Monolithically integrated InP-based photonic chip development for O-CDMA systems

66 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 11, NO. 1, JANUARY/FEBRUARY 2005

Monolithically Integrated InP-Based Photonic ChipDevelopment for O-CDMA Systems

Chen Ji, R. G. Broeke, Y. Du, Jing Cao, N. Chubun, P. Bjeletich, F. Olsson, S. Lourdudoss, R. Welty, C. Reinhardt,P. L. Stephan, and S. J. B. Yoo

Abstract—This paper discusses photonic integration efforts to-ward developing an InP-based monolithically integrated photonicchip for optical code-division multiple-access (O-CDMA) systemapplications. The chip design includes the colliding pulsed mode(CPM) locked laser, the Mach–Zehnder interferometer-basedthreshold detector (MZI), and the monolithic O-CDMA en-coder/decoder chip based on array-waveguide-gratings and phasemodulator arrays. The compact 4 1 cm monolithic chip canreplace a complex and large O-CDMA setup based on bulk optics.The integration technique involves active-passive integration usingdry etching, metal organic chemical vapor deposition growth, andlateral hydride vapor phase epitaxy regrowth technologies. Thefabricated CPM showed stable 1.54 ps modelocked laser output,the MZI showed excellent O-CDMA threshold detection, andthe O-CDMA encoder showed Walsh-code O-CDMA encoding.Further, the fabricated devices showed excellent planarity,which accelerate our progress toward monolithic integration ofO-CDMA systems.

Index Terms—Arrayed waveguide grating, colliding pulse mode(CPM) locked laser, InP, Mach-Zehnder interferometer (MZI),monolithic integration, nonlinear threshold detector, opticalcode-division multiple-access (O-CDMA), photonic integration.

I. INTRODUCTION

THE RECENT surge in Internet traffic has propelled rapiddeployments of wavelength division multiplexing (WDM)

and time division multiplexing (TDM) technologies in metroand wide area networks. In local area networks (LANs), themarket emphasis is on the high volume adoption of simple andinexpensive components, and, as a consequence, the high hard-ware cost and complex protocols associated with WDM andTDM technologies are the main hurdles in their widespread de-ployment in LAN’s.

The optical code-division multiple-access (O-CDMA) tech-nology is potentially well suited for providing very flexible andhigh-capacity access to the vast networking capacity availablein all-optical LANs. Unlike WDM and TDM networks that use

Manuscript received August 5, 2004; revised October 29, 2004. This workwas supported in part by DARPA and SPAWAR under agreement numberN66001-02-1 by JOP-OIDA supported by DARPA and NSF, and by generousequipment loan from Pritel, Inc.

C. Ji, R. G. Broeke, Y. Du, J. Cao, N. Chubun, P. Bjeletich and S. J. B. Yooare with the Department of Electrical and Computer Engineering, University ofCalifornia, Davis, California 95616, U. S. A. ([email protected]).

F. Olsson and S. Lourdudoss are with the Department of Microelectronics andInformation Technology, Royal Institute of Technology, KTH-Electrum 229,S-16440 Kista, Sweden.

R. Welty, C. Reinhardt and P. L. Stephan are with Lawrence Livermore Na-tional Laboratory, Livermore, CA 94550, U. S. A.

Digital Object Identifier 10.1109/JSTQE.2004.841710

Fig. 1. Basic operation principle of (a) an O-CDMA encoder and (b) anO-CDMA decoder.

wavelength and time-slot channels, O-CDMA networks con-figure or reconfigure such access by assigning or reassigning ofcodes. In particular, local nodes can independently choose suchcodes in the optical layer without involving central nodes, andthus LANs become very easy to manage. Further, rapid codereconfigurations allow security enhancements in the physicallayer without posing limitations in data rates.

In the past two decades, there has been active research onO-CDMA based on a number of different schemes [1]–[7].While the O-CDMA technology offers numerous unique ad-vantages, it must overcome two challenges mainly pertainingto integration and multiple user interference (MUI). TypicalO-CDMA systems employ a number of discrete componentsincluding encoders, decoders, optical sources, and receivers. Inmany cases, O-CDMA systems include carefully aligned bulkoptical components in free space. Monolithic integration andminiaturization are essential for reliable operation and low-costmanufacturing of such systems. While O-CDMA utilizes codesto separate common channels, multiple users in the same LANwill cause interferences (MUI) even if they employ differentcodes. As a result, the O-CDMA network will experiencedifferent levels of MUI depending on the O-CDMA methodsand codes utilized in the network. Based on how these codes areimplemented in time, space, and wavelength domains, a numberof different O-CDMA schemes exist. This paper focuses onan O-CDMA implementation based on the spectral phase en-coded time spreading (SPECTS) technique [5] which achievesencoding and decoding by applying phase shift to individualslices of the optical spectrum. Fig. 1 shows the basic operationprinciple of (a) the O-CDMA encoder and (b) the O-CDMAdecoder. In the encoder, the first diffraction grating spatiallyspreads the ultrafast optical pulse and maps it on the Fourierplane where the spatial light modulator (SLM) applies phasemodulation (e.g. 0, ) to individual slices of the spectral profile.Another diffraction grating recombines the spectral slices intothe time domain pulse, thus completing the “pulse shaping”

1077-260X/$20.00 © 2005 IEEE

JI et al.: MONOLITHICALLY INTEGRATED InP-BASED PHOTONIC CHIP DEVELOPMENT FOR O-CDMA SYSTEMS 67

Fig. 2. O-CDMA encoder–decoder operation based on monolithicAWG-phase modulator-AWG semiconductor chips.

by applying optical codes. The “time-spread” pulse shape inthe time domain is a result of spectral phase encoding in thefrequency domain. The principle of the O-CDMA decoder isidentical to that of the encoder, and the conjugate code (e.g.0, ) applied to the SLM restores the original pulse. In anO-CDMA network, there can be a large number of O-CDMAnodes (encoders and decoders), and a proper choice of a codeset, such as the Walsh code or the -sequence code set withorthogonal or pseudoorthogonal properties, can greatly reducethe MUI effects. As stated earlier, proper decoding takes placewhen the decoder employs the conjugate code set for a givenencoding pulse, thus recovering the original pulse. The samedecoder will improperly decode a pulse encoded differently,resulting in another time-spread pulse which resembles a noiseburst, and the orthogonality of the codes will provide lowinterference (MUI) between the noise burst and the properlydecoded pulse. It is important to note that both improperly andproperly decoded pulses will have the same integrated powerbut with different pulse shapes at fine time scales, thus a con-ventional square-law optoelectronic detector cannot distinguishthem. Hence, nonlinear thresholding detection is necessaryin SPECTS O-CDMA methods, and the same is true for themajority of other O-CDMA methods. A recent publication [6]reported error-free 4 10 Gb/s multiuser SPECTS O-CDMAnetwork operation in a free-space testbed incorporating ahighly nonlinear fiber (HNLF)-based threshold detector. Thetestbed also included a number of bulk components includinga pico-second mode locked laser, diffraction grating pairs, andSLMs.

The O-CDMA system can possibly achieve monolithicintegration by incorporating advanced semiconductor compo-nents. First, encoders and decoders can utilize a monolithicallyintegrated arrayed waveguide grating (AWG) pair and phasemodulators. Fig. 2 illustrates such O-CDMA encoder anddecoder in analogy to those in Fig. 1. Even further integrationcan take place by integrating a colliding pulse mode lockedlaser (CPM) with the O-CDMA encoder, and by integratinga Mach–Zehnder interferometer (MZI)-based threshold de-tector with the O-CDMA decoder. The former completes anO-CDMA transmitter, and the latter completes an O-CDMAreceiver. Fig. 3 shows a futuristic view of the monolithicO-CDMA transceiver. Realized on a semiconductor wafer(e.g., InP–InGaAsP wafer), the monolithic transceiver willoccupy an extremely small footprint ( 4 1 cm), allow rapidcode reconfigurations, and will offer very robust operation.This paper discusses a roadmap and progress toward realizingthe monolithically integrated InP O-CDMA microsystem.Section II provides an overview of the integrated O-CDMA

Fig. 3. Schematic block diagram of O-CDMA integrated photonic chip.

chip and its operation. Section III describes semiconductorO-CDMA chip integration process common to all componentsin the microsystem. Section IV demonstrates performance ofthe subsystems and components constituting the O-CDMAsystem, which includes monolithic and reconfigurable InPO-CDMA encoders and decoders (Section IV-A), InP CPMlasers (Section IV-B), and InP Mach-Zehnder threshold detec-tors (Section IV-C). Section V summarizes the paper.

II. INTEGRATED PHOTONIC CHIP OVERVIEW

The integrated AWG-phase modulator array-AWG chip onan InP substrate (Fig. 2) can achieve O-CDMA data encodingand decoding functions (Fig. 1). The ultrashort laser pulse en-tering the first AWG will spatially spread each spectral slice ofthe incident pulse according to the well-known AWG character-istics. The phase modulator array allows programmable phasecoding across the spectrum by applying a reverse bias voltageon each of the phase modulator thus achieving a refractive indexchange through electrooptical effect [8], according to the de-sired phase code (e.g., Walsh code). The second AWG recom-bines the phase shifted spectral slices and generates the encodedtime domain pulse. The chip contains a reversed biased phasemodulator array and passive waveguide components, and it ex-pects to consume very little electrical power.

Fig. 3 shows a block diagram schematic of the monolithic InPO-CDMA transceiver. Again, the O-CDMA transmitter sectionincludes the CPM laser providing subpicosecond optical pulses,the data (electroabsorption) modulator, and the O-CDMA en-coder consisting of AWG-phase modulator array-AWG. TheO-CDMA receiver section includes the O-CDMA decoder con-sisting of AWG-phase modulator array-AWG, and the thresholddetector consisting of differential Mach-Zehnder wavelengthconverter and the photodiode. The entire chip is to be realizedin the InP–InGaAsP material system with a target operationwavelength of 1550 nm.

In the transmitter section of the chip, the CPM laser pro-vides an ultrashort optical pulse train synchronized to externaldrive sources using the electrical hybrid mode locking tech-nique. High-contrast distributed Bragg reflectors (DBR) can de-fine the CPM laser cavity on chip while providing broad reflec-tion bands ( 1 THz) centered at 1550 nm. The integrated elec-troabsorptive (EA) modulator modulates data on the CPM laseroutput. The mode locked optical pulses propagates through theAWG-phase modulator array-AWG encoder section with proper


bias voltages for O-CDMA encoding. The encoded pulses cou-ples into the lensed output fiber.

On the receiver side, the incoming encoded signal couplesfrom an input optical fiber into the AWG-phase modulatorarray-AWG decoder section. Application of a decoding codethat is conjugate to the encoding code results in properly de-coding the optical pulse. Subsequently, a differential MZI-basedthreshold detector with a laser and a photodiode integrated onchip will allow O-CDMA detection. Section IV-C furtherdiscusses the operation principle of the differential MZI as anO-CDMA threshold detector.

The monolithic photonic chip will share one thermoelectricalcooler for temperature control, and a butterfly package will in-clude proper electrical, optical, and thermal considerations. Theelectrical packaging will involve wirebonding to electrical con-tacts for encoding, decoding, data modulation, and data detec-tion. The optical packaging includes lensed fibers for couplingoptical signals in and out of the chip. The integrated photonicchip and the package completely replace the bulk free-spaceO-CDMA optical system [6], [7] while providing relatively ro-bust and compact operation. The estimated chip size is 41 cm for a 32-channel 50-GHz channel spacing AWG-basedO-CDMA transceiver, which requires 64 wirebondings for theoptical coding and decoding, and additional wirings for datamodulation and detection. Sections III and IV discuss the de-tails of the device fabrication, operation, and packaging.

III. ACTIVE–PASSIVE INTEGRATION AND DEVICE FABRICATION

The O-CDMA transceiver chip requires monolithic integra-tion of various discrete semiconductor components. The CPMlaser and the MZI threshold detector contain both active and pas-sive waveguides, while the AWGs and phase modulator arraysconsist of passive waveguides. Passive waveguides also inter-connect all components on chip. What is critical for realizing amonolithic chip microsystem is the photonic integration methodcombining active and passive waveguide sections that providesmooth active–passive interfaces with minimal coupling lossand low reflection at the interface.

The O-CDMA microsystem utilizes a common initial epi-taxial structure shown in Fig. 4(a). The initial wafer, grown onan -type wafer by metal organic chemical vapor deposition(MOCVD) consists of a 1.5- m-thick -type InP layer, a0.5- m-thick weakly -type InGaAsP waveguide core layerwith photoluminescence wavelength at 1150 nm (1.15Q),a 20-nm undoped InP etch stop layer, an undoped multiplequantum well (MQW) structure with six quantum wells, and a20-nm undoped InP etch stop layer. Fig. 4(a) indicates dopinglevels of each layer. The MQW active region in the structurecontains six 6-nm-thick undoped In Ga As QWs with7-nm-thick InGaAsP (1.25Q) barriers, adjusted to provide itsphotoluminescence peak wavelength at 1570 nm. Through thefabrication steps described in the following, the active sectionof the chip retains the initial structure and obtains additionallayers grown on top. These layers are a 0.5- m-thick -typeInP layer (Zn doping at 2E16/cm ), a 0.5- m-thick -typeInP layer (Zn doping at 1E17/cm ), a 1- m-thick -type InPlayer (Zn doping at 5E17/cm ), and a 0.1- m-thick -type

Fig. 4. (a) Starting epitaxial structure for the active–passive integrationprocess (b) epitaxial structure for the active section of the chip after regrowth.The passive section does not have the MQW active region.

In Ga As contact layer (Zn doping at 5E18/cm ). Like-wise, the passive section will also have identical layers grownbut the MQW section will be etched off before this growth.Consequently, the final structure for the active section of the in-tegrated chip is as shown in Fig. 4(b) and the passive section ofthe chip is identical except that the multiquantum well (MQW)active region is not present. This fabrication process keeps thewaveguide core aligned and continuous across the active-pas-sive interface, thus minimizing coupling losses and reflectionsrising from the modal mismatch. The calculated confinementfactor for the MQW is still quite reasonable at 4% despite thefact the peak of the optical mode is off-center due to the pres-ence of the waveguiding layer below the MQW active region.The doping profiles for the 2- m-thick -type InP claddinglayer above the MQW active region and the 1.5- m-thick

-type InP cladding layer below the waveguiding core layerconsist of a series of steps that increase monotonically awayfrom the waveguiding core layer. This doping profile tries tosatisfy the conflicting requirements of minimizing propagationlosses from free carrier absorption in the passive section and ofavoiding excessively high junction series resistance inthe active section.

The processing flow for fabricating the CPM lasers, theAWG-phase modulator array-AWG encoders and decoders, andthe MZI threshold detectors follows the active-passive integra-tion process described as follows. Starting with the epi-waferstructure of Fig. 4(a), we deposited 270-nm PECVD SiOpatterned to mask the active sections of the chip and selectivelyremoved by wet etching the MQW active region in the passivesections, stopping on the 20-nm InP layer protecting the 1.15 Qwaveguiding core layer. After the removal of the SiO maskinglayer, a thin (100 nm) undoped InP layer, the 2- m-thickp-type InP cladding layer with the incremental doping profile,and the 100-nm-thick highly doped p-type InGaAs contactlayer were MOCVD regrown across the whole wafer. Fig. 5shows a scanning electron micrograph (SEM) cross-sectionthrough the active-passive interface with the MQWs delineatedwith chemical staining. The SEM photo reveals the smoothinterface with excellent planarity after the MOCVD regrowth


Fig. 5. SEM cross-section of the active-passive interface after regrowthshowing apparently very smooth regrowth interface.

free from apparent defects. The continuous 1.15Q waveguidingcore layer throughout the active and passive sections ensuresthat the optical mode cross the active-passive interface withminimal disturbance, the only discontinuity being the removalof the 85-nm-thick MQW active region in the passive sections.Modal propagation simulations estimate the mode couplingloss of 0.3 dB and the reflectivity of 3E-5 at the active-passiveinterface. To reduce the reflection at the interface even further,the waveguide fabrication also included designs where theactive-passive interface was laterally tilted at 45 relative to thewaveguides oriented along the direction. The fabricationrevealed, again, a smooth and clean active-passive interfacesimilar to that in Fig. 5. Theoretical estimations indicate that the45 interfacial tilt effectively suppresses the back reflection atthe interface propagating backward in the waveguide, assuminga transverse electric (TE) waveguide mode consisting of manyplane waves.

After the completion of the active-passive integration re-growth, the fabrication process continued to laterally definethe active and passive waveguides by a 270-nm PECVD SiOfilm deposition, photolithography, and a CF reactiveion etch (RIE) etching. Using the patterned SiO as a mask,a CH H -based RIE process vertically dry etched the

-type InGaAs contact layer, the -type InP layer, and past the1.15Q waveguide core layer. Subsequently, a lateral regrowthprocess [9] based on a low-pressure hydride vapor phaseepitaxy (HVPE) achieved planarizing regrowths of Fe dopedsemi-insulating InP seeded by the waveguide sidewalls. Thelateral regrowth process is characterized by the highly en-hanced growth rate in the lateral direction. Fig. 6(a) shows thecross-sectional view of the buried waveguide after the lateralregrowth. The lateral regrowth resulted in a gently sloped mesaaround the deep-etched waveguide with an excellent planarprofile close to the etched waveguide, creating essentially aburied heterostructure (BH) waveguide using a single HVPEregrowth step. The HVPE-based BH waveguide fabricationtechnique has many advantages compared to conventional ap-proaches used for realizing ridge, rib-loaded, or traditional BHcurved waveguides. In general, BH waveguides offer lower lossthan dry-etched ridge waveguides, and more repeatable resultsthan dry-etched rib-loaded waveguides. The HVPE-based BHfabrication technique requires one less regrowth step than the

Fig. 6. SEM of the waveguide defined by deep etching and HVPE lateralregrowth of Fe doped InP. (a) Zoomed-in cross-sectional view. (b) Portion ofthe AWG-phase modulator array-AWG encoder/decoder.

traditional BH fabrication process. In addition, the combinedBH fabrication technique requires no critical control of theetch depth. This is necessary for a “shallow-etched” waveguidewith the shallow-etching stopping at exactly 100 nm below theinterface between the core and the upper cladding layers. Thedemonstrated HVPE-based fabrication technique provides amechanically robust planarization technique that passivates theetched waveguide sidewall and significantly reduces scatteringlosses from the typical sidewall roughness after the dry etching.Furthermore, since the regrowth time is short, the possibilityof dopant redistribution in the basic structures comprising theintegrated components is minimal. Fig. 6(b) shows a portionof an integrated AWG-phase modulator array-AWG encoderafter the HVPE lateral regrowth showing an excellent profileof planarized waveguides. Such high-quality planarizationis essential for realizing integrated photonic microsystemsrequiring further metallization and fabrication steps.

Subsequently to the HVPE regrowth, we proceeded with awet etching process to remove the SiO masking layer, and abisbenzocyclobutene (BCB) process to provide electrical insu-lation and to reduce the bond pad capacitance in the saturableabsorber (SA) section of the CPM laser. This process includesspinning of a 2- m-thick BCB polymer layer, curing at 250 C,and planarization etching in a CF O RIE process. Finally, thecompleting steps of the integrated chip fabrication includes pho-tolithography, -metal (Ti/Pt/Au) e-beam evaporation, a liftoffprocess using a bi-level resist, backside lapping of the waferto 200 m, the -metal (AuGeNi/Au) e-beam evaporation, andrapid thermal annealing (RTA).

Section IV demonstrates individual modules and devices (theO-CDMA encoder and decoder, the CPM laser, and the MZIthreshold detector) fabricated and tested individually.

IV. O-CDMA INTEGRATED CHIP COMPONENT DEVELOPMENT

A. Integrated AWG-Phase Modulator-AWG Encoder/Decoder

The key element in a SPECT O-CDMA system is the spec-tral phase encoder and decoder. Realizing the integrated encoderand decoder on an InP platform is especially interesting, becauseof the combined capabilities of high resolution AWGs, elec-trooptical waveguide phase modulators, optical sources, and de-tectors all available on the InP platform. Recent reports includedAWG-based encoders using silicon with external phase modula-tors [10] and polymer on silicon with chip bonded phase modu-lators [11]. This section presents a monolithically integrated InP


Fig. 7. O-CDMA integrated encoder/decoder chip showing multiplexingAWG1, eight channel phase modulator array with electrical bond pads, delaylines for path length equalization, and multiplexing AWG2.

O-CDMA encoder (or decoder) including AWGs and an elec-trooptical phase modulator array for fast reconfigurability andnegligible power consumption.

Operation Principle and Design: Fig. 7 shows the layout ofthe monolithic AWG-phase modulator array-AWG O-CDMAencoder chip. The chip consists of all passive sections. A sub-picosecond optical pulse enters the first AWG (AWG1) throughone of the input waveguides. The multiple (five) input ports ofthe AWGs allow discrete tuning of the spectral responses ofthe AWG so that optimal choice of the one providing the bestoverlap between the spectral characteristics of the two AWGs(AWG1 and AWG2) can achieve the best throughput. Whilethe photolithography mask incorporates identical AWG designs,fabrication and material growth tolerances will inevitably pro-duce a finite misalignment between the two AWG spectral re-sponses when the input and output ports are selected symmetri-cally.

AWG1 decomposes the incoming optical pulse into eightspatially separated spectral slices. Then, the optical wave ineach slice undergoes phase modulation in an electroopticalwaveguide. Additional delay lines incorporated between theelectrooptical waveguide array and AWG2 compensate forintrinsic optical path length differences between channels dueto the geometry of the AWG pairs. Therefore, the O-CDMAencoder chip is designed to achieve equal optical path lengthfor any wavelength channel within the O-CDMA spectral bandwhen there is no applied bias on the electrooptical waveguides.In practice, the nonuniformity in fabrication and materialgrowth will add phase errors, which will result in unequaloptical path lengths across the spectrum. Applying additionalbias voltages across the electrooptical waveguides allows finiteamount of corrections for the phase errors. Finally, the eightwavelength slices are multiplexed by AWG2 and transmittedoff-chip through one of the output waveguides.

Fig. 8(a) shows the measured transmission spectrum of asingle AWG using the central input waveguide. The AWGdesign for the O-CDMA encoder included an identical pair ofAWG’s, and the transmission spectra exhibited eight-channeltransmission with a 180-GHz channel spacing, a measuredcentral wavelength of 1549-nm (1550-nm design), and a freespectral range (FSR) of 12-channel spacings (12 180 GHz).The total wavelength span covered by the eight channels isapproximately 1.4 THz, which is sufficient for encoding pulsesdown to 300 fs. The phase modulator array consists of eight1-mm-long electrooptical phase modulators with reverse biasedacross the PIN junctions with negligible power consumption

Fig. 8. Transmission spectra of (a) single eight-channel AWG with 180-GHzchannel spacing and (b) back-to-back AWG pair.

Fig. 9. Packaged encoder/decoder chip with electrical contact pins for thephase modulator and TE cooler for temperature control.

( 1 mW). The test passive waveguides of 3 m width on theAWG-phase modulator array-AWG chip measured 2.5 dB/cmpropagation loss. The AWG transmission spectrum showedsignal-to-background ratio (SBR) of 18 dB and excess loss of5 dB compared to a straight 3 m-wide passive waveguide.The eight-channel AWG passbands exhibited less than 2 dBnonuniformity in the peak transmission values. Two AWGs inan encoder or decoder chip had typically a center wavelengthmismatch of up to 2 nm. As stated earlier, selecting of a dif-ferent input waveguide of the AWG will provide a shift in thespectral response, which can partially compensate for this mis-match. Fig. 8(b) shows the spectral response of the AWG-phasemodulator array-AWG monolithically integrated chip whenan optimal selection of the input waveguide resulted in theleast amount of misalignment. Compared to the passbands ofthe single stage AWG, the AWG-phase modulator array-AWGchip showed significantly narrower passbands resulting fromthe cascaded filter characteristics and from the residual passband mismatch between the two AWG spectral responses.This periodic amplitude structure in the frequency domain willaffect the impulse response as ringing with 5 ps ( 1/180GHz) periodicity in the time domain even without any phasemodulation.

The phase modulator response determined from a wavelengthshift in the Fabry–Perot resonance measurements is approx-imately 0.10 rad/V/mm, which is approximately half of thetheoretically predicted value. This discrepancy is presumablydue to -type dopant (Zn) diffusion, causing a displacementbetween the optical mode center and the PN juntion. Based


Fig. 10. Measurement setup for on-chip O-CDMA encoding with packaged AWG-phase modulator array-AWG encoder chip.

on the measured data, a phase shift in a 1-mm-long phasemodulator requires a 30-V reverse bias. These estimates arein agreement with the O-CDMA encoder measurement resultsexplained below.

The overall size for the 8-channel O-CDMA encoder chip was12 4 mm. Fig. 9 shows the finished encoder chip packagedin a butterfly package with separate electrical connections tothe individual phase modulator elements, and a thermoelectricalcooler for temperature stabilization.

Walsh Code Encoding Results: Fig. 10 shows the measure-ment setup for characterizing the encoder chip. The fibermode-locked laser generated a 10-GHz pulse train, which wascompressed to 0.4 ps and passed through a polarization con-troller (PC) which maintained the input signal polarization toTE. Tapered lensed fibers then coupled the pulse into and outof the encoder chip. The encoder chip output was amplifiedwith a dispersion compensated erbium doped fiber amplifier(EDFA) and coupled into a cross-correlator for time domaincharacterization.

The encoder chip was characterized using a set of orthogonal8-b Walsh codes. It is difficult to completely avoid phase er-rors in the O-CDMA encoder chip, which essentially requiresbetter than 1E-4 optical phase control ( 0.25 n/L, where

m, , 1.5 mm). Fortunately, phase mod-ulation across the phase modulator arrays can be dynamicallyadjusted and compensating voltages can be easily applied to thephase shifters to correct for the initial phase coding offset. Moresophisticated approaches have been reported in literature, butthe most direct approach which we currently employ involvesmaking fine adjustments to the phase modulator array codingin real time until we obtain the sharp single peak output, cor-responding to the unencoded pulse. The phase coding extractedthis way then constitutes the phase error offset in the encoderchip. The solid curve in Fig. 11(a) shows the measured cross-correlation trace from the encoder chip output with proper phasecompensation (effectively approximating a W8 Walsh code at3.5-V reverse bias). A single strong peak is visible with someringing side peaks at a 5.6-ps period (corresponding to the180-GHz AWG channel spacing) due to the nonflat-top spectralresponse of the cascaded AWG filters. The encoder output main-tained the short input pulse quite well with a pulse full-widthat half-maximum (FWHM) of 900 fs compared to 500 fs forthe original laser pulse. The dotted curve in Fig. 11(a) clearlyshows pulse spreading for Walsh W8 [10 010 110] coding with14-V reverse biasing for the “1”s and 0 V for “0”s respectively,where ideally the “1” and “0” in the Walsh code should corre-spond to a and 0 rad phase shift, respectively. Fig. 11(b) showsthe corresponding simulation results of the pulse shapes forthe unencoded pulse (solid line) and a W8 coded pulse (dottedline). The simulation results took into consideration the ini-

Fig. 11. ( a) Measured cross-correlation and (b) simulated pulse shape resultsfor code W8 [10 010 110]. (c) Measured cross-correlation and (d) simulatedpulse shape results for code W4 [10 011 001] (solid curves represent unencodedpulses and dotted curves represent encoded pulses).

tial phase coding offset, and also included the effects of non-flat-top AWG channel spectral response generating the ringingside peaks, which can be clearly observed in the unencodedpulse.

Fig. 11(c) and (d) shows that measurement and simulationresults are well matched for the Walsh code W4 [10 011 001].The slightly asymmetric profile of the experimental W8 and W4curves is possibly due to small uncompensated residual phaseerrors in the encoder chip and power transmission nonuniformi-ties in the eight spectral channels.

Overall, the simulation and measurement results are in rela-tively good agreement. There are still some deviations betweenthe experiment and simulation due to residual phase errorswhich can be further improved upon through incremental re-finements to the process flow, but the significant time spreadingof the encoded pulse and good contrast between the encodedand decoded pulses in our latest data shows that our design isquite promising for O-CDMA applications.

AWG-Phase Modulator-AWG-Based Encoder/DecoderDesign for O-CDMA: Our eight-channel AWG-phase mod-


ulator-AWG encoder results demonstrate reasonable timespreading effect, which is the key point to SPECTS O-CDMA.However, the current eight-channel encoder/decoder may sufferfrom the MUI due to limited spectral resolution. The MUIeffects in O-CDMA will be much reduced by adopting AWGdesigns with higher channel counts. Increasing the numberof AWG channels for the given FSR can further improve thetime spreading effect and provide larger code space. In themean time, the reduced AWG channel spacing will cause theringing peak to appear further away. In the case of a 32-channelAWG with 50-GHz channel spacing, the side peak caused byAWG spectral modulation will be about 20 ps away from thecenter peak and can be sufficiently rejected by the nonlinearthresholding device (MZI). Currently, we have finished thedesign of 16- and 32-channel AWG-phase modulator-AWGencoder and are working toward the successful fabrication ofsuch devices. We believe AWG based encoder/decoder canindeed meet the requirement of the OCDMA application basedon our experimental and simulation results.

B. Colliding Pulse Mode Locked Laser

One of the key components for the integrated O-CDMA trans-mitter chip is the semiconductor-based mode locked laser forgenerating ultrashort optical pulses. The colliding pulsed modelocked laser (CPM) is suitable for this application. In the middleof the CPM laser cavity is a saturable absorber (SA) with a re-verse-biased MQW. Under appropriate biasing conditions, twocounter-propagating short pulses collide directly at the saturableabsorber, which allows deeper and more stable absorber satura-tion and elimination of self-pulsation effects commonly seen inregular mode locked lasers [12]–[14], [16].

Major issues with passively mode locked lasers are the largetiming jitters and difficulties in synchronization with externalclock sources. External hybrid mode locking through electricalmodulation of the saturable absorber region [12]–[15] or op-tical injection locking using an external master mode lockedlaser [16], [17] are two possible approaches explored for re-solving these issues. The electrical hybrid mode locking is thesimpler approach without requiring sophisticated optical setups.Subharmonic optical injection or electrical modulation can alsoachieve modelocking [15]–[17] while reducing the bandwidthrequirements on the external modulation source.

Fig. 12 (inset) shows the fabricated CPM laser with the gainand saturable absorber sections clearly identified. The metalcontact pads for the gain and the EA sections sit on top of a2- m–thick BCB polymer layer, which was spun, cured, andpatterned after the laser ridge formation. The BCB layer pro-vides planarization and reduces the contact pad capacitance as-sociated with the SA section. The contact pad capacitance for a100- m SA is estimated to be 0.1 pF, allowing efficient elec-trical modulations without parasitic roll-off effects at higher fre-quencies.

Fig. 12 shows the experimental setup for characterizing theCPM lasers stabilized with a thermoelectric cooler. A bias-Teeand a ground-signal-ground (GSG) microwave coplanar probeapplied the DC reverse bias voltage and microwave modulationsignal to the SA section. The laser output couples into a lensedfiber, passes through an optical isolator, and split into a various

Fig. 12. Experimental setup for characterizing CPM laser, and (inset) SEMpicture of fabricated CPM laser showing gain and saturable absorber sections.

number of instruments. They included the OSA for optical spec-trum measurements, the microwave spectrum analyzer with a40-GHz p-i-n detector for electrical power spectrum measure-ments, and the 50-GHz digital sampling scope and the auto-cor-relator for time domain measurements.

We characterized in detail a 3000- m-long CPM laser with a3- m-wide ridge and a 100- m-wide SA section. The InGaAscap layer in the 15- m-wide gap between the gain and SA sec-tions was selectively removed after processing by wet etching,resulting in electrical isolation of approximately 5 k betweenthe two sections. The laser threshold was 115 mA with boththe gain and the SA sections forward biased. The device oper-ated in the passively mode locking regime when the gain sectionwas forward biased at 219 mA and the SA section was reversebiased at 2.28 V. The 40-GHz microwave spectrum analyzerconnected to 40-GHz p-i-n detectors showed a single intensitypeak at 28.1 GHz, the expected CPM frequency. The peak had a3-dB bandwidth of 900 kHz, corresponding to the considerabletiming jitter associated with the passive mode locking condition.Fig. 13(a) shows the microwave spectrum with a 23-dBm mi-crowave signal at 14.1 GHz applied to the SA section. It showsa strong intensity peak at 28.1 GHz, the CPM frequency, whilea very weak peak 29 dB below appears at 14.1 GHz, the sub-harmonic modulation frequency. This indicates that the ampli-tude modulation (AM) distortion is quite low [15]. Fig. 13(b)shows the zoom-in view at the CPM frequency with 10-kHzresolution bandwidth (RBW). The 3-dB bandwidth was only 20kHz, indicating a sharp reduction in the timing jitter comparedto the passive mode locking case. Fig. 14(a) shows the time do-main output on the 50-GHz sampling scope. The sampling scopetriggered by the external microwave source clearly displayeda stable pulse train at the CPM frequency of 28.1 GHz, con-firming that second–order electrical subharmonic hybrid modelocking (SHML) was indeed taking place synchronized to themicrowave source. Due to the 50-GHz sampling scope band-width limitation, only the first two harmonics of the 28.1-GHzpulse train were present, resulting in the apparent pulsewidth ofapproximately 13 ps and the characteristically small secondarypulse peaks between the main peaks.


Fig. 13. (a) Electrical power spectrum of a 3000-�m CPM laser undersecond-order electrical SHML (RBW = 1 MHz). (b) Zoomed in view at theCPM frequency (RBW = 10 kHz).

Fig. 14. (a) A 50-GHz sampling scope trace of a 3000-�m CPM laseroutput under SHML showing synchronization to electrical clock source (b)autocorrelation trace under SHML.

The true pulsewidths were measured using an auto-correlatorwith the second harmonic generation (SHG) signal generatedby a 0.5-mm–thick LiNO3 crystal. Assuming a Gaussian pulseshape, the autocorrelation trace shown in Fig. 14(b) translatesinto a pulse width of 1.54 ps. The optical spectrum of the pulsehas a FWHM of 3.3 nm at wavelength of 1590 nm. The timebandwidth product is 0.604 in this case, indicating that the modelocked pulses were slightly beyond transform-limited. The ma-terial wavelength was slightly longer than the targeted 1550 nm,primarily due to over-compensation of the design for intendedregrowths. The calculated pulse transit time through the 100- mSA region along with the two 15 m gaps between the gain andthe SA sections was 1.52 ps, implying that the pulsewidth wasmost likely limited by the SA dimension [12]. To obtain shortermode locked pulses with correspondingly wider spectral widthfor O-CDMA applications, a new CPM laser fabrication is inprogress with a reduced SA section dimension.

For 10-Gb/s O-CDMA applications, the desired CPM lasercavity length is 8600 m, and passive waveguide sections can beadded to both ends of the active section of the CPM laser usingthe active-passive integration technique discussed in Section III.The introduction of the passive waveguide sections reduces the

Fig. 15. Fabricated 8600-�m CPM laser with integrated active-passivesections.

Fig. 16. (a) Schematic diagram of a MZI operating as threshold detector. (b)Operation principles of the MZI threshold detector based on the differentialoperation.

drive current requirement compared to the all-active CPM de-sign for the relatively long 8600- m laser application. Fig. 15shows a fabricated 8600- m-long CPM laser with integrated ac-tive-passive waveguides, consisting of a 1000- m-long activesection centered at the saturable absorber and two sections of3800- m-long passive sections on either side. The typical con-tinuous wave (CW) for such device is approximately 150mA, compared with 70 mA for a 1000- m-long laser withfull active waveguide. As measured from on-chip test structures,the passive waveguide loss is approximately 5 dB/cm, which re-sulted in the somewhat higher for the 8600- m laser. Nev-ertheless, the usage of passive sections implies much lower cur-rent requirement than if the entire 8600- m waveguide needsto be pumped. This result demonstrated that uniform and highquality 8600- m-long laser cavities can indeed be successfullyfabricated with our current process. Characterization of the ac-tive-passive CPM laser structures is currently underway.

Ultimately the realization of an integrated O-CDMA trans-mitter chip including a CPM lasers and the AWG-phase modu-lator array-AWG based encoder requires on-chip reflectors suchas broadband distributed Bragg or photonic crystal structures[18]. Electron-beam lithography based photonic crystal fabri-cation development is in progress.

C. MZI-Based Threshold Detector

An important component for the integrated O-CDMA en-coder/decoder chip is the MZI-based ultrashort pulse thresholddetector (MZI-TD). This subsection discusses ultrashort pulsedetection principles demonstrated using a commercial all-ac-tive MZI and also presents characteristics of MZI’s fabricatedwith our active-passive integration technology. In the SPECTSO-CDMA system, correctly and incorrectly (time spread)decoded pulses have the same integrated total energy. Althougha conventional square-law photoreceiver cannot distinguish be-tween them, the MZI detection allows nonlinear and time-gateddetection which selectively distinguishes the properly decodedpulse. Previous detection schemes proposed include either peak


Fig. 17. Experimental setup for characterizing MZI based threshold detection.

intensity discriminators (threshold detection) or time gatingdevices. The peak threshold detectors utilize highly nonlinearfiber (HNLF) [19] or second harmonic generation in a period-ically poled LiNO crystal (PPLN) [4]. Time gating devicesinclude nonlinear optical loop mirrors (NOLM) [20], terahertzasymmetric optical demultiplexers (TOAD) [21], and ultrafastnonlinear interferometers (UNI) [22]. The advantage of thetime gating schemes is that the locally generated control pulsestrigger nonlinearity required for detection, which reduces thepower requirement on the signal pulses. However, it requiresstrict synchronizations between the transmitted pulse and thecontrol pulse. The peak threshold detection schemes, on theother hand, require no synchronization, but require higher peakpower in the signal pulses, which eventually limit the systemscalability. In this paper, we use the semiconductor opticalamplifier (SOA)-based MZI fabricated in InP/InGaAsP as anultrashort pulse MZI-TD. This device requires relatively lowinput power (theoretically estimated 40-fJ pulse energy) dueto the high sensitivity of the phase response inside the SOAwhile simultaneously providing thresholding and time-gateddetection.

UltraShort Pulse Threshold Detection Principle: Fig. 16shows the MZI schematic and its operation principle as athreshold detector based on the differential operation [23]. Thedevice is very compact 800 200 m , with low switchingpower, and monolithically integratable with other on-chipcomponents. Each arm of the MZI was equipped with anintegrated semiconductor amplifier. Incoming ultrafast opticalpulses, which will be split into the two arms, will saturatethe gain and alter the phase of two arms at two offset times.While the gain saturation will take place rapidly due to theenhanced stimulated emission when the signal is present, thegain recovery is slow due to the relatively long carrier lifetime(100–400 ps depending on the current injection level) when thesignal power is low. The MZI exploits the offset in the fast risetime while canceling out the slow gain recovery component tocreate a narrow pass window for the sharp decoded signal pulse.At the same time, the low peak power of a broad undecodedpulse implies that the MZI reduces its transmission, creating

the thresholding effect. The probe beam will be CW and will befixed at a TE polarization. The wavelength of the probe beamwill be set at a short end of the SOA transparency window(e.g., 1520 nm) to exploit the superior performance of theMach–Zehnder device in the “up-conversion” mode comparedto the “down-conversion” mode, and to allow spectral filteringof the output probe wavelength from the background ultrafastsignal spectrum (1535–1565 nm).

MZI-based ultrashort pulse thresholder demonstra-tion: Fig. 17 shows the experimental setup for demonstratingthreshold detection employing a commercial four port MZI.A synchronized LiNbO Mach-Zehnder modulator provided2 PRBS data modulation to the ultrashort optical pulsetrain generated by a commercial mode locked fiber laser,with FWHM 400 fs at center wavelength of 1550 nm andrepetition rate of 9 GHz. The SPECTS O-CDMA test-bedindicated in Fig. 17 encodes and decodes the modulated opticalpulses using an optical pulse shaper, which is then amplifiedby a dispersion compensated EDFA. A 3-dB coupler splits theamplified signal pulse and variable attenuators enabled separatepower control of both signal paths. A variable optical delay lineintroduces a relative delay between the two MZI input signalsbefore the pulse trains entered the two input signal ports of theall-active Mach-Zehnder interferometer. CW output generatedby a distribute-feedback laser diode (DFB-LD) coupled intothe center probe port of the MZI provides the probe signal. A1-nm bandwidth optical tunable filter blocks the residue inputsignal of at 1550 nm coming out of the MZI and transmitsthe 1560-nm probe signal to a 10-GHz optical receiver and abit-error rate (BER) tester.

Fig. 18 shows the measured BER curves for the back-to-back laser signal, the correctly and incorrectly decoded SPECTSO-CDMA signals before the MZI thresholder, and the correctlyand incorrectly decoded signals after the MZI thresholder, alongwith the corresponding electrical eye diagrams. One can seethat responses of the slow optical receiver to the correctly andincorrectly decoded signals without the nonlinear thresholderare almost identical, since the receiver does not have sufficientbandwidth to respond to the OCDMA pulse spreading effect.


Fig. 18. BER results and eye diagrams for the different test cases including,correctly decoded signal without MZI thresholder, incorrectly decoded signalwithout MZI thresholder, correctly decoded signal with MZI thresholder, andincorrectly decoded signal with MZI thresholder.

With the MZI thresholder added, a clear open eye can be ob-served for the correctly decoded signal, by contrast the eye isalmost closed in the incorrectly decoded case. The differencebetween the correctly and the incorrectly decoded signal BERcurves is small without the MZI thresholder, which implies vir-tually no discrimination between the correctly and incorrectlydecoded signals. With the MZI thresholder, the correctly de-coded signal demonstrates a BER lower than 1E-9 while the in-correctly decoded signal saturates at an error floor above 1E-3.This demonstrates that the MZI operating in the differentialmode is indeed capable of thresholding detection in the SPECTSO-CDMA scheme. Due to the high sensitivity of the cross phasemodulation mechanism in the MZI device, the average requireddecoded pulse energy is only about 200 fJ in the current exper-imental setup.

Integrated active-passive MZI devices for thresholding op-eration: We have fabricated MZI elements based on the active-passive integration technology with waveguides formed by deepetching/lateral InP regrowth, as described in Section III. Fig. 19shows a photograph of the finished chip. The MZI consists ofall passive waveguide elements with the exception of the SOA-based phase modulators in each of the MZI arms, which havethe active section structure with MQW’s. The 3 dB couplers inthe MZI employed 2 2 multimode interference (MMI) cou-plers, which have the advantage of providing both the invertingand noninverting probe outputs simultaneously. The SOAs inthe MZI arms were 500 m long and 2.0 m wide. The inter-connecting waveguides were also 2.0 m wide. Power splittingerrors in the 2 2 MMI were approximately 0.5 dB and excesslosses were less than 1 dB.

We performed wavelength conversion experiments with thefabricated active-passive MZIs to quantify their performance.The experimental setup is shown in Fig. 19. Tunable lasersources (TLS), each followed by a polarization controller (PC),generated the probe and signal. An optical spectrum analyzer(OSA) characterized the chip output.Fig. 20(a) shows theMZI electrical switching curve with variation in SOA currentinjection with the probe laser at 1555 nm either switched

Fig. 19. Fabricated MZI with MMI-based 3-dB coupler and integratedactive-passive sections. Also shown is the MZI characterization setup foroptical switching experiments.

Fig. 20. (a) MZI electrical switching curves with signal input at 14.5 dBm andturned off. Arrows indicate possible biasing points for wavelength conversion.(b) Optical switching and wavelength conversion in the noninverting mode.

on or off. The injection current into SOA1 was kept fixed atI1=105 mA. The shift in the electrical switching curve minimaidentified the optimal biasing points for optical switching.Fig. 20(b) shows the corresponding optical switching curve inthe noninverting mode, with an extinction ratio of 15 dB. Thisdemonstrated that our in-house designed and fabricated MZIswith integrated active-passive sections and utilizing the HVPEbased regrowththeir technology are performing quite well asexpected. Experiments are currently underway to operate themin the differential threholder mode as shown in Fig. 16 usingtestbed setup shown in Fig. 17. Based on the good opticalswitching results, we expect these devices to perform at a levelcompatible to the commercial four port MZIs for thresholdingoperation, and satisfy OCDMA thresholder requirements aswell.

V. CONCLUSION

The O-CDMA networks open exciting new possibilitiesin high-performance local area networking including flexiblecapacity access, simplified network control and management,and physical level security. The deployment of the O-CDMAnetwork relies strongly on the integrated systems technology.This paper discussed the advanced integrated photonics tech-nologies toward realizing a monolithic InP O-CDMA systems.Dry etching and HVPE lateral regrowth in addition to standardwet etching and MOCVD growths achieved active and passivedevice integration with excellent planarity. The fabricationprocess applied to individual devices and modules resulted inhigh-performance CPM lasers, MZI wavelength converters,and the monolithic AWG-phase modulator-AWG O-CDMAencoder/decoder. In particular, O-CDMA encoding operationusing eight-chip Walsh code has been demonstrated for the


first time using a single integrated encoder chip, with rapidcoding reconfiguration and negligible power consumption. It isexpected that the O-CDMA photonic chip development effortwill demonstrate the feasibility of large scale, cost effectivedeployment of O-CDMA technology in LANs. The componentand integration technologies investigated are also highly rele-vant with respect to other integrated photonic chip applications,with the goal toward on-chip optical systems with higher levelof functionalities and dramatically reduced packaging cost.

ACKNOWLEDGMENT

The authors would like to thank Dr. K. Okamoto for the ini-tial design of the O-CDMA encoder and numerous enlighteningdiscussions.

REFERENCES

[1] J. A. Salehi, A. M. Weiner, and J. P. Heritage, “Coherent ultrashort lightpulse code-division multiple access communication systems,” J. Lightw.Technol., vol. 8, no. 3, pp. 478–491, Mar. 1990.

[2] H. P. Sardesai, C. C. Chang, and A. M. Weiner, “A femtosecondcode-division multiple-access communication system test bed,” J.Lightw. Technol., vol. 16, no. 11, pp. 1953–1964, 1998.

[3] H. Tsuda, H. Takenouchi, T. Ishii, K. Okamoto, T. Goh, K. Sato, A.Hirano, T. Kurokawa, and C. Amano, “Spectral encoding and decodingof 10 Gb/s femtosecond pulses using high resolution arrayed-waveguidegrating,” IEE. Electron. Lett., vol. 35, no. 14, pp. 1186–1188, 1999.

[4] Z. Jiang, D. Seo, S.-D. Yang, D. E. Leaird, A. M. Weinder, R. V.Roussev, C. Langrock, and M. M. Fejer, “Four user, 2.5 Gb/s spectrallycoded O-CDMA system demonstration using low power nonlinearprocessing,” presented at the Proc. Optical Fiber Communication Conf.,2004, Postdeadline Pap. PDP29.

[5] J. P. Heritage, A. M. Weiner, and R. N. Thurston, “Picosecond pulseshaping by spectral phase and amplitude manipulation,” Opt. Lett., vol.10, no. 12, pp. 609–611, 1985.

[6] R. P. Scott, W. Cong, K. Li, V. J. Hernandez, B. H. Kolner, J. P. Heritage,and S. J. B. Yoo, “Demonstration of an error-free 4� 10-Gb/s multi-userSPECTS O-CDMA network testbed,” IEEE Photonics Technol. Lett.,vol. 16, no. 9, pp. 2186–2188, Sep. 2004.

[7] Z. Jiang, D. S. Soe, S.-D. Yang, D. E. Leaird, R. V. Roussev, C. Lan-grock, M. M. Fejer, and A. M. Weiner, “Low-power high-contrast codedwaveform discrimination at 10 GHz via nonlinear processing,” IEEEPhoton. Technol. Lett., vol. 16, no. 7, pp. 1778–1780, Jul. 2004.

[8] C. Rolland, R. S. Moore, F. Shepherd, and G. Hillier, “10 Gbit/s, 1.56�m multiquantum well InP/InGaAsP Mach Zehnder optical modulator,”IEE. Electron. Lett., vol. 29, no. 5, pp. 471–472, 1993.

[9] S. Lourdudoss and O. Kjebon, “Hydride vapor phase epitaxy revisited,”IEEE J. Select. Topics Quantum Electron., vol. 3, no. 3, pp. 749–767,Jun. 1997.

[10] H. Tsuda, H. Takenouchi, T. Ishii, K. Okamoto, T. Goh, K. Sato, A. Hi-rano, T. Kurokawa, and C. Amano, “Spectral encoding and decoding of10 Gbit/s femtosecond pulses using high resolution arrayed-waveguidegrating,” IEE Electron. Lett., vol. 35, no. 14, pp. 1186–1188, 1999.

[11] J. T. Ahn, S. Park, J. Y. Do, J.-M. Lee, M.-H. Lee, and K. H. Kim,“Polymer wavelength channel selector composed of electroopticpolymer switch array and two polymer arrayed waveguide gratings,”IEEE Photon. Technol. Lett., vol. 16, no. 6, pp. 1567–1569, Jun. 2004.

[12] Y. K. Chen and M. C. Wu, “Monolithic colliding pulse mode-lockedquantum well laser,” IEEE J. Quantum Electron., vol. 28, no. 10, pp.2176–2185, Oct. 1992.

[13] H. Fan, C. Wu, M. El-Aasser, N. K. Dutta, U. Koren, and A. B. Piccirillli,“Colliding pulse mode locked laser,” IEEE Photon. Technol. Lett., vol.12, no. 8, pp. 972–973, Aug. 2000.

[14] H. K. Lee, V. Loyo-Maldonado, B. C. Qiu, K. L. Lee, C. Chu, S. Pinches,I. G. Thayne, A. C. Bryce, and J. H. Marsh, “Efficient direct locking ofcolliding pulse mode-locked lasers on semi-insulating substrate at 1.5�m,” IEEE Photon. Technol. Lett., vol. 14, no. 8, pp. 1049–1051, Aug.2002.

[15] T. Hoshida, H. F. Liu, M. Tsuchiya, Y. Ogawa, and T. Kamiya, “Subhar-monic hybrid mode-locking of a monolithic semiconductor laser,” IEEEJ. Select. Topics Quantum Electron., vol. 2, no. 5, pp. 514–522, Sep.1996.

[16] S. Arahira and Y. Ogawa, “480-GHz subharmonic synchronous modelocking in a short-cavity colliding-pulse mode-locked laser diode,” IEEEPhoton. Technol. Lett., vol. 14, pp. 537–539, April 2002.

[17] H. Kurita, T. Shimuzu, and H. Yokoyama, “Experimental investi-gations of harmonic synchronization conditions and mechanisms ofmode-locked laser diodes induced by optical-pulse injection,” IEEEJ. Select. Topics Quantum Electron., vol. 2, no. 3, pp. 508–513, Sept.1996.

[18] T. D. Happ, A. Markard, M. Kamp, M. Forchel, S. Anand, J.-L. Gentner,and N. Bouadma, “Nanofabrication of two-dimensional photonic crystalmirrors for 1.5 um short cavity lasers,” J. Vac. Sci. Technol. B, vol. 19,no. 6, pp. 2775–2778, 2001.

[19] C.-C. Chang, H. P. Sardesai, and A. M. Weiner, “Code-division mul-tiple access encoding and decoding of femtosecond optical pulses overa 2.5-km fiber link,” IEEE Photon. Techno. Lett., vol. 10, no. 1, pp.171–173, Jan. 1998.

[20] J. H. Lee, P. C. Teh, P. Petropoulos, M. Ibsen, and D. J. Richardson,“Reduction of interchannel interference noise in a two-channel grating-based O-CDMA system using a nonlinear optical loop mirror,” IEEEPhoton. Technol. Lett., vol. 13, no. 5, pp. 529–531, May 2001.

[21] I. Glesk, V. Baby, C.-S. Bres, L. Xu, D. Rand, and P. R. Prucnal,“Experimental demonstration of a 2.5 Gbps incoherent 2D O-CDMAsystem,” presented at the Conf. Lasers Electro-Optics (CLEO), 2004,Pap. CWH5.

[22] V. J. Hernandez, R. P. Scott, W. Cong, K. Li, V. B. H. Kolner, J. P.Heritage, and S. J. B. Yoo, “6 user, 10 Gb/s time slotted SPECTSO-CDMA system incorporating a UNI based demultiplexer,” IEEEPhoton. Technol. Lett., submitted for publication.

[23] S. Bischoff, A. Buxens, S. T. Fischer, M. Dulk, A. T. Clausen, H. N.Poulsen, and J. Mork, “Comparison of all optical co- and counter-prop-agating high-speed signal processing in SOA-based Mach-Zehnder in-terferometers,” Opt. Quantum Electron., vol. 33, pp. 907–926, 2001.

[24] Y. Du, S. J. B. Yoo, and Z. Ding, “Non-uniform spectral encoding for en-hanced multi-user performance in optical CDMA networks,” presentedat the Conf. Lasers Electro-Optics, 2004, paper CWH2.

[25] K. I. Kang, T. G. Chang, I. Glesk, and P. R. Prucnal, “Comparison ofSagnac and Mach-Zehnder ultrafast all-optical interferometric switchesbased on a semiconductor resonant optical nonlinearity,” Appl. Opt., vol.35, pp. 417–426, 1996.

Chen Ji received the B.S. degree in physics from University of Illinois, Urbana-Champaign, in 1993, and the Ph.D. degree in electrical engineering from CornellUniversity, Ithaca, NY, in 1999.

He is currently a Postdoctoral Research Scientist with the Department of Elec-trical and Computer Engineering, University of California, Davis. His researchinterests include semiconductor lasers, optical integrated devices, and systemintegration for next-generation optical network.

R. G. Broeke, photograph and biography not available at the time of publication.

Y. Du received the M.S. degree in physics from Oklahoma State University,Stillwater, in 1998. She is currently working toward the Ph.D. degree in elec-trical engineering at the University of California, Davis.

Her research interests include device simulation and testing in optical code-division multiple-access networks.

Jing Cao (S’01) received the B.S. and M.S degrees from the Department ofElectronics Engineering, Tsinghua University, Beijing, China, in 1997 and2000, respectively. He is currently working toward the Ph.D degree at theDepartment of Electrical and Computer Engineering, University of California,Davis.

His research interests include optical integrated devices and system integra-tion for next generation optical network.

Mr. Cao is the student member the Optical Socity of America.


N. Chubun, photograph and biography not available at the time of publication.

P. Bjeletich, photograph and biography not available at the time of publication.

F. Olsson, photograph and biography not available at the time of publication.

S. Lourdudoss (M’97–SM’99) received the M.Sc. degree in chemistry fromMadras University, Madras, India, in 1976 and the Ph.D. degree in chemistryfrom Faculté Libre des Sciences de Lille, Lille, France, in 1979.

As a Postdoctoral Fellow at KTH, Stockholm, Sweden, he worked on thethermochemical aspects of thermal energy storage and absorption heat pumpsuntil 1985. Afterwards, he started to work on the epitaxial growth and charac-terization of III-V semiconductors for device fabrication at Swedish Institute ofMicroelectronics, Kista, Sweden. Along with his colleagues, he developed an at-tractive semi-insulating regrowth technology by HVPE for integration and highspeed device fabrication. Currently, he is a Professor of semiconductor materialsat KTH and is involved in undergraduate and graduate education. His researchinterests include MOVPE of GaN and related materials and the issues related tothe integration of III-V materials on Si by HVPE.

R. Welty, photograph and biography not available at the time of publication..

C. Reinhardt, photograph and biography not available at the time of publica-tion.

P. L. Stephan, photograph and biography not available at the time of publica-tion.

S. J. B. Yoo (S’91–M’92–SM’97) received the B.S. degree (with distinction) inelectrical engineering, the M.S. degree in electrical engineering, and the Ph.D.degree in electrical engineering with a minor in physics, from Stanford Univer-sity, Stanford, CA. His Ph.D. thesis was on linear and nonlinear optical spec-troscopy of quantum well intersubband transitions.

He is currently serving as Professor and the University of California (UC),Davis Branch Director of the Center for Information Technology Research inthe Interest of Society (CITRIS). Prior to joining UC Davis, he was a SeniorScientist at Bellcore leading technical efforts in optical networking researchand systems integration. His research activities at Bellcore included optical-label switching for the Next Generation Internet, power transients in reconfig-urable optical networks, wavelength interchanging cross-connects, wavelengthconverters, vertical cavity lasers, and high-speed modulators. He also partici-pated in the Advanced Technology Demonstration Network/ MultiwavelengthOptical Networking (ATD/MONET) systems integration, the OC-192 SONETRing studies, and a number of standardization activities. Prior to joining Bell-core in 1991, he conducted research on nonlinear optical processes in quantumwells, four-wave mixing study of relaxation mechanisms in dye molecules, andultra-fast diffusion driven photodetectors. During this period, he also conductedresearch on life-time measurements of intersubband transitions and on nonlinearoptical storage mechanisms at Bell Laboratories and at IBM research Laborato-ries, respectively. His current research involves advanced photonic technologiesand optical communications systems. In particular, he is conducting research onarchitectures, systems integration, and network experiments of all-optical labelswitching and optical code-division-multiple access networks.

Prof. Yoo received the Bellcore CEO Award in 1998 and the DARPA awardfor sustained excellence in 1997. He is a senior member of IEEE Lasers andOpto-Electronics Socity, and a member of Optical Society of America and TauBeta Pi.

Monolithically integrated InP-based photonic chip development for O-CDMA systems

Documents