UNIVERSITA’ DEGLI STUDI DI PAVIA

DIPARTIMENTO DI ELETTRONICA

DOTTORATO DI RICERCA IN INGEGNERIA ELETTRONICA, ELETTRICA

ED INFORMATICA – XXII CICLO

RING‐CAVITY AND QUANTUM‐DOT

SEMICONDUCTOR LASERS:

ALL‐OPTICAL SIGNAL PROCESSING

AND PARAMETERS EXTRACTION

Tutore: Ing. Guido Giuliani

Tesi di Dottorato di Maria Jose Latorre Vidal

Anno Accademico 2008/2009

He aprendido que la inteligencia es saber adaptarse a todos los medios sin dificultad

y alcanzar las metas propuestas...la sabiduria viene con el tiempo, pasando obstculos y aprendiendo de las caidas y tropiezos...No es posible medir cuan

inteligente eres con un simple IQ factor o con una simple materia, todos tenemos diversas zonas desarrolladas en el cerebro y lo que para ti es difícil para otro

puede ser fácil.

El tesoro mas precioso es lo que hemos alcanzado y aprenido, es algo que se queda siempre con nosotros y nadie podra quitarlelo... hay que vivir feliz con lo que se tiene, intentando aprender cada dia, hay que vivir la vida intensamente

porque no sabes nunca si al otro dia te despertaras...

Maria

TABLE OF CONTENTS

1. SUMMARY 1.1 Executive Summary 5 1.2 Acronyms 6 1.3 Acknowledgements 7 2. INTRODUCTION 9

Part 1. SEMICONDUCTOR RING LASERS

3. THE STATE OF ART OF SEMICONDUCTOR RING LASERS 11

4. FUNCTIONALITY OF THE SEMICONDUCTOR RING LASERS 13 4.1 Fundamental Concepts 14

4.1.1 Propagating Modes 14 4.2 Optical Bistability in Semiconductor Ring Lasers 15 4.3 IOLOS Project 17 4.4 All-Optical Functions with SRL 18 4.5 Device Design and Fabrication 20

5. STATIC EXPERIMENTS BASED ON SEMICONDUCTOR RING LASERS 23

5.1 Introduction 24 5.2 Experimental Measurements of Static Characterisation 25 5.2.1 Burn-in Procedure 25 5.2.2 Photo-Diode Characterisation 26 5.2.3 Optical Lensed Fibre Characterisation 27 5.3 Results of the Static Characterisation of the SRL 29

5.3.1 Light-Current Curves 29 5.3.2 Bistability Properties 32 5.3.3 Spectral Properties 33 5.3.4 Effects of Optical Feedback from the end-Facets 37 5.4 Experimental Measurements of the SRL Lindwidth 39 5.4.1 Theoretical Linewidth 39 5.4.2 Linewidth Measurements. 40

5.4.3 Results of the Linewidth Characterisation 42

5.5 Conclusions about the Static Behaviour of the SRL 45 6. ALL-OPTICAL DYNAMIC EXPERIMENTS BASED ON SRL 47

6.1 Introduction 47 6.2 All-Optical Set-Reset Flip-Flop based on SRL 48

6.2.1 Slow-Pulse Optical Trigger Injection 49 6.2.2 Slow-Pulse by electrical/Optical Trigger Pulse Pattern Generator 54 6.2.3 Results of the All-Optical Flip-Flop Measurements by PPG and Slow Pulse 55 6.2.4 Fast and Ultra-fast Optical Trigger Injection Pulse 42 6.2.5 Experimental Setup to Generate the Ultra-Fast Injection Pulse 45 6.2.6 Results of the All-optical Flip-Flop Operation by PPG and Ultra Fast Pulse 46

6.4 Conclusions of Dynamic Behaviour of the SRL

Part 2. SEMICONDUCTOR QUANTUM DOT LASERS 7. THE STATE OF ART OF THE QUANTUM DOT LASERS 70 8. THEORETICAL BEHAVIOUR OF THE QUANTUM DOT LASER 71

8.1 Introduction 71 8.2 Properties of the Semiconductor Quantum Dots 72 8.2.1 The Quantum Dot in a Laser 72 8.3 Structure of the Quantum Dot Laser 73

8.3.1 Self-Assembly of Quantum Dots 74 8.3.2 Design and Fabrication of the used Quantum Dot Lasers 75

8.4 Static Characterisation of the Quantum Dot Lasers 77 8.4.1 Experimental Set-ups of the Static Characterisation 78

8.5 Results of Static Characterisation of the QDLs 79

8.5.1 Light – Current Curves 79 8.5.2 Spectral Measurements 80

8.6 Conclusions about the Static Characterisation of the QDLs. 81

9. DYNAMIC EXPERIMENTS BASED ON QUANTUM DOT LASERS 82 9.1 Introduction 82 9.2 Frequency Response on Quantum Dot Laser 83

9.2.1 Relaxation oscillations 84 9.2.2 Modulation Response

9.3 Frequency Response on Quantum Dot Laser 9.3.1 Measurement Set-up of Frequency Response by an all Optical Modulation Technique

9.3.2 Results of the Frequency Response on QDLs. 9.4 Linewidth Enahancement Factor Measurements 9.4.1 Fiber Transfer Function Method. 9.4.2 Set-up of Measurements of the α-Factor 9.4.3 Linewidth Enhancement Factor Results.

10. CONCLUSIONS A. ANNEXE A B. ANNEXE B Scientific Production Bibliography

SUMMARY

1. Summary

1.1 Executive Summary The next generation of Optical Networks must to be able to carry more bitrate, generating more capacities and useful opportunities to increase the quality of transmission service to the users. To manage in an efficient way such capacity, optical digital functions need to be applied by implementing all-optical devices. In this context, this thesis focuses on optical devices which enable to operate in all-optical domain in digital processing signal in transmissions systems. This work has two objectives: i) the assessment of all-optical digital functions on monolithic devices in static and dynamic conditions, to explore the possibility to enhanced the traffic signal processing into the optical networks based on Semiconductor Ring Lasers (SRL); and. ii) To analyse the behaviour of Quantum Dot Lasers (QDL) as high speed sources in optical telecommunications. The first part of the work is carried out to explore the bistability of the Semiconductor Ring Lasers (SRL) devices and by examining their optical digital functionalities by measuring the bit error rate under experimental system operation conditions. The second part contains the analysis of the static behaviour of the QDLs and the measurements of their frequency response. This document is organised as follows: The first part presents an introduction and describes the state of the art of the SRL; the fourth chapter describes the functionality, design and fabrication and fundamental concepts of the SRL as an opportunity to generate digital functions to be applied to optical telecommunications systems. The fifth chapter describes experimental results of static characterisation exploring the behaviour of the device in continuous wave. The sixth chapter describes experimental results of the dynamic characterisation of SRLs. The second part presents an innovative laser based on quantum dot structure that represents the first step to nano-devices in optical transmission sources; semiconductor quantum dots are a promising way for the realization of high performance lasers, therefore this part shows experimental characterisations to describe the expected QDL behaviour. The lasts chapters are focused on the measurements of QDL behaviour under static and dynamic conditions. In generally to evaluate the behaviour of the devices, at first the static characteristics of the SRL were investigated, including a detailed analysis of the stability of unidirectional and single longitudinal mode operation under static conditions. The results were extremely good, showing high robustness to optical reflections. Results on dynamic conditions such as frequency response and linewidth enhancement factor have been demonstrated but not including in this work, the measures achieving bandwidths in excess of 13 GHz and alpha factor as low as 1.4. The possibility of an All-optical Set-Reset Flip-Flop (SRFF) operation was investigated in detail, using three different types of optical trigger signals, that helped in assessing the device performance, and gaining further insight into the physics of the devices. The SRFF operation was easily and reliably demonstrated for each type of optical trigger and a good BER of 3.10-12 has been achieved by Pseudo Random Bit Sequence (SRBS) of 2-31-1 bits. The most interesting results cam from the use of trigger pulses of 5 ps duration, revealing high-speed operation of the SRL. We reported smaller rise and full time. The SQLs are good promising devices exhibiting high temperature independence, low threshold current and a bandwidth around 3Ghz.

1.2 Acronyms

Acronym Significance

1R Re-shaping 2R Re-shaping + Re-timing 3R Re-shaping + Re-timing + Re-amplification AM Amplitude Modulation

AOFF All-Optical Flip-Flop DSA Digital Signal Analyzer DCF Dispersion Compensating Fiber DER Directional Extinction Ratio DFB Distributed FeedBack CW Clock Wise

CCW Counter-Clock Wise ECL External Cavity Laser ER Extinction Ratio

ESA Electrical Spectrum Analyzer FM Frequency Modulation FTF Fiber Transfer Function method

FWHM Full Width Half Maximum GPIB General Purpose Interface Bus HB Holding Beam

MZM Mach Zehnder Modulator OPO Optical Parametric Oscillator OSA Optical Spectrum Analyzer P.C Polarization Controller PPG Pulse Pattern Generator PIC Photonic Integrated Circuit

PRBS Pseudo-Random Bit Stream QD Quantum Dot

QDL Quantum Dot Laser SRFF Set-Reset Flip-Flop SRL Semiconductor Ring Laser

SMSR Side Mode Suppression Ratio TFPF Tunable Fabry Perot Filter WDM Wavelength Division Multiplexing

1.3 Acknowledgements First I would like to express my gratitude to my professor at L’Ecole National d’Ingenieurs of Bretagne (France): Yan Boucher, who has given me the opportunity to know the subject of PhD based in optical devices for transmission systems. During this period of investigations I have been learning and producing Scientifics knowledge by the help of my appreciated tutor who accompanied me all along my stay in the Università degli Studi di Pavia into the electronics department: Guido Giuliani who introduced me into the world of semiconductor devices, assisted me all along my researches and participated actively at all stages of my work, pushed me to improve my English. His support in difficult moments was highly appreciated. I acknowledge in particular, Mauro Benedetti for his immense patience to teach me the techniques of Matlab tools which allowed me seeing problems from different points of view and gave me many new ideas, and Andrea Fanzio for his accurate manner to produce my specific tools and pieces of work. This thesis would not be as rich if it wasn't for my colleague, Marco Zanola who introduced me into the Italian slang and taught to me the secrets of the optical instrumentation. My friends in Pavia: Claudia Trentin from Argentina, Luca Garofolis and Davide Raimondo who were always giving to me a great support and true friendship. Their presence refreshed my spirit and left me with unforgettable memories. Finally I would like to thank my boyfriend Romek for his support, which gave me the strength to continue. Without them this thesis would not have reached its present level. Pendant mon séjour en Italie, j’ai eu toujours le contact et le support de tous mes amis français, ils ont été a mon cote même si on na pas eu beaucoup des opportunités de nous rencontrer, un jolie e-mail a été la manière plus mignon de me monter son amitié. Le mie attività extralavorative erano concentrate nella musica e nella palestra, dove ho fatto la conoscenza di Madalina Simona chi é diventata la mia cara amica, come sempre succede in ambiente internazionali (lei originaria di Bucarest): in mezzo le partiture de musica classica, tra due stranieri parlando l’italiano, è nata una splendida amicizia. Alla fine del mio ultimo anno di dottorato ho avuto il piacere di conoscere a Fabrizio Magni, anche lui ingegnere, ma nelle costruzioni mi ha dato il supporto necessario per finire gli ultimi esperimenti, e tra multe ore di lavoro era sempre benvenuto una pausa per mangiare bene, almeno una sera su sette. Justo un mes antes de la sustentacion, mi querida hermanita ha llegado al continente europeo para visitarme y estar a mi lado, que linda, ella no tenia idea que hacia yo aqui, pero siempre se imaginaba qualquier invento cientifico.

INTRODUCTION

2.Introduction

Since 1962, the Semiconductor lasers have been playing an important roll into the industry. The telecommunication industry has been advancing very fast and the performance of the transmitters and receivers in the current wavelength communication systems are not enough able to support the modern exigencies. Most of the serious problems lasers for 1.3 and 1.5 μm communications as temperature performance, low the threshold current at room temperature, the necessity of thermal and so on have been resolved. In this work are reported two types of semiconductor lasers interesting for transmission systems and all-optical signal processing. Those devices provides low threshold currents and reduced dimensions for Photonic Integrated Circuits. The first one has a ring cavity and a bistable behavior working at 1.5 μm of wavelength, this special feature gives it an important issue for the development of future fiber-optic telecommunication systems in all- optical processing of digital light signals, it can open new ways to control and manipulate single bits of information, more efficiently and fast, without converting information from the electrical to the optical domain, or vice-versa. Here we report for the first time a monolithic photonic device capable of realizing all-optically the function of a Set-Reset Flip-Flop in a practical, reliable, and technologically reproducible manner. The second semiconductor laser reported, is a quantum dot laser working into the O-band of transmission exhibiting multiple modes at 1.3 μm of wavelengths with low temperature dependence and low threshold current.

Part 1

SEMICONDUCTOR RING LASERS

3. The State of the Art of Semiconductor Ring Lasers

Since the eighties the idea to produce integrated semiconductor optical devices for telecommunication system applications have been attractive into the scientific domain. Monolithic integration of semiconductor lasers is becoming very important for Photonic Integrated Circuits (PICs) where the development of lasers that do not use cleaved facet mirrors is highly recommended. Ring type lasers would be effective candidate not only for their monolithic integration capability but also for their low-threshold operation potential. Several types of semiconductor ring lasers, such as triangular, square and circular geometries types, have been demonstrated during the last decade. The circular ring cavities are attractive sources for fibre-to-the-home systems that require affordable components. A SRL provides an active area with coupling waveguides, where the performance of the device must to be accurate by controlling the coupling output efficiency. At the final of the nineties the construction of ring lasers of straight waveguides have represented an advantage for low cavity loss that would be a significant factor for realizing small size lasers [5]. Integrated SRL made in GaAs/GaAlAs heterostructure material and waveguide structures with output stripe waveguides coupled to the rings via Y-junction have been demonstrated in the last years, this type of structure can be use in monolithic integrated devices for optical applications [2]. One important parameter of SRLs is their unidirectional operation; this feature is desired because it offers the advantage of enhanced mode purity and higher single beam power, and also improves the mode stability and single-frequency performance reducing the sensitivity to feedback into the laser cavity [7]. Unidirectional operation has been reported at the end of the nineties, in square and circular ring lasers by providing preferential cross coupling for one circulating direction over the other [14]. Several kind of cavity geometries as illustrate in Fig. 3.1 have been fabricated during the last decade improving the appreciated unidirectional behaviour.

(a) (b) (c)

Figure 3.1. Evolution of the geometries of the SRL. (a) Triangle cavity SRL, [28] 1997. (b) Racetrack cavity, [24], 2000, (c) GaAs–AlGaAs circular SRL with single-mode ridge waveguide built to operate in bidirectional regime. Unidirectional operation is demonstrated in a square-shaped laser using an optical feedback mechanism for the first time at the early nineties, observing bistable switching phenomenon between the unidirectional and bidirectional operation synchronized with the switching of the longitudinal lasing mode. The fundaments of unidirectional behaviour of the SRL is illustrate in Fig.3.2.

The bistablility of the laser are expected to play an important role into the optical digital functions in transmission systems because the switching by optical manner has been a problem, therefore at lot of techniques have been studied to improve the operating networks performance, indeed experiments on an all-optical flip-flop operation have been realized by using the directionality of different kind of Semiconductor devices, the mean problem is to overcome the difficulty of “optical-reset” operation, to resolve this problem, laboratories and universities are studying the dynamic behaviour of the semiconductor ring lasers taking advantage of their directional switching and mode competition, also testing their digital functionalities for optical signal processing.

Figure 3.2 Propagating light into the SRL device in unidirectional operation The increased traffic on the networks open the way of the realisation of digital functions into an all-optical level employing optical gates, optical flip-flops, recently experiments on SRL have demonstrated the possibility to implementing all-optical photonic logical gates in telecommunication networks [37]. The future photonic networks on Wavelength Division Multiplexing (WDM) will require optical digital functions such as packet buffering, bit-length conversion, retiming, reshaping, time-division multi/demultiplexing, and wavelength conversion, therefore, a successful realization of optical networks requires compact devices to implement wavelength dependent digital functions. Having in mind this aim, the first step is to produce a compact cavity of the device, provides of a ring active area and coupled waveguide, this is typically achieved using shallow etching ridge geometry. Presently, a robust and interesting SRL device with small radius have been fabricated using this technique at the University of Glasgow and tested at the University of Pavia.

Interference Pattern Couterpropagating light Propagating light

4. Functionality of the Semiconductor Ring Lasers

The SRL is a monolithic photonic device capable of realizing all-optically the function of a Set-Reset Flip-Flop in a practical, reliable, and technologically reproducible manner. The device exhibits robust bistability among the two possible lasing direction within the cavity. The lasing direction can be reverted by injecting optical pulses of moderate energy into the device and the response times for switch-off and switch-on can be observed, while the device is continuously operated at room temperature. This device can offer the solution of operating speed into the fibre optic communications, optical signal processing and optical computing because can lead to the realization of complex all-optical systems as a analog-to-digital and digital-to-analog converters, sequential logic circuits, storage and processing bits o ultrafast all-optical storage. Those devices must to be tested under static conditions in order to verify their unidirectional behavior, during the experiments bistability properties and spectral characteristics based on Directional Extinction Ratio (DER) and Side Mode Suppression Ratio (SMSR) have been verified achieving good results for single mode operation. This chapter contains the static and dynamic results based on SRL devices as an All-optical Flip-Flop.

4.1 Fundamental Concepts

The ring laser is an active device, which generates the light and sustains it by incorporating laser excitation in the path of the light. The generated continuous wave propagate in two opposite directions along the perimeter of the ring cavity of the laser, i.e. Clock-Wise (CW) and Counter-Clock-Wise (CCW). This process is in Fig. 4.1a.

To understand the principle of operation of the SRL, it is helpful discuss the process of laser emission with continuous wave generation; this process can be analysed through the theoretical L-I curves (light Intensity vs. current intensity), as illustrated in figure 4.1b. When increasing gradually the current, is observed at the beginning the spontaneous emission zone, where basically the noise is amplified. Just beyond the threshold current value, two modes simultaneously propagate in the ring cavity in the bi-directional regime. Finally the laser enters into the uni-directional regime, where only one mode is propagating in one of the two directions CW or CCW, while the other is suppressed. Both modes in this region are in a constant competition between them to reach the lasing mode condition. This feature of SRL enables applying changes of the propagation direction within the ring and thus makes it working as a digital optical memory or an optical switch.

Counter Clock Wise Mode

Spontaneous Emission CCW

Clock Wise Mode

CW

(a) (b)

Figure 4.1. In (a) the propagation directions CW and CCW into the ring cavity. In (b) the theoretical L-I curve of propagating light of the SRL exhibiting the directional regimes of emission.

The bistability of the SRL is caused by the fact that the propagating modes in the laser cavity are in competition between them, until one of them acquires the gain necessary to reach the lasing condition. The uni-directional operation is favored by the fact tat the cross-gain saturation (a non-linear phenomenon within the active medium) is twice as large as the self-gain saturation [Balle art]. 4.1.1 Propagating Modes

The longitudinal mode spacing depends of the length of the cavity and is defined as:

cavitynLc

=Δν , (1.1)

where cavityL , corresponds to the cavity perimeter of the SRL, and n the refractive index of the semiconductor.

R

L

Cavity

Coupled Waveguide

Figure 4.2. Geometry of the SRL device showing the racetrack shape active area of radius R and two straight output waveguides of length L

The ring laser used have a racetrack-shaped cavity which is illustrated in Fig. 4.2; the length of the perimeter corresponds to LRLcavity 22 += π , where R is the radius of the ring and L the length of the coupler waveguide. The frequency space between consecutive transmission modes are defined as the Free Spectral Range (FSR). The values the Mode Spacing (MS) are reported in tables 1 and 2.

4.2 Optical Bistability in Semiconductor Ring Lasers

The optical bistability is an attribute of certain optical devices where two stable states are possible. In optical devices, the bistability can be achieved by different mechanisms. In absorptive bistability, an absorber is used to block the light allowing the two states appear, in this way if the light is absorbed the state is OFF, and the ON state is reached when the light saturates the absorber. Another method is the refractive bistability, that utilizes an optical mechanism that changes the refractive index allowing to get two states, the first state resides at a given intensity where no optical mechanism is used; the second state resides at the point where a certain light intensity causes resonance in a cavity, by changing the refractive index. This phenomenon has been applied in optical transmitters, memory elements and so on [].

Any reasonably practical all-optical memory must include an optical bistability characteristic within an integrated structure where light is confined by waveguides or other means. A broad division can be made between all-passive structures and structures including an active medium. In all-passive devices, bistability is achieved via the exploitation of optical non-linearity in various types of medium, including semiconductors. Despite the potential advantage of being an “electrically cold” device (i.e., one that does not require to be supplied by electrical currents/voltages), a passive bistable has a main drawback in that it requires one or more holding optical beams to exploit the non-linearity and generate the desired bistability. As a consequence, passive bistable cannot be regarded as stand-alone devices, because the bistability is not self-sustained, and one or more external laser sources are essential for its operation. Based on the above observation, the all-passive approach offers poor prospective for integration, and has low potential impact on the architecture of future all-optical logic networks/systems.

Active bistable devices are essentially based on semiconductor lasers, either in the form of a single laser, or two mutually / asymmetrically coupled laser oscillators, or two coupled laser modes of a single laser in appropriate regimes. The variety of optical bistability appearing in semiconductor lasers is extremely wide, and numerous bistable laser diodes (BLDs) have been proposed and demonstrated [5]. Research on application of BLDs to all-optical signal processing was very active in the later half of the 1980s and the beginning of the 1990s. A SRL bistable is also an attractive device for signal regeneration, as it is capable in principle of extinction ratios well in excess of 25 dB [21]. Moreover, directional switching in a bistable SRL can occur without variations of carrier density population, so the output pulse is expected to have low chirp, hence increasing propagation distance along an optical fibre. A complete new variety of applications recently emerged with the demonstration of directional bistable operation between the CW and CCW modes of the SRL [30]. The dynamics of SRL devices are defined by two distinct effects inside the cavity: the mode cross gain compression within the laser medium, and the direct mode coupling effects via both the laser medium and via non-ideal conditions in the optical waveguide (such as scattering) [23]. Both effects can favour directional bistability illustrate in Fig. 4.3a. (i.e., stable operation in the unidirectional regime with one CW or CCW is operating into the ring cavity), the exciting feature is that switching between the two modes can be optically induced where the switching speed has the potential to be extremely fast, because the two directions have equal status, SRLs do not rely on different mechanisms governing the switch-on and switch-off transitions. In Fig.4.3b is illustrate the mechanism of un-stability (i.e., operation in bidirectional regime) where both modes are present into the cavity propagating at the same direction CW or CCW.

(a) (b) Figure 4.3. Propagation of the modes into the laser cavity. In (a) the bistability is reached in unidirectional regime with one operating mode CW (red-line) while CCW is suppressed (blue-dotted line) and in (b) the unistability operation. As long as the operation of bistable SRL is considered as an optical memory or an AOFF, no results can be found in the available literature to-date. Hence, the work performed within this work and IOLOS project is the first systematic attempt to investigate the use of SRL as optical digital memory that can be addressed all-optically.

Unistability

WEST EAST WEST EAST

WEST EAST

or

Bistability

WEST EAST

or

4.3 IOLOS Project The goal of the IOLOS project is to build monolithic chips capable to work as an optical memory. This is possible due to the bistability behaviour of the devices. Those devices are able to work as an AOFF allowing to generate all-optical digital functions. The IOLOS Project (Integrated lOgic and memory using uLtra-fast micrO-ring bistable Semiconductor laser) is funded by the European Commission within the Sixth Frame Program (FP6) and its goal is to build and to verify the behaviour of photonic integrated devices in all-optical signal processing. Partners of the project are European Enterprises and Universities: The University of Bristol (U.K), Intense Ltd (U.K.), Universidad Illes Balears (Spain), Vrije Universiteit Brussels (Belgium), Nokia-Siemens (Portugal), University of Glasgow (U.K.) whose goal is to fabricate the devices, and Università degli Studi di Pavia whose aim is to analyze the behaviour of the devices. More information can be read in www.iolos.org

4.4 All-Optical Digital Functions with SRL The SRL can be tested for many different digital functions that can be realized entirely in the optical domain. These digital functions include the regeneration functions of Re-Shaping (1R), Re-Timing (2R) and Re-Amplification (3R), and basic logic functions such as all logical gates as AND, OR, NOR, XOR. The directional bistability can make digital functions available in optical domain for signal processing applications. The basic operation to realise all-optical digital functions of the devices as an AOFF is illustrated in Fig.4.4.

CW

fibre fibre

CCW

The last propagation direction remains, in this case CW.

WEST EAST

Figure 4.4. Schematic behaviour of the SRL as an all-optical flip-flop

The SRL can work as an AO-FF controlled by the injection of one external pulse signal at each port of the waveguide West and East. This pulse must has the same wavelength as that emitted by the laser in bistable regime. Due to the directional stability of the SRL, the FF operation can be exploited. Following the arrows in Fig. 4.4 the FF operation of the SRL can be understood. In the example, starting in CW direction operation, the pulse signal is injected at the port on the west side of the device changing the propagation to CCW mode. In the other case if the pulse signal is injected at the east output port side of the device, while a CCW mode propagation is present, this CCW propagation mode changes to CW direction. The last step in the figure means that even without an injection pulse, the last state of propagation remains into the cavity for an infinite amount of time. The illustrated SRL behaviour allows to implement a digital memory of one bit.

The SRL device can also be used in wavelength converting transponders as 1R, 2R or 3R functions in WDM systems. Signal regeneration in optical networks becomes important because the transmitted pulses lose their shape during the propagation into the fibres, and the optical signal must

to be re-shaped and re-timed. In bidirectional regime, the SRL can operate as an all-optical 2R regenerator because it is able to re-shape the signal as illustrated in Fig. 4.5a. The pulse is injected at the port (west) while the SRL is operating in bi-directional regime, favouring the CCW propagation mode. In Fig. 4.5b. the pulse is injected at opposite port (east), then the transmitted signal is favoured on the CW propagation mode. In this manner the signal is regenerated and can be successful re-transmitted.

Figure 4.5. In bidirectional regime, re-shaping and re-timing the signal as a 2R regenerator. (a) injection at west side in CCW, (b) at east side in CW. To enhance the switching speed of the device, the favouring propagation direction can be produce by an external Continuous Wave (CW) optical beam acting as a pump, in this way the optical digital functions can operate faster.

Figure 4.6. In unidirectional regime, re-shaping function using holding beam

WC holding beam WC holding beam

(b)

0 0

0

(a)

EAST

WEST

4.5 Device Design and Fabrication Several geometries of cavities of the ring laser have been demonstrated since the early 1990’s in GaAs and InP based compounds. A lot examples of fabrication and characterization of SRLs have been reported in the last two decades. Different cavity geometries have been proposed, such as circular [22, 23], racetrack [24, 25], square [26, 27], and triangular [28], employing various light guiding mechanisms such as pillbox structures using the “whispering-gallery” effect [22], deep-etched or rib-waveguides [24], shallow-etched ridge waveguides [23], buried hetero-structures [29]. Despite the appealing potential for monolithic integration, the mode and frequency stability of SRLs was always limited by the coexistence of the two counter-propagating modes in the lasing cavity. Even if several geometries have been proposed to introduce asymmetry in the cavity and thus induce unidirectional behaviour, SRLs never reached the maturity and device performance to stand as real competitors against conventional linear cavity lasers for good quality light sources for optical fibre transmission. The SRLs tested in this work are shallow-etched ridge-geometry devices and were fabricated at Glasgow University in Multi-Quantum Well (MQW) InGaAs/InGaAlAs/InP material, using electron-beam lithography technique. The shallow-etched waveguides design ensures single transverse mode operation. To enhance the coupling factor, each device has a racetrack-shaped active area with straight output waveguides used to inject or to collect the light signal. In Fig.4.7 it is shown the laser structure. In Fig.4.7(a) the old generation of the SRL with only one coupled waveguide and 2 ports used to injected/extract the optical signals, in 4.7(b) the new generation of the SRL with 2 coupled waveguides and 4 all-active ports that can act as Semiconductor Optical Amplifiers (SOAs) and are named North West (NW) SOA, South West (SW) SOA, North East (NE) SOA, South East (SE) SOA. In Fig. 4.7(c) the 10° tilted and tapered waveguide of the new generation of devices are shown.

+10dB SE SOA

NE SOA

SE SOA

NW SOA

SW SOA

(a) (b) (c)

Racetrack Active Area

Coupled Waveguide

10°

SRL

Figure 4.7. Illustration of the geometry of two generations of SRL devices showing the racetrack active area. In (a) the device provides 2 ports. In (b) the device provides 4 all-active ports (SOAs), (c) tapered output waveguide. The prototype devices are made into a small bar of size 1x15mm, each including 7 devices of different sizes for the new batches or 2 arrays of 4 devices for the old batches (see figure 4.8 (a) and (b)). In the following tables are specified the devices of each bar with their corresponding dimensions of radius R and length of the waveguide L. Fig.4.8 (a) shows the bar of the first generation (as the 4th and 5th) and 4.8 (b) devices of the new generation.

Figure 4.8. Picture of the evolution of the device bars. (a) First Generation. (b) New generation.

Table 1. First generation dimension devices. The values of the L and R with their corresponding perimeter.

Device L [µm] R[µm] Length Cavity [µm] Mode Spacing [nm] FSR [GHz]

1-2 150 150 1242.47 0.56 23.5 3-4 200 150 1342.47 0.52 21.7 5-6 150 200 1556.63 0.45 18.7 7 200 200 1656.63 0.42 17.6

Table 2. Dimensions of the new generation (3G) of devices. The values of the L and R with their corresponding perimeter.

The new generation provides two electrical contacts at the end facet that can act as absorbers when are left unbiased, or as SOA when are forward biased. To reduce the effect of reflections from the end-facets and to increase the fiber-device coupling efficiency, the output waveguides are up-tapered as shown in Fig. 4.7(c). According with the tables 1 and 2, the old and the new generation of devices have the specified dimensions. Those devices work in continuous wave at room temperature with emission wavelength into the C-band. Table 3 resumes the description for each generation of devices tested during this work.

Generation Bar Description Tilt 1G 4 and 5 8 devices per bar 1 waveguide; 2 ports 5° 2G New 7 devices per bar 2 waveguides; 4 ports 5° 3G F, G and R 7 devices per bar 2 waveguides; 4 ports with SOAs 10°

Table 3. Description of the Generation of devices.

The ridge waveguide geometry was chosen to reduce the radiactive recombination, to have low back-scattering, better heat dissipation and better processing control. The directional coupler between the active area and the waveguides has 10% of efficiency, the tilted waveguides allow to decrease the back-reflections. However to build devices with smaller radius of curvature (less than 100µm) it is necessary to use deep-etched geometry. Both shallow and deep etched ridge-geometries are illustrated in Fig.4.9.

Device L [µm] R[µm] Length Cavity [µm] Mode Spacing [nm] FSR [GHz] 1 100 100 828.32 0.84 35.3 2 300 100 1228.32 0.57 23.8 3 100 300 2084.95 0.33 14.02 4 300 300 2484.95 0.28 11.76

RRL L

1 2 3 5 6 7 1 2 3 4 (a) (b)

Figure 4.9. Transversal geometry of the waveguide. (a) Ridge type, (b) deep-etched type.

The devices have a coupling factor of 0.1 - 0.3 due to the long interaction length of the coupler section between the ring and the waveguide. In Fig. 4.10, it is pictured the transversal output beam section of the laser device with dimensions of 3µm x 1µm, the emitting power is in around 10mA with fibre coupling efficiency of 10%. From the tilted waveguides of the 1G devices the out power reached is around -25dBm, indeed for the new generation tilted 10° the value of the collected power increase in more than -20dBm, decreasing the back-reflections and enhancing the behavior of the device, the collected power reached values around -1dBm.

Figure 4.10. SEM picture of the directional coupling

The back-reflections have been reduced during the fabrication of the device, by using Reactive Ion Etching (RIE) technique that allows to reduce the imperfections of the lateral surface of the ridge waveguide making it exactly vertical. The Quantum Well-Intermixing technique was used to obtain passive waveguides that are that are transparent to the laser radiation and show low absorption of the emitted signal.

(a) (b)

2-3 μm < 1 μm

y

x

Waveguid

Outgoing

STATIC EXPERIMENTS ON SEMICONDUCTOR RING LASERS

5.1 Introduction The goal of the static characterization is to investigate the directional emission properties of the SRL, together with the spectral properties, in order to assess the bistability of the SRL, and its suitability to be used as optical source in fiber telecommunications. The static characterization is also essential to assess the quality and the yield of the fabrication process. In this section, all the experimental techniques that have been used will be presented and commented.

5. Static Experiments based on Semiconductor Ring Laser

5.2 Experimental Measurements of the Static Characterisation To test the behaviour of the SRLs, the device must be characterised by continuous wave techniques carried out using different setups. The static properties of all the devices have been analysed by injecting a dc forward current into the contact of the ring cavity at constant temperature. The temperature was controlled by means of a thermistor and a Peltier cell, and during all the experiments it was kept at the constant value of 17°C. The aim of this procedure is to know the laser behaviour for each one of the devices into the bar, in terms of Light-Current curves, Directional Extinction Ratio, Side Mode Suppression Ratio and other characteristics that will be explained along this chapter. The setups of characterisation have been arranged on an optical table; the bar under test is fixed to a brass support, in this way the devices are protected. To connect the SRLs with all the necessary instruments of measurement, the ensemble of bar-brass support named SB1 B in Fig. 5.4(a) has been positioned on a big aluminium mount named SB2 B. The static characterisation procedure is varied depending on the type of bar to be tested. For example the 1G devices needed to undergo a burn-in procedure before the characterisation. The burn-in procedure will be explained in the next paragraph. 5.2.1 Burn-in Procedure The 1G-devices showed originally a very high contact resistance when the pump current was increased. Consequently, the lasers can undergo sudden death failure mechanisms, because at high voltage the insulating oxide can break down and become a short circuit or an open circuit, preventing further operation of the laser. The reason for this is most likely due to the bad waveguide sidewall coverage of the p-contact of the semiconductor device. To avoid these failures, it is necessary to burn-in the device in case it was never tested before. For this procedure is convenient to use two electrical contact probes: one probe for the inner part of the ring, and other one outside, both connected in parallel, (see Fig. 5.1). To reach an optimum contact between the bar of devices and the mount, an indium sheet has been embedded to fill the space between them. This sheet of indium will be used in all the experiments.

Figure 5.1. Burn-in setup

The burn-in procedure has been followed step by step for around 60 minutes with the device temperature of 15°C controlled. Starting at very low current values, the voltage increase to around 3.2V. Then it is slowly decreases during time, i.e. by 0.2 to 0.6V in 5-10 minutes. The reason for this is the contact resistance slowly drops, because the locally increased temperature helps to

Current PumpGenerator

TemperatureController Peltier Cell

SRL

Probes

achieve a better annealing of the top metal-p contact due at he Joule effect. Increasing in small steps the pump current, it is possible to reach the threshold current of the laser, it is important not to increase the value of the voltage more than 3.4-3.5V during the burn-in. If the voltage does not drop anymore, the pump current must be increased only, checking always the voltages and the temperature until the threshold is reached.

SRL Imax [mA] Ith [mA] Voltage [V] 1 37 Open Circuit 3.01 2 50.3 - 2.9 3 130 78 3.1 4 200 89 3.2 5 13 V error limit - 6 28.4 Open circuit - 7 150 75 3.09 8 80 V error limit -

Table 4. Burn-in of the bar number 4, at 15°C temperature controlled and maximal voltage of 3.5V. The second device no really present threshold current. The second column reports the maximum current reached.

SRL Imax [mA] Voltage [V] Result

1 30 2.99 ok 2 60 2.9 ok 3 150 2.9 ok 4 170 3.11 V error limit 5 30 2.99 ok 6 50 2.99 ok 7 130 3 V error limit 8 150 3 ok

Table 5. Burn-in of the bar number 5, at 25 °C (10KΩ), and max voltage 3V

The results of the burn-in reported in tables 4 and 5, show that more than 50% of devices survived this procedure. The next generation devices (2G and 3G) of devices are ready to be used without burn-in. 5.2.3 Photodiode Characterisation The 1G devices have been characterised using large area photodiodes to collect the light at the end-facets of the waveguide; the corresponding setup is illustrated in Fig.5.2. With this procedure, the output light is easily collected even if the photodiode is not so near to the output ports of the device, but the drawbacks are that the experiment must be carried out into a dark room. The measurements with this procedure have been done by injecting dc current in the cavity using 2 electrical probes; this current is increased until a maximum of 150mA and 200mA. The results show only the L-I curves with some diode effect revealed by the slop when the device reach the unidirectional regime (see results in paragraph 5.3).

Figure 5.2. Photodiode Characterisation Setup

5.2.4 Optical Lensed Fibre Characterisation To enhanced the measurements taking advantage of the spectra of each device, the setup has been adapted to use lensed optical fibres to collect the SRL light from both output ports. This setup is illustrated in Fig.5.3. The SRLs have a small emission spot, the lensed fibre allows to collect 10-20% of the emitting light. One important point in order to maximize the collected power, is the position of the fibres; therefore those fibres have been fixed onto a support and positioned using 3 axis a micro-positioner because it allows to arrive closely at the end facet of the device. The maximum fibre coupling angle between the waveguides and the air has been calculated by the diffraction conditions, i.e. using the Snell Law 2211 Sin n Sin n θθ = ; where nB1B is the refractive index of the of the semiconductor equal to 3.41 and, and nB2 B the refractive index of the air equal to 1, the angle θB1 B corresponds to the tilt waveguide, that for 1G and 2G devices it has a value of 5°, and for 3G devices a value of 10°, in this manner is obtained the needed angle of the fibre position θB2 B. So, for devices tilted 5°, this angle θB2 Bis 17° and 36° for devices tilted 10°.

Figure 5.3. Lensed optical fibre setup of characterisation.

In Fig. 5.4. it is pictured the setup showing in (a) the probes position required to pump the laser with external dc current, the angled fibre supports in which is arranged the position of the fibres by the help of micro-screws. The Peltier cell used to dissipate heat, a thermistor used to measure the temperature. In Fig.5.4(b) it is pictured the SRL bar of devices the tilted position of the fibres.

Current Pump Generator

Temperature Controller

Peltier Cell

SRL

ProbesOSA

34

1

Current PumpGenerator

TemperatureController

Peltier Cell

SRL

Probes

Multimeter 1 Multimeter 2

PD1 PD2

Figure 5.4. Picture describing the setup. (a) Position of the probes, Fibre supports East (EF) and West (WF), Peltier cell and the thermistor, ensemble of supports SB1 B and SB2 B, the tilt position of the device into the setup. (b) Up 1G devices, down 3G SRL devices and the tilt fibres at West and East sides (WF and EF). The complexity of this micro-setup is that the probes and fibres need to be positioned by the help of a micro-position screw and the microscope, therefore is necessary to be carefully when the probes are positioned on the device avoiding to stumbling, even with the fibres when are closely to the end facets because it would damage the device. Finally, all the measurements have been done from the instruments by the help of the GPIB interfaces implemented in LabView Software and subsequently processed using Matlab tools.

Probes

Peltier cell S2

S1

EF WF

3G-devicesDevices

WF

EF θB1 B

Thermistor

Micro-screw

(b)(a)

5.3 Results of the Static Characterisation of the SRL The results of the Static Characterisation are used to classify the performance and fundamental parameters of each SRL device at continuous wave operation. The emission characteristics of the devices have been analysed in terms of the light-current curves and the correspondent optical spectra. 5.3.1 Light-Current Curves The emitted light of the SRL is measured as a function of the device current pump and is represented in light-current (L-I) curves. This measurement is strongly temperature dependent; therefore it has been carried out at constant temperature. In the L-I curves it is remarkable the break point at which the light abruptly starts to increase, this point represents the threshold current value. Further increasing the dc current, the device enters in unidirectional regime exhibiting their directional switching behaviour. This particular bistability allows to classify the devices to be explored in further experiments. This static characterisation has been done on 4 types of bars. The first L-I curves have been done by using large area photodiode on the 1G devices. The Figs. 5.5 and 5.6 are showing the L-I curves of 1G devices. The background with logarithmic-like behaviour is caused by parasitic light emission from the side of the device waveguide. In Figs.5.7 and 5.8 are shown the behaviour of the 3G-devices and the L-I curves aren’t exhibiting the background light term. The cause is maybe the new geometry of the devices with 4 output ports.

0 50 100 150 200-0.5

0

0.5

1

1.5

2

2.5

3 x 10-4

Laser Current [mA]Inte

grat

ed W

ES

T /

EA

ST

Pho

tode

tect

or C

urre

nt [

A]

File: BAR4_04_Photodiode_T15_n01.dat

Iphd westIphd east

(a) (b) Figure 5.5. L-I curves of 1G-devices (batch 4) at 15°C using photodiodes. (a) Device 3, Imax of 140mA, (b) device 4, Imax of 200mA. In Fig. 5.5 (a) the break point corresponds at the threshold current value of 78mA approximately by the third device of 1G-bar 4, and in (b) 90mA of threshold current by the forth device, the devices show small and poor directional switching. In Fig.5.6, the devices tested corresponds at the 1G-bar 5 and are exhibiting better switching behaviour at threshold currents of 90mA almost.

(a) (b)

Figure 5.6. L-I curves of 1G-devices (bar 5) using photodiodes. (a) Device 3, (b) device 6.

The L-I curves of the 2G of devices that did not need to be burnt-in are shown in Fig. 5.7, and exhibit better spontaneous directional switchings. The measurements reveal more sharp curve when the propagating mode changes the direction from CW to CCW.

(a) (b) Figure 5.7 . L-I curves of the 2G-devices. (a) Device 1, (b) device 3.

The 3G-devices with 4 outgoing ports have been characterised using one photodiode capturing the light of only one propagating mode; those measurements are illustrated in Fig.5.8: (a) the device 5 is exhibiting 4 switchings; in (b) the device 7 is exhibiting 3 switchings with better Extinction Ratio (ER). Those devices have a larger size creating more changes of propagation direction, both devices represents a good candidate to be tested with lensed optical fibres; indeed this will be the method adopted to carried out the further measurements.

(a) (b) Figure 5.8. L-I curves of 3G-devices (a) device 5, (b) device 7.

In conclusion, the L-I curves using photodiodes can reveal approximately the device behaviour because a strongly LED effect of the photodiode is detected with some spontaneous emission from the laser. This explain why, when the propagating mode are suppressed, the photodetected current doesn’t reach zero. In the L-I curves measured using lensed optical fibres, the collected light corresponds exactly to the lasing mode without spontaneous emission from the side of the waveguide. The figures show a good bistability of the devices. The Fig.5.9 shows the lensed fibre characterisation of the third device of the 1G with very good unidirectional switchings that were revealed by the photodiodes.

Figure 5.9. Lensed fibre characterisation of 1G-3device (bar 5). In Fig.5.9 it can be observed the effect of vibrations, due to the sensitivity of the fibres to external vibrations. Figs. 5.10 and 5.11 show the behaviour of the 3G-devices, showing robust bistability in the laser of changes of propagation direction CW and CCW. The measurements exhibit more switchings for bigger devices (see Figs. 5.10 (b), and 5.11(b)).

(a) (b) Figure 5.10. L-I curves by lensed fibre characterisation of 3G-SRL devices (bar F) with different cavity length (LBcavityB) (a) Device 4 with LBcavityB of 1342.4 μm, (b) device 5 with cavity length of 1556.6 μm

(a) (b)

Figure 5.11. L-I curves by lensed fibre characterisation of the 3G-SRL devices (bar G). (a) Device 2 with LBcavity B of 1242.5 μm , (b) device 7 with LBcavityB of 1556.6μm In conclusion, the L-I curves reveal that devices with shorter length cavity show less spontaneous directional switchings, while for devices with larger cavities. For example in Fig. 5.10 (a) the device has a cavity length of 1342.4 µm it is exhibiting 3 changes of propagation direction up to a dc current of 160mA, while in Fig. 5.10 (b) the device has a large cavity of 1556.6µm and is showing 7 changes in propagation direction up to a dc current value of 150mA. In devices of bar G with the same geometry, it is observed up to 4 switchings for small devices and 5 switchings for big devices up to a dc current of 150mA. More experimental measurements of L-I curves are reported in annex A. 5.3.2 Bistability Properties The bistability has been analysed on 3G devices. Those SRLs have been tested by injecting a dc forward current into the ring contact, at constant temperature of 18°C. The property of bistability is observed in the L-I curves when the device is working in unidirectional regime exhibiting the changes of propagation direction, when one mode is active while the other one is highly suppressed. In this regime, the operating direction can be reversed when the injected current is increased. The SRL has an infinite directional bistability, that is, it remains into the same state until another spontaneous switching occurs when the dc current is increased. This robust bistability has been observed in each device. The total number of tested devices during this work of static characterisation amounts to 30-40. To assess the quality of bistable behaviour of the devices, the Directional Extinction Ratio (DER) was measured, that is defined as the ratio of the intensities between the two counter-propagating modes CW and CCW. The DER tells about the degree of suppression of the non-lasing mode, and it

is considered good when is larger than 10 dB. The Fig.5.12 shows a DER up to 20 dB from the 3G-devices of bar F and G.

(a) (b)

(c) (d) Figure 5.12. DER of 3G-devices. (a-b) bar F-devices 4 and 5, (c-d) bar G-devices 2-6, injecting dc current until 160mA and 200mA exhibiting a DER≥20dB. 5.3.3 Spectral Properties The optical spectra typically are the range of wavelengths corresponding to the operating bandwidth of the devices typically, from 1550 to 1570nm. The spectra contains all the information about the lasing wavelengths, and very according to the dc current. The spectral properties of the devices shown in the follow figures are exhibiting the changes of wavelength lasing modes for both propagation directions West and East in function of the output power when the dc current is increased. Figs. 5.13 and 5.15 show the spectra of the 3G-devices exhibiting a SMSR more than 30dB, which is by far enough to ensure single mode laser operation. The spectral contour curves shwon in Figs. 5.14 and 5.16 are showing with better detail the bistability feature of the devices. The arrows indicate when the active mode are present in one direction while it is suppressed in the opposite direction. The Mode Spacing is evaluated by zooming the contour curves and it is illustrate in Fig. 5.17. In the examples from the 3G-devices (bar F), the Mode Spacing is around 0.48 nm for the devices number 4, 5 and 6, while for small devices as 2 the Mode Spacing is 0.78 almost.

Figure 5.13. Optical spectra exhibiting full directional switchings, and 4 single longitudinal lasing modes (3G-bar F-device 5). (a) West direction, (b) East direction with SMSR ≥20dB each directional lasing modes.

Figure 5.14. optical Spectral Contours for West and East propagation direction. The arrows are indicating the changes of propagation direction of the lasing modes when the current is increasing. Fig. 5.15 shows the spectral characteristics of 3G-devices (bar G), the device is exhibiting SMRS larger than 30dB. Their corresponding spectral contours are shown in Fig. 5.16.

(a) (b)

Figure 5.15. Optical spectra exhibiting 4 single lasing modes of 3G-device 2 (bar G). (a) East direction, (b) West direction, with SMSR≥20dB.

Figure 5.16. Optical Spectral Contours for West and East propagation direction (3G-bar G). The arrows are indicating the change of direction of the lasing modes.

(a) (b)

(a) (b)

Figure 5.17. Mode Spacing of 3G-devices (bar F).

With a SMSR larger than 20dB in the unidirectional regime, these devices can be considered good single mode laser, similarly to Distributed Feed-Back (DFB) lasers. Each spontaneous directional switching is accompanied by a jump in the emission wavelength, i.e., by a longitudinal mode-hop when the ring current is increased, this behavior is observed in Figs.5.18 and 5.19.

(a) (b)

(c) (d)

Figure 5.18. Spectral properties of 3G-devices-bar F, (a-b) SMSR≥20dB by 4 P

thP and 5 P

thP device

respectively. (c-d). Jump Emission Wavelength when one propagation direction is active.

Δλ= 0.58 nm

m = 0

m = +1 m = -1

Δλ= 0.48 nm

m = 0

m = +1 m = -1

Δλ= 0.48 nm

m = 0m = +1 m = -1

Δλ= 0.49 nm

m = 0

m = +1 m = -1

(a) (b)

(c) (d)

It can be concluded that each directional switching is accompanied by a wavelength discontinuity; this is a hint that the spontaneous directional switching are triggered by differences in the optical gain for the CW and CCW modes, that arise due to the peculiar ring cavity geometry. Fig. 5.19 shows the SMSR and the peak of lasing wavelength from the device 6-bar G measured.

(a)

(b) Figure 5.19. Spectral properties of 3G-bar G-device 6. (a) SMSR≥20dB at each propagating direction, (b) their corresponding jump emission wavelength. 5.3.4 Effects of Optical Feedback from the end-Facets The 3G devices have been fabricated with tapered waveguide to reduce the reflectivity from the uncoated end-facets and to increase the fibre-device coupling efficiency. To investigate the effects of the biased SOAs, it was gain SOA increased when it is biased at large current values, but does not a negatively effect to the directional switchings nor the DER. There are two remarkable effects on the unidirectional behaviour due to the biased SOAs that have been investigated and will be explained long this paragraph. These effects can produce a poor SMSR.

(a) (b)

TFigure 5.20. First result of the effect of forward bias of SOA output waveguides of 4P

thP device (bar

F) for submount temperature of 18° C. The L-I curves of East/West powers collected by the fiber are shown together for SOA currents of 2.5mA in (a) and 10mA in (b). The bistability in (b) is observed with only one operating directional mode, hence the device is tending to work in unidirectional regime, but without spontaneous directional switchings.

Typically, the optical gain provided by a SOA biased with 10-15 mA is around 10-13 dB. If the SOAs are unbiased the reflectivity of the uncoated tilted end-facets can be estimated to be 0.5⋅10P

-3P,

while the effective reflectivity with the SOAs forward biased reaches 10P

-2P. A first result is presented

in Fig.5.20, where it is shown the L-I curves by two directions for varying SOAs bias. In Fig.5.20

(a) the out power is increased by almost 2dBm by SOAs bias of 2.5mA and the bistable condition are always present. In Fig.5.20 (b) at high values of SOAs bias of 10mA, the reflectivity is affecting the behavior of the device allowing this to work only in un-directional condition suppressing the spontaneous switching. In all those measurements Tthe fluctuations in the L-I curves are due to vibration of the fibre tips. T

A comparative result is presented in Fig.5.21 that reports L-I curves for the two directions for varying SOAs bias. In some cases, the active mode for a certain range of ring current changes from East to West, and vice-versa. At higher SOA bias, only one mode remains active, while the other is suppressed for any ring current value. This is probably due to the asymmetry in the device facets and waveguides, in such a way that one specific direction is favored. This asymmetry effectively helps in making the operation direction deterministic for each device, allowing the use of the SRL as a viable laser source for telecom applications in PICs where the laser cavity shall not be defined by etching or by Distributed Bragg Reflector (DBR).

TFigure 5.21. Effect of forward bias of SOA output waveguides of 4P

thP device (3G-batch F) for sub-

mount temperature of 18° C. The East/West powers collected by the fibre are shown. There is experimental evidence that the increased reflectivity from the end-facets due to the SOA gain does not negatively affect the directional bistability of the SRLs, nor the DER. The only noticeable effects when increasing the SOA forward bias were: 1. A change in the sequence of spontaneous directional switchings when the ring current was

increased. 2. At large SOAs currents (> 10 mA) only one directional mode tended to lase in the unidirectional

regime, with the other mode being suppressed for almost any value of the ring current. 3. To reduce the effects of reflections from the end-facets to a minimum, the SOAs output

waveguides are kept unbiased. In this condition, it can be estimated that the SOAs single-pass attenuation is around 10 dB, yielding effective optical feedback strength from the facets into the laser cavity smaller than 10P

-5P. The power emitted in air can reach 20 mW (with SOAs forward

IWEST = 10mA, IEAST = 10mA

II SSOOAA WWEESSTT

IWEST=0mA, IEAST=0mA I WEST = 2.5mA, IEAST = 0mA IWEST = 5mA, IEAST = 0mA IWEST = 10mA, IEAST= 0mA 10-3

10-3 IWEST = 0mA, IEAST = 10mA

IWEST = 0mA, IEAST = 5mA

IWEST = 0mA, IEAST = 2.5mA IWEST = 2.5mA, IEAST = 2.5mA

IWEST = 5mA, IEAST = 5mA

II SSOOAA EEAASSTT

Unidirectional EAST

biased), and 3mW at 250mA with SOAs unbiased. The typical device-to-fibre coupling efficiency can reach 20%.

It can be concluded that the SRL devices have a very robust directional bistability, and, given the large reported value for the SMSR also in presence of forward SOA bias, they make good CW lasers for use in optical transmission systems. In this respect, the possibility of increasing the SOA current without negative effects on the spectral and unidirectional emission allows to achieve a power level coupled into the optical fibre that is sufficient for telecom applications. 5.4 Experimental Measurements of the SRL Lindwidth The linewidth is one of the main features for a laser in optical telecommunications because it is related to the coherence time. The methods for linewidth characterization of un-modulated single mode lasers are often defined in terms of the Full-Width Half-Maximum (FWHM) of the optical field power spectrum. For those measurements, it was implemented a heterodyne technique using an External Cavity Laser (ECL) and the two optical fields (ECL and SRL) are incident on the photodiode. This procedure will be explained in the follow sub-chapter. 5.4.1 Theoretical Linewidth The dependence between the coherence time and the linewidth is illustrated by the theoretical curves in Fig. 5.22 (a) and (b). Where the coherence time varies inversely with respect to the

FWHM according to FWHM

tc *1

π= . The FWHM linewidth is a measure of the spectral purity of

the laser frequency over time.

(a) (b) Figure 5.22. Theoretical description of the linewidth Δν. (a) Coherence time dependence, (b) light emission Lorentzian shape of a laser useful to measure the FWHM (∆ν@-3dB). In semiconductor lasers, the FWHM values change as a function of the variation of the instantaneous emission frequency of the laser, and are proportional to the variation of the refractive index (Δn) in the cavity, that changes during the emission of photons when the electron in the conduction band fall down to the valence band. A variation of carrier densities into the cavity (Δn ≈ ΔN ≈ Δν), then causes a variation in the emission wavelength of the laser. The FWHM linewidth of the laser is often measured with respect to an assumed Lorentzian spectral shape SBEB(ν) illustrated in

Fig. 5.22, the Lorentzian optical power spectrum is centred at a frequency νB0 B and it has a functional form given by:

( ) 2

0

21

1

⎥⎥

⎦

⎤

⎢⎢

⎣

⎡

Δ−

+

≈

ννν

νES , (1.2)

Where ν is the optical frequency. The usual case of a Lorentzian-shape spectrum and the correspondence between the full-width at a specific power level width respect to the peak is shown in the table 6.

Measured Full-Width Point Corresponding Width -3 dB Δν -10dB 9 Δν -20dB 99 Δν -30dB 999 Δν

Table 6. Linewidth relations

5.4.2 Linewidth Measurement As the Optical Spectrum Analyser (OSA) doesn’t offer sufficient frequency resolution to display the details of each longitudinal mode of the laser, the measurement of linewidth was carried out by an alternative technique. The typical values of linewidth for semiconductor laser are between 1 to 50MHz, and the OSA has a resolution of the order of GHz, then it was necessary to use the heterodyne technique to analyse the optical signals and measure the linewidth of our SRL. The heterodyne analysis was carried out by using ECL to produce a beating between the ECL and the SRL, this technique offers exceptional sensitivity and resolution when a narrow linewidth reference laser is used and can be also used to characterized non symmetrical spectral lineshapes. The setup used is illustrated in Fig.5.23: the goal is to produce the beating between the ECL and the SRL signals, this is a heterodyne method because both optical signals incident on the high speed photodiode which detects the interference beat tone, converting it to an electrical tone. Hence it is necessary that the linewidth from ECL is much narrower than SRL linewidth, (FWHM ECL << FWHM SRL). The light from the ECL is combined with the signal of the SRL by a 50/50 fibre coupler, collecting the power of the SRL by the help of a lensed fibre, both signals SRL and ECL have been aligned using the OSA. The ECL is tuned to a frequency just lower than the average frequency of the SRL then this created a heterodyne beat tone between the ECL and each correspondent frequency of the SRL signal spectrum. When the mixing product falls within the bandwidth of typical detection electronics, this beat signal is shown in Fig. 5.24. The overlap polarisation between ECL and SRL is optimized using the Polarisation Controller (P.C) before the photodiode. A High Speed Photodiode (HSP) detects the interference beat tone, converting it to an electrical tone, this electrical signal is analysed and measured by a 20 GHz Electrical Spectrum Analyser (ESA) in the frequency domain.

Figure 5.23. Optical Heterodyne setup to measure the SRL linewidth using an ECL as a local oscillator to produce the beat.

Figure 5.24. Mixing process of beating between ECL and SRL that can be analyzed with ESA. The spectra SBoB(ν) in optical domain and SBEB (ν) in electrical domain. This measurement can be physically analysed by writing the two field functions described by:

))](2[)()(

))](2[)()(

ttExpitPtE

ttExpitPtE

ECLECLECLECL

SRLSRLSRLSRL

φπν

φπν

+=

+= (1.3)

Where EBSRLB represents the field signal of SRL and EBECLB the field signal of ECL in terms of propagation. The interference between both signals on the photodiode is then: ))()(2cos(2)()()( ttPPtPtPti ECLSRLECLSRLECLSRL φννπ Δ+−++= (1.4)

With central frequency given by ECLSRL νν − , and a phase noise )(tφΔ . If the phase noise of the SRL is bigger than ECL (φ BSRL B>> φ BECLB), then the produced beating is evidently due to the SRL noise phase and consequently the linewidth measurement of the SRL. In the ESA the spectral density is described by ( ) ( )ECLSRLECL SPRS ννν −∝ 22 , where ( )ECLSRLS νν − represents the emission spectra of the SRL in baseband and RP

2P [A/W] the responsivity of the photodiode, in this way the FWHM

measured by the ESA corresponds to the FWHM from the SRL.

ν «

200THz 10GHz

ECL

SRL

S BoB(ν)

Beat tone

SE(f)

f 10GHz

ECL + SRL Beat tone

L-bandEDFA

RF SPECTRUM

ANALYZER

ECL

HIGH-SPEEDPHOTODIODE

50:50 Coupler

ISOLATOR

LENSED FIBRE

SRL

P. C

λ BECL B≈ λBSRL B

OSA

5.4.3 Results of the Linewidth Characterisation The first measures have been taken on 1G-bar 5 of the SRLs, which static behaviour in the P-I curves shown a high threshold current. The linewidth measurements were difficult to obtain because the Lorentzian emission spectra in the ESA was widened, the inconvenient is due at the fact to find the accurate resonance between one mode of the SRL and the ECL wavelength to produce the beat of both signals. The first measurements on 1G-bar 5 devices, shown high values of the linewidth, this can be due to the fact that the directional switchings are by competing to obtain the gain for lasing condition. Indeed if the SRL signal is beating with the external signal, this competition can produce a small signal with repeated commutations altering the behaviour of the laser. The measurements of the linewidth can be complicated in the case that not enough optical power from the device is collected; this generates a small and noisy beat signal. The linewidth measurements are strongly dependent of forward dc current: at current values near to threshold the linewith is wide; when the current is increased the linewidth values decrease down to a few MHz. This can be observed in the follow figures, where the measurements have been taken on 3G devices, from 95mA to maximum current value of 145mA, the large optical output power allowed to obtain better beating. The bar works well in continuous wave allowing to obtain reasonable FWHM measurements. The linewidth of the SRL decreases for increasing emitted power, except for a small jump occurring in correspondence with a directional switching. The minimum linewidth value is measured as 2.2 MHz at 150mA. Larger device exhibit larger linewidth, of order of 10 MHz (see Table 2.2).

3G-Device Lasing Wavelength [nm] FWHM [MHz] 02 1550-1560 2.4 04 1550-1560 11 07 1560-1570 10

TTable 7. Summary of FWHM results for different devices sizes.

In order to analyze the dependence of linewidth with the small jump produced by the directional switchings, the linewidth measurements have been done on 3G-bar F-device 2, increasing and decreasing the dc current from threshold to maximum, illustrated in Fig. 5.25. This device exhibits a widening of the linewidth around 125mA, probably caused by instabilities arising from the vicinity to a directional switching. The FWHM measurements of the bar F from the small device number 2 are shown in table 8. This linewidth measurement can be obtained in equal conditions when the lasing mode are CW or CCW operation at West or East direction.

Current [mA]

Mode Δν [GHz] Linewidth [MHz]

70 West 8.4 2.8 90 West 8.9 2.9 110 East 7.1 2.3 120 West 8 2.6 140 East 6 2.0 150 East 7.1 2.3

TTable 8. FWHM measurements front ha bar F-device2 at different current values.T

I BSRLB = 95mA IBSRLB = 105mA IBSRLB = 110mA

I BSRLB = 115mA IBSRLB = 120mA IBSRLB =125mA

I BSRLB = 130mA IBSRLB = 135mA IBSRLB = 145mA Figure 5.25. RF Spectra of one small device at different current values from threshold current of 95mA to maximum current of 145mA, exhibiting the Lorentzian function and the maximum peak of power. The RF Spectra let see that the linewidth of the device 2 from the bar F exhibit widening at current values of 125mA and 130mA due to the mode competition effect. On this device, a narrow

Frequency [GHz]

RF Spectra [dBm]

linewidth was obtained at 115mA. Fig. 5.26 reports a summary of linewidth RF spectra directly from the Network Analyser exhibited by devices of 3G of different size.

(a) (b) (c)

Figure 5.27. Summary of RF Spectrum for different size of devices of the bar F, exhibiting the Lorentzian function and maximum peak of power. (a) Small size at 115mA, exhibiting more power, (b) medium size device, (c) big size device. In conclusion, the linewidth decreases for increasing emitted power, except for a small jump occurring in correspondence with a directional switching as illustrate in Fig. 5.28. The minimum linewidth value is measured as 2.2 MHz. Larger device exhibit larger linewidth, of the order of 10 MHz (see Table 7).

TFigure 5.28. Measured linewidth vs. ring current for BAR_F_device 2.

5.5 Conclusions about the Static Behaviour of the SRL

In conclusion of the static measurements on the SRL, directional bistability with spontaneous directional switchings have been demonstrated in all the measured devices, confirming that the directional bistability is an instrinsic characteristics of the SRLs. Those spontaneous switchings are triggered by differences in the optical gain for CW and CCW modes due to the ring cavity geometry and occur in correspondence with a wavelength discontinuity. A good SMSR (larger than 30dB from some devices) together with satisfactory DER values (larger than 20dB) were reported. 3G devices exhibited robust directional bistability with large DER and SMSR values also in presence of forward currents in the terminal SOAs, showing that SRLs can make good CW laser sources to be integrated into Photonics Integrated Circuits for use in optical transmission systems.

DYNAMIC EXPERIMENTS ON SEMICONDUCTOR RING LASERS

6. All-Optical Digital Functions for Transmission Systems based on Semiconductor Ring Laser

6.1 Introduction The realisation of digital functions directly in the optical domain is becoming a key issue for the development of future optical fibre transmission systems, where the implementation of all-optical digital signal processing with Photonic Integrated Circuits (PICs) will bring more flexibility and security into the networks enhancing the system performance. Indeed above is well in agreement with the goal of the IOLOS project.

The SRL devices have demonstrated to be a viable solution for monolithic implementation of all-optical digital functions, thanks to the inherent directional bistability behaviour and compact size, as explained in previous chapters. In particular, the bistable SRL can be used as an all-optical Set-Reset Flip-Flop (SRFF) triggered and switched by the injection of an external optical signal. This chapter describes the experiments of AO-SRFF carried out to investigate the dynamical behavior of the SRL devices used as all-optical SRFF. The procedure of dynamic characterization has the goal of confirming that the device can be operated as a SRFF at different speed and using different optical switching trigger signals. This chapter reports on experiments about the demonstration of all-optical SRFF operation under repetitive injection of Set and Reset signals using different optical trigger pulses.

6.2 All-Optical Set-Reset Flip- Flop and Bit Error Rate based on SRL The dynamic experiments were carried out using variations of the set-up shown in Fig.6.1, where the source to generate the optical pulse to be used as Set-Reset trigger has been changed accordingly to the specific measurement, going from very slow pulses of 20 ns duration to ultrafast 5 ps pulses. To test the SRFF functionality it is essential that the Set and Reset trigger pulses be alternatively injected into the SRL from both sides. Those injection pulses must have a good ER between ‘high’ and ‘low’ states of the injected signal to ensure a successful result. The pulses are generated by M.Z.Ms, optimizing the ER by P.C and synchronizing the arrival times of the pulses by the equal lengths of the optical and electrical transmission lines. The SRL output is acquired by using high speed photodiodes and digital oscilloscope.

Figure 6.1. General setup for the dynamic experiments based on SRL SRFF.

The characteristic table 9 defines the state of the flip-flop as a function of inputs and previous state where Q refers to the present state that corresponds at one propagation mode direction, for example when CW is ON. The Q(t+1) refers to the next state. The characteristics of an all-optical SRFF show that the next state is equal to the present state when both inputs Set (S) and Reset (R) are equal to 0, for example when the CW operation is present into the active area working in Q state, this state remains when there is not injection pulse and in this case the SRFF can be represented as a bit of memory. In other case, when R=1 and S=0, the next clock pulse clears the flip-flop and the state Q (t+1) is equal to 0. When S=1 and R=0, the flip-flop output is set to 1 changing the propagation to CCW. The condition of S and R are both equal to 1 represents an indeterminate the next state (X).

Q(t) Q (t+1) S R Action 0 0 0 X No change: CW 0 1 1 0 SET:CCW 1 0 0 1 RESET:CW

S R Q (t+1) 0 0 Q(t) 0 1 0 1 0 1 1 1 X

Table 9. Set-Reset Flip-Flop operation. Left: characteristics. Right: excitation

To generate the SRFF behaviour of the SRL, it is necessary to reach the follow conditions:

The clock signal needed to triggered the Set-Reset pulses must be shape accurate, this trigger is generate by an external laser modulated fundamentally by MZM. The external wavelength of the laser must be on one of the directional modes of the SRL.

The Set-Reset injection pulses must be even shaped accurate and good synchronized in complementary operation with enough ER.

The SRL response at those Set-Reset injected pulses determines their SRFF operation, as shown in Fig.6.2. To ensure this operation, the collected output light from the SRL device must be enough, typically around 1mW (-3dBm).

Figure 6.2. Scheme of the signalling process, Clock-SET-RESET and SRL answer.

The general setup of dynamic experiments based on SRL-SRFF is illustrate in Fig.6.1. Basically, the P.Cs are used to accurate the trigger before to be modulated by the M.Z.Ms. This trigger is used to generate the Set-Reset injection pulses by different manners depending of the bit rate required. 6.2.1 Slow-Pulse Optical Trigger Injection The first experiment was implemented using a “slow” optical trigger pulse of 20 ns duration using a repetition time of 200 ns. This pulse is shown in Fig.6.3: in (a) it is schematized the pulse and in (b) the real generated pulse. The “slow” optical trigger pulse refers to the intrinsic characteristic response time of the SRL that, in the worst case, is around 0.5 ns.

CLOCK

SET

RESET

CW

CCW

time [ns] 5 10 15 20

SRL output: ‘1’

‘1’

‘0’ ‘1’

‘0’ ‘0’

Figure 6.3. a) Schematic representation of the “slow” pulse, that was generated either by a MZ modulator driven by a 10 MHz function generator, or by an SOA directly modulated through the same signal generator. b) Time-domain trace of the optical signal injected as Set and Reset pulses. Time scale: 10 ns/div. The FWHM duration is 18 ns. At the beginning it was used an arbitrary pulse generator working at 500 MHz with a function generator of 20 MHz and 10% of duty cycle (5 MHz) but it was difficult to accurate the trigger pulse, then the experiment was carried out by using MZM driven by 10MHz function Modulator and illustrated in Fig. 6.4. This pulse has been generated before by an SOA directly modulated but with negligible results. Indeed further experiments were carried out by generated the injection pulse by a Pulse Pattern Generator (PPG). The rise time of this pulse is much longer than the response time of the SRL; this slow pulse is the optimum choice to test the bi-stable SRFF capabilities in quasi-static conditions, that are of interest for an ultimate test of the real SRFF operation. The quasi-static pulse is essential to verify the effective directional bistability of the SRL, and to check whether the SRL acts as an optical bistable with an optical threshold. The SRFF operation is very stable, and it can be achieved at any pump current value within the unidirectional regime, provided the optical trigger wavelength matches one of the longitudinal modes of the SRL. Also, when the output of the SRL is observed on a real-time or sampling oscilloscope with infinite persistence time, it appears that the directional switchings that realize the FF operation take place for each and every injected trigger pulse. This experiment is illustrated in Fig.6.5 and it confirms that the SRFF operation is stable and reliable.

20 ns

200 ns

ER > 30 dB

(a) (b)

≈18 ns

Figure 6.4. First setup to generate pulses of 20 ns duration by MZM.

During the experiment it was noticed that the ER of the optical trigger pulse is an important parameter for successful operation. When the ER is not sufficiently large, the FF operation is not guaranteed, especially for large input powers. The reason for this can be that an incomplete extinction of the trigger can enforce the stability of the SRL directional mode opposite to the one that should be turned on by the trigger pulse producing a poor SRFF response observed with the oscilloscope. To avoid any residual effect caused by the incomplete extinction of the trigger the optical trigger was generated using an external SOA used as amplitude modulator via direct current modulation, for slow-speed operation, a SOA is capable of providing an ER as large as 30-40 dB, but it was difficult to produce a good trigger clock signal; for this cause in further experiments this optical trigger have been generated by a MZ modulator controlling the ER by the bias of the MZ and the P.C at both sides East and West. When the SRL is in unidirectional regime the experimental time-domain trace shown on the digital oscilloscope reveals that the SRL has a preferred unidirectional mode, and it is switched to the opposite direction only when the external pulse is injected. If the ER is good enough the SRFF operation is produced and the devices goes on a bi-stable situation, but if the ER is not enough, it causes a coupling unbalance of the East and West optical injection pulses, this situation resembles that of the monostable operation, where a CW holding beam forces the SRL to operate in one preferred direction in which the holding beam is injected. It is the demonstration of monostable operation that can be used for 2R or 3R regeneration where the preferred pulse is reshaped and retimed. This operation is illustrated in Fig.4.5. Using the above method, it was also investigated the effect of varying the polarization state of the optical trigger by the help of P.C.Iit was confirmed that the polarization state have some influence, and in particular that injecting a trigger with polarization different from TE is equivalent to a reduction of the effectively injected power. However, the polarization dependence can be practically eliminated by injection a trigger with sufficient power.

Figure 6.5. Experimental time-domain trace of the SRL east output (bottom) that confirms the operation as SRFF in dynamic conditions when the SRL is biased in the unidirectional regime. The residual Set-Reset pulse can be seen on top of the trace of the SRL response. Device used is 3G-batch_F_05. The SRL is biased in the unidirectional regime at 130 mA, and the peak optical trigger power injected into the SRL waveguide is 10 μW. Another important parameter s the rise-time, illustrated in Fig. 6.6, showing details of the rise and fall times of the output signal of the SRL. Both rise and fall times are around 700 ps, with negligible dependence upon SRL biasing condition and injected trigger pulse power. The above switching speeds can be interpreted as SRL intrinsic spontaneous switching times, i.e. the time it takes for the SRL to change the direction of operation without injection of an extra amount of external photons. This should coincide with the switching speed for the directional switchings that occur spontaneously when the SRL current is increased in quasi-static conditions (i.e., the switchings that are observed in P-I curves). It is define a “spontaneous” switching as a switching event that can be triggered by a variation of external conditions (current change, photon injection), but without the injection of too many photons in the new direction of lasing. The reason for the slow speed of these switchings is that in the new lasing direction (i.e., a new mode to be switched on) there is available a large carrier density ready to provide optical gain, but there are no photons at the beginning, and hence the effective carrier lifetime is long. Conversely, when a large number of photons is externally injected in the new lasing direction, the stimulated emission processes induced by these photons helps reducing the effective carrier lifetime, thus making the directional switching much faster. Experimental evidence for the above mechanism will be given in the following sections. When detuning effects are taken into account, it is interesting to check whether the SRL can be switched by an optical trigger that is resonant with a non-lasing longitudinal mode, instead of the lasing mode. Experiments proved that the SRFF operation is guaranteed even when the injection occurs on a side mode. In particular, we observed switchings when injecting on modes ranging from mode -3 to mode +3 (where modes are indexed with positive sign when they are at longer wavelength with respect to the lasing mode). When the injection occurs on a mode that was a side mode for the non-injected SRL, the injected mode then becomes the new lasing modes, and it can be maintained for an indefinite amount of time. In Fig.6.7 it is shown the optical spectra for the unperturbed SRL, and for SRFF operation with injection on mode 0 and mode -3. When the optical trigger is detuned from a longitudinal mode, the power required for the directional switching increases. The requirement that the optical trigger must be resonant with a SRL longitudinal mode does not prevent proper operation in a telecommunication system, provided the

West/

SRL East output

East Input trigger with peak of -20dBm

wavelength of the optical trigger is known. In fact, the longitudinal modes of the SRL can be finely tuned by changing the operating ring current and the SRL can thus be tuned according to the wavelength of the incoming signal. When the trigger pulse is ultrashort and has an optical spectrum that overlaps with more than one cavity resonance, the directional switching occurs irrespective of the bias current and the specific center wavelength of the pulse, and it can thus be said that the SRL can operate with broadband optical trigger without any major concern for accurate wavelength tuning.

(a) (b)

Figure 6.6. Time-domain traces showing the rise time and fall time of the SRFF when triggered by a slow optical trigger pulse (20 ns duration). a) rise time and; b) fall time around 700ps.

(a) (b) (c) Figure 6.7. Optical spectra of the SRL used as SRFF with injection on different longitudinal modes. A) no injection (free running); b) injection on mode 0 (i.e., originally free running lasing mode); c) injection on mode -3 (negative detuning by 3 longitudinal mode spacings)

6.2.2 Slow-Pulse by Electrical/Optical Trigger using Pulse Pattern Generator As explained before, the experiment with slow pulse can be carried out generating the trigger pulse by the help of a PPG: this instrument can generate an electrical clock signal that can be transformed into an optical trigger by a MZM through E/O conversion. Following the setup illustrate in Fig.6.8, the optical clock after MZM is amplified to reach maximum of power around +10dBm by a high-power EDFA (up to +27dBm), the optical signal goes through the fibres and is 50/50 divided by the coupler. In this point the light must be accurately path-guided: indeed, the optical path must be adjusted to equal values to avoid that the injected pulses from West and East side are not synchronised in time, because this can generate false SRFF operation or simply the SRL cannot switch when the pulses are injected. In this setup the optical paths have been adjusted on the east side using a fibre prolongation of 304cm to ensure the synchronisation of the injection pulses and accurates by the delay knob of the PPG instrument. A crucial point for a successful experiment it is that the polarisation of the light must be accurately adjusted before each MZM, in order to attain the best possible ER. In fact, a poor ER can prevent proper SRFF operation. The polarisation is adjusted by the P.C IN, then the optical trigger signal goes to the P.C West 1 and P.C East 1 for ER adjustment before being modulated by the West/East MZMs to generate the Set-Reset injection pulses. At this point, is important that the electrical paths from the PPG to the West/East MZMs have the same length. Therefore it was used a RF cable of 1m length at each side, but sometimes during the experiment it is was required some adjustment by a short cable extension. In this way, the injection pulses can be perfectly synchronized. The polarisation state of the West/East pulses are adjusted by the help of the P.C west 2 and P.C East 2 before being injected into the SRL.

Figure 6.8. Setup using PPG to test the function of an all- optical Flip-Flop on SRL.

A

B C

D E

The PPG is able to generate arbitrary digital words of 2 to 16 bits duration or it can generate standard in Pseudo Random Bit Sequence (PRBS). Operatively, after the accurate conditions to get the commutation function of the SRL as an all-optical SRFF, are found the injection of the Set-Reset pulses are tested at frequencies of 100 or 200MHz with a short digital word of 2 or 4 bits. The word can be changed during the experiment procedure observing the SRL working as a SRFF. To be sure that the observed output signal is really that of the SRL, the dc current pump of the SRL is turned off, observing a flat response on the oscilloscope. The injected wavelength is aligned to that of the mean lasing mode of the SRL by the help of the OSA. Sometimes it was observed that the best position was the second mode (-2). This is shown in Fig.6.9 and previously in Fig.6.7. At this point is possible to measure approximately the commutation time, which was around 100ps for almost all the devices tested.

m = 0

m = +1 m = -1

m = +2 m = -2

The all-optical SRFF was tested alternatively triggered by optical Set and Reset pulses in the form of bits or PRBS sequence, by generating two RZ PRBS sequences: follow the setup in Fig. 6.8, the Set signal (using DATA from the Pattern Generator) and the Reset signal (using DATA). In this way, during each bit time-slot, the SRFF is injected by either a Set or a Reset signal. The response of the SRFF is a NRZ signal with the same data sequence of the Set or Reset bit-stream, depending on the side from which the output signal is extracted. An error-free operation is a proof that the SRFF can indeed work with an arbitrary Set and Reset data sequence, without missing any externally triggered switching, nor exhibiting spontaneous switchings of its state. The PPG is able to generate arbitrary digital words from 2 to 16 bits duration, or it can generate standard Pseudo Random Bit Sequence (PRBS).

6.2.3 Results of the All-Optical Flip-Flop Measurements by PPG and Slow Pulse The results show that the SRL operating in unidirectional bistable regime can be used as an AO-SRFF. In this configuration, an externally injected signal pulse can reverse the direction of operation of the SRL. Experimental measurements have been carried out with a peak power coupled into the SRL waveguide of 0.64mW for large size devices (as the device 4 and 5 of the bar F) and 1mW to 10mW for smaller devices. To start the experiment is useful to know the mode separation between the m0 and the consecutives modes m±1 because it is convenient to match the injection pulses at the same wavelength, in this

λSRL

λinjection

Figure 6.9. Sketch of SRL modes with injection wavelength in mode +2.

way the measurements have been carried out injecting at the mean lasing mode m0, or at adjacent modes preferably +1 and +2.

Figure 6.10. Optical Trigger after MZM in A on the setup.

The measurements have been taken by detecting the light with a photodiode with 1GHz Bandwidth that inverts the sign of the signal. The clock signal must be good-shaped. As is illustrated in Fig. 6.10 the optical trigger used is a square wave signal. The clock signal in this experiment had 10ns duration. To ensure commutations, the optical trigger is optimised by biased MZM adjusting the ER with the P.C IN, schematised in the setup in Fig.6.8. The shape of the Set-Reset injection pulses have been accurately adjusted by the help of the P.C West 1 and P.C East 1 and the respective bias of the MZMs. Those pulses must be complementary with respect to each other. In this way there is good possibility to get a commutation of the SRFF following the required signalling process schematised in Fig.6.2. In order to obtain a good ER between ‘0’ and ‘1’ from the directional switching of the SRL the bias of the MZM bias and the P.C West 2 and P.C East 2 have to be carefully adjusted. The Set-Reset injection pulse have been modulated using word of 4 bit length (as 0011 or 1010) and shown in Fig.6.11 (a) and (b). The shape waveform of the Set and Reset signals is not so accurate even if the ER between ‘1’ and ‘0’ has been well defined to get a commutation of the device.

4_Bits 0011 4_Bits 1010

Complementary 1100 Complementary 0101

(a) (b)

Figure 6.11. Injection pulses of 4 bits taken before to be injected in the point B (West) and C (East) on the setup. (a) 0011 Length of bits and their complementary 1100, (b) 1100-0011length of bits. The experiment has been carried out changing the injected signal wavelength by tuning the ECL. It was observed that when the injection wavelength overlaps with m-2 mode of SRL, the device shows

switchings with better ER. A satisfactory SRFF operation was observed with small size devices, exhibiting commutation with signalling of more than 3 minutes of error free; Fig.6.12 shows a comparison of SRL output when the injection wavelength coincides with that of mode m-2 and mode m+2 .

(a) (b)

1 0 0 1

20ns

Figure 6.12. Output SRL traces when the injected wavelength corresponds to (a) m-2 and (b) m+2 modes. Those measures have been taken at 113mA ring current and high SOA bias from 25mA to 30mA, at 1556nm of wavelength at bit rate of 2Mbps. [Note: the photodiode inverts the sign of he pulse].

Fig.6.12b (mode +2) shows that the bits ‘0 0’ have a hole, probably due to fibre reflections or because the overlap of the external wavelength not coincides that of the SQDL. Other possible reasons are the synchronisation that sometimes is difficultly obtained, an accurate bias of the MZM and a good polarised signal at West and East side. Generally, the mean success of the experiment with better ER switchings has been done when the mode to injection coincides with the m-2 and m+2 of the SQDL but it was often hard to acquire. In Fig. 6.13 is shown a good example of high ER switchings.

Other better measurements have been acquired by low biased SOAs of 4mA and ring current around 190mA, the result is shown in Fig.6.13 that I showing better ER. This behaviour let the interest to measure the Bit Error Rate (BER).

10ns

01

Figure 6.13. SRL switchings using a word of 4 bit length ‘0101’, injecting at the same mode +2 with ring current of 190mA and SOAs of 4mA. Bit rate of 2Mbps and bit period of 10 ns (3G_BatchF_device 4). Further measurements have been taken at bit rate of 200MHz and 1900MHz. Where the injection of Set-Reset pulses required to produce the SRFF functions must be with high ER. The follow figures shown the shape wave of the injection Ser-Reset pulses at 200MHz, these pulses are appropriately complementary with an acceptable ER.

0 10 20 30 40 50-40

-20

0

20

40

60

80

100

120

Time [ns]

Sig

nal

[mV]

File: COUPLER_0011_200MHz.txt

0 10 20 30 40 50-40

-20

0

20

40

60

80

100

Time [ns]

Sig

nal

[mV]

File: COUPLER_0101_200MHz.txt

(a) (b)

0 10 20 30 40 50-100

-50

0

50

100

150

Time [ns]

Sig

nal

[mV]

File: DELAY_0011_200MHz.txt

0 10 20 30 40 50

-60

-40

-20

0

20

40

60

80

100

120

Time [ns]Sig

nal

[mV]

File: DELAY_0101_200MHz.txt

(c) (d)

Figure 6.14. Set-Reset injection pulses of 4 bit length generated at 200Mbps. (a) Sequence 0011 and (c)complementary, (b) Sequence 0101 and (d) complementary. Fig.6.14 shows the Set-Reset injection pulses, collecting the light signal at West in D and East in E sides into the setup after the coupler and delay (see Fig.6.9). In Fig.6.14a is remarkable the amplitude difference between the first ‘0’ and the second one, this effect is probably due to the MZM bias or not enough light polarisation. The experiment has been carried out on devices of 150µm and 200 µm of radius, with high output power. The results of SRFF operation shown in Fig.6.15 reveal the switching response of the device functioning as a SRFF injecting the Set-Reset pulses shown in Fig.6.14. The Fig.6. 15 shows the commutation behaviour of the 3G_bar F_ device 4 of the bar F using Set-Reset pulses, Fig.6.15a shows a short bit length of ‘1010’ with their complementary ‘0101’ and Fig. 6.15b shows a SRBS of 27-1. These results have been taken with coupled output power nearly to -8dBm at ring current of 178mA. The experimental results exhibited an error-free operation for 3 minutes with Bit-Rate of 200 Mbps.

0 10 20 30 40 50-30

-20

-10

0

10

20

30

Time [ns]

Sig

nal

[mV

]File: F_04_0101_200MHz_1559nm_-9dBm.txt

0 10 20 30 40 50

-20

-15

-10

-5

0

5

10

15

20

Time [ns]

Sig

nal

[mV

]

File: F_04_PRBS_7POS_200MHz_1558nm_-12dBm.txt

(a) (b)

0 10 20 30 40 50-30

-20

-10

0

10

20

30

Time [ns]

Sig

nal

[mV

]

File: F_04_1010_200MHz_1559nm_-9dBm.txt

0 10 20 30 40 50

-20

-15

-10

-5

0

5

10

15

20

25

Time [ns]

Sig

nal

[mV

]

File: F_04_PRBS_7NEG_200MHz_1558nm_-12dBm.txt

(c) (d) Figure 6.15. 3G_bar_F_Device 4 operating as a Optical SRFF. Injecting Set-Reset pulses with Bit-Rate of 200Mbps at 1559nm of wavelength and SOAs bias of 11mA at West and 16mA at East. (a) Word of 4 bits length (0101) and (c)the complementary. (b) PRBS Sequence of 27 -1 and (d) the complementary. The SRL output was around from -9dBm to -12dBm. Fig. 6.16 shows the SRFF operation for a bit symmetrical sequence of 4 bits ‘1100’, where for this kind of word the FF response wave shows better shape. It is evident that the Set and Reset Flip-Flop operation is going on with good enough ER even for SRBS Sequence, these results of have been observed with large error free letting the possibility to BER measurements at higher Bit-Rate. The experiment was repeated for PRBS sequence of 27 -1, increasing the Bit-Rate until 1900MHZ and the time repetition of 200ns, the result is shown in Fig.6.17, exhibiting satisfactory SRFF operation.

0 10 20 30 40 50-60

-50

-40

-30

-20

-10

0

10

Time [ns]

Sig

nal

[mV]

File: F_04_159mA_1559nm_1100_200MHz.txt

(a)

0 10 20 30 40 50-60

-50

-40

-30

-20

-10

0

10

Time [ns]

Signa

l [m

V]

File: F_04_159mA_1559nm_0011_200MHz.txt

(b)

Fig.6.16 3G_bar_F_Device 4 exhibiting Set and Reset FF response using a symmetrical word of 4 bit length ‘1100’ and Bit-Rate of 200 Mbps. Ring current of 159mA and SOA bias of 11mA. Injection on 1559nm. (a) Bit Sequence ‘1100’. (b) Complementary bit Sequence ‘0011’.

(a)

(b)

Fig.6.17 3G_bar_F_Device 5 exhibiting Set and Reset FF response using PRBS sequence of 27-1 and Bit-Rate of 1900 Mbps. Ring current of 169mA and SOA bias of 11mA at west and 16mA at East. Injection on 1569nm. (a) PRBS positive. (b) Complementary PRBS.

After to try hardly to carried out the experiment the best result has been done with a Bit-Rate of 1Ghz measuring error free time, this result is shown in Fig.6.18. it is exhibiting good SRFF operation using PRBS sequence of 231-1.

0 5 10 15 200

0.5

1

1.5

2

2.5

Time [ns]

Sig

nal

[mV]

Fi le : 1GHZ-PRBS-POS.txt

Fig. 6.18. Output of the SRL Set-Reset Flip-Flop device (NRZ signal) at Bit-Rate of 1Gbps. Red trace: output for positive Set and Reset data. Blue trace: output for negative (complemented) Set and Reset data. At this point the Bit-Error-Rate (BER) has been measured from the best device (3G_Batch F_device 2) when it was repeatedly and alternatively triggered by optical Set and Reset pulses in the form of a PRBS sequence, using a smaller device with radius and length of 200µm and injecting pulses at Bit rate of 1Gbit/s with 231-1 PRBS, the error-free operation was for times longer than 5 minutes, corresponding to a BER of 3.10-12. The figure 17 shows the SRFF output signal for two cases by the smaller device, when the Set and Reset data streams were complementary. The SRL exhibits complementary output and good aye opening can be inferred. The successful experiment bring the possibility to increase the Bit-Rate until 2.5GHz. Fig.6.19 are showing a SRFF operation using the same ring and SOA currents conditions of the experiment shown in Fig.6.17. Injecting Set-Reset pulses of 6 bit duration during 20ns the error free was sometimes around 10 minutes, the response of the SRFF device shows good ER for short bit sequence and repetition time of 10ns shown in Fig.6.19a and 20ns shown in Fig.6.19b.

(a) (b)

Figure 6.19. Set-Reset Flip Flop operation response injecting 6 bit length using a Bit-rate of 1Gbps. SOA bias of 16mA at high dc current. Set-Reset sequence of 6 bits ‘111000’. (a) SRFF time repetition of 10ns. (b) SRFF time repetition of 20 ns. Red line (111000), and their complementary (000111).

The experiment at Bit-Rate of 2.5Gbps has repeated using a PRBS sequence of PRBS 27-1. At this Bit-Rate the SRFF device response is exhibiting some burst errors probably due to degradations of the fibre coupling. Fig.6.20 shows the SRFF operations with some signal degradations.

0 5 10 15 20-150

-100

-50

0

50

100

150

Time [ns]

Sig

nal

[mV

]

Fi le : 1GHz_PRBS7POS_SOAs _168m Aring_1565nm .txt

NEGPOS

Figure 6.20. Set-Reset Flip Flop operation response injecting Set-Reset pulses of PRBS 27-1. at Bit-rate of 1Gbps. SOA bias of 16mA at high dc current. SRFF time repetition of 20ns. red line the positive Set FF and the complementary in blue line. 6.2.4 Fast and Ultra-Fast Injection Pulses Before injecting any fast optical trigger into the SRL, some specific experiments have been carried out to investigate the effects caused by the presence of SOAs in the input and output waveguides of the device. In the experiment, a square optical pulse of 3 ns duration with 130 ps rise and fall time was injected into the SRL waveguide. The shape of the pulse emerging from the opposite side of the SRL waveguide was analyzed by means of a 15 GHz photodiode and a 20 GHz sampling oscilloscope. Traces were recorded for different pulse power and bias conditions of the waveguide SOAs. The results are summarized in Fig.6.21. It clearly turns out that the SOAs play a role in reshaping the input pulse depending on the SOAs pump current and the pulse power. This shall be kept in mind for all SRFF and logic gate experiments, as a distortion of the pulses may occur for the trigger pulse that effectively enters into the SRL after propagation through the SOA waveguides, and also for the SRL output signal after being directionally switched it. Is useful remember that the “coupler” that has no electrical contact, and it can only be optically pumped by the two neighboring waveguide SOAs. A very critical phenomenology was observed for input powers around the value that sets the saturation power and/or saturation energy of the SOAs (these parameters also depend on the SOA bias). The results can be summarized as follows: The major effect is played by the unpumped “coupler” SOA, that tends to act as a saturable

absorber and the device shows the typical behavior. When the waveguide SOAs pumping is low, and the input signal is weak, the signal experiences absorption until the coupler SOA is saturated. This causes a slow upward transient (1 to 3 ns), as shown in Fig.21a, by injecting 1.8mW with bias SOAs of 8mA, at lower injection power and more bias current of SOAs the upward transient peak decrease.

When the waveguide SOAs pumping is large, the coupler SOA is optically pumped, and no longer absorbs the signal. Rather, gain saturation effects occurs, and the output signals exhibits a peak followed by a rapid decrease, on a time scale of 200-300 ps.

When the input pulse is strong, its distortion is reduced, except for the rapid falling edge of the initial peak as illustrate in Fig.6.21c by injecting high power of 3mW and biased at moderate or high current SOAs of 14mA. The initial peak decrease when the injected current decreases and the current SOAs increase.

In general, the interest is in: i) injecting a fast trigger pulse into the SRL, and ii) observe a possibly fast response of the SRL to the injected pulse. Hence, the best conditions are met when the SOAs are strong forward biased and the injected pulse is strong. For all SRFF experiments the SOAs have been forward biased with at least 15 mA each.

Injection Power and SOA current increases

(a) (b) (c)

Figure 6.21. Effect of waveguide SOAs on the shape of a trigger pulse injected into the SRL waveguide with 3 ns duration and 130 ps rise and fall times, with the SRL biased below threshold (I = 37 mA) and identical East/West SOA currents. Time scales are in ps. Injecting SOA current increasing from left to right: a) Moderate input power, low SOA bias: P = 1.8 mW, I = 8 mA; b) Moderate input power, moderate SOA bias: P = 1.8 mW, I =12 mA; c) High input power, moderate SOA bias: P = 3.0 mW, I = 14 mA.

SRL

inj,peak,WG

SOA inj,peak,WG SOA

inj,peak,WG SOA The fast optical trigger pulse is depicted in Fig.6.22, it has 400 ps duration and 130 ps rise and fall times. It is generated by a MZM driven by a 500 MHz square wave generator. When this pulse was used as Set and Reset trigger signal, it is bserved proper SRFF operation. The time traces are shown in Fig.6.23, where it is clearly visible in (a), the optical trigger pulse that passes through the SRL output waveguide (optical trigger signal - soft line), the total signal before subtraction of the injected pulse. This could be eliminated by looking at the output port located on the opposite side (East/West and North/South). As the SRL response time is now comparable to the optical trigger rise time, by subtracting the trigger pulse it was obtained the pure SRL response (thick line). The rise time is 150 ps, while the fall time is 100 ps. In (b) is illustrated the falling edge of SRL output, yielding a fall time of 100 ps. This fall time could be measured directly without post-processing, as the trigger pulse that switches off the observed direction does not appear at the same output waveguide.

Figure 6.22 a) Schematic representation of the “fast” pulse of 400 ps duration, that was generated by a MZ modulator driven by a 500 MHz square wave generator. b) Time trace of the optical trigger pulse as acquired by a 15 GHz photodiode and a 20 GHz sampling oscilloscope. The pulse has 130 ps rise and fall time. The SRFF time delay could not be assessed for this step-like optical trigger, because the rise time of the trigger was comparable to the FF rise and fall time, and the delay measurement was not enough precise. The rise and fall times are much shorter than those obtained with the slow trigger. As already explained, this is due to the fact that the fast trigger injects a large amount of photons in the new direction over a short time interval, thus helping in reducing the carrier lifetime, and speeding up the switching mechanism.

Figure 6.23. Time-domain traces for the SRFF function with fast 400 ps optical trigger. SRL current = 160 mA; SOA current = 17 mA, peak trigger power injected into SRL waveguide = 1.2 mW. a) rising edge of SRL output, with subtraction of trigger pulse, yielding a rise time of 150 ps. b) falling edge of SRL output, yielding a fall time of 100 ps. The ultrafast optical injection pulse is generated by an Optical Parametric Oscillator (OPO), with a FWHM time duration of 5 ps and a repetition frequency of 200Mhz, hence the pulses are separated by 5ns. For the BER experiments, the clock is generated by the electrical trigger of the OPO and the Set/Reset signal is generated by PPG using the electrical trigger of the OPO connected externally. The behavior of the pulses is schematically represented in Fig.6.24, in (a) the OPO pulse train before being modulated; (b) shows the OPO modulated by Set/Reset signals. In Fig 24.c is shown the expected SRFF operation at West and East side.

400 ps

200 nstime [ns]

(a) (b)

rise/fall time130ps Set/Reset delay 100ps

(a) (b)

Figure 6.24. a) Schematic representation of the “ultrafast” pulse generated with OPO. b) Set/Reset injection pulses. The FWHM duration is 14 ps. c) SRFF operation at West and East sides. 6.2.5 Experimental Setup to Generate the Ultrafast Injection Pulse

Figure 6.25. Sketch of the experimental setup employed to demonstrate the SRFF using ultrafast trigger pulse of 5ps. The OPO has two cavities and can works at 1064nm, the laser light of the mean cavity is used to generate a clock signal by the help of a photodiode, this clock must to be adjusted to obtain a good

OPO

SET

RESET

SRL East

SRL West

(a)

(b)

(c)

5 ps

t [ns]

5 ns

square shape by a pattern generator with 5.3 ns of period and 2.6 of width because in this way can be use as an external clock of the PPG. The OPO signal can be tuned into a range of wavelength between 1500 nm to 1620 nm, this operation is do in the second cavity and handily adjusted by temperature variations. The experimental setup sketched in Fig.6.25. and shows that the OPO light is focalised by an optical fibre coupler and divided by an 50:50 coupler in two pulses. To test SRFF operation triggered by ultrashort optical pulses, the optical pulses are injected at different time instants from the two opposite directions by pulses of 5ps of width generated by the OPO. The two OPO pulses are sent through two arms of the setup and modulated by MZM to generate the Set/Reset injection pulses. After the modulators, the length of the optical path must be the same to avoid the bad synchronization of the injection pulses that produce a bad FF operation. The modulation is controlled by output signals of PPG using DATAand DATA . Following the setup, each arm has two P.C. to adjust the ER of the pulses before being injected into the device waveguides. Finally two circulator at both sides West and East allow to inject the pulses and collect the light output from the SRL. The signal outputs from the two circulators are revealed through a fast photodiode connected to a sampling oscilloscope, or sent to the error detector for BER measurements.

6.2.6 Results of the All-Optical Flip-Flop Operation by PPG and Ultrafast Pulse

The experiment using ultrashort optical trigger pulse is interesting for two reasons: to check whether the SRL SRFF can be triggered by such a short pulse; and in case of positive response to the previous point, to analyze the time dynamics stimulated into the SRL. After proper alignment of the set-up, the SRFF operation using the 5 ps pulse as optical trigger separated 5 ns was successful by using short words of 6 or 8 bits. The sample time-domain traces for west and East directions are shown in Fig.6.24. The operating conditions in which the SRL can be operated as SRFF triggered by the 5 ps pulse are the following: Accurate forward bias of the waveguide SOAs, optimizing the current between 15 to 20 mA. SRL currents ranging from 80 to 270 mA. As the spectral width of the trigger Gaussian shaped

pulses larger than the mode spacing of the SRL (1.2 nm vs. 0.6 nm), there is no need to tune the optical trigger to a specific longitudinal mode of the SRL. This is a great advantage of the use of short pulse trigger. Correct SRFF operation was observed for detunings between the free-running lasing wavelength of the SRL and the peak wavelength of the OPO as large as 6 nm, with a preference for the situation where the OPO lied on the long wavelength side of the SRL.

The polarization of the trigger pulses has to be slightly adjusted before being injected into the SRL by P.C west and East 1 illustrate in figure ?? of the setup, but the experiment exhibited a large tolerance.

A typical value for the peak pulse power injected into the SRL waveguide is around 30-100 mW.

Successful SRFF operation have been observed on 3G devices of the batch B. The optical clock pulse power before being modulated to Set/Reset was around 30mW, the Set/Reset injected pulses exhibited good shape and good ER. The injection Set-reset pulses have a short and symmetrical bit sequence of 6 bits (111000) generate by PPG with a low Bit-Rate of 180Mbps, the results are illustrate in Fig. 6.26 showing the switch operation of the device for different repetition time.

(a) (b) (c)

Figure 6.26. SRFF operation of the SRL triggered by 5 ps optical trigger. The Set-Reset injection pulse is a word of 6 bits length ‘111000’ with Bit-rate of 180Mbps. For different repetition time of: (a) 50 ns, (b) 100ns, (c) 200ns. The device was tested under 18°C until 25°C the results have been obtained with ring current of 177mA and biased SOAs (IEast of 17.8mA and IWest of 18.8mA). The SRFF device response has been acquired at both West and east sides of the device showing correctly SRFF operation. Fig.2.27 shows the switch of the device taking the signal from West side. the peaks and low ER probably is due to photodiode effect.

(a) (b)

Figure 6.27. SRFF operation of the SRL. The Set-Reset injection pulse is a word of 6 bits length ‘111000’. For different repetition time of: (a) 100 ns, (b) 200ns, (c) 200ns. Fig.6.28 shows the SRFF behavior of the device getting only the first 3 bits of the bit word. The bit period is around 15ns.

(a)

(b)

1 1 1

0 00

15 ns

Figure 6.28. Complementary West/East SRFF device operation by the 5 ps optical trigger and 180Mbps Bit-rate. With 3 bits the repetition time is 15ns.

The experiment has been repeated for a word length of 8 bits ‘11001010’, increasing the ring current to 189mA, whit SOA bias (IEast of 17.8mA and IWest of 18.8mA). The result is illustrates in Fig.6.29.

Figure 6.29. SRFF device operation triggered by the 5 ps optical trigger using a word length of 8 bits (11001010). The SRFF rise and fall time as well as the response delay of the device by 5ps injection pulse have been acquired by a 1 GHz photodiode and 40 GHz oscilloscope, and is illustrated in figure 25, where the rise time reach 300 ps and fall time of 500 ps. All the measurements have exhibited a error free long as 10 minutes at low Bit-Rate of 180Mbps. In conclusion, preliminary measurement confirms that the SRFF can be operated at high frequency (5 Gb/s). The experiment can be repeated by short injection pulses at high Bit-Rate.

(a)

trise

tfall

(b)

Figure 6.30. Time-domain traces showing the rising and falling edge response of the SRL SRFF to 5 ps optical trigger pulse. Device 6_ Batch_B.

Part 2

SEMICONDUCTOR QUANTUM DOT LASERS

7. The State of the Art of the Quantum Dot Lasers

Before dealing with the subject of Semiconductor Quantum Dots Lasers (SQDL), it is useful to understand what a Quantum Dot (QD) is a material of few atoms implicating a small size, which can thus exhibit their own individual behaviour showing interesting characteristics. In QD materials, electrons act discretely thanks to the quantization of energy and 3-dimensional confinement of movement. During the past few years, research in semiconductors has developed nanofabrication technologies capable of creating structures almost atom by atom, opening the study and the development of QD semiconductor lasers. One such example is illustrated in Fig.7.1.

In [Indium] As [Arsenide]

Pump current

Laser light

Electrode

Electrode

p-doped QD active

InAs QDs

p-type GaAs

(a) (b)

Figure 7.1. The QD laser. (a) QD structure based on InAs compound, (b) QD Laser structure. Recently, semiconductor QD laser devices have been fabricated with the goal to exploit their quantized emission energy and their particular features, like a weak temperature sensitivity and symmetric gain function, representing a concrete possibility to enhance the functionality of the devices in optical system applications. The development of QDL has been done in the last decade, with QD cavities that due to the tight confinement of carriers can exhibit an electronic structure similar to atoms that avoid some of the negative aspects of device performance associated with traditional semiconductor lasers. Thus, room temperature and continuous-wave operation lasers with a high modal gain, low lasing threshold current and high output power were expected. Improvements in modulation bandwidth and linewidth enhancement factor have been observed. The QD cavities may also be engineered to operate at different wavelengths by varying dot size and composition, an opportunity that is not possible using conventional bulk or Quantum-Well semiconductor laser technology. Recently, as in Fig.7.1, Self-Assembled technologies based on InAs/InGaAs QDs have been developed, finding commercial application in medicine, display technologies, spectroscopy and telecommunications. In the last year 1.3 μm-lasers with high-speed operation, insensitive to temperature and working at 10 Gbit/s (ideal data transmission in optical LANs and metro-access systems) have been demonstrated.

8. Theoretical Behaviour of the Quantum Dot Lasers

8.1 Introduction Since the invention of semiconductor lasers, huge improvements in device performance have been achieved, and a large variety of specialized designs for different applications were conceived. Two major steps have played a key role in the improvement of device properties. The first step was the application of semiconductor heterostructures that allowed to separate the optical and carrier confinement. The second step was the introduction of quantum films, also called quantum wells, in the carrier recombination zone permitting a strong reduction of threshold current due to an increased density of states at the laser energy. Further quantum mechanical effects in semiconductor lasers have been improved by the fabrication of a laser material with quantum mechanical carrier confinement in all three dimensions, called quantum dots. The breakthrough came with the application of self-assembly techniques on quantum dots during the material epitaxial growth. A quantum dot is a nanocrystal that can be made from a semiconductor nanostructure that confines the motion of electrons and holes in all three spatial directions. The confinement can be due to electrostatic potentials (generated by external electrodes, doping, strain, impurities), interfaces between different semiconductor materials, presence of a semiconductor surface (e.g. semiconductor nanocrystal), or a combination of these. A quantum dot has a discrete quantized energy and the corresponding wave functions are spatially localized within the dot and extend over many periods of the crystal lattice. The small size of the materials using in the quantum dot fabrication behave differently, giving to the quantum dots an unprecedented tunability and possibly enabling new applications in science and technology. An other important feature of the quantum dots is their spectral and spatial separation of the gain. Therefore it is possible to expect that they cover a wider wavelength range, bringing the possibility of emission and amplification at different wavelengths simultaneously.

8.2 Properties of the Semiconductor Quantum Dots In bulk semiconductor, the energy levels are very close together, so close that they are described as a continuum, meaning there is almost no energy difference between them. The bandgap of the bulk is fixed, therefore the transition when the electron is falling back across the bandgap results in fixed photon emission frequencies. The quantum dot has a different level separation between electrons and holes for each material; this can offer the unnatural ability to tune the bandgap and hence the emission wavelength. In quantum dots it is possible to make the size of a semiconductor nanocrystal small enough that it approaches the size required to make the electron energy level become discrete. Those discrete energy levels imply that the addition or subtraction of just a few atoms in the quantum dot alter the boundaries of the bandgap, changing the geometry of the dot surface and their size, causing also changes of the bandgap energy and the effects of quantum confinement. The size of the bandgap is controlled simply by adjusting the size of the dot; therefore the emission frequency is linked to the dot size with extreme precision.

8.2.1 The Quantum Dot in a Laser

The unique property of zero-dimensional or 3D carrier confinement, in theory, make the quantum dot lasers have a high optical gain and a symmetric gain curve as shown in Fig.8.1. It is a high efficiency material, because most of the carriers generates photons at the same energy. They should exhibit low threshold current due to their weak temperature dependence compared to quantum-well devices; as a consequence the Peltier cooler is not required. Because of their symmetrical gain, the quantum dot lasers can show a low chirp, i.e., low shift of the lasing wavelength with injection current. This property is ideal for a low linewidth enhancement factor (or commonly know as α-factor) because there is not variation of refractive index at the maximum of the gain, a low α-factor means a low sensitivity to optical feedback. As the sensitivity to optical feedback scales is proportionally to √1+α2. Thus, QD lasers can be operated without the optical isolator, implying a great reduction of the package complexity and cost.

With QD materials it is possible to fabricate lasers with high speed at a low cost. As shown in Fig.8.1, light emission at 1.3µm wavelength can be achieved using a substrate of GaAs, that is more cheapers than InP. Unfortunately, using quantum dots the fabrication of lasers at 1.5µm useful in optical telecommunications is more difficult.

1.2 1.71.41.3 1.5 1.6 λ [µm]

g [cm-1]

10-4

10-3

10-2

Bulk

Wire [10nm]

Well [10 nm2]

Dot [nm3]

Figure 8.1. Comparison of the curve of material gain vs wavelength between Quantum Wire. Well, Dot and Bulk Semiconductor.

8.3 Structure of the Quantum Dot Laser

Quantum dots are also known as nanocrystals, they are a special class of semiconductor composed of periodic groups of II-VI, III-V, or IV-VI materials. The quantum dots are an unique class of semiconductor because they are so small, ranging from 2 to 10 nanometers (10-50 atoms) in diameter. The devices based on QD semiconductors are energetically working as shown in Fig. 8.2, where there is a comparison between a bulk and a QD energy behaviour.

Figure 8.2. Energy-band difference between the Bulk and the Quantum Dot Semiconductor. Conduction Band (CB), Conduction Energy (CE), Valence Band (VB), Valence Energy (VE). Increased Exciton Bohr Radius (EBR) for Quantum Dot Semiconductor. In bulk semiconductor material, an extremely small percentage of electrons occupy the CB, the majority of electrons occupy the VB filling it almost completely. The only way for an electron in the VB jump to the CB is to acquire enough energy to cross the big bandgap, and most electrons in bulk simply don’t have enough energy to do so. Applying a stimulus such as heat, voltage, or photon flux can induce some electrons to jump the forbidden gap to the CB leaving a temporary "hole" in the VB electron structure. Sufficiently strong stimulus will cause a VB electron to take residence in the CB, causing the creation of a positively charged hole in the VB. Both electron and hole taken as a pair in the CB are called an Exciton. There is a minimum energy of radiation that the semiconductor bulk can absorb for raising electrons into the CB, that corresponds to the energy of the bandgap. The bangap of a given composition of bulk semiconductor is fixed; this means that there is a continuous electron energy levels as well as the number of atoms in the bulk. The electrons in QD have a range of energies and the excitons have an average physical separation between the electron and hole, referred to as the Exciton Bohr Radius (EBR). This physical distance

EBR VB

CB

Band Gap

EBR

CB

VB

CE

VE

E1

E2

CE

VE

E1

E2

Continuous Energy

Discrete Energy

Bulk Semiconductor

Quantum Dot Semiconductor

Band Gap

is different for each material. In bulk, the dimensions of the semiconductor crystal are much larger than the EBR. To see the difference between the bulk and the QD semiconductor material is necessary to think in dimensions. For example, the bulk is much longer than 10nm, then the EBR are very close together, as shown in Fig. 8.2. However, if the size of a semiconductor crystal becomes small enough that it approaches the size of the EBR, the electron energy levels can not be treated as continuous but must be treated as discrete, meaning that there is a small and finite separation between energy levels. This situation of discrete energy levels is called quantum confinement, and under these conditions, the semiconductor material instead can be called a QD. In this way the carrier confinement has 0-degrees of freedom and it is possible to think about it as a bulk of semiconductor material closed into a box, this mechanism is illustrated in Fig. 8.3.

Figure 8.3. Confinement and degrees of freedom of. (a) Quantum Well. (b) Quantum Wire. (c) Quantum Dot. Keeping in mind all the properties of lasers based on QD, this part of the work is dedicated at to static and dynamic behaviour of QD lasers. 8.3.1 Self-Assembly of Quantum Dots Self-Assembled means spontaneous organization of the nucleation under certain conditions during the Molecular Beam Epitaxy (MBE) procedure and Metallorganic Vapor Phase Epitaxy (MOVPE), when a material is grown on a substrate to which it is not lattice matched. The result are coherently strained islands on top of a layer that can be subsequently buried to form the QD. The main limitations of this method are the cost of fabrication and the lack of control over positioning of individual dots.

In Fig. 8.4 is schematically represented how is possible to create an individual dots, using from two-dimensional electron or hole present in doped quantum wells or semiconductor heterostructures. The sample surface is coated with a thin layer of resist, a lateral pattern is then defined in the resist by Electron Beam Lithography. This pattern can be transferred to the electron or hole by etching or by depositing metal electrodes that allow the application of external voltages between the electron and the electrodes. Such Self-Assembled technique is mainly of interest for experiments to create different applications based on QDs.

(a)

1D Carriers Confinement

Planar Propagation Direction

1

2

3

(b) (c) 1

3

2

2D Carries Confinement

Propagation Direction

1

2 3

3D carrier confinement 0D dimensionality

8.4. Self-Assembled Technique. (a) Deposition layers, (b) definition by lithography and (c) transfer of pattern by etching or electrodes. The realization of device-quality based on QD structures became possible by the introduction of self-organized growth using both MBE and MOVPE techniques, because the deposition of a fraction of an atomic monolayer can be controlled. The variation of the dot size and geometry is controlled by the temperature during the formation process. A thermally activated variation of the dot size and geometry occurs during the formation process due to high deposition temperatures. Therefore this process must be controlled. The variation of the dot size leads to a broad distribution of transition energies and huge number of dots necessaries in the amplification process of the propagating wave in optical structures as the laser, amplifiers and so. Then, in order to start the study about static and dynamic characteristics of a SQDL, the following step is to describe the design of the devices used. 8.3.2 Design and Fabrication of the used Quantum Dot Lasers The devices used in the experiments were fabricated using Self-Assembled technique by the NNL National NanoTechnology Center - CNR, Lecce, Italy. The SQDL are positioned in a bar, those devices are pictured and described in the following figures. In Fig. 8.5 (a) it is pictured a bar of 5 devices, each one provides of 3 square electrical contacts and two mirror at the end facets.

Bar 4E

Characteristics: Type: Fabry–Perot Laser Operation: Multilongitudinal – Mode Material: InAs/GaAs Dimensions:L2 = 600µm

W1 = 2µm

Figure 8.5 (a). Description of the E-type bar of QDL with 5 devices.

In Fig. 8.5 (b) it is pictured the second bar of SQDL with 2 devices which 5 square electrical contacts.

Resist layerelectrode

+

-

+

-

+

-

+

-

+

-

+

-

+

-

+

-

+

-V

electrode

Resist layer

Depositing Layers

SC heterostructure

Lateral PatternPattern transferred to e-, holes

Etching Lithography

1:

2:

(a) (b) (c)

Bar 1E

Characteristics: Type: Fabry–Perot Laser Operation: Multilongitudinal – Mode Material: InAs/GaAs Dimensions:Device 1: Device 2: L1 = 1mm L1 = 1mm W1 = 2 µm W2 = 1.5 µm

Figure 8.5 (b). Description of the E-type bar 1E with 2 devices.

Those prototypes have been fabricated in two different laser cavities lengths L1 and L2, and two different width W1 and W2, the size of the E type are around 600µm with 5 devices named 4E and 1mm with 2 devices named 1E, according to pictures 8.5 (a) and (b). To characterize these devices it was necessary to sold them onto a copper support. The size of the support is comfortable to include a semiconductor thermistor of 10kΩ and the ground contact. The device are put into the micro-setup pictured in Fig. 8.6. The SQDLs are fixed and positioned on the support. The following sub-chapter contains the procedure of static characterisation based on QD Lasers.

8.4 Static Characterisation of the Quantum Dot Lasers

After sticking together the bar of devices with the support by the conductive epoxy, a set-up for temperature control similar to that used for the SRLs was implemented.

Electrical contact

lensedfibre support

Thermistor Probe

Lensed fibre

Devices

(a) (b)

Figure 8.6. Picture of the quantum dot devices onto the setup. (a) Thermo-setup with devices positioned. (b) QDL device close together to the lensed optical fibre to collect the light.

In Fig.8.6a. it is shown the microprobe, that is positioned by a micrometric screw. When the microprobe is on the square electrical contact of the QDL, it can be forward biased with dc current. Is necessary to be careful when the microprobe is put on the small square because it can damage the device. To avoid damage a microscope was used. The static characterisation has been carried out using a large area photodiode and lensed optical fibres, that focalises the light optimising the coupling. The fibre is manipulated by a micropositioner with the goal to arrive close to the end-mirror facet of the laser avoiding to damage it. The QDL waveguide is not tilted, so the fibre has been positioned straight in front of the facet. The maxim power emitted in air by the QDK amounts to 3-10mW. The coupling efficiency to the optical fibre is around 30% and a maximum power of 2mW can be launched into the fibre.

8.4.1 Experimental Setups of Static Characterisation All the static characterisations on QD devices have been carried out using the thermo-setup. The first static characterisation measurement has been carried out using a large area photodiode to collect the laser light, this method is easily implemented and is illustrated in Fig.8.7. Whit this method the L-I curves of the QD Lasers are measured observing the threshold and maximum de current biased. The goal was to verify the temperature independence and low threshold current values varying the temperature.

Figure 8.7. Set-up of measurements of L-I curves of quantum dot lasers using a large area photodiode. The second set-up of static characterisation measurement is illustrated in Fig. 8.8, and it has been implemented using lensed optical fibres with 5µm of spot to improve the fiber coupling efficiency. This method was carried out with the goal of investigating the spectral properties of the device by the help of the OSA instrument.

Figure 8.8. Set-up to measure the spectra from the quantum dot lasers using lensed fibre

The following sub-chapter contains the results of static characterisation based on QDL verifying the spectral properties and the low threshold current in continuous wave operation.

Current Source

Temperature ControllerPeltier Cell Termistor

Micro Probe

±

Batch of devices

OSA Lensed fiber

Current Source

Temperature ControllerPeltier Cell Termistor

Multimeter

Micro Probe

PD

±

Bar of devices

8.5 Results of the Static Characterisation of QDLs 8.5.1 Light –Current Curves The Fig. 8.9 is illustrating the L-I curves of the first QD device tested by collecting light by large area photodiodes, where the threshold current value results constant around 15mA by changing the temperature from 23°C to 31°C, confirming the low temperature insensitivity of the QDL. The device shows thermal roll-over above 60mA. The QDL emits a few mW in air.

Figure 8.9. L-I curves of the device 4th - bar 4E of Semiconductor Quantum Dot Lasers, changing temperature. The static characterisation measurement on the bar type 1E has been carried out also using a lensed optical fibre; the results are shown in Fig. 8.10 (a) for device 1 and (b) for device 2, revealing the threshold and maximum current before roll-over. The stumble-line of the P-I curves is caused by the fibre that is sensitive to the external vibrations.

(a) (b) Figure 8.10. L-I curves for 1E bar of QDL using lensed fiber exhibiting a threshold current around 15mA. (a) device 1, (b) device 2.

8.5.2 Spectral Measurements The spectral properties of the QDL have been measured by a high resolution OSA using the lensed optical fibre. Fig. 8.11 shows the contour spectra exhibiting multi lasing modes around 1.3 μm wavelength up to 19 lasing modes are observed simultaneously in currents values between 40mA to 45mA.

Figure 8.11. Spectral behaviour of the 4th device bar 4E of QDL.

Fig.8.12 shows the spectra of both QDL of the bar 1E, exhibiting multiple lasing longitudinal modes. For both devices are the zoomed contour shows the mode spacing that is resulting around 0.23 nm.

(a) (b)

0.23nm

Figure 8.12. Spectra of the batch 1E of QDL. (a) Device 1(b) Device 2. The static characterisation of the devices of type E has been carried out as follows: pumping dc current until 70mA and temperature 24.7°C, devices of 4E have exhibited threshold currents of 15mA. The bar 1E has been tested at 19°C temperature pumping until 80mA. Both have exhibited low threshold current nearly 13mA. 8.6 Conclusions about the Static characterisation of the QDLs The analysis on semiconductor lasers based on QD has demonstrated that 1.3μm devices fabricated with Self-Assembled InAs/GaAs can operate at very low threshold current due to a high modal gain. At room temperature and continuous-wave operation those QDL reported a good stability over temperature and high laser power around a few mW interesting in optical sources for transmission systems.

9.1 Introduction One important experiment to quantify the performance of optical sources is the frequency response measurements, because it explores the dynamical behaviour of the laser sources. The frequency response is the measure of the response at the output of the system for an input signal of varying frequency, and it is typically characterized by the magnitude of the system response, and is measured in dB versus frequency giving the -3dB bandwidth of the device. In this chapter it is shown how it is possible to modulate the semiconductor quantum dot lasers to obtain the measurement of the intrinsic frequency response by eliminating the unwanted effects of electrical parasites that can alter or limit the response.

9.2 Frequency Response on Quantum Dot Laser Before illustrating the measurements, it is better to describe what is happening into the laser when it is operating in dynamic conditions, i.e. when the laser gain is modified by current modulation or by the injection of an external light; both causes generate a modulation the carrier population into the cavity. In dynamic process, the carrier density N and photon density NP populations exhibit temporal variations while the current pump is changed from below to above threshold value Ith, as theoretically illustrate in Fig.9.13 for different modes of a laser.

1.0

0.8

0.6

0.4

0.2

0.0 2 4 6 2 4 6 2 4 6

N N

Figure 9.13. Sketch of the evolution of the N and NP exhibiting relaxations below and above threshold during the transient response. The figure shows the relaxation oscillation for (a) the mean lasing mode m=0, (b) the modes m=±1, and (c) for m=±2 modes. The turn-on delay td is also shown. In Fig.9.13 it is shown that the photon population exhibits a turn-on delay time td, showing that the stimulated emission does not occur until the carrier concentration has reached the threshold value Nth. The most important feature of the transient response shown is that the N and Np oscillate before attaining their steady-state when the laser has reached the threshold current. This are referred to as ‘relaxation oscillations’ because there is a resonance in the transferred the energy in the system between the carriers and the photons. Each variation of photon population Np(t) is sensitive to small variations in carrier populations N(t). 9.2.1 Relaxation Oscillations In Fig.9.13, the semiconductor laser exhibit damped periodic oscillations before reaching the steady-state. Such relaxation oscillations are due to an intrinsic resonance in the nonlinear laser system. The decay rate of the oscillations can be obtained using the small-signal analysis, because the values of the NP and N populations are perturbed by a small amount of ΔNP and ΔN. The dynamic effects in the laser can be studied by a differential analysis of the rate equations. This analysis is measurable in terms of the calculation of relaxation resonance frequency and the -3dB modulation bandwidth. In the application of an above-threshold dc current I, with a small ac current i into the laser; under steady-steady conditions the carrier density N and photon density NP of the laser would have similarly response. Assuming that the spontaneous emission is neglected, then with mathematical manipulations the frequency domain equations can be saved and the transfer function Np(ω)/I(ω) that represents the relationship between the photon density and the carrier

0(t) N1(t) 2(t)

td td td

time (nsec) time (nsec) time (nsec) (a) (b) (c)

Np0

m = 0

Np1

m = ± 1 m = ± 2

Np2

density into the cavity can be found and the frequency response function denoted H(ω) can be obtained. The direct modulation of the carriers of a laser can be realised implementing the set-up shown in Fig.9.16. The modulation produces a natural resonance into the laser cavity which shows up as a ringing in the output power of the laser in response to sudden changes in the input dc current. This frequency of oscillation is referred to as relaxation resonance frequency ωR (relaxations means the attempt by the photons and carriers to relax to their steady-state values) that depends directly on the differential gain a and the average photon density into the cavity NP0, and inversely proportional to the photon lifetime into the cavity τp, it is represented by the follow equation:

p

pgR

aNτ

υω 02 = (2.1)

The small-signal carrier and photon densities in function of the current can be found, and the modulation transfer function can be expressed as:

( )ωγωω

ωωωω

jISH

R

R

+−== 22

2

)()( (2.2)

Where, γ represents the damping factor. The general behaviour of the H(ω) is shown in Fig.9.14, and it represents essentially a second-order low pass filter with damped resonance appearing near to the cut-off frequency.

Pout

≈ ωP Figure 9.14. Theoretical behaviour of the modulation transfer function by increasing values of relaxation resonance, peak frequency ωP and damping factor γ. The intensity modulation can follow the current modulation up to frequencies near ωR , the damping peak decrease step by step when the pump current and/or the laser power is increased. The peak frequency ωP depends on ωR and γ, as is explained in the following equation,

⎥⎥⎦

⎤

⎢⎢⎣

⎡⎟⎟⎠

⎞⎜⎜⎝

⎛−=

222

211

RRp ω

γωω .

The definition of the modulation bandwidth corresponds to a value where the electrical response attains a value 3dB bellow the response at ω=0, and it can be expressed a: 4422

3 RPPdB ωωωω ++= .

The 3dB frequency can be improved by increasing the photon density or laser power. The so-called k-factor describes the damping of the response, and as such is an important parameter in the characterization of high speed lasers, that depends of the differential gain, photon lifetime into the cavity and the confinement factor. In practice, k and γ0 are used to define the modulation bandwidth capabilities of the laser. The modulation bandwidth can be determined for low damping as well as the possible maximum bandwidth with the follow relationships: RdBf ω76.93 ≈ , with (γ/ωR <<1),

and kf dB88.8

max3 = . The modulation bandwidth increase linearly with the relaxation resonance

frequency and remains about 50% larger than ωR until the damping becomes strong. The optimum damping and maximum bandwidth occur when ωP = 0 and ωR = 3dB seen in Fig.9.14. 9.2.2 Modulation Response One important advantage of semiconductor lasers is that they can be directly modulated by changing the pump current. Another characteristic of the semiconductor laser is that their intensity or Amplitude Modulation (AM) occurs simultaneously with Frequency Modulation (FM), this dependence between AM and FM under direct current modulation is governed by the linewidth enhancement factor or α-factor because the refraction index changes in response to variations in the carrier population, this variation of carriers can be interpreted as a gain and phase modulation into the cavity of the laser.

9.3 Frequency Response Measurements on Quantum Dot Lasers The devices have been tested by dynamic characterisation with the aim to check their frequency response. The frequency response experiment has been carried out implementing an alternative technique. In fact, those devices cannot be current-modulated efficiently up to high frequency because electrical parasitic can strongly alter the intrinsic frequency response of the laser. For this reason the QD laser was optically modulated to achieve the modulation of its carrier density. Therefore, the intrinsic frequency response of the laser can be measured. 9.3.1 Measurement Set-up of Frequency Response based on an All Optical Modulation

Technique It was theoretically analysed that to measure the frequency response is necessary to modulate the carrier density into the cavity. Then, an easy manner is to modulate the pump current at small-signal because in this way the variation of the carrier densities can be produced. Fig.9.16 shows a conventional setup to modulate directly the current.

I

P

Figure 9.16. Setup of electrical modulation used in frequency response.

The small-signal AC is taken from the Network Analyser (NA) instrument by the help of a capacitor C. At the input of the NA it is connected a high speed photodiode. To carry out the alternative technique of laser modulation, it is necessary the calibration of the system to eliminate the residual response of all the others instruments, in this way the frequency response obtained corresponds accurately to the device under test. The procedure of calibration is done by connecting the external modulated signal with the high speed photodiode. This external modulation is used to modulate the QDL optically. This modulation is direct and must be done at frequencies more higher of the relaxation frequency of the QDL, this frequency can be around 10GHz. Whit this procedure is ensure that electrical parasitic elements don’t disturb the optical modulation of the QDL at high frequencies. The calibration of the photodiode is obtained in relationship with the in-out signals of the NA where must be well plate at the reference value. In annexe D is shown this procedure. The lasers are connected using dc electrical microprobes, hence the modulation of the device at high frequencies larger a few GHz becomes very difficult, because the microprobes can irradiate most of

the electrical power. In addition, the electrical parasitics of the QDL also alters the frequency response. Keeping in mind all those drawbacks, an alternative technique has been implemented, basically modulating optically the gain of the QDL by injecting an external light from a Distributed FeedBack (DFB) laser at 1.3 μm that is amplitude-modulated by a high speed 10GHz MZM.

Figure 9.17. Gain modulation procedure on SQDL. (a) Spectra choosing one mode, (b) gain of the QDL vs. wavelength, (c) gain decreased with the DFB injection. Therefore, according with Fig. 9.17, after to have chosen the wavelength of one active mode of the QDL preferably that exhibiting high out power guaranteed the maximal gain, the light from a modulated DFB laser at 1.3 μm is injected into the cavity of the device. In this way some carriers of the QDL are depleted by the additional stimulated emission caused by the injected DFB laser, and the optical output power from the SQDL decreases. When the DFB is amplitude modulated at high frequency it is obtained directly the amplitude modulation of the SQDL. The direct modulation of the carrier density into the active area allows to measure the intrinsic response of the QDL and is only limited by the modulation technique of the external laser. All the setup of modulation is described in Fig. 9.18.

Figure 9.18. Frequency Response setup for QDL using direct modulation technique.

According with the setup illustrated in Fig. 9.18, the DFB was modulated by a MZM in lithium-niobatium capable to obtain high frequencies (f–3dB ≅10GHz). The modulated signal must to be controlled in polarisation before being injected into the QD laser cavity, therefore by the help of two polarisation controller (P.C 1 and P.C 2) the signal can be always maximised, in this manner the efficiency of the internal modulation of the QDL is enhanced. The modulated DFB signal pass through the ports 1 to 2 of the circulator; the modulated signal of the QDL goes through the ports 2 and 3 of the circulator and it is then filtered by a Tunable Filter (TF) to eliminate the unwanted

B

QDL

Temperature Controller

DC Current Generator

1

32

M-Z DFB

NETWORK ANALYSER0-20 GHz

in out

Tunable F-P Etalon Filt er

HIGH-SPEED PHOTODIODE

P.C 1 P.C 2

λQD = 1295nm

λDFB = 1309nm

gQDL

λλQDL

λDFB

λ λQDL

gQDL

(a) (b) (c)

λ ≈1290nm

contribution produced by the DFB signal. To maximise the response of modulation is necessary to center the filter on one active mode of the QDL, at this point the response of modulation is detected by a high-speed photodiode (f–3dB ≅15GHz) that sends the electrical response to the NA with 20GHz of bandwidth. The electrical output has been used to drive the MZM and the input of the NA to capture the electrical signal of the photodiode.

Figure 9.19. Calibration setup

The calibration setup is illustrated in Fig. 9.19. This procedure is required before to starting the real QDL bandwidth measurement because it is necessary to obtain one accurate calibration of the response of the not ideal frequency response from the MZM, photodiode and cables. In the procedure of calibration, the photodiode is positioned after the MZM, and the P.C 2 to enhance the M-Z modulation, the results of this procedure are illustrated in annex B. At this point the measurements of frequency response of the QDL can be carried out, and are shown in the following section. 9.3.2 Results of the Frequency Response on Quantum Dot Lasers The first measurements of frequency response on QDL have been taken on devices of the bar 4E illustrated in Fig. 9.5, maintaining a constant temperature (at the value of 11.83KΩ of the thermistor). The principal variables taken in to account were the follow:

• DFB laser driven at 149mA, so that after the MZM the power was -8dBm at the port 1 of the circulator.

• Power launched by the QDL into the lensed fibre of -1dBm. • Power of QDL after circulator in port 3 -6dBm. • SPAN of the NA between [130MHz to 5.01GHz], and reference level -50dB. • The calibration of the NA -18dB of reference. • The –3dB bandwidth of frequency response of the QDL was around 3GHz.

Therefore at low dc current values of 25mA and 30mA the bandwidth is poor. For currents of 32 to 35mA (red and blue lines) the bandwidth reaches 3GHz. In other case, for lowest and highest values of dc current the bandwidth decreases. This first result is illustrated in Fig.9.20.

P.C 1 P.C 2 DFBM-Z

NETWORK ANALYSER0-20 GHz

in out

HIGH-SPEED PHOTODIODE

Figure 9.20. Frequency response of device 3, bar 4E varying the pump current

Another interesting experiment was the measurements of the frequency response by changing the temperature keeping the same values of dc current. The measurements have been carried out varying the resistance value of the thermistor as follows: 11KΩ, 12 KΩ, 13 KΩ, 14 KΩ, 15 KΩ, by the help of the Temperature Controller. The results are illustrated in Fig.9.21.

(a) (b) Figure 9.21. Frequency response results changing dc current at resistance values of (a)11 KΩ exhibiting 3GHz of bandwidth at -3dB and (b)12 KΩ exhibiting 3.4GHz of bandwidth for highest values of current. At 11KΩ of resistance (22.5°C), the frequency response just after threshold current (25mA) shows a low relaxation frequency with very strong damping. If the pump current is increased the dumping factor increases and the peak of modulation frequency increase as is shown by red line at 30mA in Fig. 9.21a. By decreasing the temperature (i.e., increasing the thermistor at 12 KΩ), the frequency response after threshold was better and is appreciated in the measurements illustrated in Fig. 9.21b, where the peak frequency is highly pronounced (see line blue at 0dB of reference), because at high values of pump current, obviously the relaxation frequency increase. Indeed, measuring at reference response value of 0dB, the bandwidth obtained when is decreasing reporting a measure of 3 GH. Both last experiments induce to decrease more the temperature. Whit the thermistor at 13 KΩ, the peak of relaxation frequency is higher than 2dB reporting a bandwidth value around 3.5GHz.

In Fig. 9.22 is shown the frequency response decreasing the thermistor resistance in 14 and 15 KΩ, the response exhibits bandwidths of 3GHz to 3.5 GHz, at low laser currents of 30mA and 35mA.

(a) (b)

Figure 9.22. Frequency response measurements of the QD FP laser using thermistor resistance of (a) 14 KΩ and (b) 15 KΩ. Driving the laser at 40mA, exhibiting bandwidths nearly to 3.5GHz at -3dB. The results observed in Fig. 9.22, have been carried out by an accurate dc current to obtain a output QDL power of 0dBm (3mW), a bandwidth of 3GHz was reached by 40mA, for highest current values the bandwidth measurement becomes immeasurable. Best bandwidth measurements of 3.5GHz were demonstrated by dc currents of 30 to 35mA at low temperature. In conclusion about the dynamic behaviour of the QDL devices, an uncooled directly modulated laser into the O-band with more than 3GHz of bandwidth has been reported. This value can be good for use in optical transmissions systems. This measurement of frequency response is useful because it allows to determine the intrinsic dynamics of the QDL. It can be interesting to test devices with different active layer structure, to check whether (and how) the intrinsic frequency response is affected.

9.4 Linewidth Enhanced Factor Measurements The linewidth enhancement factor also referred to as α-factor, is an important parameter to understand the dynamic behavior of a semiconductor laser and it affects other fundamental aspects, such as linewidth, modulation-induced chirp, mode stability and so on. Theoretically, the α-factor is defined as the ratio of the partial derivatives of the real and complex parts of the complex susceptibility χ = χr + iχi with respect to carrier density N, expressed in the following equation:

α = −∂χ r /∂N∂χ i /∂N

= −4πλ

dndg (2.3)

where dn and dg are the refraction index and optical gain variations that occur for a carrier density variation dN. This definition is used to assess frequency chirp effects caused by variation of the refractive index. A lot of methods to measure the α-factor have been developed for single longitudinal mode lasers that have a high degree of stability of the mode. Those methods cannot be applied to Fabry-Perot QD Lasers because the response of all the lasing modes are superposed. In this work, it was implemented an alternative technique of α-factor measurement based on the standard technique of Fiber Transfer Function (FTF) that consist to propagate the optically modulated light of the laser under test though a dispersive medium. This method is directly linked to the frequency response measurement method used before, where a direct modulation of the QD laser is not possible due at the electrical parasitic effects and an all-optical modulation of the QD laser light was necessary. The complexity of this alternative technique is related to the narrow mode spacing of the multiple lasing modes, this alternative method has been proof on single mode lasers at 1.5µm but never in multiple modes at 1.3µm. Indeed a filtering of a single longitudinal mode has been required. The filter must be accurate taking exactly a single longitudinal mode, the filter chosen was a F-P in air and the collimation is difficult also it is necessary to collect high values of power to ensure that the power of the external laser not over-cover the power of the QDL. During the modulation of the QD laser, an AM and FM are simultaneously occurring, where the mode frequency of a directly modulated semiconductor laser shifts periodically producing a chirp, this phenomenon is a limiting factor in the performance of optical transmission systems. The QDL should in principle overcome this drawback because there should be not variations in the refractive index thanks to their symmetric gain function. Hence, the α-factor of QDL should be zero.

9.4.1 Fiber Transfer Function (FTF) Method

The interaction of chirped light with fiber dispersion can be used to measure the α-factor from the measured fiber transfer function. With the aid of a NA, the transfer function of a dispersive medium, such as an optical fiber, can be measured. The interaction of dispersion and phase modulation due to the laser chirp when the laser bias current is modulated, will configure a frequency response expressed by the Srinivasan equation [26]:

θαθ sin1cos)( ⎟⎟⎠

⎞⎜⎜⎝

⎛−−=

ff

jfH c (2.4)

with θ=f2πλ2DL/c, where fc is the chirp frequency, accounting for adiabatic chirp, D is the fiber dispersion coefficient, and L is the fiber length. A calibration of the system is made detecting directly the light emitted by the QDL before propagation in the fiber. The dispersive medium is inserted afterwards to get the transfer function of the fiber when the chirped light from the QDL is launched. 9.4.2 Set-up of Measurements of the α-Factor The setup shown in Fig.9.23 was used to carry out the experiment of α-factor measurements. The external DFB laser at 1.3 µm is directly modulated by a MZM, maximizing the signal by P.C. This signal enters the QDL cavity through the lensed optical fiber modulating the gain of the QDL. A variation of the carrier density into the QDL cavity is generated, producing the optical modulation of the gain and of the emitted light. After the calibration and an accurate frequency response, the signal is propagating through a Dispersion Compensated Fiber (DCF) that is connected to the high speed photodiode to compare both responses.

Figure 9.23 Linewidth Enhancement Factor set-up. Due to the multiple modes behaviour of the QDL, It was used a F-P air-filter to avoid the inconvenient of lasing modes superposition. This filter has a FSR of 300GHz, finesse of 40 GHz and FWHM of 7.5GHz that shall be compared to the Mode Spacing of the QDL of 0.25nm (i.e., corresponding to 44 GHz). Thus, it is result convenient to select a single mode of the laser. An OSA

is used to increase the power of the signal after to get a good filtering of a single mode. A picture of the QDL support and the FP air filter is shown in Fig.9.24.

(a) (b)

QDL

Microprobe

Lensed Fibre

Thermal Support F-P Filter

FilterFiber Collimator IN

Fiber Collimator OUT

Figure 9.24. Pictures of the QDL α-factor setup. (a) QDL drive setup, (b) F-P Air Filter to selection of the QDL single mode In Fig.9.25. is shown the typical behavior of the transfer function of α-factor, that for an value 0 of α, the electrical spectra of transmission shows a hole in 7GHz, that for positives values of α, the hole moves at high frequency values and the curve of transmission goes up at maximum of 0GHz before arrives to the hole. For negative values of α, the hole moves at low frequency values and the curve goes down 0GHz before the hole. The frequency response from the QDL after the propagation through the dispersive fiber is analyzed doing a comparison with the theoretical transfer functions of Fig. 9.25. Finally, the extraction of α-factor parameter can be obtained by fitting, using the Srinivasan equation (2.4). The length DCF fiber is 33Km with a dispersion of -50 ps/nmKm and 8 dB of loss.

Figure 9.25 Transfer function describing the behaviour of the α-factor.

9.4.3 Linewidth Enhancement Factor Results

The alternative method used to measure the α-factor is a variation of the FTF method, this was never tested before for 1.3μm multimode QDL. The problems with this method are related to the fact that, in theory, the laser under test must be single mode. So, an accurate filtering was necessary. The narrow mode spacing of the QDL makes the experiment more difficult because it is necessary to select a mode into 44GHz, the filter chosen has 7.5GHz of FWHM and FSR of 300GHZ that almost reach the goal, to increase this filtering is possible to add anther filter with larger FSR of 3000GHz (FWHM of 75GHz). The second problem was the dispersion of the fiber at 1.3μm, a SMF fiber has 0 dispersion at this wavelength, then it was used a DCF with dispersion around -50ps/nmKm. During the experiment, it was difficult to reach the power necessary to acquire a frequency response from the QDL. According of the setup shown in Fig. 2.23, SOAs at 1.3μm have been implemented after the air filter because high power from the DQL is needed before the high speed photodiode to compare the frequency response between the external DFB laser and the QDL. In this way a preliminary result obtained have demonstrated that the alternative technique can work. The RF spectra is shown in Fig.2.26 and is exhibiting a hole before 7GHz and a maximum between 0GHz and the hole frequency, confirming a positive value of α according with Fig.9.25. After, by fitting using the Srinivasan equation (2.4), the value of α extracted is around 1.

D -50 ps/nm.km L 33 Km fc 0.09246427 dB Fiber loss 8 dB Mean error 6.52E-04 α ≈ 1

Figure 2.26. RF Spectra of Optical Modulation and FFT. Exhibiting the typical maximum (nearly to 3GHz), and a hole around 7GHz yielding a positive value for the α-factor.

10. Conclusions In generally, the SQDLs exhibit a typical modulation bandwidth between 3GHz and 5GHz, therefore for our knowledge more than 3GHz of bandwidth have been demonstrated with temperature stability and low threshold current values representing a useful opportunity for optical sources in access and local networks. Improvements in terms of modulation speed on SQDL can be achieved by exploiting p-doping heterostructures of the active region, according with literature the limited performances recorded from QD lasers in terms of maximum bandwidth are attributed to the large damping and gain saturation due to the limited QD density and difficult control of their size homogeneity.