguides. Optimizing the performance of RPE waveguides is based on trade-off between propagation losses and conversion efficiency, taking into account the negative effects of the proton exchange (e.g., “dead layer” of suppressed nonlinearity as well as possible structural defects). The best results we have obtained so far in RPE waveguide OPA are shown on Fig.1. The waveguides were fabricated by proton exchange at 185 °C for 18.75 hours followed by annealing at 333 °C for 15.67 hours and RPE at 310.5 °C for 17.53 hours. The waveguide widths ranged from 7 to 18 µm with length dimensions of the mode filters, tapers and mixing regions identical to those of the APE waveguides above. Up to 21.5 dB of gain was observed with 20-µJ 200-ns input pump pulses. This result suggests internal conversion efficiency 2.75 times that of the best APE waveguides. We think that the internal conversion efficiency of the RPE waveguides is 2-2.5

Table 1. Mode filter design and performance for the APE and RPE waveguide OPA. The QPM periods are presented in microns. Pp stands for pump power.



times that of the APE waveguides, with the rest of the observed enhancement due to better taper performance. The fabrication parameters and loss measurement results are included in Table 1, indicating a still sub-optimal taper design but with significantly better performance than that in APE waveguides. The signal losses in the waveguides with 3-µm-wide mode filters appeared to be between 0.21 and 0.3 dB/cm by the FP measurement, only slightly above the accurate loss estimate of 0.13 dB/cm for single mode 3 µm-wide waveguides without tapers, indicating that the performance of the 2-mm-long tapers for that mode filter width was close to adiabatic.

In flat-polished waveguides we also observed change in transmission of the signal during the pump pulses even away from phase matching. The effect was much more pronounced for longer pulses (1-30 µs) and was observed in both APE and RPE waveguides. It was most likely caused by heating of the waveguide due to absorption of pump photons. Pump-absorption induced temperature changes in the waveguide may limit the conversion efficiency and gain by altering the QPM condition. The origins of absorption in the waveguides need to be addressed in order for higher output pulse energy to be possible. Absorption beyond that typical for bulk congruent lithium niobate may be present in the waveguides due to structural defects generated during proton exchange [6]. We speculate that such defect states may be the reason for the reduction of the observed gain below theoretical prediction by causing absorption of idler photons by the OH--group stretching resonance. Even though the OH- absorption peak around 2.86 µm in APE waveguides is strongly polarized orthogonal to the guided polarization, waveguides proton exchanged in pure benzoic acid also exhibit a broad unpolarized absorption peak around 3.2 µm before annealing [7]. After annealing this peak vanishes, but trace concentration of remaining defects could cause unpolarized absorption by the OH--group of several dB/cm, which could explain the significant discrepancy between expected and observed gain, and may be reduced by further annealing.

With flat polishing, parametric oscillations were observed in some of the more efficient APE waveguides at pulse energy of 10-15 µJ, while in the good RPE waveguides parametric oscillations started around 5µJ.

Besides the observed gain, the phase-matching bandwidth of the APE and RPE waveguides was also monitored. Typical full-width half-maximum bandwidths for the signal varied between 12 and 50 nm⋅cm depending on waveguide width. The bandwidth for the idler around 3.2 µm is approximately 4 times as large indicating mid- IR tunability of tens to hundreds of nanometers simply by tuning the 1.6 µm signal wavelength.

With the demonstrated RPE waveguide conversion efficiency we expect to achieve saturation of a PPLN RPE waveguide OPA with 60 mm long poling using a 10-mW continuous-wave input signal at 1.6 µm and 50-100 µJ pump pulses with duration of 1 µs if the thermal effects are kept under control. This will amplify the signal peak power during the pump pulse from 10 mW to the 10 W level, which can be used to seed a high-power bulk OPA.



More accurate characterization of the waveguides can be done with a quasi-continuous wave pump. Current efforts concentrate on fabricating and testing longer devices with better tapers, as well as understanding the issues with propagation loss and absorption. Sources of proton exchange other than benzoic acid will be tried in the near future in effort to address the problem of structural defects in the waveguides as possible cause of idler absorption.

References: 1. D. J. Bamford et al.,”Generation of broadly-tunable mid-infrared radiation in periodically-poled

lithium niobate waveguides”, CLEO 2003, CTuM8 2. J. L. Jackel and J. J. Johnson, Electron. Lett. 27,1360-1361 (1991); Yu. N. Korkishko et al.,

J.Opt.Soc.Am. A, vol. 15, No. 7, July 1998 3. M. H. Chou et al, Opt. Lett. 21, 794-796 (1996) 4. R. Roussev et al, “Accurate semi-empirical model for annealed proton exchanged waveguides in

z-cut lithium niobate”, submitted to LEOS 2003 5. R. Regener and W. Sohler, Appl. Phys. B 36, 143-147 (1985) 6. S. Chen et al, “Loss mechanisms and hybrid modes in high-δne proton-exchanged planar

waveguides”, Opt. Lett. 18, No.16, pp. 1314-1316, 1993. 7. J. L. Jackel and C. E. Rice, “Short- and long-term stability in proton exchanged lithium niobate

waveguides”, SPIE Vol. 460 Proceeding of Guided Wave Optoelectronic Materials (1984)

Fig.1 Measured phase-matching wavelength and gain with 20 µJ pulses of duration 200 ns. Input pump peak power is approximately 100W. Power inside the waveguide can be estimated using the throughput results from Table 1, ranging from 16 to 23 W for the



Fabrication of photonic crystal AlGaAs microcavities for non-linear optical applications.

L. Scaccabarozzi, Z. Wang, X. Yu, S. Fan, M. M. Fejer, J. S. Harris

Applied Physics Department

For applications such as fiber optic networks, it is required to switch from one wavelength to another, or to extract information from a determined channel. Present systems, involving electronic converters, are relatively slow and power hungry, especially at high bit rates. All-optical systems offer interesting alternatives. By taking advantage of non-linear properties of Gallium Arsenide (GaAs)[1] it is possible to convert light from a generated frequency into the desired frequency. Though GaAs has very large nonlinear susceptibilities (d14 = 90pm/V) and mature device fabrication technologies, techniques must be developed for solving the problem of phase-matching in the absence of birefringence.

Frequency conversion has already been demonstrated in GaAs quasi-phasematched (QPM) waveguides[2], but their conversion efficiency so far has been limited due to weak confinement and high losses. We present here a device based on a photonic crystal microcavity which should greatly improve the conversion efficiency due to the tightly confined waveguide and to the photonic crystal structure. The theoretical conversion efficiency is 4 %/mW (of input power), compared to 0.01 %/mW for 1 mm long QPM waveguide. Moreover, because of its short length (~ microns), this device is intrinsically phasematched and integrable.

A schematic of the device is shown in figure 1. It consists of an AlGaAs core supported by an aluminum oxide cladding. The waveguide is used to couple light in a cavity, which is realized by introducing a defect in a 1-dimensional array of holes (photonic bandgap crystal), so that a defect mode is created. The cavity is resonant at the wavelength of 1.55 µm (input fundamental wave) and designed to have a high cavity Q. The waveguide core and cladding are respectively 200 nm and 2.5 µm thick, the pitch is 450 nm, the hole diameter varying from 100 to 200 nm, the central hole spacing is 700 nm. The depth of the structure must be at least 1um (this issue will be discussed in detail later). This structure has been designed to be resonant at 1.55 um with a Q value as high as 2000-4000.

The fabrication process starts with the growth of the structure by Molecular Beam Epitaxy (MBE), as shown in fig. 2.1. The cladding is Al0.93Ga0.07As and will be oxidized to AlxOy to increase the refractive index difference with the core and then the confinement. The core has been chosen to be Al0.5Ga0.5As, since at lower concentrations of Al, the material starts to be absorbing at the second

Figure 1: Schematic of the microcavity device

AlGaAs core

AlxOy

GaAs substrate

1.55 µm SHG 775 nm



harmonic (775 nm). Next, we have to define the etching mask. Given the submicron size of the features, ebeam lithography on PMMA is used (fig. 2.2). Since PMMA has very poor dryetching resistance, the pattern is transferred to a nickel mask by liftoff. It has been shown that Nickel has excellent selectivity vs. GaAs (~50:1) in chlorine-based etches[3]. This allows us to use a very thin layer (~20-30 nm) as mask, and consequently a very thin layer of PMMA (80 nm), which gives high resolution and decreases proximity effect. The proximity effect is a critical issue in this design, because the liftoff technique implies that the positive pattern must be exposed (the shaded area in fig. 2.2).

After exposing and developing the PMMA pattern, ~30 nm of nickel are evaporated on the sample (fig. 2.3). Nickel is then lifted off in acetone (fig. 2.4). The result is a very hard, thin and smooth mask (<10 nm peak-to-peak measured sidewall roughness). The sample is then dry-etched in a RIE-ECR reactor (PlasmaQuest) in a Ar:BCl3:Cl2 plasma. Profiles are vertical and smooth (no extra-roughness added during etching. The main issue of this step is the much slower hole etch-rate vs. ridge etch-rate. Moreover, it is steeply decreasing with the aspect ratio. For our geometry, the hole depth is only 35-45% of the ridge depth, for ~ 200 nm hole diameter and ~1.5 µm ridge depth. This is a well known problem, due to a transport effect: neutral species in the plasma reach with more and more difficulty the bottom of high aspect ratio holes, slowing down the etching process. However, since the mode in the cavity extends ~500 nm (1/e2) outside the core, as long as the hole-depth is at least ~500 nm, the cavity Q should still be high. FDTD simulations confirm this expectation, giving a Q of ~3000.

The next step is the nickel removal. Different methods have been tried: 1)Direct etch in nickel etchant, but this etchant attacks also AlGaAs 2)Use an intermediate mask, like SiN or SiO2, but this increases the complexity of the process and add extra-roughness 3) Use a sacrificial layer to liftoff nickel. This last technique, though not fully tested yet, proved to be doable and promising. The sacrificial layer is simply the GaAs cap layer on top of the structure. Citric:H2O2 solutions have a selectivity >260 in etching GaAs vs. Al0.5Ga0.5As[4] and can be used for this purpose. The only disadvantage is that a thicker cap is required (then requiring a slightly deeper etch).

Core: 0.2 um

Cladding: 2.5 um

GaAs substrate

AlGaAs 50%

AlGaAs 93%

1) MBE growth 2) Ebeam lithography 3) Nickel evaporation

4) Nickel lift off 5) Dry etching

6) Cladding oxidation

Shaded area = exposed area

Figure 2: Schematic of the fabrication process



Oxidation is the last step. The sample is oxidized in a furnace at 430°C for 1 hour. We noted that samples oxidized immediately after etching produced very smooth surfaces and negligible expansion, whereas samples left in air for a 1-2 days before oxidation produced cracks and an uneven surface. Finally, we lap the backside of the wafer and cleave the sample.

In conclusion, we designed an ultra-compact non-linear microcavity for optical frequency conversion, which does not require any phasematching technique. The theoretical conversion efficiency of this device is estimated 40 times greater than the one of previous QPM waveguides. At this stage of the project, we have fully tested all the major process steps, and though some small issues are still to be solved, a testable sample has been fabricated and is ready to be tested. We are now working on the experimental setup in order to measure loss and transmission spectrum of the device and perform second harmonic generation as a simple test of the nonlinear response.

References:

[1] S.J.B.Yoo, R. Bhat, C. Caneau, M. A. Koza, App. Phys. Lett. 66(25), pp.3410-12. [2] T. J. Pinguet, L. Scaccabarozzi, O. Levi, T. Skauli, L. A. Eyres, M. M. Fejer and J. S. Harris, Jr.,

LEOS, pp. 376-377, Nov. 2001. [3] A. Scherer, J.L. Jewell, Y.H. Lee, J.P. Harbison, L.T.Florez, App. Phys. Lett. 55(26), pp. 2724-

2726, Dec. 1989 [4] E. Moon, J. Lee, H. M. Yoo, J. App. Phys.,84(7), pp. 3933-3938, Oct. 1998

Figure 3: SEM picture of a) Nickel mask, after liftoff and b) device at the end of the fabrication process



Parametric Oscillation of Ultrashort Pulses in Quasi-Phase-Matched Nonlinear Materials

Andrew M. Schober, Mathieu Charbonneau-Lefort, Martin M. Fejer E. L. Ginzton Laboratory, Stanford University

Recent interest in time-resolved spectroscopy has renewed a long-standing interest in the reliable generation of wavelength-tunable ultrashort optical pulses over spectral ranges pushing farther into the mid-infrared. As this need grows, so does the need for techniques for the generation and amplification of such pulses. With the large nonlinearity and significant engineering flexibility of more advanced quasi-phase-matched nonlinear materials, parametric frequency conversion is an attractive alternative to existing laser sources. This paper describes the design and testing of a synchronously-pumped optical parametric oscillator based on Periodically-Poled Lithium Niobate (PPLN) as the parametric gain medium and recent progress towards understanding advantages of quasi-phase-matched nonlinear materials in the amplification and oscillators for short optical pulses in the presence of high gain and/or a depleted pump field.

The frequency conversion of short optical pulses presents additional challenges as compared to the conversion of continuous-wave signals. The large bandwidth of ultrashort pulses conventionally requires the use of very short crystals (and correspondingly low net gain) during parametric interactions, since the mismatch of group velocities between interacting frequencies limits the available bandwidth in parametric interactions. This limitation may be avoided by operation at wavelengths for which the dispersion of the interacting nonlinear medium affords a zero mismatch in group velocities, but in general this is not true for most wavelengths in the transparent window of any nonlinear material. This bandwidth limitation can be overcome more generally through the use of engineered quasi-phase-matching structures [1,2], or noncollinear geometries designed to compensate for the mismatch of group velocities [3, 4]. Furthermore, in single-pass second-harmonic and difference-frequency generation experiments, engineered grating structures have been used to manipulate the phase response for short pulses. Though recent results indicate that such engineered QPM devices can offer distinct advantages inside a parametric oscillator [5] as well as in high-gain and depleted-pump interactions, these advantages have yet to be thoroughly investigated.

Figure 1: PPLN Synchronously-pumped optical parametric oscillator. Cavity mirrors are labeled M1 through M5. M1 and M2 are focusing mirrors with radii of curvature equal to 22.5 cm and spaced by 25 cm. M2 through M5 are flat mirrors. M5 is the output coupler, with transmittivity of 4-8% at the signal wavelength. The cavity high-reflector (M4) can be translated to maintain synchronicity with the Ti:Sapphire pump laser pulses.



The synchronously-pumped optical parametric oscillator (SPOPO) in this experiment was built with a variety of practical and experimental applications in mind. Most obviously, it exists as a tunable source of short-duration pulses in the infrared to be used for probing other nonlinear interactions, including those in engineered PPLN waveguides and for testing developing engineered nonlinear materials including stoichiometric and Magnesium-doped PPLN, and orientation-patterned quasi-phase-matched Gallium Arsenide. Furthermore, it exists as a laboratory for direct testing of engineered nonlinear materials inside such a short-pulse OPO.

A schematic diagram of the PPLN-SPOPO is given in Figure 1. The pump laser is a Spectra-Physics Tsunami CW-modelocked Ti:Sapphire laser operating at a wavelength of 810 nm and emitting pulses 1.4 picoseconds in duration with a maximum pulse energy of 15 nJ at a repetition rate of 81 MHz. We use 1.2 nJ to pump our SPOPO. The pump laser is focused through the focusing (dichroic) mirror M1 to match the spatial mode size of the SPOPO cavity inside the PPLN crystal (70 µm 1/e intensity radius) to achieve optimum spatial overlap and the highest available gain. The PPLN crystal is Brewster cut at the signal wavelength to minimize Fresnel losses. In a short-pulse OPO, it is important to maximize not only the transverse spatial mode overlap but the temporal overlap as well. This is accomplished by matching the cavity length of the OPO to the length of the Ti:Sapphire pump laser so that the circulating signal pulses upon passing through the crystal on each successive round trip are co-incident with a pump pulse injected from outside the SPOPO. In our cavity, the high reflector (M4) is mounted on a translation stage to fine-tune the synchronous pumping.

The PPLN SPOPO tunes continuously from 1400 to 1600 nm, corresponding to idler wavelengths from 1640 to 1900 nm, as shown in Figure 2. The tuning range is limited by the optical coatings of the cavity mirrors; the coatings must have high reflectance at the signal wavelength while cutting off very sharply the reflectance of the idler near degeneracy. (Due to the flexibility of quasi-phase-matching technology, we expect that with the proper cavity optics, we could build a PPLN SPOPO with the same Ti:Sapphire pump laser to provide any wavelength from 900 nanometers to 5 microns).

The measured threshold pump power of 65 mW is consistent with other previously obtained results [6], and near to the record-low threshold for a PPLN-SPOPO [5]. The output power versus pump power is shown in Figure 2 below, with a slope efficiency of 25%. It should be noted the OPO is undercoupled with a very low output coupling of 4%.

The development of a theoretical model describing the behavior of a short pulse OPO is a rather involved problem because the nonlinear growth is dictated by the overlap of three pulses that walk off each other due to group velocity mismatch. As a first step, have undertaken the study of single-pass optical parametric amplification with aperiodic QPM gratings. Because of their larger amplification bandwidth and their phase response [5], chirped gratings could be used inside OPO cavities to generate light pulses having desirable features. At this stage, we have obtained theoretical results [2] only in rather simplified conditions, but the formalism developed up to now constitutes a solid basis for a theoretical framework explaining the behavior of SPOPOs using aperiodic QPM gratings.



Figure 2: LEFT: Signal power versus pump power (average). Crosses are measured points, while the solid line is a linear fit. The measured slope efficiency is 25% (50% quantum slope efficiency), with an oscillation threshold of 65 mW. RIGHT: Temperature tuning of SPOPO. The quasi-phase-matching period is 20.4 µm. Crosses represent the measured signal wavelength, while the theoretical signal/idler wavelengths (as calculated according to the Sellmeier of lithium niobate determined by Jundt et. al.) are shown in the solid and dotted lines, respectively. The inset shows a representative spectrum of the signal wave, with a full-width half-maximum width of 2.2 nm corresponding to a transform-limited pulsewidth of 1.7 ps.

We thank Spectra Physics Lasers, the Air Force Office of Scientific Research, and Crystal Technology, Inc. for their continued support in this research.

References

[1] Imeshev, G., Arbore, M. A., Fejer, M. M., Galvanauskas, A., Fermann, M., and Harter, D., JOSA B, 17, p. 304-318 (1999).

[2] M. Charbonneau-Lefort, M. M. Fejer and B. Afeyan, poster presented in the SPRC 2003. [3] Ashihara, S., Shimura, T., and Kuroda, K., JOSA B, 20, p. 853-856 (2003). [4] Di Trapani, P., Andreoni, A., Solcia, C., Foggi, P., Danielius, R., Dubietis, A., and Piskarskas, A.,

JOSA B, 12, p. 2237-2244 (1995). [5] Tillman, K. A., Reid, D. T., Artigas, D., Hellstrom, J., Pasiskevicius, V., and Laurell, F., Optics

Letters, 28, p. 543-545 (2003). [6] Burr, K. C., Tang, C. L., Arbore, M. A., and Fejer, M. M., Optics Letters, 22, p. 1458-1460

(1997).



Low Threshold and Near-transform-limited Pulse Generation with Cascaded Optical Parametric Generation in Reverse Proton Exchanged Lithium Niobate Waveguides

Xiuping Xie, Andrew M. Schober, Carsten V. Langrock, Martin M. Fejer Ginzton Laboratory, Stanford University

Low power and compact sources of tunable ultrashort near and mid-infrared pulses are necessary in many applications. The single-pass optical parametric generator (OPG) offers inherent simplicity, robustness and very wide tunability. However, it has a high threshold and material-dependent temporal properties. One current challenge for practical OPG systems is to further lower the OPG threshold to levels attainable directly from laser oscillators while keeping good temporal properties. In this report we present the results of OPG in reverse-proton-exchanged (RPE) PPLN waveguides, showing record-low threshold energies and near transform limited output. In our experiment, a pump pulse with FWHM length of 1.8ps at 769.6nm yielded an OPG threshold of 200pJ. The quasi-phase-matched OPG demonstrated up to 33% saturated photon conversion efficiency (internal) with 1nJ pump. The single-pass OPG was tunable from 1.15um to 2.3um for pump between 770nm and 789.5nm. Moreover, when the OPG outputs are related to the sum frequency generation (SFG) process (we call them cascaded OPG, or COPG products), they could be near transform limited, even in the presence of significant group velocity mismatch.

In the experiments we use uniform QPM gratings and transform-limited pump pulses. Under these conditions, the OPG signal-pulse length is proportional to the group-velocity walkoff (GVW) between the signal and the pump. The signal bandwidth, however, is inversely-proportional to the GVW between the signal and idler, as required by the phase-matching condition. These properties are similar to SHG and OPO in QPM devices, where a combining chirped grating and a chirped pump would result in a compressed output pulse. [1] The vacuum ‘seed’ for OPG introduces more complexity. Numerical simulations with fast Fourier transform (FFT) split-step method showed that cascaded processes instead of chirps could bring in the possibility of transform-limited pulse generation from OPG.

In conventional OPG, the time-bandwidth product (TBP) of the signal has a minimum, which is proportional to the ratio between the group index differences nsp/ nsi, with s, i and p denoting signal, idler and pump. It is possible to obtain transform-limited signal when |nsp/nsi| < 1. [2] In more general configurations, transform limited output could be obtained by engineering the effective group velocities of the interacting waves. In materials like BBO, this can be realized by collinear birefringent phase matching. [3] However, in congruent lithium niobate waveguides all three waves are in TE polarization and such possibilities were eliminated. If the cascaded OPG process (OPG and SFG between signal/idler and pump) is present, because of the higher group index of the sum frequency than the pump, group-velocity matching is automatically realized. By engineering the QPM gratings, the limitation is the transparent range of the material rather than the dispersion. Similar effects would appear in other nonlinear materials. The dynamics of the cascaded process is revealed by the numerical simulations: fixing the grating length and increasing the pump power, the conventional OPG will reach threshold first, with dips appearing at wavelengths where the SFG QPM conditions are satisfied; with pump power increasing more, the cascaded OPG will reach its threshold and grow up



exponentially faster than the conventional OPG products, because of the group velocity matching. If the QPM grating is long enough, the COPG products become dominant, and the whole output pulse would be near transform limited. These results were verified in the experiments with a setup shown in Fig. 1.

pump idler

signal

OPG

SFG

Variable attenuator Input

objective

RPE PPLN at 120°C Mode-locked

Ti:Sapphire laser760~830nm, 82MHz, 2ps

AOM with 1% duty cycle

Output objective

Beam separation with dichroic mirrors and filters

Figure 1: Diagram of cascaded OPG and experimental setup of optical parametric generation

The conventional OPG pump throughput and conversion efficiency curves (internal) are shown in Fig. 2(a). The signal wavelength was centered at 1350nm, with a bandwidth of 10nm. For this device, the QPM grating period was 15.75um in a 40mm-long grating, which was about 6 times the signal-pump GVW length. The OPG threshold of 200pJ matched well with the simulated value of 190pJ. The maximum saturated photon conversion efficiency was 33%. The typical OPG tuning behavior of the RPE waveguide devices is shown in Fig. 2(b). The theoretical prediction fits the measurements well. In other samples, the output wavelengths ranged from 1.15um to 2.3um for a pump wavelength between 770nm and 789.5nm, limited by the cutoff wavelength of the waveguide.

0 0.5 1 1.50

25

50

75

100

Pump power (nJ)0 0.5 1 1.5

0

10

20

30

40

Pum

p po

wer

thro

ughp

ut (%

)

Sig

nal p

hoto

n co

nver

sion

(%)

764 766 768 770 772 774 776 778

1200

1500

1800

2100

Sig

nal/I

dler

Wav

elen

gth

(nm

)

Pump Wavelength (nm)

(a) (b)

Figure 2: (a) The pump throughput and the signal power conversion efficiency in a 40mm single pass conventional OPG. The pump was 1.8ps (FWHM) pulse at 769.6nm and the signal was centered at 1350nm. (b) Pump tuning curve of a RPE PPLN waveguide device (120°C). The dotted curve is from simulations based on theoretical waveguide dispersion. The open circles are measurements.

The presence of cascaded OPG changed the conversion efficiency, as well as the time-bandwidth product dramatically. Although less signal power conversion was recorded with the presence of



COPG, more pump depletion was observed because of the sum frequency component. The dynamics of the cascaded OPG process is clearer from the temporal properties, as shown in Fig. 3. The shape and length of the pulse was measured with a cross-correlator. If we consider only the COPG component, the TBP of the generated signal was 0.51 with a pulse length of 1.9ps and a bandwidth of 1.4nm. The conventional OPG product had a TBP of 4.4. Transform-limited TBP for the supposed secant pulse shape is 0.315. By making use of the cascaded process, we have therefore generated signal with much better temporal properties. Numerical simulations of cascaded OPG predicted that more of the output power would be in the cascaded product with a longer QPM grating length. This was verified by the measurement on the 40mm devices which yielded a signal with little conventional OPG component. The TBP of the square shaped signal pulse was 0.89, close to the value of 0.886 for a transform limited square pulse. More studies on the devices with long QPM gratings are in progress.

1240 1250 1260 1270 1280 1290Wavelength(nm)

Nor

mal

ized

OSA

trac

e (a

rb. u

nits

)

0.3nJ

1.2nJ Pump power

-10 -5 0 5

0

0.05

0.1

0.15

0.2

0.25

Time(ps)

Cro

ss-c

orre

latio

n tra

ces

(arb

. uni

ts)

1.2nJ

0.3nJ

Pump power

(a) (b)

Figure. 3: (a) The OSA traces (b) the cross-correlation traces with different pump powers from 0.3nJ to 1.2nJ. Note shift of spectrum from direct OPG to cascaded output with increasing pump energy. Cross-correlation also shows both direct and cascaded pulses.

To improve the cascaded OPG process, we are trying to design amplitude and phase-modulated QPM gratings to control the cascade processes. A grating with periodically interlaced OPG and SFG sections are also shown by simulations to generate transform limited output. After a better knowledge of these approaches is obtained, an integrable tunable device might finally be fabricated and find its applications.

References [1] G. Imeshev, M. A. Arbore, M.M. Fejer, A. Galvanauskas, M G. Fermann, and D. Harter, J. Opt.

Soc. Am. B, 17, 304 (2000) [2] T. Sudmeyer, F. Brunner, R. Paschotta, etc, Long Beach, CA, USA, CTuO4, CLEO, May 19-24,

2002. [3] P. Di Trapani, A. Andreoni, C. Solcia, P. Foggi, R. Danielius, A. Dubietis, and A. Piskarskas, J.

Opt. Soc. Am. B, 12, 2237(1995).

Lasers

SPRC 2002 – 2003 H-1 Annual Report

Research Program

H. Lasers

A. E. Siegman, “Gain-Guided Optical Fibers for High-Power Fiber Lasers”......................................H-3

Arun Sridharan, “High Pulse Energy Yb:YAG MOPA and Non-Linear Frequency Conversion Module for Remote Sensing Applications” .....................................................................H-6

Lasers


Lasers


Gain-Guided Optical Fibers for High-Power Fiber Lasers

A. E. Siegman Ginzton Laboratory, Stanford University ([email protected]).

Fiber lasers could deliver much larger power outputs while staying below the threshold for nonlinear effects and maintaining single transverse mode operation if one could increase the core diameter of the fiber from a few tens of microns to a few hundreds of microns while still restricting propagation to only a single lowest-order mode. One way to achieve this might be to substitute pure gain guiding for the usual index guiding in the fiber, or alternatively to combine strong enough gain guiding with index antiguiding in the fiber core, so as to maintain single-mode propagation even at much larger than usual core diameters.

As a first step toward this objective we have analyzed the mode propagation properties of a slab waveguide or optical fiber such as that shown schematically in Figure 1, in which the core region is assumed to have a uniform amplitude gain coefficient ∆α in place of or in addition to the usual refractive index step ∆n. A structure of this type might be made by doping a large density of active laser atoms into the core region, and then pumping these atoms heavily using cladding pumping techniques. The modes of structures of this type turn out to be both interesting and potentially useful for large-diameter, high-power, but still single-mode fiber lasers, although substantial practical problems also still remain. A more detailed description of this analysis and of the mode propagation results can be found in a recent publication.1

n0, ∆n , ∆α

n0

n0

-a

a

Figure 1: Schematic drawing of a symmetric slab waveguide with slab thickness 2a or a cylindrical optical fiber with core diameter 2a having a uniform gain step ∆α as well as a positive or negative index step ∆n in the core region.

Optical fibers are normally characterized by the parameter v = (2πa/λ)(2n0∆n)1/2 where ∆n is the index step in the core region of the fiber. For systems with mixed index plus gain guiding, however, the square of this parameter, that is v2 = (2πa/λ)2 2n0 [∆n + j (λ/2π) ∆α], is more useful since the real part of v2 then corresponds directly to the real index step ∆n in the fiber core while the imaginary part corresponds to an imaginary refractive index step with the value (λ/2π) ∆α.

Figure 2 shows, for example, the regions in the complex v2-plane corresponding to the two lowest-order modes of a symmetric slab waveguide. Conventional index-guided fibers with ∆n>0 and ∆α=0 are located on the positive real axis to the right of the origin in this diagram, while purely gain-guided fibers with ∆n=0 and ∆α>0 lie on the heavy solid line along the vertical axis.

Lasers


- 6 -4 -2 0 2

1

2

3

4

5

m = 1

m = 2

25%

50%60%

70%

80%slab

waveguidemodes

Re[v2] (~∆n)

Im[v2](~∆α)

Figure 2: Regions occupied by the two lowest-order modes of a symmetric slab waveguide in the complex v-squared plane in which the horizontal and vertical axes are directly related to the core index step ∆n and the core gain step ∆α, respectively. The lowest-order mode exists everywhere above and to the right of line labelled m=1; the next higher-order mode exists everywhere above and to the right of the m=2 line. The lighter lines indicate the mode filling factor, i.e. the fraction of the total mode energy that propagates within the core, for the lowest-order mode at various locations within the complex v2 plane.

This figure shows that one can obtain a single purely gain-guided lowest-order mode (i.e., only ∆α, no ∆n in the waveguide slab) for any non-zero value of ∆α if the slab thickness is large enough. Moreover this mode can confine more than 80% of the energy propagating in the mode within the slab itself before the next higher-order m=2 mode begins propagating. More interestingly, with comparable values of ∆α gain-guided lowest-order mode propagation with excellent filling factor can also be obtained over an apparently indefinite range in the left-hand portion of the figure, corresponding to ∆n<0, i.e. to index anti-guiding by the central slab.

Figure 3 shows the comparable but slightly different mode boundary curves in the v2 plane for the lowest-order LP01 and LP11 modes in a cylindrical optical fiber. Again pure index guided fibers are located along the positive real axis, and again adding increasing amounts of gain guiding ∆α leads to operating points that move up and to the left in the v2 plane at constant filling factor values. In this case, however, pure gain guiding for the lowest-order mode does not turn on along the positive real axis until the product of ∆α × a2 rises above a finite value corresponding to j v2 ≈ 2. Also in contrast to the slab waveguide case, increasing values of ∆α can actually cause the fiber to cease propagating at very low values of positive index guiding ∆n, a region where the lowest-order fiber mode is known to be very weakly confined within the fiber.

As a practical matter, the index steps in conventional optical fibers may range from ∆n~0.1 down to ~0.001. A large but realizable power gain coefficient of 2∆α = 1/cm in a very heavily doped and heavily pumped fiber laser will correspond by contrast to an imaginary index component (λ/2π)∆α~10-5. A purely gain-guided fiber with a gain coefficient of this magnitude will then fall within the purely gain-guided single-mode region along the vertical axis in Figure 3 for core diameters in the range of 100 to 200 microns, which is an attractive possibility.

Lasers


- 8 - 6 - 4 - 2 2 4 6

2

4

6

8

20%

40%

60%

80%

LP01mode LP11

mode

70%

Cylindricalfiber modes

Im[v2] (~∆α)

Re[v2] (~∆n) Figure 3: Regions occupied by the LP01 and LP11 modes of a cylindical optical fiber in the complex v-squared plane. The line along the vertical axis corresponds to pure gain guiding, with no index step or index guiding in the core.

Real-world considerations do intrude, however. This single-mode behavior can be maintained only if the real part of the index step can be kept below the same value, i.e. ∆n<10-5, a condition which may be difficult to achieve in initial fabrication, and even more difficult to maintain as the fiber temperature increases in high-power operation. On the other hand, it may be possible to use a fiber which begins with a substantial amount of index antiguiding in the core, with this antiguiding decreasing as the temperature increases.

Weak gain guiding will also imply a large sensitivity to bending losses. On the other hand a gain coefficient as large as that indicated above will only require fiber amplifier or oscillator lengths on the order of a few tens of cm. Fibers this short with the desired core doping and core-cladding index matching can perhaps made using older rod-and-tube pre-form technologies, making it easier to achieve the closely matched refractive index values that will be needed. The gain guiding effects will also decrease with increasing signal level due to gain saturation, requiring a more detailed mode analysis at higher power levels. On the other hand, the fiber will function to some extent as a switch when the pump power is turned off, which make be of some practical interest.

The author appreciates sustained past support from the Air Force Office of Scientific Research for his earlier research; helpful courtyard discussions with Ginzton Lab colleagues, notably Marty Fejer; and discussions with Hiroshi Komine and Steve Brosnan of TRW, Redondo Beach, California, concerning the real-world thermal problems associated with large-diameter fiber lasers.

References [1] A. E. Siegman, "Propagating modes in gain-guided optical fibers," J. Opt. Soc. Am. A, vol. 20, pp.

1617--1628, August 2003.

Lasers


High Pulse Energy Yb:YAG MOPA and Non-Linear Frequency Conversion Module for Remote Sensing Applications

A.K. Sridharan, K. Urbanek, R. Roussev, C. Voss, S. Saraf, T.S. Rutherford, M.M. Fejer, R.L.Byer Electrical Engineering, Ginzton Laboratory, Stanford University

Efficient energy storage and extraction from a solid state laser is important for pulsed laser applications such as laser remote sensing of stratospheric ozone concentration or tropospheric wind velocities. Our approach to meeting the remote sensing wavelength requirements involves developing a high pulse energy Yb:YAG MOPA system followed by non-linear frequency conversion to 1.5 µm for global wind sensing[1] or to the 305 nm and 320 nm UV wavelengths for DIAL based detection of stratospheric ozone concentration.

Our laser system design is based on the master oscillator, power amplifier (MOPA) configuration. The MOPA configuration provides the ability to tailor the pulse width, beam divergence, and spectral width of the output pulses with low power components prior to the power amplifier. The scaling to high output pulse energy is determined by the power amplifier design. Further, the advantage of the MOPA architecture as opposed to an oscillator is that it offers a soft failure mode and allows scaling by implementing power-amplifiers in the chain.

However, for efficient energy extraction in each power amplifier the fluence at the input must be at least equal to the Yb:YAG saturation fluence (Jsat = 9.6 J/cm2). Because the optical damage fluence for 10 ns Q-switched pulses is approximately 10 J/cm2, it is not possible to extract the stored energy from Yb:YAG amplifiers efficiently without causing optical damage. However, since the optical damage fluence for a 1 micro-second pulse is 100 J/cm2, it is possible to saturate and extract the stored energy without optical damage using micro-second signal pulses.

In this µs pulse MOPA system, our choice of Yb:YAG as the gain material as opposed to a more established material like Nd:YAG was partially motivated by the smaller emission cross-section of Yb:YAG. The smaller emission cross-section of Yb:YAG compared to Nd:YAG allows storage of more energy for a given small signal gain coefficient(g0l). Depending on the specifics of the amplifier design, the smaller gain cross-section of Yb:YAG allows us to store ~ 10 times more energy per unit volume then Nd:YAG. Thus, by careful choice of gain material and signal pulse width, it is possible to saturate and extract efficiently a significant fraction of the stored energy without optical damage to the crystal.

The present all solid-state, laser-diode pumped, µs pulse Yb:YAG MOPA system is designed to generate a one micro-second (∆τ = 1 µs ) 100 mJ pulse at a 10 Hz repetition rate. High brightness laser-diodes are used to pump the Yb:YAG MOPA system which consists of a master-oscillator, pre-amplifier, and a 100 mJ power amplifier. The Yb:YAG amplifiers are designed with the end-pumped zig-zag slab geometry. The Yb:YAG MOPA can be made to operate with 1 MHz transform limited linewidth with the use of etalons in the master-oscillator to make it operate in a single longitudinal mode. Demonstration of 100 mJ pulse output is a critical first step toward the demonstration of a 10 J/pulse Yb:YAG MOPA. In addition to meeting the long term laser transmitter goals for remote wind

Lasers


sensing, we have shown that a 10 J/pulse MOPA meets the demands of an all solid state laser illuminator that achieves S/N = 10, with a target resolution of 20 cm at distances up to 4000 km [2].

High Pulse Energy Yb:YAG Laser Design Specifications

Yb:YAG micro-chip Master Oscillator Module

At the heart of a coherent MOPA system is a highly coherent cw master oscillator. Based on the work of Taira, we have built a 1 W cw master-oscillator that is shaped into 1 µs pulses at a 10 Hz repetition rate with the aid of an acousto-optic modulator. These pulses are then amplified in a pre-amplifier and a 100 mJ slab amplifier.

Pre-Amplifier Module and 100 mJ slab amplifier module

Pre-amplifier slabs 1 and 2 are double-passed as shown in Figure 1, where the 1 micro-Joule pulses shaped by the acousto-modulator are amplified to the 7 mJ level. The end-pumped slab dimensions are 0.4 mm x 0.4 mm x 11 mm. Each slab is pumped with 120 W of cw power with cw fiber-coupled LDs, modulated at 10 Hz with a 1 % duty cycle (i.e. 0.120 J per pulse, pump energy).

Figure 1: Schematic of pre-amplifier stage of MOPA system

The average pump power is thus 1.2 W. The small signal gain, g0l, in each slab is conservatively set at 2.25 to prevent the onset of parasitic oscillations. As a consequence, the pulse output of the first slab is 87 uJ pulse, and the pulse output of the second slab is 7 mJ.

In the next stage of amplification, we expect the 7 mJ pulses to be amplified to the 100 mJ level in an end pumped double-pass, slab amplifier of dimensions width = 1.05 mm, length = 11 mm and thickness = 1.05 µm. The pump power required is approximately 750 W cw, or about 7.5 W of average power when operated at the 1 % duty cycle.

On a practical note, parasitic oscillations can limit the gain achievable in a slab lasers and amplifiers and therefore limit the stored energy. We propose to suppress parasitic oscillations in each amplification stage of the MOPA system by choosing the aspect ratio (i.e. width to thickness ratio) of the slab carefully and polishing the edge faces with a wedge angle. This approach has been demonstrated to be effective in the first edge-pumped Yb:YAG zig-zag slab laser oscillator assembled by Rutherford [3].

1 µJ pulse @ 10 Hz 7 mJ pulse

@ 10 Hz

2-slab double passed pre-amplifier

Slab 1 Slab 2

Lasers


Another key to improved energy extraction with reduced thermal distortion in MOPA laser systems is the use of super-gaussian spatial profile pulses. Propagation of super-gaussian pulses through saturated amplifiers results in better extraction of the stored energy because the incoming beam more uniformly fills the slab and utilizes a greater fraction of the inverted population. Mansell has developed micro-miniature optical device that converts a gaussian intensity profile beam into a super-gaussian profile beam and we propose to use these devices in the Yb:YAG power amplifiers[4].

Further average power scaling of this MOPA system to higher pulse energy levels and higher repetition rates will likely involve multi-passed edge-pumped Yb:YAG slabs[3]. The edge pumping innovation allows the realization of a compact, conduction cooled, reliable, laser engine that will permit operation at pulse energy levels that will meet the requirements for global remote wind sensing and stratospheric ozone concentration measurement applications. Edge-pumped slab amplifiers inherently allow energy to be stored and extracted in a variety of pulse formats depending on the application. By careful engineering it is possible to extract both high pulse energies and high average power from an edge-pumped slab amplifier system [5].

Nonlinear Frequency Conversion

The high pulse energy 1.03 µm Yb:YAG MOPA system under development has applications in remote wind sensing if the output can be converted to the longer infrared wavelengths. We are investigating frequency conversion using optical parametric amplifiers (OPAs) based on both waveguide and bulk periodically poled lithium niobate (PPLN). The goal is to down-convert to the telecommunications bands to take advantage of the tunable laser oscillators in the C and L band region. This band also allows the use of fiber pre-amplifiers, quantum noise limited detectors, and fiber based components for coherent detection of signals. To meet the source requirements for ozone detection, frequency conversion to the ultraviolet is possible by harmonic and sum generation steps. As illustrated in Figure 2, the key is to utilize the advantages of the Yb:YAG system for energy storage and extraction followed by efficient frequency conversion to meet the various remote sensing application needs.

Figure 2: Schematic of non-linear optics approaches for various applications

Lasers


However, since the Yb:YAG MOPA is still under development, we have chosen to pursue demonstration of some of the non-linear optics modules in parallel, using existing Nd:YAG laser sources. Since the output radiation of a Nd:YAG MOPA system at 1.064 um is close to the expected 1.03 um output of the Yb:YAG MOPA system, the non-linear optics module will need to be only slightly modified to meet the final energy and wavelength requirements of the various remote sensing applications. Using flash-lamp pumped Nd:YAG rods as amplifiers, we have amplified a modulated non planar ring oscillator (NPRO) to the 100 mJ/µs level. Figure 3 shows a schematic of a 1.064 um MOPA that we have built in the laboratory.

Using a small fraction of this 100 mJ/µs MOPA energy, we are at present working on a 1.5 um band wave-guide based periodically poled lithium niobate (PPLN) optical parametric amplifier (OPA). This system schematic is shown in Figure 4.

In the first test versions of the PPLN waveguide devices, gains of ~ 21 dB at the signal wavelength have been measured. By fabricating devices with longer interaction lengths and consequently higher conversion efficiency, we expect to measure ~30 dB signal gain in this OPA.

Lasers


We will present updated results at the conference. The waveguide based PPLN OPAs will be followed by bulk PPLN OPAs to scale the ~ 1.6 µm energy to higher energy levels suitable for remote sensing of wind velocities.

References: 1) A.K.Sridharan, T.S.Rutherford, W.M. Tulloch, R.L.Byer, Proceedings of 10th ClRC, 1999. 2) R.L Byer. Report on "Research on Laser Illuminators," submitted to BMDO, July 15, 2000. 3) T.S. Rutherford, W. M. Tulloch, S. Sinha, and R. L. Byer, “Yb:YAG and Nd:YAG edge-pumped slab lasers,” Opt. Lett. Vol 26, Issue 13, pages 986-988, 2001. 4) J. D. Mansell, T. Rutherford, W.M. Tulloch, M. Olapinski, M. Fejer, and R. L. Byer. CLEO

Digest, pages, 406-407, 2000. 5) T.S. Rutherford, W.M.Tulloch, E.K. Gustafson, R.L. Byer. pages 205-219, JQE February 2000.

Optical Materials

SPRC 2002 – 2003 I-1 Annual Report

Research Program

I. Optical Materials

Junxian Fu, Seth bank, Mark Wistey, Homan Yuen, James S Harris, Jr , “Solid-Source Molecular Beam Epitaxy Growth of GaInNAsSb with Photoluminescence at 2.04µm”...................... I-3

M. Katz, R. Route, D. Hum, R. Roussev, K. Parameswaran, V. Kondilenko, G. Miller, M. Fejer, “Near-stoichiometric 1% Mg-doped LiNb03 and stoichiometric LiTa03 fabricated by vapor transport equilibration for freqeuncy conversion” ................................................................. I-5

P. S. Kuo, X. Yu, K. L. Vodopyanov, M. M. Fejer, J. S. Harris, D. Weyburne, D. Bliss, C. Yapp, K. O’Hearn, “Growth and Characterization of Thick-film Orientation-patterned GaAs” ................................................................................................................................................... I-8

Vincenzo Lordi, Homan Yuen, Seth Bank, Mark Wistey, James S. Harris, “Electroabsorption from GaInNAs and GaInNAsSb Multiple Quantum Wells”................................ I-11

Ji-Won Son, Yin Yuen, Sergei S. Orlov, Bill Phillips, Ludwig Galambos, Vladimir Ya. Shur, and Lambertus Hesselink, “Direct e-beam domain engineering in LiNbO3 thin films grown by liquid phase epitaxy”.................................................................................................. I-14

Mark A. Wistey, Seth R. Bank, Homan B. Yuen, Lynford L. Goddard*, and James S. Harris“GaInNAs(Sb) Growth Improvements Leading to Record Long Wavelength Lasers”................................................................................................................................................. I-17

Optical Materials


Optical Materials


Solid-Source Molecular Beam Epitaxy Growth of GaInNAsSb with Photoluminescence at 2.04µm

Junxian Fu*, Seth Bank, Mark Wistey, Homan Yuen, James S Harris, Jr. *Department of Applied Physics, Solid State and Photonics Lab. Stanford University

Low noise near infrared photo-detector within the wavelength range of 2-2.5µm is of great interest to bio-sensing[1] (like non-invasive blood glucose analysis) and environmental gas detection. Integration of photo-detector, light source and circuits on the same semiconductor substrate can not only reduce the overall costs but also increase the system response performance.

As a promising material for opto-electronic devices grown and fabricated on GaAs substrate[2], GaInNAs has been extensively studied in the optical communication field[3]. The small size of nitrogen atoms[4], the bandgap bowling effect of GaNxAs1-x

[5], and the strain in the active region layer grown on GaAs substrate significantly reduce the bandgap[6] and make possible long wavelength optoelectronic devices operating at 1.3µm and 1.55µm[7]. Latticed-matched Ga0.47-3xIn0.53+3xNxAs1-x grown on InP substrate has the potential to reach that wavelength with only a little bit of nitrogen introduced[8].

GaInNAsSb with room-temperature photoluminescence peak wavelength at 2.04µm was successfully grown on InP substrate using solid-source molecular beam epitaxy with a nitrogen plasma source. In-situ reflection high-energy electron diffraction during crystal growth was used to observe the surface reconstruction. X-ray diffraction, room-teperature and low-temperature photoluminescence were utilized to characterize the material quality and optimize the crystal growth condition.

The samples were grown in a GenII MBE system with indium and gallium SUMO-cell from Applied Epi. By controlling the temperature of the tip and base of the cells, we can get very stable source flux during the MBE growth so that the lattice match condition can be pertained. The arsenic and antimony source have two regions: the arsenic flux was monitored by adjusting the needle valve position with the sublimation region of arsenic cell being set to 390C and the cracking zone set to 800C. The antimony source does not have a valved cracker, the flux was controlled by changing the sublimation zone temperature with a cracking zone temperature of 850C. The nitrogen source is a redesigned SVT plasma source with highest rf-power of 300W and ultra-clean nitrogen gas flow of 20sccm.

The desorbing temperature of phosphorous from InP substrate is around 300C, while the oxide layer on the surface remains untouched until InP decompose at the temperature of around 500C. The oxide blowoff process varies based the substrate history and blowoff environment.

The substrate was first heated to 300C quickly without arsenic overpressure. Then in arsenic overpressure environment the substrate was heated to 350C and ramped to 510C with ramping rate of 20C/min in the arsenic overpressure environment of 1e-6Torr. After 2×4 surface reconstruction was clearly recognized, the substrate was cooled to the growth temperature. Lattice matched In0.53Ga0.47As buffer layer was grown with mismatch strain less than 0.05%.

After the buffer layer growth, the transition layer of lattice matched InGaAs was grown with substrate temperature gradually cooled down to 370-390C, at which temperature long wavelength material

Optical Materials


GaInNAsSb was grown. The use of transition layer can reduce the non-radiative defects in the SQW layer when the substrate temperature change for cladding layer and Single-Quantum-Well (SQW) layer growth. A capping layer of lattice matched InGaAs layer was also grown on the top of the SQW layer.

In conclusion, narrow bandgap material of GaInNAsSb was successfully grown on InP substrate with room-temperature photoluminescence at the wavelength of 2.04µm. NIR photodetection devices can be fabricated and integrated on the same chip with light source and supporting circuits to make use of bio-sensing and chemical gas detection.

0.00E+00

5.00E-12

1.00E-11

1.50E-11

2.00E-11

2.50E-11

1500 1700 1900 2100 2300

Wavelength (nm)

PL In

tens

ity (a

.u.)

100

1000

10000

100000

1000000

10000000

31.5 31.6 31.7 31.8 31.9 32 32.1 32.2

Omega/2Theta

Cou

nts/

s

Figure 1: Room-temperature photoluminescence measurement and XRD measurement of InGaAs/GaInNAsSb/InGaAs SQW sample.

We are grateful to Sumitomo for support of this research with InP wafer supply. This work was partly supported by DARPA.

References [1] Kamal Youcef-Youmi and Vidi A. Saptari, Progree Report 2-4, 1999 MIT Home Automation and

Healthcare Consortium [2] Kondow, M; Uomi, K; Niwa, A, et.al., J. J. App. Phys. Part 1, Feb 1996; v.35, no.2B, p.1273-1275 [3] Harris, JS, Semi. Sci. Tech.; Aug 2002; v.17, no.8, p.880-891 [4] Shan, W; Walukiewicz, W; Ager, JW, J. App. Phys., Aug 15 1999; v.86, no.4, p.2349-2351 [5] Lindsay, A; O'Reilly, EP, Sol. Sta. Comm., 1999; v.112, no.8, p.443-447 [6] Chow, WW; Jones, ED; Modine, NA, App. Phys. Lett., Nov 8 1999; v.75, no.19, p.2891-2893 [7] Nakahara, K; Kondow, M; Kitatani, T, et.al., IEEE Photon. Tech. Lett., Apr 1998; v.10, no.4,

p.487-488 [8] Gokhale, MR; Wei, J; Wang, HS; Forrest, SR, App. Phys. Lett., Mar 1 1999; v.74, no.9, p.1287-

1289

Optical Materials


Near-stoichiometric 1% Mg-doped LiNb03 and stoichiometric LiTaO3 fabricated by vapor transport equilibration for frequency conversion

M. Katz, R. Route, D. Hum, R. Roussev, K. Parameswaran, V. Kondilenko, G. Miller, M. Fejer E. L. Ginzton Laboratory, Stanford University, Stanford CA 94305-4088

Frequency conversion of light via periodically-poled lithium niobate (PPLN) and periodically-poled lithium tantalate (PPLT) has widely been demonstrated during the last decade. Two material issues limit the performance of congruent PPLN and PPLT crystals for frequency conversion at room temperature with high intensities and visible light: photorefractive damage (PRD) and green-induced infrared absorption (GRIIRA). Recent publications on as-grown near-stoichiometric MgO:LiNbO3 [1,2] crystals and as-grown near-stoichiometric LiTaO3 and MgO:LiTaO3 [3], show a remarkable decrease in photorefraction and GRIIRA in these crystals. However, it is much more difficult to grow stoichiometric or near-stoichiometric LiNbO3 and LiTaO3 crystals than congruent crystals.

In our laboratory the vapor transport equilibration (VTE) method [4] has been used to fabricate near-stoichiometric, 1%-MgO-doped LiNbO3 and stoichiometric LiTaO3. The origin crystals were 0.5-mm-1-mm thick, z-cut crystals 1%-MgO-doped congruent LiNbO3 and undoped congruent LiTaO3 substrates, which, unlike the near-stoichiometric crystals, can be grown by straightforward Czochralski methods. The congruent plates were heated to 1100oC (the 1%-MgO-doped LiNbO3) and 1360oC (the LiTaO3) in a closed crucible containing a mixture of MgO, Li2O, and Nb2O5 for LiNbO3 and a mixture of Li2O, and Ta2O5 for LiTaO3, pre-reacted into a two-phase powder. Additionally, the LiTaO3 was electric-field poled at 185oC and was subsequently annealed at 610oC to make the crystal single domain.

Several measurement techniques were used in order to investigate the properties of the vapor- transport-equilibrated crystals. By direct measurement of the photogalvanic current and the photoconductivity we found the saturated space charge field, which determines the photorefractive damage sensitivity. A 514.5 nm laser with 16 kW/cm2 intensity and a 60-µm diameter (1/e2) served as the illumination beam for the current-voltage study along the z-axis of the 1-mm-thick crystal at room temperature. The near-stoichiometric MgO:LiNbO3 and LiTaO3 crystals were compared to their congruent counterparts and in both cases the near-stoichiometric crystals had smaller saturated space charge fields than the congruent crystals. The near-stoichiometric MgO:LiNbO3 had a measured saturated space charge field of 1.2 V/mm compared to a congruent value of 2500 V/mm. The LiTaO3 reduced from a saturated space charge field of 90 V/mm in its congruent state to 1.6 V/mm in its near-stoichiometric state. In each crystal, the saturated space charge field was reduced by several fold. A proportional reduction would be expected in the saturated photo-induced change in the refractive index, and hence in the photorefractive damage. Measurements of the wavelength dependence of the saturated space charge field were performed at available argon laser wavelengths and showed no significant dependence of the field upon wavelength.

Using Photothermal Common Path Interferometry, the Green-induced infrared absorption (GRIIRA) of the VTE’d and as-grown crystals were measured in both the ordinary polarization and the extraordinary polarization [2]. In each case, the 514.5 nm (green) laser was focused to a beam diameter of 80 µm with 0.8 W of optical power resulting in a peak intensity of 16 kW/cm2 at room

Optical Materials


temperature. Both crystals in both orientations showed reduced GRIIRA compared to their congruent counterparts and in most cases had lower overall absorption in the infrared.

Figure 1. Left: VTE’d 1% Mg0-doped LiNbO3 crystal with 20 micron periodic poling. Right: VTE’d LiTaO3 crystal with 8 micron periodic poling.

Both VTE’d crystals can be periodically poled by the electric field poling technique. The coercive field for the VTE’d 1% MgO-doped LiNbO3 was found to be 2-4kV/mm while the coercive field of the VTE’d LiTaO3 was in the range of 100-180V/mm. Periodic electrodes were patterned by well-developed photolithographic techniques and the electric field was applied to the crystal via a liquid electrolyte. Small numbers of samples have been poled with varying results. Figure 1 shows a VTE’d 1% MgO-doped LiNbO3 crystal which has been poled with a 20 µm period on the left and a VTE’d LiTaO3 crystal which has been poled with an 8 µm period on the right.

Figure 2. Left: a) SHG power tuning curve as a function of temperature controller setting. Right: b) SHG power as function of fundamental input power.

The 17-mm-long VTE’d LiTaO3 crystal was used for second harmonic generation (SHG) of 1064 nm. By appropriately focusing 18 W of single frequency 1064 nm CW light into the crystal, 1.5 W of 532 nm green light was produced at a temperature controller setting of 43.3oC. This results in a normalized efficiency of 0.3%/Wcm as compared to the theoretical value of approximately 1%/Wcm. Most of the reduction in efficiency can be attributed to the small duty cycle of reversed domains as illustrated in Figure 1. This can easily be corrected by patterning future devices with larger electrodes. Figure 2a demonstrates experimental and theoretical tuning curves for the SHG power as a function of temperature controller setting. Clearly one can see that the theoretical and experimental results match quite nicely. Figure 2b shows SHG power as a function of incident power. Note that the experimental points taken at 43.3oC deviate from the theoretical curve at low input powers, due to temperature rise of the crystal resulting from absorption of the infrared at high incident powers and scant thermal

0 2 4 6 8 10 12 14 16 18 200

200

400

600

800

1000

1200

1400

1600

T=45.10C

T=43.90C

T=44.10C

T=44.50C

T=44.80C

SHG

(Gre

en) p

ower

(mW

)

Fundam ental pow er (W )

T=43.7 0C

T>43.7 0C Theoretical curve

36 38 40 42 44 46 48 50 520

200

400

600

800

1000

1200

1400

1600

1800

Pfund=18W

E xperim enta l po in ts Theoretica l curve

SHG

(Gre

en) P

ower

(mW

)

Tem perature (0C )

Optical Materials


design. This results in a slightly shifted temperature for maximum SHG power at lower input powers. Experimental points of the maximum SHG power at a given input power were then measured by adjusting the temperature controller setting. To fully explore the limits of this material two simple tests have been performed. The first test was to measure the degradation of SHG power as a function of time using the same 18 W 1064 nm CW laser for 300 minutes. Shown in Figure 3a, the crystal showed no significant signs of aging over the duration of the test. The second test was to measure the degradation of SHG power over a longer period using a pulsed 1064 nm laser. The laser delivers 150 ns pulses at a repetition rate of 10 kHz at an average power of 15 W. Focusing the beam into the crystal, 5.5 W of 532 nm light was produced and tested for over 450 hours. Figure 3b demonstrates significant lifetime of the material over an extended period. The sudden drop in SHG power at 40 hours resulted from a failure of the temperature controller.

Figure 3. a) Right: SHG power shows no significant decrease in output power over a 300 minute span. b) Left: SHG Power over 450 hours of continuous output. Drop out at 40 hours resulted from a failure in the temperature controller.

Near-stoichiometric crystals produced by VTE have been shown to have potential as an economical alternative to the melt-grown near-stoichiometric counterparts. Moreover, the crystals show improved resistance to photorefractive effects compared to their congruent counterparts. Thus, these materials may prove to be a viable solution for efficient, high power frequency converters perhaps even at or near room temperature. Further studies of both the VTE’d MgO-doped LiNbO3 and LiTaO3 crystals are still ongoing. In particular, processing of larger crystals, poling of off-axis cut VTE’d materials and continuing lifetime tests are of interest.

References [1] Y. Furukawa, K. Kitamura, S. Takekawa, A. Miyamoto, M. Terao, and N. Suda “Photorefraction

in LiNbO3 as a function of [Li]/[Nb] and MgO concentration” Appl. Phys. Lett. 77, 2494 (2000). [2]. Y. Furukawa, K. Kitamura, , A. Alexandrovski,, R. K.Route, M. Fejer, and G. Foulon “green-

induced infrared absorption in MgO doped LiNbO3” Appl. Phys. Lett. 78, 1970 (2001). [3]. Takaaki Hatanaka, Koichiro Nakamura, Tetsuo Taniuchi, Hiromasa Ito, Yasunori Furukawa,

Kenji Kitamura “Quasi-phase-matched optical parametric oscillation with periodically poled stoichiometric LiTaO3” Opt. Lett., 25, 651-653 (2000)

[4] D.H. Jundt, M. M. Fejer, and R. L. Byer “Optical properties of lithium-rich lithium niobate fabrication by vapor transport equilibartion” IEEE J. Quant. Electron. 26, 135 (1990).

0 50 100 150 200 250 300 350 400 450 5000

1

2

3

4

5

6SHG Lifetest

Time (hours)

Pow

er (w

atts

)

0 50 100 150 200 250 3000

200

400

600

800

1000

1200

1400

1600

1800

Pfund = 18W

T=43.70C+/-0.30C

SHG

(Gre

en) P

ower

(mW

)

Time (minutes)

Optical Materials


Growth and Characterization of Thick-film Orientation-patterned GaAs

P. S. Kuo, X. Yu, K. L. Vodopyanov, M. M. Fejer, J. S. Harris E.L.Ginzton Laboratory, Stanford Univerity

D. Weyburne, D. Bliss, C. Yapp, K. O’Hearn Air Force Research Laboratory, Hanscom AFB

GaAs is a promising material for nonlinear optics. It has a large nonlinear coefficient and a wide transparency range. Optical parametric oscillator and DFG devices based on GaAs could be used as a source of coherent radiation in the near- and mid-infrared, which would have useful applications in spectroscopy, remote sensing and infrared countermeasures. However, GaAs is not birefringent so quasi-phasematching must be used for practical nonlinear optical devices. Recently, a method has been developed to grow GaAs using molecular beam epitaxy (MBE) with lithographically defined, periodically inverted domains; this material is called orientation-patterned GaAs (OP-GaAs).1

Because of the very low growth rate of molecular beam epitaxy, only thin films of OP-GaAs can be grown using MBE. Thick films, deposited with Hydride Vapor Phase Epitaxy (HVPE) on MBE-grown templates, are being developed for use in “bulk” interactions. Previous thick film growths, performed by Thomson-Thales in France on templates fabricated at Stanford, had thicknesses up to 500 µm with successful gratings periods down to ~27 µm.2 These samples have been used to demonstrate efficient mid-IR second-harmonic generation3 and difference-frequency generation4 in orientation-patterned GaAs. The Thomson samples were of fair quality; the domain walls grew vertically, but there were several pyramid-shaped defects and missing single domains, as shown in figure 1. Also, the growth was interrupted several times in order to remove parasitic growth on the walls of the reactor. Defects that were incorporated into the film during the interruptions are visible when the in-plane transmission of the sample is viewed; the defects give the sample a layered appearance. Even with these defects, good nonlinear-optical results were produced using these samples.

Figure 1: Cross-section of 60µm-period, ~520µm thick film OP-GaAs previously grown by Thomson-Thales. A portion of a pyramid-shaped defect is visible.

In an effort to explore alternative HVPE thick film growth, a collaboration was established between the Air Force Research Lab (AFRL) at the Hanscom AFB in Massachusetts and Stanford University.

Optical Materials


Our growth at AFRL differs in several ways from the Thomson growth. The AFRL HVPE reactor is designed for low pressures (~4 torr) whereas the Thomson reactor ran at close to atmospheric pressure. The growth rate using the AFRL reactor is between 80-100 µm/hr as compared to ~10 µm/hr by Thomson. We are observing parasitic growth on the reactor walls, which limits the film thickness in a single run (patterned films as thick as 700 µm have been grown), but it is less than the parasitic growth at seen by Thomson that necessitated the multiple growth interruptions.

Figure 2 shows a cross-section through a ~150 µm-thick, HVPE film that we grew on an OP-GaAs template. The grating period is 80 µm. In this growth, the domain walls grew vertically and the top surface of the growth developed the characteristic “flat-triangle” top. Similar growth is observed in grating periods down to 20 µm. We do not see any pyramid-shaped defects that were present in the high-pressure growths. However, the thickness of this film is limited by parasitic growth on the reactor walls, but because of the higher growth rate in the low-pressure HVPE system, we believe that we can achieve thickness suitable for “bulk” interactions without interrupting the HVPE growth.

Figure 2: Cross-section of ~150µm thick, 80µm period grating. The domain walls are nearly vertical and the top surface of the growth has developed the characteristic “flat-triangle” shape.

We are exploring ways to get to thicker growths. One method is to move the wafer very close to the Ga source in the HVPE reactor and as a result, increase the GaAs growth rate. The parasitic growth becomes substantial after the same amount of time, but with the increased growth rate, a thicker film can be deposited. Growths under these conditions have reached over 600 µm thick, but domain walls do not stay vertical. Also, gratings with periods shorter than about 130 µm did not survived to the top of the growth. Figure 3 shows a cross-section of the 168 µm period grating with a 635 µm film thickness. The domains are shaped like cigars, which leads to grating duty cycle that varies as a function of height. However, the duty cycle remains close to 50% during the first ~450 µm thickness of growth, and this material may by useful for nonlinear optics. In-plane transmission of this film was observed to be uniform. There were no layered structures, but we did observe a slight difference in

Optical Materials


transmission between regions with gratings and regions that were single-domain. This difference is likely caused by increased scattering in the grating-filled regions as compared to the single-domain regions.

Figure 3: Stained cross-section of 635 µm thick, 168 µm period grating. The domains are “cigar-shaped”, but the grating has close to 50% duty cycle in the lower ~450 µm thick of growth.

We are also examining other methods to grow thick, high quality films using HVPE. We will be implementing addition HCl gas flow near the reactor walls to try to suppress the parasitic growth. Also, we are considering altering the temperature profile in the reactor to further favor deposition on the wafer.

Thick film, orientation-patterned GaAs continues to be an exciting material with many possible applications in “bulk” nonlinear optics. We are making progress towards growing thick films of OP-GaAs using low-pressure HVPE on MBE-grown templates. Thickness up to 600 µm with good domain quality in the first ~450 µm of growth in large period gratings have been demonstrated, with no evidence of the pyramidal defects or layered absorption that were present in earlier, high-pressure HVPE films. We are now conducting optical tests of the existing films, as well as continuing development of thicker high quality OP-GaAs films.

References [1] C. B. Ebert, L. A. Eyres, M. M. Fejer, and J. S. Harris, J. Cryst. Growth, 201, 187 (1999). [2] L. A. Eyres, P. J. Tourreau, T. J. Pinguet, C. B. Ebert, J. S. Harris, M. M. Fejer, L. Becouarn, B.

Gerard and E. Lallier, Appl. Phys. Lett. 79, 904 (2001). [3] T. Skauli, K. L. Vodopyanov, T. J. Pinguet, A. Schober, O. Levi, L.A. Eyres, M. M. Fejer, J. S.

Harris, B. Gerard, L. Becouarn, E. Lallier and G. Arisholm, Opt. Lett. 27, 628 (2002). [4] O. Levi, T. J. Pinguet, T. Skauli, L. A. Eyres, K. R. Parameswaran, J. S. Harris, Jr., M. M. Fejer,

T. J. Kulp and S. E. Bisson, B. Gerard, E. Lallier, and L. Becouarn, Opt. Lett. 27, 2091 (2002).

Optical Materials


Electroabsorption from GaInNAs and GaInNAsSb Multiple Quantum Wells

Vincenzo Lordi, Homan Yuen, Seth Bank, Mark Wistey, James S. Harris Solid State and Photonics Research Laboratory, Stanford University

GaInNAs and GaInNAsSb are novel active-layer materials for potential use in a variety of optoelectronic devices operating in the telecommunications wavelength band of 1300–1600 nm, particularly lasers and modulators. The materials can be grown nearly lattice matched on GaAs, allowing all of the advantages of working on this less expensive and better developed substrate, compared to the alternative of InP. Electroabsorption modulators operating in the wavelength range of 1300–1600 nm are important not only for optical fiber communications, but also for use in optical interconnects to replace the electrical lines limiting the future speed of microelectronics. The design of long wavelength optical interconnects allows lower voltage operation as well as seamless integration with optical networking. We have studied the electroabsorption properties of GaInNAs and GaInNAsSb quantum wells to determine their suitability for application in such optical modulators.

Samples consisted of GaAs p-i-n diodes grown on semi-insulating GaAs substrates. The p-type regions were 1.38 µm thick, doped 5x1017 cm–3 with Be, while the n-type regions were 1.00 µm thick, doped 1x1018 cm–3 with Si. The intrinsic regions were 0.50 µm thick. The quantum wells (QWs) were each 8.0 nm thick with 20.0 nm GaNAs barriers and were placed in the center of the intrinsic region. GaInNAs active regions contained up to 9 QWs, while GaInNAsSb active regions contained up to 3 QWs due to greater lattice-mismatch strain. Circular mesa test devices were fabricated, with two concentric top-side metal ring contacts formed to the p and n regions. The top ring contact also served as an optical aperture to define the area of interaction of the probing light beam. Finished devices were wirebonded into special packages that fit inside a liquid He cryostat that could be cooled down to ~20 K while maintaining external electrical contact to the devices for biasing.

Material growth was accomplished using solid source molecular beam epitaxy, except with atomic N supplied by a radio-frequency plasma source. Both As and Sb were supplied by cracked sources. Growth of the active region materials (GaInNAs, GaInNAsSb, and GaNAs) was done at 400–425 °C to prevent phase segregation. Post-growth rapid thermal annealing was performed on some samples to improve the luminescence properties. For GaInNAs samples, annealing was done at 720 °C for 60 sec, while for GaInNAsSb samples, annealing was done at 760–800 °C for 60–180 sec. For 1.3 µm material, the nominal GaInNAs composition was ~1.6% N and ~30% In, with GaNAs barriers containing ~2% N. For 1.55 µm material, the nominal GaInNAsSb composition was ~3% N, ~2% Sb, and ~40% In, with ~9% N in the GaNAs barriers.

Absorption spectra were measured at various temperatures from room temperature down to 26 K, with applied electric fields up to ~200 kV/cm (up to 9.0 V reverse bias), using the photocurrent method and a monochromated quartz tungsten halogen white-light source. Values of absorption coefficient were extracted using the full thickness of QWs plus barriers for the interaction length. The input power as a function of wavelength was carefully calibrated to accurately extract the absorption coefficient, which was determined to within ~15% confidence.

Optical Materials


Photocurrent spectra taken from the GaInNAs samples, shown representatively in Figure 1, clearly show quantum confined Stark effect behavior. Sharp excitonic resonances are observed at the band edge, with ~20 nm full-width at half-maximum (FWHM) at low fields, and persist to high fields (broadened to ~30 nm FWHM). Although Figure 1 shows spectra taken at 70 K, similar peak widths were observed even up to room temperature. The sharp peak at the band edge corresponds to the heavy-hole–to–electron exciton, while the broader peak at shorter wavelengths probably corresponds to the light-hole–to–electron exciton. At temperatures above ~200 K, the shorter-wavelength peak was broadened so much as to not be resolvable.

Figure 1: Photocurrent spectra of GaInNAs/GaNAs 9 QW sample at 70 K. The sample was annealed at 720 °C for 60 sec after being removed from the growth chamber.

Figure 2(a) shows the electroabsorption spectra taken at room temperature from the annealed GaInNAs sample, for reverse biases of 0–5 V (with corresponding electric fields across the QWs shown in the figure legend). Here, the measured photocurrent was converted to absorption coefficient by assuming an interaction length equal to the total QWs plus barriers thickness. A peak absorption coefficient of ~8500 cm–1 was measured at 1250 nm with 0 V bias. Figure 2(b) shows the calculated change in absorption coefficient from the measured data in Figure 2(a), relative to the 0 V bias point. Several possible operating wavelengths are indicated by dashed vertical lines in the figure, which correspond to points where either the peak absorption is maximized (1250 nm) or the change in absorption is maximized (1292 nm), as well as the 1300 nm point. In a modulator application, a large absolute value of absorption might be desirable in some configurations, although the insertion loss is high in this case. On the other hand, a large contrast ratio is usually more desirable, which is achieved with a large change in absorption. In any case, the particular performance is dictated by both the absolute values of absorption coefficient and the change in absorption coefficient, as well as other system parameters. From our data, we obtain single-pass on-off ratios of 2.3:1, 6.9:1, and 9.1:1, respectively, for operation at 1250, 1292, and 1300 nm, with maximum absorption coefficients greater than 3000 cm–1 in all cases. These figures are at least sufficient for digital modulation, and can be further enhanced by using some kind of resonant cavity, such as in an asymmetric Fabry-Perot reflection modulator.

Optical Materials


Figure 2: (a) Absorption spectra of GaInNAs/GaNAs 9 QW sample at room temperature, with possible modulator operating wavelengths indicated by dashed vertical lines. The values of applied electric field correspond to 0, 1, 2, 3, and 5 V reverse bias, respectively, for increasing field. (b) Change in absorption coefficient for various voltage swings, relative to 0 V bias, calculated from the spectra in (a). The reverse voltages correspond to the electric fields indicated in the legend in (a) and depend on the intrinsic region thickness, which is 0.5 µm in this case. The sample was annealed at 720 °C for 60 sec subsequent to growth.

Early measurements on GaInNAsSb QWs yielded at least similar values of absorption coefficient at 1550 nm as measured for GaInNAs QWs at 1300 nm. However, the excitonic resonances in the spectra were not as sharp as for the GaInNAs samples, and the changes in absorption with bias were also much less. More careful measurements and further study of the GaInNAsSb samples is ongoing. The early results are optimistic toward the development of a high-performance material for electroabsorptive modulation at 1550 nm on GaAs.

In summary, very encouraging electroabsorption data was measured from GaInNAs multiple quantum wells, clearly demonstrating the quantum confined Stark effect in this material and exhibiting sharp excitonic resonances at the band edge. Sufficiently high values of absorption coefficient up to ~8500 cm–1, combined with relatively large changes in absorption with moderate reverse bias swings, make the fabrication of an electroabsorption modulator operating at 1300 nm feasible. A single-pass on-off ratio of ~9:1 was measured at 1300 nm for a device with 0.50 µm intrinsic region, which is enough for digital operation. An actual device could be designed for reflection modulation using an asymmetric Fabry-Perot cavity configuration, which would significantly enhance the contrast ratio. In addition, reducing the intrinsic layer thickness to 0.25 µm would enable comparable performance to be achieved with only a 2.5 V swing, which is readily accessible from a 1.25 V supply using voltage doubling circuitry. Furthermore, early measurements on GaInNAsSb multiple quantum wells showed reasonable absorption at 1550 nm, with work continuing toward the development of material suitable for electroabsorption modulation at this longer wavelength.

Optical Materials


Direct e-beam domain engineering in LiNbO3 thin films grown by liquid phase epitaxy

Ji-Won Son, Yin Yuen, Sergei S. Orlov, Bill Phillips, Ludwig Galambos, and Lambertus Hesselink Center for Integrated Systems, Stanford University, Stanford, CA 94305

Vladimir Ya. Shur Institute of Physics and Applied Mathematics, Ural State University, Ekaterinburg 620083, Russia

Domain engineering in LiNbO3 has been studied intensively for various applications such as quasi-phase matched optical parametric oscillators [1] and electro-optic Bragg gratings [2]. Recently, sub-micron domain engineering in waveguides has drawn much attention because of the possibility of implementing it in tunable Bragg grating structures [3]. To date, electric field poling is the most popular technique to produce periodically poled lithium niobate (PPLN). On the other hand, a direct-write electron beam poling technique appears as a promising alternative method since it does not include lithography and has a very high spatial resolution (5nm) [4].

For a more efficient and stable function of the devices, it is advantageous to fabricate waveguides rather than using bulk LiNbO3. Techniques based on diffusion processes such as proton exchange [5] have been extensively studied in combination with a PPLN structure. Thin film waveguide fabrication techniques, however, provide attractive advantages such as the ability to create a step index profile and suitability for integration with other processes. Among them, liquid phase epitaxy (LPE) is a very promising technique to produce high quality thin film waveguides of several-micron thickness with excellent crystallinity and surface morphology [6,7].

In this paper, we report domain engineering in LPE LiNbO3 thin films grown on LiNbO3 and LiTaO3 substrates for waveguide applications using direct-write electron beam poling. To our knowledge, it is the first attempt using an e-beam to periodically pole Z-oriented LPE LiNbO3 thin films for waveguide applications.

LiNbO3 thin films with several-micron thickness were grown using a flux melt of 20 mol% LiNbO3-80 mol% LiVO3 [6]. The films grown on LiTaO3 substrates are planar waveguides supporting both TE and TM modes at λ=632.8nm, and exhibit a step index profile.

To engineer domain structures in LPE LiNbO3 thin films, direct-write electron beam poling was implemented [4]. The opposite side of a surface exposed by an e-beam exposure was coated with 50nm Au for the ground electrode, and LPE LiNbO3 films were exposed to an e-beam with different doses ranging from 50µC/cm2 to 400µC/cm2. The domain structure of the specimens were investigated after etching in a HNO3:HF=1:1 mixture at room temperature.

The domain orientation of LPE grown LiNbO3 films was investigated by etching the cross section of the samples. The films were single domain, Z- oriented on both +/- Z surface of LiNbO3 substrates, and also on LiTaO3 substrates. The domain structures of LPE grown LiNbO3 thin films on substrates are shown in figure 1. The single domain layer beneath a Z- oriented LiNbO3 film on a LiTaO3 substrate is a Z+ oriented layer of a LiTaO3 substrate, which appears at the surface of a LiTaO3 crystal

Optical Materials


fter heat treatment above the Curie temperature (~ 605oC) [8]. Except for the surface layer, the LiTaO3 substrate is multi-domain.

(a) (b) (c) Figure 1: Domain structures of the LPE LiNbO3 films (a) on a LiTaO3 substrate; (b) on a Z+ surface of LiNbO3 substrate; (c) on a Z- surface of LiNbO3 substrate.

Since the LiNbO3 films are Z- oriented and the whole structure including substrates have a complicated multi-layered domain structure (especially the films on LiTaO3 substrates), it is not suitable to use conventional periodic poling techniques which are preferably done on the Z+ surface of LiNbO3. On the other hand, e-beam direct scanning is performed on the Z- surface, hence it is a more proper approach to engineer domains in LPE LiNbO3 thin films.

(a) (b) Figure 2: Inverted domain structure of the LPE LiNbO3 films (a)on a LiNbO3 substrate and (b) on a LiTaO3 substrate by direct-write e-beam poling.

Figure 2 shows the domain structures of LPE LiNbO3 films on a LiNbO3 substrate and a LiTaO3 substrate after direct-write e-beam poling with a 6.6µm period, respectively. For the LiNbO3/LiNbO3 homoepitaxial case, both the film and the substrate are Z- oriented, therefore the inverted domain structure penetrates the substrate. On the other hand, the inverted domain structure is isolated in the LiNbO3 thin film in the LiNbO3/LiTaO3 heteroepitaxial system. It is shown that we can engineer the domain structure of LPE LiNbO3 thin films by using direct e-beam poling, although the whole domain structure through the thickness is multi-layered as in the LiNbO3/LiTaO3 system. The domain engineering in the LiNbO3/LiTaO3 system is particularly interesting since it produces a step-index waveguide.

It is also worth noting that the written section consists of narrow domains with a width around 500nm. To discern whether the structure is determined by the e-beam scanning parameters or the intrinsic materials properties of LPE films, both single crystals and LPE films were experimented with under

LiNbO3 film LiTaO3 substrate

Ps

Multi-Domain Ps LiNbO3 Substrate

LPE LiNbO3 Film

Ps

LPE LiNbO3 Film

LiNbO3 Substrate

LiNbO3 film

LiTaO3 substrate

LiNbO3 film

LiNbO3 substrate

Optical Materials


the same e-beam scanning conditions. Figure 3 shows the difference in the poling behavior between a single crystal and a homoepitaxial LPE LiNbO3 film. Using the same e-beam scanning parameters, the definition of the domain structure is enhanced in LPE films compared with domains in single crystal LiNbO3. Moreover, merging of domains is prevented in LPE LiNbO3 films, resulting in submicron domain structures. We obtained structures with a 1µm period consisting of ~400nm wide domains extending 50µm in the LPE LiNbO3 films, which could not be obtained in a single crystal. We propose that the defect structure, such as the misfit between the film and the substrate and point defects in the LPE films can possibly explain these positive results. Further optimization of the domain structures by changing scanning parameters and materials is presently under study.

(a) (b) Figure 3: Surface domain structures after e-beam irrdiation of a 0.9 µm period pattern on (a) a single LiNbO3 crystal; (b) a LPE LiNbO3 film.

In summary, we report domain engineering results in Z- oriented LPE grown LiNbO3 epitaxial thin films using direct-write electron beam poling. It is shown that domain engineering is possible using an e-beam even in LiNbO3 waveguide thin films on a LiTaO3 substrate which has a complicated multi-layered domain structure. The resulting domain structure consists of submicron domains, which cannot be obtained in single crystals under the same scanning conditions.

Acknowledgement: Support from Hitachi metals is gratefully acknowledged.

References [1] L. E. Myers, R.C. Eckardt, M. M. Fejer, R.L. Byer, W. R. Bosenberg, and J. W. Pierce, J. Opt.

Soc. Am. B, 12(11), 2102-2116 (1995) [2] M. Yamada, M. Saitoh, and H. Ooki, Appl. Phys. Lett. 69(24), 3659-3661 (1996) [3] A. C. Busacca, C. L. Sones, V. Apostolopoulos, R. W. Eason, and S. Mailis, Appl. Phys. Lett.

81(26), 4946-4948 (2002) [4] C. Restoin, C. Darraud-Taupiac, J. L. Decossas, J. C. Vareille, J. Hauden, and A. Martinez, J.

Appl. Phys. 88(11), 6665-6668 (2000) [5] M. Chou, Ph.D. Dissertation, Department of applied physics, Stanford University, Stanford, CA

(1999) [6] S. Kondo, S. Miyazawa, S. Fushimi, and K. Sugii, App. Phys. Lett. 26(9), 489-491 (1975) [7] A. Yamada, H. Tamada, and M. Saitoh, J. Crystal Growth 132, 48-60 (1993) [8] Y. Cho, K. Matsuura, S. Kazuta, H. Odagawa, and K. Yamanouchi, Jpn. J. Appl. Phys. 38, Pt.1,

No.5B, 3279-3282 (1999)

Optical Materials


GaInNAs(Sb) Growth Improvements Leading to Record Long Wavelength Lasers

Mark A. Wistey, Seth R. Bank, Homan B. Yuen, Lynford L. Goddard*, and James S. Harris Electrical Engineering and *Physics Departments, Stanford University.

Next-generation communications networks will use optical fiber not only for long-haul networks, but across town (metro/campus area networks) and from PC to PC (local area networks). Furthermore, the SIA roadmap predicts bus bandwidths within computers which are greater than can be carried by conventional electronic wiring, so some sort of optical interconnection will be needed for chip-to-chip buses and clock distribution. Until recently, the lasers for such applications were bulky, expensive, and difficult to integrate with CMOS. We have made several significant advances which enable inexpensive fiber networking and interconnects. The results of these new techniques will be outlined in this paper.

Our optical devices are based on gallium arsenide, a mature semiconductor which can be flip-chip bonded for integration with CMOS. Our active region is made of a thin layer of GaInNAsSb grown by MBE. We have produced some of the longest-wavelength emitting gallium arsenide based cw devices, covering the range from 1200-1500nm, from CWDM to DWDM ranges of fiber communication, as well as new wavelengths near 1400nm for Raman amplification in fiber. The chief difficulty in making long-wavelength, vertical cavity surface emitting lasers (VCSELs) on GaInNAs(Sb) has been the lack of gain in this material. We have solved the gain shortage.

Figure 1: Normalized photoluminescence intensity vs. anneal temperature from two significant, independent growth improvements. The control samples for each case are on a par with the best published device wafers, and each option presents a significant improvement. Option A and Option 1 were simultaneously implemented for our laser results shown elsewhere.

Of course, the substantial proof of any supposed growth improvements is not found in PL plots but in demonstration of an actual working device such as a laser. We have created several such lasers at

3

2

1

0

PL Intensity 800760720680640

Anneal Temperature (C)

Option A Control

None

2-3x Increase6

5

4

3

2

1

0

PL Intensity 800760720680640

Anneal Temperature (C)

Option 1 Option 2 Control

None

5x Increase

Optical Materials


~1500nm using the key techniques above. These lasers have established far longer wavelengths and higher average power than any other conventional, cw lasers reported on GaAs. The laser structure will be reported in a separate paper, easily identified by the following plot, of which we are inordinately fond:

Figure 2: Comparison of recent lasers (blue triangle) with other reported long-wavelength lasers on GaAs. In addition to low threshold Jth and long wavelength, it should be noted that several of the devices plotted in red only work in pulsed mode, while our lasers operate cw at room temperature.

Until this year, there seemed to be an inherent, severe nitrogen penalty in dilute nitrides: adding nitrogen to gallium arsenide always produced poor material, especially at long wavelengths. However, with the discoveries above, soon to be published, we have established that high-quality material can be grown even with the dilute nitrides such as GaInNAs(Sb).

1.20 1.25 1.30 1.35 1.40 1.45 1.50

0.5

1.0

1.5

2.0

2.5

3.0

3.5

J th (k

A/cm

2 )

λ (µm)

<1.3 µm GaIn(N)As >1.3 µm GaInNAs(Sb) This Study GaInNAsSb

Sensing Applications

SPRC 2002 – 2003 J-1 Annual Report

Research Program

J. Sensing Applications

Sameer R. Bhalotra, Helen L. Kung, Junxian Fu, Noah C. Helman, Ofer Levi, Jon Roth, Yang Jiao, Ryo Urata, David A. B. Miller, and James S. Harris, Jr., “Adaptive optical microsensors and spectra-selective imaging”........................................................................................ J-3

H. Chin, R. Chen, and D. A. B. Miller, “Linear electro-optic conversion of current pulses for a photonic-assisted electrical analog-to-digital converter”.............................................................. J-6

O. Levi, S. E. Bisson, T. J. Kulp, J. S. Harris, and M. M. Fejer, “Mid IR Cavity Ring Down spectroscopy using broadly tunable OP-GaAs light source”...................................................... J-9



Adaptive optical microsensors and spectra-selective imaging

Sameer R. Bhalotra, Helen L. Kung, Junxian Fu, Noah C. Helman, Ofer Levi, Jon Roth, Yang Jiao, Ryo Urata, David A. B. Miller, and James S. Harris, Jr. Edward L. Ginzton Laboratory, Department of Applied Physics, and Department of Electrical Engineering, Stanford University, Stanford, CA, USA

In recent years there has been an explosion of research in miniature sensors of all types, for application in a new generation of compact, portable systems. This has been driven by increased interest in persistent sensors (low-power, low-footprint sensors that can continuously take relatively simple measurements, e.g. gas monitoring in manufacturing facilities, or environmental monitoring in remote locations), sensors for mobile platforms (e.g. hyperspectral detection systems on helicopters or aircraft for surveillance, or integrated biochemical material sensors in man-portable systems for quick-response to emergencies), and inexpensive, mass-produceable sensors for personal use (e.g. medical diagnostic systems for the home or field-office).

Whereas significant effort has gone into miniaturizing sensors with tried-and-true architectures, many research groups, especially in the academic world, have developed completely new types of sensors. Leveraging our experience with optical and optoelectronic devices, the Stanford / DARPA PWASSP / Miller Group team has focused on sensors with novel optical and optoelectronic architectures, attempting to develop sensors with unique and useful characteristics. The grand theme of our research is that ideal sensing systems should be adaptive to a wide variety of sensing tasks; they should be simply and rapidly programmable to changing requirements or environments. We have focused our work in basic research of adaptive optical microsensors, as well as selective filtering techniques that allow the device to “look for” a signal (or combination of signals) of interest. These two research thrusts are complimentary, as these sensors and filtering technologies together enable efficient, portable, adaptive decision-oriented sensing systems that require minimal computing power to generate actionable information.

Research progress is reported in a number of detailed publications and project reviews, available on the Miller Group website or by request. Here we summarize recent research progress in four areas: the first three sections review new microsensors that we have designed, developed, prototyped, and tested; and the fourth section reviews a novel method of spectral filtering:

1. The first integrated standing-wave spectrometer. This represents the third generation of optical sensors using the standing-wave architecture, with near-infrared sensitivity in a 17 x 13 x 1 mm compact package. For the first time, the entire spectrometer is fully integrated, including Si MEMS mirror, thin-active-region GaAs photodetector, and requisite structural and electrical components (Fig. 1). Each component, notably the mirror and detector, have been completely redesigned for higher performance and compatibility in the integrated system. Device performance is vastly improved, for the first time matching or surpassing the spectral resolution of commercial microspectrometers, while preserving the multiplexing and SNR advantages of a transform spectrometer. Spectral analysis is demonstrated from 633 - 866 nm, with a spectral resolution of 4 nm (λ = 633 nm).



Figure 1. Left: spectrometer schematic. Light enters from below the detector holder, traverses the photodetector, and reflects off the mirror pillar back toward the detector. Right: photos of MEMS mirror-actuator and GaAs photodetector components.

2. The first microsensor for selective optical detection based on spectral coherence. Using the integrated standing-wave design, we demonstrate selective detection of optical sources of different bandwidths, by direct real-time interferogram analysis. We introduce the concept of a tunable coherence cutoff, for adaptive discrimination / selective imaging of different optical source types, without full spectral analysis and computation-intensive data processing. We present experimental results directly comparing a near-IR laser (coherence length 495 µm) and near-IR LED (coherence length 17 µm).

3. A microspectrometer with adaptive spectral resolution tuning. By real-time adjustment of the MEMS mirror drive voltage in a standing-wave spectrometer, the device resolution is varied, thus optimizing the SNR for a wide range of desired sensing tasks. This is the first standing-wave microspectrometer with sensitivity across the visible range, due to implementation of a new silicon photodiode with thin active region and transparent conductive contacts. We demonstrate spectral analysis from 488 – 665 nm, with continuously tunable spectral resolution of 72 – 6 nm (Fig. 2).

4. A novel method of real-time bias modulation of a photoconductive detector to implement our method of interferogram filtering by time-domain inner product. This method simplifies the task of discriminating among optical signals by minimizing data generation, handling, and processing. This is the simplest implementation yet, with only two key components required (detector and integrator) to accomplish filtering; the previously separated detector and multiplier stages have been combined in one 20 x 20 µm GaAs photoconductor stage. We demonstrate 2-D spectra-selective imaging in real-time, accomplished without digital processing.

Photodetector

13 mm

13 mm

17 mmDetector Holder

Pillar FlexureMirror Frame



Figure 2. Top: Interferograms for mirror scan lengths of a) 3.8 mm, b) 14 mm, and c) 31 mm. Middle: Derived optical spectra with resolutions of a) 72 nm (low), b) 14 nm (medium), and c) 5.6 nm (high). Bottom: Spectral resolution and scan length vs. actuator drive amplitude Vpp.

References [1] S. R. Bhalotra, H. L. Kung, J. Fu, N. C. Helman, O. Levi, D. A. B. Miller, and J. S. Harris, Jr.,

“Integrated standing-wave transform spectrometer for near infrared optical analysis,” in Proc. of IEEE LEOS 15th annual meeting, Glasgow, Scotland (10-14 Nov 2002). Paper ML5.

[2] S. R. Bhalotra, H. L. Kung, J. Fu, N. C. Helman, O. Levi, D. A. B. Miller, and J. S. Harris, Jr., “Standing-wave microsensor for adaptive analysis of spectral coherence,” in Proc. of OSA Conference on Lasers and Electro-Optics, Baltimore, MD (1-6 Jun 2003). Paper CThA.

[3] S. R. Bhalotra, H. L. Kung, D. Knipp, H. Stiebig, and D. A. B. Miller, “All-silicon standing-wave microspectrometer with tunable spectral resolution,” in Proc. Of IEEE International Conference on Optical MEMS, Waikoloa, HI (18-21 Aug 2003). Paper MC2.

[4] S. R. Bhalotra, J. Roth, Y. Jiao, H. L. Kung, R. Urata, and D. A. B. Miller, “Adaptive spectra-selective imaging by real-time photoconductor bias modulation,” to be published in Proc. of IEEE LEOS 16th annual meeting, Tucson, AZ (26-30 Oct 2003). Paper WZ4.

16 18 20

665 nm & 633 nm

Wavelength (nm)

Opt

ical

Spe

ctru

m (

arb)

Wavenumber (103 cm-1)

700 600 500

0 200 400 600 800

Time (µs)

16 18 20

633 nm

Wavelength (nm)


700 600 500

0 200 400 600 800

Time (µs)

16 18 20

665 nm 633 nm

Wavelength (nm)


700 600 500

0 200 400 600 800

Inte

rfer

ogra

m (

arb)

Time (µs)

Low Res. Med Res. High Res.a b c

a b c

0 20 40 60 80 100 120 140

10

100

S

pect

ral r

esol

utio

n (n

m)

Voltage Vpp

0

10

20

30

40

50

Mirror scan length (µm

)



Linear electro-optic conversion of current pulses for a photonic-assisted electrical analog-to-digital converter

H. Chin, R. Chen, and D. A. B. Miller Electrical Engineering, Stanford University

Performance improvements of electrical analog-to-digital (A/D) converters have been relatively slow in recent decades. For example, from 1989 to 1997, the sampling rate of A/D converters only increased by about 20%. One limiting factor has been the jitter of the clock used in the initial sampling of the analog signal to be converted [1]. Consequently, increased attention has been given to photonic-assisted A/D conversion of electrical signals, since mode-locked lasers, with reported timing jitters of less than 10 fs, can serve as the sampling clocks in such systems [2].

One concept for such an A/D converter was proposed by Professors David Miller, James Harris, and Bruce Wooley of the Electrical Engineering Department at Stanford University. Figure 1 shows the basic concept. An electrical transmission line carries the analog signal to be converted. Photoconductive switches, which are turned on by optical pulses, sample the electrical signal onto hold capacitors. Each switch would be triggered at a rate of 1 Giga-sample per second (GSPS). If we had 100 switches attached to the transmission line, the triggering of the switches would be interleaved such that the aggregate sampling rate would be 100 GSPS. A more conventional CMOS A/D converter is connected to each photoconductive switch. This converter, which need only operate at 1 GSPS, digitizes the sampled voltage stored on the hold capacitor.

Figure 1: Basic architecture of photonic-assisted electrical analog-to-digital converter.

Figure 2 depicts an alternative to electrically connecting the CMOS A/D to the sampling switch. The sampled signal is converted into a differential optical signal, sent to photodectors, and then digitized by the CMOS A/D. This electrical isolation of the switches from the CMOS circuitry, which we term “optical remoting,” increases the noise immunity of the switches from electrical noise. In addition, with direct electrical connections, the pitch spacing of the sampling gates must match the spacing of the relatively large CMOS circuits. This could prove problematic, since the switches would then not be able to sample the same position on the transmission line with respect to the highest frequency signals. With optical remoting, the pitch of the switches need not match the pitch of the CMOS A/D converters.

To enable this idea, a device which performs linear electro-optic conversion of sampled analog signals is needed. We have been testing this concept by operating traditional GaAs-based multiple quantum-

photoconductive sampling gates

conventional CMOS A/Ds

digital output

staggered optical trigger pulses analog electrical

input



well (MQW) optical modulators in a novel way. Though the modulators have nonlinear electro-absorption curves, the optical output is linearly proportional to the input current. More specifically, each electron’s worth of input charge causes one photon’s worth of optical energy in the differential optical signal [3].

Figure 2: Optical remoting for a single channel on the A/D system.

While this linear conversion process has been demonstrated in the past under DC operating conditions, our current work investigates this idea for sampled input signals. Figure 3 shows the optoelectronic circuit we are using. All three devices A, B, and C are MQW diode modulators. The differential electro-optic converter consists of devices A and B connected in series in a totem-pole fashion. For the purpose of this experiment, the input charge pulses are supplied by a third modulator C, which is driven with mode-locked optical pulses that have a FWHM of approximately 150 fs.

Figure 3: Optoelectronic circuit used for testing linear electro-optic conversion.

To begin the conversion process, the input charge packet is injected into the center node by exciting device C with a laser pulse. This packet of negative charge temporarily lowers the voltage V, so that the voltage bias on device A increases and the voltage bias on device B decreases. Relative to the situation just before the pulse excitation, A now absorbs more light while B absorbs less light. Consequently, the photocurrent through device A is now larger than that through device B, which causes the voltage V to increase. Eventually, V returns to its initial state, and the device is ready to convert the next electrical pulse. The time constant of this recovery is given by τ=hωC/(2πeγPin), where hω/2π is the photon energy, C is the total capacitance at the center node, e is the charge of an electron, Pin is the total input CW power, and γ=δΑ/δV is the partial derivative of the absorption with respect to voltage [4].

Figure 4(a) shows the difference in the powers of the output optical beams. Since this result was obtained by subtracting two beams with small swings on large DC backgrounds, the y-offset of the curve is subject to a large fractional error. The height of the peaks, however, is accurate to an experimental error of about 5%. The fitted time-constant of the device recovery is 3.5 ns. With our experimental conditions, a measured capacitance of 60 fF, and an absorption sensitivity of about 0.05

photoconductive sampling gate conventional

CMOS A/D

optical trigger pulse analog electrical

input

V+

V– V–

V+

digital output

continuous optical beams

Vbias

PA-in

PB-in

PA-out

PB-out

PpulseA

B

C



V-1, the predicted time constant is on the order of nanoseconds and hence consistent with the measured result.

We determine whether the electro-optic conversion is linear by varying the power of the pulsed laser driving device C. This has the effect of changing the amount of charge in the charge packet injected into the center node. By measuring the average current flowing through the power supply providing Vbias, we can deduce the amount of charge in each charge packet. We also measure the transmitted power through devices A and B with a high-speed photodetector and oscilloscope, and then calculate the difference of the average powers. The result is shown in Fig. 4(b), where a linear relationship is evident. Two features are worth noting. First, the slope of the line is close to unity, as expected for 100% internal quantum efficiency. Also, the last data point indicates that at the highest charge injection, fewer photons are present in the signal than anticipated. This is because for that specific case, so much charge has been injected into the center node that device A has reached a flat portion of the electroabsorption curve. Since the absorption sensitivity is low here, the device slows down and cannot complete the electro-optic conversion process within the repetition period of the input charge pulses.

Figure 4: Experimental data showing (a) time-resolved behavior and (b) linear conversion.

In summary, we have demonstrated the concept of electro-optic conversion of pulsed currents into a differential optical signal, using self-linearized MQW SEED modulators. The measured time constant is in agreement with approximate calculations using a simple first-order model. We have also shown that the electro-optic conversion is linear, even though the input signal is of a much higher bandwidth than the modulator itself. Hence, this device holds great promise for high-sensitivity electronic systems, and in particularly for our proposed photonic-assisted analog-to-digital converters.

References:

[1] R.H. Walden; IEEE J. on Sel. Areas in Comm. 17 (1999) 539 [2] T.R. Clark, T.F. Carruthers, P.J. Matthews, and I.N. Duling III; Elect. Lett. 35 (1999) 720 [3] D.A.B. Miller, D.S. Chemla, T.C. Damen, T.H. Wood, C.A. Burrus, Jr., A.C. Gossard, and W.

Wiegmann; IEEE J. of Quant. Elect. QE-21 (1985) 1462 [4] E.A. De Souza, L. Carraresi, G.D. Boyd, and D.A.B. Miller; Appl. Optics 33 (1994) 1492

Differential Optical Signal versus Input Charge Packet

(slope = 1)

0.00.10.20.30.40.50.60.7

0 0.2 0.4 0.6 0.8Input Charge (10^6 electrons)

Diff

eren

tial O

utpu

t pe

r In

put C

harg

e Pa

cket

(1

0^6

phot

ons)

Data

Linear fit

(a) (b)



Mid IR Cavity Ring Down spectroscopy using broadly tunable OP-GaAs light source

O. Levi1, S. E. Bisson2, T. J. Kulp2, J. S. Harris1, and M. M. Fejer1 1E. L. Ginzton Laboratory, Stanford University, 2Sandia National Laboratories, Diagnostics and Remote Sensing Dept., Livermore, CA

Mid IR spectroscopy in the 7-12 µm spectral region enables sensitive detection and identification of various biological and chemical molecules [1]. Laser sources optimized for spectroscopic measurements ideally should have sufficient power for high signal to noise ratio, narrow linewidth for high sensitivity and selectivity, high beam quality, large wavelength tuning, and low source noise [2]. One class of sources includes those which generate tunable mid-IR laser radiation directly from gain in the laser media which include lead-salt lasers, narrow gap Sb based lasers, and quantum cascade diode lasers tuned to operate at the desired spectroscopy wavelength. Another class of sources are based on nonlinear optical frequency conversion of near-infrared (≈ 0.9 to 2µm) laser sources [2]. Figure 1 shows some of these laser sources and their wavelength coverage.

Figure 1: laser sources and typical wavelength coverage [2]

Cavity ring down spectroscopy (CRDS) is a high sensitivity spectroscopy technique which measures small absorption changes inside high finesse cavities [3]. The decay rate is determined by the cavity mirror transmission, absorption and scattering as well by the wavelength dependent absorption loss of the measured sample gas. This decay rate is independent of laser amplitude, thus allowing for an accurate absorption measurement even in the presence of laser power fluctuations. Recently, we demonstrated tunable light source based on nonlinear difference-frequency generation (DFG) in orientation-patterned GaAs (OP-GaAs) crystals [4]. In this report we review progress in light source realization, and demonstrate continuous tuning over a large frequency band around 8 µm, and the coupling of this tunable light source to a high finesse cavity for CRDS.

The experimental setup is shown in Figure 2. Pump and signal wavelengths from tunable external cavity diode lasers (ECDLs) are mixed in the OP-GaAs crystal to generate a difference-frequency Idler frequency tunable between 7.9 and 8.5 µm. The tunable Idler beam is collimated and mode matched to couple into a high Finesse laser cavity for CRDS absorption measurements. The



transmitted light was collected into an MCT detector and sampled into a computer by a digitizing card. A detailed explanation of CRDS absorption measurements is shown in Ref. 3.

Figure 2: Experimental setup. The pump source (ECDL amplified by Pr amplifier, 1.3 µm ) is mixed with a signal source (ECDL mplified by Er amplifier, 1.55 µm) to generate idler wavelength radiation (8.4 µm). The polarizations of the three interacting beams relative to the crystal lattice are indicated. The pump and signal are combined and focused into the OP-GaAs crystal. The idler beam is colimated and mode matched to the cavity (L = 46 cm). The transmitted Idler power is collected by a MCT detector and sampled by a digitizer to extract cavity decay time.

N2O gas was chosen for demonstrating CRDS using the OP-GaAs light source. The OP-GaAs crystal was heated to 700 C, to allow scanning the Idler frequency in the 1185 – 1230 cm-1 frequency range (8.13 – 8.43 µm). This CW Mid IR light source demonstrates broad tuning, narrow linewidth (< 1 MHz), and small step size (< 0.001 cm-1). Figure 3 shows N2O absorption spectrum overlaid with DFG output power at crystal temperature of 700 C. The continuous tuning of the Idler beam allows for scanning of narrow absorption features while obtaining large scanning range. At present, useful ring down signals can be recorded at 1185-1220 cm-1 due to low signal to noise ratios beyond this spectral region. Future improvements of the spectroscopy system including multi-grating OP-GaAs crystal systems, reduction of detector noise, increase of DFG source power, and optical system throughput optimization will further increase the spectral region realized by this spectroscopy system.

Figure 3: N2O absorption spectrum overlaid with DFG output power at crystal temperature of 700 C. Vertical lines represent the region between which useful cavity ring down signals can be obtained. These limits are imposed (left) by low DFG power values, and (right) by high mirror reflectivity (low cavity transmission).

0

1

2

3

4

1170 1180 1190 1200 1210 1220 1230 1240Frequency (cm-1)

DFG

pow

er ( µ

Wat

t)

BS OP-GaAsL1 F1

MCTdetector

]110/[]011[

]100[ ]011/[]110[

pump

signal

idler

External cavity DL and Er fiber amplifier (PM) λ=1.55 µm, 1.1 W

L2

L3

Ge filter

External cavity DL and Pr fiber amplifier λ=1.3 µm, 40 m W

CavityBS OP-GaAsL1 F1

MCTdetector

]110/[]011[

]100[ ]011/[]110[

pump

signal

idler

External cavity DL and Er fiber amplifier (PM) λ=1.55 µm, 1.1 W

L2

L3

Ge filter

External cavity DL and Pr fiber amplifier λ=1.3 µm, 40 m W

Cavity



CRDS absorption was measured for N2O gas at the 1185 – 1192 cm-1 frequency range with ~ 0.007 cm-1 step size. Higher resolution CRDS scans were also measured with step sizes as low as 0.001 cm-1. The high resolution OP-GaAs DFG source (< 1MHz) allows for this high sensitivity and selectivity demonstrated in the spectral scanning. Figure 4 shows spectra of the first overtone of N2O, with water vapor and methane (CH4) impurities. Theoretical Hitran absorption spectra of these species are overlaid with experimental data. These spectra show the potential of a broadly tunable narrow frequency Mid IR DFG source for sensitive Mid IR spectroscopy.

Figure 4: Cavity ring down spectra of the first overtone of N2O, with water vapor and methane (CH4) impurities. Theoretical Hitran absorption spectra of these species are overlaid with experimental data to illustrate the excellent agreement. The theoretical Hitran absorption values of the individual species were reduced by a constant to improve presentation clarity.

In conclusion, we demonstrated a CW light source based on OP-GaAs with broad tuning, narrow linewidth (< 1 MHz), and small step size (< 0.001 cm-1). CRDS of N2O was demonstrated with this light source around 1185 cm-1 (8.44 µm). The combination of a tunable MID-IR DFG light source with CRDS will enable identification and highly sensitive detection of a wide range of species. This system will find many applications in chemical and biological sensing.

References [1] P. Werle, F. Slemr, K. Maurer, R. Kormann, R. Mucke, and B. Janker, “ Near- and mid-infrared

laser optical sensors for gas analysis”, Opt. Lasers in Eng., 37, 101 (2002). [2] F. K. Tittel, D. Richter, and A. Fried, "Mid-Infrared Laser Applications in Spectroscopy," Solid-

State Mid-Infrared Laser Sources, Eds. I.T. Sorokina, and K.L. Vodopyanov, Springer Topics Appl. Phys. 89, 445-510 (2003).

[3] P. Zalicki, and R. N. Zare, “Cavity ring-down spectroscopy for quantitative absorption measurements”, J. Chem. Phys., 102, 2708, (1995).

[4] O. Levi, T. J. Pinguet, T. Skauli, L. A. Eyres, K. R. Parameswaran, J. S. Harris Jr., M. M. Fejer, T. J. Kulp, S. Bisson, B. Gerard, E. Lallier, and L. Becouarn, “Difference-frequency generation of 8-µm radiation in orientation-patterned GaAs crystal”, Opt. Lett. 27, 2091 (2002).

-2000

-1000

0

1000

2000

1184 1186 1188 1190 1192 1194

Frequency (cm-1)

Abso

rptio

n (p

pm)

H2OCH4

N2OFit

Experiment

SPRC Annual Meeting 2003

Speaker Presentations (.pdf) Each presentation will open as a separate file; some have .ppt versions available also

Barnholt

Belikov (.ppt)

Bhatnagar (.ppt)

Campbell (.ppt)

Denman (.ppt)

D’Orgeville

Giesen

Gopalan

Hum (.ppt)

Jopson

Kaertner (.ppt)

Kahn

Kalogerakis (.ppt)

Kane

Lemoff (.ppt)

Matteo (.ppt)

Novotny (.ppt)

Santori (.ppt)

Saraf

Schuele (.ppt)

Sinha (.ppt)

Suh

Wickham (.ppt)

Xie

Yablonovitch

Post-Deadline Talk Files (.pdf) Scott Bisson, Ofer Levi, Thomas J. Kulp, James S. Harris, Martin Fejer; “Long-Wave IR Chemical Sensing based on Difference Frequency Generation in Orientation Patterned GaAs” Jason M. Eichenholz, Ceyhun Akcay, Jannick Rolland; “Spectral Shaping to Inhibit Sidelobes of Point Spread Function in Optical Coherence Tomography” .ppt version Monica Minden, Hans Bruesselbach, Jeffrey Rogers, Cris Jones, Dave Hammon John Solis,Metin Mangir, A. Siegman, "Coherent Combining og Fiber Amplifiers and Lasers"

Poster Files (will open as .pdf unless otherwise noted; some have .ppt versions also) Seth R. Bank, Mark A. Wistey, Homan B. Yuen, Lynford L. Goddard, Wonill Ha, and James S. Harris Jr.; “Progress Towards High Performance 1.55 mm Lasers Grown on GaAs for Metro Area Networks and Optical Interconnects” (.ppt only) Aparna Bhatnagar, Christof Debaes, Ray Chen, Noah Helman, Gordon Keeler, Salman Latif and David A. B. Miller; “Optical Clocking using Mode Optical Clocking using Mode-locked locked Lasers” .ppt version Mathieu Charbonneau-Lefort and Martin M. Fejer; “Optical Parametric Amplification with Chirped Quasi-Phase-Matching Gratings” C. H. Cheng, A. S. Ergun, B. T. Khuri-Yakub; “Electrical Through Silicon Wafer Interconnects for High Frequency Photonic Devices” H. Chin, R. Chen, and D. A. B. Miller; “Linear electro-optic conversion of current puleses for a photonic-assisted electrical analog-to-digital converter” .ppt version Eleni Diamanti, Edo Waks and Yoshihisa Yamamoto; “Generation of photon number states” D.P. Fromm, A. Sundaramurthy, K.B. Crozier, P.J. Schuck, G.S. Kino, C.F. Quate, W.E. Moerner; “Development of Electromagnetically Enhanced Au “Bowtie” Nanostuctures” .ppt version J. Hwang, M. M. Fejer and W. E. Moerner; “Exploring Novel Methods of Interferometric Detection of Ultrasmall Phase Shifts” .ppt version D.B. Jackrel, and J.S. Harris; “Advanced LIGO Photodiode Development” .ppt version Yang Jiao, Shanhui Fan, and David A. B. Miller; “Designing for beam propagation in periodic and nonperiodic photonic nanostructures: extended Hamiltonian method” .ppt version In-Sung Joe, Kyoungsik Yu, and Olav Solgaard; “Flexible, scalable 100Tbps Internet router with an optical switch fabric using optical tunable filters” Alireza Khalili, James S. Harris; “Side-coupled in-line fiber-semiconductor laser” .ppt version

Uma Krishnamoorthy, Il Woong Jung, Patrick Lu, Yves-Alain Peter, Emily Carr, Robert L. Byer, Olav Solgaard; “Development of Segmented Deformable Mirrors for Free Space Communication” Paulina S. Kuo, Xiaojun Yu, Konstantin L. Vodopyanov, Prof. Martin M. Fejer, Prof. James S. Harris; “Growth and Characterization of Thick-film Orientation-patterned GaAs” Carsten Langrock, Rostislav V. Roussev, and Martin M. Fejer; “Counting Photons PPLN Style…Sum-frequency generation in a PPLN waveguide for efficient single-photon detection at communication wavelengths” WahTung Lau, Prof Shanhui Fan; “Creating large bandwidth line defects byembedding dielectric waveguides into photonic crystal slabs” Rostislav Roussev, Arun Sridharan, Karel Urbanek, Robert Byer and Martin Fejer; “Parametric Amplification of 1.6 mm Signal in Annealed- and Reverse- Proton Exchanged Waveguides” Vijit A. Sabnis, Hilmi Volkan Demir, Jun-Fei Zheng, Onur Fidaner, James S. Harris, Jr., and David A. B. Miller; “Novel optically-switched electroabsorption modulators for wavelength conversion” Shally Saraf, Supriyo Sinha, Arun Kumar Sridharan and Robert L. Byer; “Power Scaling of Diffraction Limited Slab Amplifiers for LIGO” L. Scaccabarozzi, Z. Wang, X. Yu, S. Fan, M. M. Fejer, J. S. Harris; “Photonic Crystal AlGaAs Microcavities for Non-Linear Optical Applitcations” Andrew M. Schober, Mathieu Charbonneau-Lefort, Martin M. Fejer; “Parametric Oscillation of Ultrashort Pulses in Quasi-Phase-Matched Nonlinear Materials” G. Schüle, Ph. Huie, A. Vankov, E. Vitkin, L. Perelman, D. Palanker; “Noninvasive determination of temperature-induced subcellular changes in cells using light scattering spectroscopy” Miroslav Shverdin, Deniz Yavuz, David Walker; “Stable Three Dimensional Raman Soliton” .ppt version Supriyo Sinha, Yin-wen Lee, Michel J. F. Digonnet, Robert L. Byer; “Scaling of High Average Power Fiber Lasers at Stanford” .ppt version

Ji-Won Son, Yin Yuen, Sergei S. Orlov, Bill Phillips, Ludwig Galambos, and Lambertus Hesselink; “Direct e-beam Domain engineering in LiNbO3 thin films grown by liquid phase epitaxy” .ppt version Wonjoo Suh and and Shanhui Fan; “Flat-top Reflection and All-pass Transmission Filter using Coupled Resonance in Photonic Crystal Slabs” Zheng Wang and Shanhui Fan; “Compact all-pass filters in photonic crystals as the building block for high capacity optical delay lines” J. Wisdom, R. Gaume, V. Kondilenko, T. Plettner, S. Wong, R. Route, M. Digonnet and R.L. Byer; “Scattering, Absorption and Laser Performance Comparisons Between Ceramic and Single Crystal Nd:YAG” .ppt version Kenneth K.Y. Wong, Michael E. Marhic, and Leonid G. Kazovsky; “Toward Practical Application of Fiber Optical Parametric Amplifiers in Optical Communication Systems”

Xiuping Xie, Andrew Schober, Carsten Langrock, Rostislav Roussev, Jonathan Kurz, Martin Fejer; “Cascaded Optical Parametric Generation in Reverse Proton Exchanged Waveguides” Mehmet Fatih Yanik, Shanhui Fan, Marin Soljačić, J. D. Joannopoulos; “High-density Low-power All-optical Logic GateS in Photonic Crystals” Kyoungsik Yu, Daesung Lee, Uma Krishnamoorthy, and Olav Solgaard; “Variable Bandwidth Optical Filter and Tunable Optical CDMA Encoder/decoder Based on MEMS Gires-Tournois Interferometer” H.B. Yuen, S.R. Bank, M.A. Wistey, V. Gambin, W. Ha, J.S. Harris Jr.; “Growth and characterization of mbe-grown GaInNAs(Sb) for long-wavelength optoelectronic devices” (.ppt only) X.J. Zhang, S. Zappe, C-C. Chen, O. Sahin, J. Harris, C. Quate, M. Scott and O. Solgaard; “Silicon microsurgery-force sensor based on diffractive optical MEMS encoders”

stanford photonics research centerstanford.edu/group/khuri-yakub/publications/02_cheng_01.pdf ·...

Documents