1

PROJECT FINAL REPORT: Publishable Summary

Grant Agreement number: 224211

Project acronym: VISIT

Project title: Vertically Integrated Systems for Information Transfer

Funding Scheme: CP (Collaborative Project)

Period covered: from 01 June 2008 to 31 October 2011

Name of the scientific representative of the project's co-ordinator, Title and Organisation:

Prof. Dr. Dieter Bimberg, Executive Director

Institute for Solid-State Physics, and Center of Nanophotonics

D-10623, Eugene P. Wigner Building, Technische Universität Berlin

Tel: +49 30 314 22082

Fax: +49 30 314 21371

E-mail: bimberg@physik.tu-berlin.de

Project website address: www.visit.tu-berlin.de/

mailto:bimberg@physik.tu-berlin.dehttp://www.visit.tu-berlin.de/

2

Table of Contents 1. Executive summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2. Summary description of project context and objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 3. Description of the main S&T results/foregrounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 4. Potential impact (including the socio-economic impact and the wider societal implications of the

project so far) and the main dissemination activities and exploitation of results . . . . . . . . . . . . . . . . . . 38 5. Address of the project public website, project logos, diagrams or photographs illustrating and

promoting the work of the project, list of all beneficiaries with the corresponding contact names . . . . . 49

3

1. Executive summary The aim of the Vertically Integrated Systems for Information Transfer (VISIT) project, a 41-month Collaborative Project funded by the European Commission’s Seventh Framework Programme (FP7) for information and communication technologies (ICT), is to position Europe as a world leader in the design, production, and implementation of strategic, high-value, affordable photonic component technologies for the next generation of advanced medium- and short-reach optical communication systems such as broadband access and local area networks. The VISIT vision is to develop three types of advanced low-cost, ultrahigh-speed micro-lasers capable of operating efficiently with both low power consumption and at blazing speeds of up to 40 gigabits per second (written as 40 Gb/s, 40 Gbit/s, 40G, or simply 40 Gbps) – 4-times faster per single fiber channel than today’s existing commercial technology. In addition to performing research, development, and characterization activities on novel types of compound semiconductor-based micro-lasers, the VISIT consortium will help to facilitate the swift commercialization of the VISIT technology by actively participating in international Standards bodies that ultimately establish the specifications for the systems that will employ the new VISIT technology, and by producing prototype transmitter optical sub-assemblies to facilitate rapid product development and implementation. The VISIT partners focused on the development of three specific laser diode device types including: 1) 850 nm-range directly modulated (DM) vertical cavity surface emitting lasers (VCSELs); 2) 850 nm-range electro-optically modulated (EOM) Bragg reflector (BR) VCSELs; and 3) 1300 nm EOM tilted wave laser (TWL) edge-emitting laser diodes, and on related standards definitions and system specifications. The work on the 850 nm DM VCSELs was an immediate and huge success, complete with a plethora of new records for data transfer rate, device intrinsic properties, performance at high temperatures, and most critically for energy efficiency. The work on the EOM-based devices proved difficult from all aspect of device development including epitaxial growth, wafer processing, and device characterization. However the partners successfully demonstrated working devices and uncovered new interesting physics such as the TWLs extreme temperature stability and the EOM BR VCSELs highly linear behavior as a single mode device that is well suited to advance modulation formats such as 16-level phase and amplitude constellations or PAM4 (pulsed amplitude modulation) operating schemes. The VISIT partners demonstrated 850 nm DM VCSELs that can operate at up to an amazing 40 Gb/s at room temperature, and these same devices passed a series of initial reliability tests despite their amateur status. In subsequent iterations and during multimode optical fiber (MMF) link tests the 40 Gb/s DM VCSEL were shown to operate “error-free” (defined as less than one bit error per one trillion transferred bits) in both a back-to-back configuration and at over 103 m of MMF. Furthermore the partners produced single-mode and quasi-single mode 850 nm DM VCSELs and demonstrated error free data transmission over a record 603 meters of MMF at 25 Gb/s, and over 1 km of MMF at 17 Gb/s with an incredibly low energy-to-data rate ratio below 100 fJ/bit. These very low power consuming devices have enabled the development of highly energy efficient peta- and zetta-scale high performance computing systems and data centers wherein the optical links typically consume up to an estimated 40% less of the power and the partners have created the exact disruptive technology to allow a continued expansion of Information technology while greatly reducing the energy consumption and costs. For this work the partners received a “Green Photonics” Award for Communications at the Photonics West conference held in January 2012 in San Francisco, CA, USA. Other DM VCSELs had record -3dB bandwidths as high as 24 GHz whereas relaxation oscillation frequencies up to 30 GHz were measured from the small spectral width VCSELs – together implying that operation toward a very bold 56 Gb/s may be possible if device resistance and capacitance can be reduced. Finally in a brilliant device physics analysis the partners found that a purposeful reduction in the VCSELs photon lifetime via a top phase-matching layer and with or without a surface relief feature can lead to greatly improved VCSEL characteristics such as an enhanced slope efficiency and bandwidth, along with a large jump in the possible length of a MMH link. More information including a list of project-inspired publications and links to press releases on the Vertically Integrated Systems for Information Transfer (VISIT) project can be followed on the website of the project leader Technische Universität Berlin at: http://www.visit.tu-berlin.de/ .

http://www.visit.tu-berlin.de/

4

2. Summary description of project context and objective Traditionally information in the form of digital data bits is transmitted over copper-based cables as electrical pulses. The natural trend in global competitive information and communication technologies is to increase this data transfer rate over time since this helps to support a model of sustained economic growth for modern countries in our current Information age. As data rates are now pushing past 10 gigabits per second (e.g. the 10G Ethernet), the copper links must decrease in length due to fundamental electro-magnetics principles and they become excessively expensive to manufacture. As a result, copper links are rapidly being replaced by more cost efficient and reliable fiber optic links. The problem is that these short-reach (generally 300 m or less) fiber optic-based links are limited by the existing micrometer-scale laser technology to operation at about 10G due to severe reliability problems at rates above 10G, whereas the serial link standards for next generation systems are calling for laser modulation rates well beyond 10G – up to aggregated rates of 40G by 2012 and beyond 100G by circa 2015. It is here by our work in the VISIT project that we address this critical need to increase the single channel data rates (to 25G, and then to 40G) of optical interconnects, notably those based on multimode optical fiber over distances of meters to a few kilometers, but at the same time develop a technology that reduces power consumption via improved energy efficiency and novel designs. Herein we describe our accomplishments toward this end. The VISIT project consortium consists of nine highly synergistic groups and five work packages, with each group working toward a common VISIT goal, yet bringing to the collaboration unique skills, perspectives, and capabilities. The team consists of the Technical University of Berlin Germany (TUB, project leader), the University of Cambridge United Kingdom (UCAM), Chalmers University of Technology Sweden (CUT), Intel Performance Learning Solutions Ltd Ireland (INT), the Ioffe Physical-Technical Institute Russian Federation (IOF), Riber S.A. France (RIB), the Tyndall National Institute Ireland (TNI), VI Systems GmbH Germany (VIS), and IQE Ltd. United Kingdom (IQE-e). The Project Objectives for the Work Packages (WPs) are as follows: WP1 Objectives:

• the financial, administrative, and legal management of the project [TUB] • the dissemination of the results [TUB]

WP2 Objectives:

• to coordinate the technical and scientific part of the project [VIS] • to create optimised EO material for EOM-BRs [VIS, TUB] • to realize 850 nm EOM-BR VCSELs, working toward operation at 40+ Gb/s [VIS, TUB] • to compare the EOM-BR VCSEL performance with DM VCSELs [VIS, TUB] • to demonstrate EOM-BR VCSEL based links for very high capacity short reach LANs [VIS]

WP3 Objectives:

• to develop an optimum design of the reference 980 nm edge-emitting tilted wave laser (TWL) exploiting the concept of wavelength stabilization through epitaxial design [IOF, VIS]

• to develop EOM media based on InGaAs-AlGaAs structures scalable to applications at 1300 nm [IOF, VIS]

• to develop 1.3 µm GaAs based wavelength-stabilized edge-emitting TWL [IOF, CUT, UCAM] • to develop 1.3 µm GaAs based TWL with integrated electro-optical modulator (EOM) targeting 40

Gbit/s output light intensity modulation up to elevated temperatures [IOF, VIS, CUT, UCAM] WP4 Objectives:

• to develop 850 nm GaAs-based VCSELs operating at 16-22 Gbit/s up to 85°C with high reliability [CUT, TNI, IQE-e, VIS, TUB, UCAM]

• to demonstrate 850 and/or 980 nm GaAs-based VCSELs operating at 40 Gbit/s [CUT, TNI, IQE-e, VIS, TUB, UCAM]

5

WP5 Objectives:

• development of device and system specifications consistent with high speed optical Ethernet/LAN requirements [INT, TNI, UCAM, VIS, RIB]

• application validation (including CMOS compatibility and integration) [TNI, UCAM] • development of optimized packaging for the high-speed devices [VIS] • evaluation of industrial impacts [RIB, IQEe] • investigations of surrounding technology (e.g. drivers) [TNI] • engagement in standardization activities (to accelerate market introduction) [VIS]

The VISIT partners target the development of three key device types as illustrated in Figure 2.0.1 including; a) the 850 nm-range electro-optically modulated (EOM) Bragg reflector (BR) VCSEL; b) the 1300 nm-range EOM tilted-wave lasers (TWLs) –a type of edge-emitting laser diode; and c) the 850 nm-range directly modulated (DM) vertical cavity surface emitting lasers (VCSELs).

(a) (b) (c) Figure 2.0.1 Schematic diagrams of the: a) EOM BR VCSEL; b) EOM TWL; and c) DM VCSEL. The VISIT partners seek to develop these three types of diode lasers to the point of optimal high-speed and energy efficient performance. The resultant design and performance advancements are targeted to overcome the limitations of modern micro-laser devices that are vital for short-reach (up to 300 m) optical data communication systems and medium to long reach optical telecommunication systems to become future-proof drivers of the European photonics industry. The development of high-speed EOM devices proved to be highly challenging as expected. These devices monolithically incorporate an EOM section (a modulator capable of extremely fast voltage-controlled modulation rates based on the quantum-confined Stark effect) that is vertically integrated with the laser diode section. The two EOM devices are sought as the basis for future access, local area networks (LANs), and storage area networks (SANs) with data rates at and beyond 40 Gb/s. Despite the difficulties the VISIT partners, in our systems testing demonstrated operation of the EOM BR VCSEL at 20 GHz for a radio-over-fiber protocol. The partners also demonstrated a QPSK protocol and a return-to-zero protocol at 7-8 and 10 Gb/s, respectively. These are encouraging for further systems development. Finally the partners through frequency response measurements and small-signal circuit modelling estimate that the EOM BR VCSEL has an intrinsic device bandwidth of 56±5 GHz – potentially leading to the cherished 100 Gb/s single channel link of the future. The development of high-speed directly modulated (DM) VCSELs proved to be continuously exciting during the entire project period, and for all project partners. Within the first year of the project the partners demonstrated optical eye diagrams at up to 40 Gb/s, record relaxation resonance frequencies as high as 30 GHz as shown in Fig. 2.0.2, and record -3 dB bandwidths of up to 23 GHz (CUT) and 24 GHz (VIS) (see Fig. 3.2.2b)). This led in the second year to demonstrations of room temperature 25-40 Gb/s error-free transmission (bit error ratios (BERs) < 1 x 10-12) using on-wafer probing at back-to-back distances (2-3 m) (see Fig. 3.4.12 and Fig. 3.5.4).

6

0 2 4 6 8 10 12 14 160

5

10

15

20

25

30

35~3 µm

~6 µm~9 µm

Res

onan

ce F

requ

ency

(GH

z)

Current (mA)

Figure 2.0.2 Resonance frequency as a function of current at different aperture sizes for 850 DM VCSELs with three different oxide aperture diameters (3, 6, and 9 µm). Clearly the peak relaxation resonance frequency varies with oxide aperture diameter. VIS packaged the devices (packages without driver IC and also with driver IC) and demonstrated a record 40 Gb/s error-free data transmission in a TOSA/ROSA link. In Figure 2.0.3 we show a VIS photodiode (PD) mounted and wire bonded on a high frequency testboard with a transimpedance amplifier (TIA). Although the PDs are not part of the VISIT work, it was critical to develop PDs as well in order to measure the high frequency characteristics of the DM VCSELs, and vice versa. This is because the commercial photodetectors at 850 nm are only available with bandwidths up to 25 GHz, and these are not capable of testing our 40G 850 nm VCSELs. Optical eye diagrams from our packaged photodiode module at 20 and 40G are shown in Fig. 2.0.4. This unique high frequency measurement capability was a key factor in the results that we now describe in the next Chapter in our quest to develop the EOM BR VCSEL and the DM VCSEL.

(a) (b) Figure 2.0.3 Optical receiver subassembly on a ceramic carrier (a); and a complete fiber-coupled receiver module in a TO-can package attached to a high frequency test board (b). For fiber coupling a mirror is placed on top of the PD’s aperture. A 45o angle can be used for VCSELs and PIN photodiodes (PDs) to couple light to and from the multimode optical fiber. For TWLs the angle can be tuned to the TWL mode angle but the package design may remain basically the same otherwise.

20 Gb/s 40 Gb/s Figure 2.0.4 Open optical eye diagrams for error-free (BER < 1 x 10-12) NRZ operation through a complete VCSEL transmitter to a PD receiver over a multimode optical fiber link at 20 and 40 Gb/s – all using VISIT project developed technology.

PIN photodiode (PD) die

Transimpedance amplifier (TIA) IC

7

3. Description of the main S&T results/foregrounds 3.1 Introduction As mentioned in the introduction to Chapter 2 the VISIT project consists of five complementary and synergistic Work Packages (WPs) aimed at the development of high speed optical transmitter devices and optical data communication systems. In addition to Management, Reporting, and Dissemination (in WP1, see the VISIT Project Year 1, Year 2, and Year 3 reports) the four other WPs cover the development of: high speed EOM BR VCSELs (in WP2), EOM TWLs (in WP3), directly modulated VCSELs (in WP4), and Packaging, System-level Evaluation, Standardization, and Exploitation (in WP5). 3.2 WP2 The goal of Work Package 2 is the development of one of the three key enabling devices at the core of the VISIT technology, the electro-optically modulated Bragg reflector (EOM-BR) vertical cavity surface emitting laser (VCSEL), and also to provide technical project management and support directly to the VISIT Project Coordinator for all technical work performed within the VISIT project. The work includes extensive process development in the TUB and cleanroom (the Center of Nanophotonics) to develop best practices in the production of the EOM BR VCSEL devices, including the design of new mask sets. The planned work also includes a first and second round of 850 nm-range VCSEL device iterations (growth, processing, and characterization) leading in the second phase to a more optimized EOM design and in the final phase to the demonstration of a first-of-its-kind EOM-BR VCSEL operating at blazingly fast serial data transfer rates of up to 40 Gb/s. Summary of most important achievements:

• Developed EOM BR VCSEL mask set in the ground-source-ground (GSG) pad layout configuration and a companion processing sequence

• Completed the epitaxial growth, device processing, and characterization of more than 30 EOM BR test structures using wafers grown by TUB

• Growth by IQE Inc. (by MBE design #1b and #2b) and by IQE Ltd. (by MOVPE design #4) of VIS EOM-BR VCSEL epitaxial wafers in three different designs

• Completion by TUB and by an external foundry of the processing of EOM-BR VCSEL wafer pieces including extensive process development work, followed by completion of EOM-BR VCSEL device characterization

• Completion by TUB of an experimental and a theoretical analysis of the operation of an earlier EOM-BR VCSEL that was processed and characterized by TUB as part of the Year 1 and 2 WP2 activity

Tasks 2.1 through 2.4: EO material design and optimization; Device Design; Growth; and Device processing and testing We first developed high-speed 850 nm-range VCSELs as a baseline for the EOM-BR VCSELs, with VCSEL material supplied by VIS. These 850 nm VCSELs were processed at TUB. The measured characteristics for an 850 nm QW VCSEL are shown in Fig. 3.2.1 and Fig. 3.2.2. We note the superb temperature performance between 20-100˚C in the linear region of the L-I curves, the high L-I slope efficiency (~0.7), and the relatively small root-mean-square (rms) spectral linewidth. Simultaneously VIS worked on the development of EOMs and prototype EOM BR VCSELs primarily in conjunction with TUB. TUB performed multiple experimental EOM test structure growths, and test structure processing and characterization using the VIS designs. The development work on more than 30 EOM test structures is described in detail in Deliverable D2.1. Schematic diagrams of a DM VCSEL and an EOM BR VCSEL are given in Fig. 3.2.3. The DM VCSEL diode consists of two metal contacts, the anode and the cathode on opposing sides of a p-i-n diode, wherein the p- and n-doped regions are essentially planar mirrors realized as multilayer distributed Bragg reflectors

8

(DBRs), and the i-region is primarily a microcavity active region composed of QWs or QD active elements. The EOM BR VCSEL is a three-terminal device composed of a VCSEL section adjacent to an EOM section, each in the form of a p-i-n diode with a common contact, thus forming for example a p-i-n-i-p structure (or vice versa) as shown. First we concentrate on the preparation of VCSELs to be used later as a sub-part of the EOM BR VCSEL.

VCSEL Day Chalmers University of Technology 24.04.2009 Confidential and Proprietary VI Systems GmbH

34

850 nm VCSEL – ultrahigh-speed design

0 2 4 6 8 10 12 140

1

2

3

4

5

6

7

8

9

0

1

2

3

Vol

tage

, V

Pow

er, m

W

Current, mA

20C 30C 40C 50C 60C 70C 80C 90C 100C

848 850 852 854 856 858 860 862 864 866 868-75

-70

-65

-60

-55

-50

-45

-40

-35

-30

-25

-20

RMS, (nm) 0,22695 0,26475 0,2584 0,2843 0,29341

4606#28 mkm

po

wer

, dBm

Wavelength, nm

20C 40C 60C 80C 100C

at thresholdSpectral resolution - 0.08nm

The deconvoluted VCSEL rise time is below 10 ps and has a very weak temperature dependence up to 100°C

Lott et al. (unpublished)

VCSEL Day Chalmers University of Technology 24.04.2009 Confidential and Proprietary VI Systems GmbH

35

850 nm VCSELs – BER tests

-12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1

-2

-14

-12

-10

-8

-6

-4

log(

BER

)

25Gbit/s

20Gbit/s

38Gbit/s

32Gbit/s

40Gbit/s

Received optical power (dBm)

Blohkin, Lott, Ledentsov, et al. Electronics Letters 2009 (in press)

The maximum data bit rate is limited by the test system (due to the rise times and jitter of the amplifiers and connectors)

8-9 psrise time20%-80%

Figure 3.2.1 Measured: (a) L-I-V and emission spectra and eye diagrams for an 850 nm QW VCSEL; and (b) measured high frequency characteristics of an 850 nm-range QW VCSEL, including selected optical eye diagrams and bit error ration (BER) tests indicating error-free operation at up to 38 Gb/s. Note: this work was presented at VCSEL Day, 24 April 2009 in Göteborg, Sweden, as part of the VISIT Project presentation given by J.A. Lott.

Figure 3.2.2 (a) Scanning electron micrograph (SEM, from December 2008) of a cleaved cross-section of the VIS (foundry-grown) high-speed 850 nm VCSEL shown with the VIS original design of two oxide aperture layers and multiple deep oxidation layers to reduce capacitance; and (b) world record -3dB bandwidth of ~23 GHz for these same VCSELs. Note: this work was presented (and thus disclosed) at VCSEL Day, 24 April 2009 in Göteborg, Sweden at CUT, as part of the VISIT Project presentation given by J.A. Lott.

Figure 3.2.3 Schematic diagrams of: (left) a directly current modulated VCSEL; (center) an EOM BR VCSEL; and (right) the refractive index profile for the device showing the electric field intensity with and without a bias voltage applied to the EOM section.

(a) (b)

(a) (b)

(a) (b)

9

The first EOM BR VCSEL epitaxial wafer structures were designed by VIS, grown by IQE-e, and processed into devices at foundry. In Fig. 3.2.4a) and b) we show SEM images of the processed EOM BR VCSELs, and in d) we show an SEM as a cross-section with the appropriate etched layers and added metal contact layers. The SEM image in 3.2.4c) is an image of the TUB mask set that was later in the second iteration used to process EOM BR VCSELs.

Figure 3.2.4 Scanning electron micrographs of: (a) a partially fabricated EOM BR VCSEL; (b) the completed VCSEL from (a); and (c) an alternative GSG pad layout configuration for an EOM BR VCSEL as used by TUB. In (d) we show a rough schematic diagram of a “fabrication concept” for an EOM BR VCSEL test structure, based on an actual SEM of an MOVPE-grown structure. The static performance of the EOM BR VCSELs is shown in Fig. 3.2.5, Fig. 3.2.6, and Fig. 3.2.7. Previously large aperture EOM DBR VCSELs have shown only weak if any EO modulation at small currents. The voltages applied to the EO section needed to cause the maximum extinction ratio are rather high (~10-12V). However, at a properly chosen bias point of 4V, peak-to-peak modulation already results in ~1-1.5dB extinction ratio. A moderate increase in the temperature increases the extinction ratio at a given modulation voltage. In Fig. 3.2.8 we show a sine-wave modulation amplitude curve of the device. One can see that the -3dB bandwidth approaches ~17 GHz. Large signal modulation experiments have been performed. Self-absorption induced effects caused parasitic traces in the NRZ eye diagrams. Truly pulsed operation with a pure sequence of “1” and “0” levels resulted in a clear eye opening at 10 Gb/s and thus the device is expected to be suitable for RZ modulation at speeds up to ~15-18 Gb/s.

(a)

(c)

(d)

(b)

10

Figure 3.2.5 CW single-mode emission spectra of the EOM BR VCSEL for an EOM bias of 0 and -2V, showing an extinction ratio of 2 dB and a chirp of less than 0.04 nm. The chirp is likely due to heating when the applied EOM voltage shifts the DBR index in-turn decreasing the light emitting out of the EOM BR VCSEL – the conservation of energy principle and the data above lead to the conclusion that the excess power is dissipated (at least partly) as heat. Figure 3.2.6 Measured: (a) emission spectrum; and (b) L-I characteristic at 25 and 85˚C for the EOM BR VCSEL with an oxide aperture diameter of 1 µm. In (c) and (d) is the same. Except the oxide diameter is 2 µm.

846.2 846.3 846.4 846.5 846.6 846.7 846.8

-35

-30

-25

D26 µm(aperture ~ 1µm)

25oC

-2V

dB

Wavelength (nm)

0V

2 dB

842 843 844 845 846 847 848 849 850

-60

-50

-40

-30

-20

-10

>40 dB SMSR

EL In

tens

ity (d

B)

Wavelength (nm)

D26 µm(aperture ~ 1µm)

847 848 849 850 851 852 853 854 855

-60

-50

-40

-30

-20

-10

D27 µm(aperture ~ 2µm)

EL In

tens

ity (d

B)

Wavelength (nm)

17 dB

0 1 2 30.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

D26 µm(aperture ~ 1µm)

85oC

Opt

ical

Pow

er (m

W)

Current (mA)

08-6D26

25oC

0 1 2 3 4 50.00.10.20.30.40.50.60.70.80.91.01.11.21.31.4

ηdiff = 27%

D27 µm(aperture ~ 2µm)

85oC

Opt

ical

Pow

er (m

W)

Current (mA)

08-6D27

25oC

(a)

(b)

(c)

(d)

11

0 1 2 30.00

0.05

0.10

0.15

0.20

0.25

ER ~1.66 (2.2 dB)

3 Ith

Pow

er [m

W]

Current [mA]CW, RTF07-6D27

VEOM= 2V 14V

5 Ith

ER ~1.8 (2.4 dB)

0 1 2 3 4 5 6 7 80.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

CW, RTF07-6D35

VEOM= 2V 14V

Pow

er [m

W]

Current [mA]

5 IthER ~1.7 (2.3 dB)

3 Ith

ER ~1.65 (2.2 dB)

(a) (b) Figure 3.2.7 L-I curves at two modulator bias voltages for the devices with: (a) a 3µm-diameter oxide aperture; and (b) an 11 µm-diameter oxide aperture.

Figure 3.2.8 Sine-wave modulation amplitude of the device evidencing a high modulation bandwidth (top right), as presented by VIS at the ECOC 2010 conference in Torino on 21 September 2010. A new structure based on a composite resonator EOM BR VCSEL design was given to IQEe for epitaxial growth and included oxide aperture layers in both the VCSEL and the EOM sections with the intention to reduce mesa capacitance and thus increase the device’s bandwidth. The partners discussed in great detail the epitaxial device design, the growth parameters, and a growth strategy including a set of calibration structures. With this detailed information IQE completed the growth of EOM BR VCSEL #4. This structure is shown schematically in Figure 3.2.9. The measured optical reflectance for this structure is given in Figure 3.2.10. Similarly VIS designed two other composite resonator EOM BR VCSEL structures for growth by MBE via the planned subcontract with IQE Inc. (USA). IQE Inc. Successfully grew these two other Gen 2 EOM BR VCSEL structures and the measured optical reflectance is given in Figure 3.2.11. Using a new EOM BR VCSEL mask set (designed and produced by TUB, see Figure 3.2.12 (right)) TUB performed a series of processing development steps seeking to successfully produce EOM BR VCSELs from the epitaxial wafers. Subsequently the TUB was closed from January through September 2011 for renovation work and it was not possible to continue processing additional EOM BR VCSEL material. As a result, the partners instead performed an analysis and further testing of the EOM BR VCSEL #3 structure that has been processing in Project Year 2.

12

Figure 3.2.9 Schematic diagram of a Gen 2 EOM VCSEL in the form of a composite resonator. Figure 3.2.10 Measured optical reflectance of EOM BR VCSEL #4 in two wafer substrate locations after growth by MOVPE by IQE-e. The device concept of the present EOM BR VCSEL aims at the monolithic integration of a modulator with a VCSEL to benefit from the established fabrication and processing technologies and a match to the industry standard 850 nm wavelength. In Figure 3.2.12 we show schematics of the device depicting the third electric middle-contact. Compared to a conventional VCSEL this additional contact electrically divides the device into two sections: on the bottom a VCSEL and on top the EOM section realized as a second cavity within the VCSEL top DBR. All DBRs consist of Al0.9Ga0.1As and Al0.15Ga0.85As λ/4 layers employing graded interfaces with spike doping levels to facilitate charge carrier transport within the device.

700 750 800 850 900 950 1000

0.0

0.2

0.4

0.6

0.8

1.0

raw

pow

er re

flect

ance

wavelength (nm)

center edge

13

Figure 3.2.11 Measured optical reflectance of EOM BR VCSEL #1b and #2b in after growth by MBE by IQE Inc. (USA).

Figure 3.2.12 Schematic diagram of an EOM BR VCSEL for use in device modelling (left and center), and a top down white light microscope image of an actual TUB-processed EOM BR VCSEL using a new mask set created for this purpose (right). A conventional design with a GaAs MQW gain-medium with Al0.2Ga0.8As barriers is chosen for the VCSEL section. Adjacent to the active cavity within the first DBR pair an AlAs layer enclosed by AlGaAs-gradings is introduced for post-growth oxidation of a crack-free aluminum-oxide current-aperture using TUB’s novel oxidation technology. The bottom high-reflectivity AlGaAs DBR is n-doped while the intermediate DBR between active-cavity and EOM-cavity is p-doped. Within this p-doped DBR a lattice matched InGaP layer adjacent to a p+-doped GaAs contact layer serves as etch-stop during processing of the middle-contact. Active QW spectral position is confirmed to be at 835 nm wavelength by photoluminescence characterization while InGaP lattice matching is verified by X-ray diffraction measurements.

The EOM section starts on top of the middle-contact layer with the remaining part of the p-DBR. To employ the quantum-confined Stark effect for light output modulation an undoped cavity is formed by a GaAs MQW

14

stack with Al0.2Ga0.8As barriers. These EOM QWs are spectrally shifted towards shorter wavelengths with respect to the active VCSEL. On top of the EOM-cavity a Si-doped 10x DBR with a GaAs cap layer protecting against oxidation finalizes the structure. Spectral positioning of emission wavelength and EOM QWs aims at a minimized absorption within the EOM element while maintaining a sufficient refractive index modulation referred to as region II in reference. Consequently EOM QWs are tuned to a higher energy with respect to the active QWs of the VCSEL section. Dynamic and static characteristics are investigated at room temperature (RT) with operational devices mounted on a copper heat-sink. For all measurements the VCSEL section is driven at constant current above lasing threshold to enable CW emission. Modulation of the optical output is realized solely by applying a reverse EOM-voltage. The current flow across the EOM section is monitored simultaneously as a measure for the photo-absorption of laser emission within the modulator. Fundamental laser characteristics and the corresponding EOM photocurrent of a multi-mode device with 28 μm EOM mesa-diameter are given in Figure 3.2.13. The data show a constant lasing threshold independent of the applied reverse EOM bias indicating good optical modulator isolation from the VCSEL section.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 1 2 3 4 5 6 7 80.0

0.2

0.4

0.6

0.8

1.0

1.2

Opt

ical

Pow

er [m

W]

VCSEL Current [mA]

reverse EOM voltage: 1V 2V 3V

EOM

Pho

tocu

rren

t [m

A]

Figure 3.2.13 Solid lines: optical output power characteristics of the EOM VCSEL are shown for different voltages applied to the EOM section. Dashed lines: corresponding photocurrent due to absorption within the EOM section. Dotted lines: guides to the eye identify the position of data.

Optical output-power and EOM photocurrent increase linearly with the VCSEL drive current up to about 3.3 mA. With increasing EOM reverse bias, however, the absorption increases and the output power decreases. Within this current range, modulation of the VCSEL output is thus primarily due to change of absorption strength in the EOM section. Upon further increase of VCSEL current, the optical output power is decreasing, likely limited by thermal rollover. Within a current range of 5 to 6.5 mA, the VCSEL output power increases again but has a strong super-linearity. Up to the narrow output peak maximum at resonance the photocurrent increases linearly and then drops sharply. Upon further increased current both values finally decrease. The second optical output power increase cannot be explained by a change of absorption within the EOM section. Instead, we attribute this feature to resonance matching of the two coupled cavities within the device. For devices sharing a similar cavity design, a resonant output power behavior is typical. We analyze the dynamics after we derive a qualitative explanation for the observed resonance peak. To distinguish between absorption related and index of refraction induced resonance effects on signal modulation we analyze now in more detail the output power and the photocurrent within the 1-3 V EOM bias range as a function of the VCSEL drive current (see Figure 3.2.9). Additionally this figure includes the sum of both values (the EOM photocurrent is converted to the corresponding optical output power at the wavelength of operation) giving the total output power Ptot of the VCSEL section. Ptot shows an almost flat behavior for 5 mA (black curve, bottom viewgraph). Thus, the modulation of the light output as seen in the center of the

15

figure is solely driven by absorption within the EOM section and a rather large EOM voltage sweep of 1.5 V is necessary to realize 3 dB optical intensity modulation. At higher VCSEL drive currents of 6.0 and 6.2 mA (red and green curves) a sudden increase of the output power is revealed. Upon a voltage change of less than 100 mV Ptot almost doubles and only a fraction of this increase is due to absorption in the EOM section. This step like increase of the total VCSEL output power directly proves the onset of coupling between the two cavities: the EOM resonance wavelength is varying as a function of voltage. This steep increase of optical output power at resonance represents a very efficient modulation of the device. We call this electro-optic resonance modulation (ERM).

Previous large-signal experiments on similar devices revealed a limited modulation bandwidth of 3 GHz. Here we present a small-signal analysis to investigate the origin of this bandwidth limitation. Opposite to conventional current modulated devices the small-signal modulation transfer function of photon lifetime τp modulating devices decreases with 1/ω instead of 1/ω2. For this investigation the device is biased ensuring resonance conditions. Direct measurements of the small-signal modulation bandwidth (S21) identify tight limitations with a sharp drop at the beginning of the frequency range. As this is usually an indication of a parasitic limit, we conducted a more detailed investigation of the device impedance (S11). The results of this investigation are presented in Figure 3.2.14.

Figure 3.2.14 Smith Chart of electrical impedance (S11) of EOM-section between 0 to 40 GHz and equivalent circuit of parasitic network. The intrinsic (ideal) EOM-section is represented as a two pole square being connected to a parasitic network. Drive current of the VCSEL section is 6.2 mA, EOM voltage is 2.8 V. The EOM-section is modeled as reverse-biased PIN-diode and the parasitic device values can be extracted with high accuracy. We split the EOM-section into the diode capacitance of the space-charge regions of the p- and n-side (which can be modeled as one lumped element Cdiode) and an additional capacitance CQW from the EOM QWs only. Each of these capacitors has a leakage resistance Rdiode and RQW, respectively. Additionally we found a spreading resistance Rspread caused by the surrounding cladding layers, a pad capacitance Cpad and a non-perfect contact resistance Rcontact. The response from the parasitic equivalent circuit perfectly describes the measured S11-values across the entire (setup-limited) frequency range from 0 to 40 GHz. Changes of device values are also consistent with changes of the EOM-voltage. The lumped element values for different EOM-voltages are given in Table 3.2.1.

Table 3.2.1 Fitting values of the equivalent circuit of the parasitic network (as shown in figure 5). The values are derived from EOM VCSEL S11 measurements at RT for different EOM-voltages. Drive current of the VCSEL section is 6.2 mA.

UEOM CQW RQW Cdiode Rdiode Rspread Cpad Rcontact 2.8 V 55 fF 1040 Ω 177 fF 6.5 kΩ 309 Ω 55.4 fF 35.9 Ω 3.0 V 56 fF 940 Ω 176 fF 8.1 kΩ 308 Ω 55.3 fF 35.9 Ω 3.2 V 60 fF 650 Ω 173 fF 10 kΩ 302 Ω 55.4 fF 35.9 Ω

16

For increasing voltage, the space-charge capacitance Cdiode decreases consistently with the increase of the depleted region. As expected the corresponding leakage resistance Rdiode becomes significantly larger. The outer parasitic network is practically invariant to changes in voltage, whereas the EOM-section itself varies its impedance significantly. With increasing photo-current the leakage resistance RQW decreases, while the capacitance CQW becomes larger due to the variation of the refractive index and photo-absorption induced heating. From this equivalent circuit model we can extract the parasitic response of the EOM-section which is clearly limiting the high-speed performance of our prototype device. Extraction of the parasitic response by means of fitting from the S21-curve yields very similar results.

( )20 p

21 2 2p 0

ω τ jω1ωτ ω ω jωγ

S+

∝ ⋅− +

(3.2.1)

The parasitic limited performance of the device is also in agreement with our estimates based on device geometries and doping levels. To determine the intrinsic device speed with resonance conditions the parasitic response is deconvoluted from the measured frequency response. The intrinsic device speed can be fitted to Eq. (3.2.1). This fit yields a photon lifetime τp of 4.1 ps, a resonance-frequency ω0 of 2π·27 GHz and a damping coefficient γ of 1.5 1011 s-1 at driving conditions of 6.2 mA and 2.8 V. The result is depicted in Figure 3.2.15, predicting an intrinsic device bandwidth of 56±5 GHz.

0 10 20 30 40 50 60 70-6

-3

0

3

6

9

12

f3dB

Rel

. Int

rinsi

c ER

M R

espo

nse

[dB

]

Frequency [GHz]

fit to Eq. (1) intrinsic S21

VCSEL current: 6.2 mAEOM voltage: 2.8 V

fit region

Figure 3.2.15 Intrinsic bandwidth of the EOM VCSEL at RT. Electrical parasitics are deconvoluted from the measured data yielding an intrinsic bandwidth of 56±5 GHz.

Such a bandwidth together with a large overshoot is typical for intra-cavity loss modulated lasers. A well-tailored parasitic response is needed to flatten out the characteristic overshoot in the device response without limiting the bandwidth. Future EOM VCSEL generations will reduce the large parasitics of this prototype by employing smaller EOM mesa diameters and additional oxide apertures for the EOM part. Additionally, efficient modulation could necessitate single mode operation. The modulation depth of the EOM cavity can be improved by employing a larger numbers of QWs, which are precisely tuned to minimize absorption while maintaining sufficient refractive index modulation. Additionally high bandwidth device processing and mounting is required to reduce the external parasitics.

17

3.3 WP3: Development of EOM TWL Work Package 3 of the VISIT Project has as its goal the development of a novel high speed 1300 nm tilted wave laser (TWL) with an integrated electro-optical modulator (EOM). The operational goal for the final packaged EOM TWL is a data transfer rate of 40 Gbit/s for potential applications in medium range access networks and local area networks. The work represents a collaborative effort between VISIT partners: the Ioffe Institute, Chalmers University of Technology, VI Systems GmbH, and the University of Cambridge. The concept of achieving wavelength stabilization in edge–emitting lasers exclusively through the epitaxial design was proposed in 2002 [N. N. Ledentsov and V. A. Shchukin, Optical Engineering 41, 3193 (2002)]. The TWL concept is based on the existence of the tilted modes originating in planar waveguides with thin cladding layers. The TWL waveguide consists of two coupled waveguides. The optical mode tunnels from the thin waveguide into the thick waveguide and vice versa (see Fig. 3.3.1). Coherent leaky light passing through multiple reflections accumulates a phase difference and interferes with the light that is propagating in the planar leaky waveguide. In case of constructive interference the phase matching condition is fulfilled and wavelength stabilization takes place since the lasing wavelength is mainly determined by the laser cavity but not the gain: dλ/dT ≈ (λ/n) dn/dT, where λ is the emission wavelength and n is the refractive index.

a) b) Figure 3.3.1 Coupled waveguide structure: a) optical fields in the cladding layer and in the thick waveguide defining the boundary conditions at the interface; and b) lasing mode profile.

The TWL concept is not limited by the material and may be transferred to edge-emitting lasers emitting in different wavelength ranges. The strategy of WP3 was to develop GaAs-based 980nm quantum well (QW) edge-emitting TWLs which served as a cost-effective reference for the design of a GaAs-based 1300 nm wavelength-stabilized edge-emitting TWLs. The followed activity focused at integrating 1300 nm TWL and EOM into a single device. The final stage was aimed at study the dynamics properties of 1300nm EOM TWLs. Task 3.1: Design and fabrication of 980 nm edge-emitting tilted wave laser A series of 980 nm laser wafers were grown by metal-organic chemical vapor deposition (MOCVD) on n-GaAs (100)-oriented substrates. The active region consisted of two InGaAs quantum wells inserted into the middle of 0.58 µm thick GaAs waveguide. The structures had 600 nm-thick p-GaAsP cladding and 300 nm-thick p-GaAs contact layers. The lower part had a 10 µm-thick n-GaAs coupled waveguide and a 2 µm-thick n-InGaP reflecting layer. The thickness of thin cladding layer varied from 200 nm to 400 nm. The threshold current changes slightly with temperature (see Figure 3.3.2a) and the estimated value of the characteristic temperature T0 exceeds 500 K. The differential quantum efficiency at room temperature was as high as 83%.

18

a) b)

Figure 3.3.2 Temperature dependence of light-current (a) and spectra (b). Figure 3.3.2b shows the temperature dependence of the TWL spectra. In the range of 20-50˚C (∆T = 30K) the temperature-induced shift of the lasing wavelength is 1.5 nm, which gives a temperature coefficient dλ/dT of 0.05 nm/K. Beyond this temperature range the lasing wavelength is discretely shifted. This fact confirms that lasing wavelength is determined by the allowed TWL modes originating from the waveguide coupling and the spectral distance ∆λ between the modes is more than 10 nm. Lasing in the TWL takes place only if the mode wavelength lies within the gain spectrum. Thus a broader gain spectrum increases the range of temperature stability. Task 3.2: Design, growth, fabrication and optimisation of EOM media for TWL edge emitters based on InGaAs-AlGaAs materials scalable to applications at 1300 nm range EO materials applicable for 850nm and 1300nm range were investigated both theoretically and experimentally. EO wafers were grown, processed, and tested using optical reflectance spectroscopy under an applied field. Angle resolved studies with temperature dependencies were carried out. EOM DBR structures designed for 850nm and 1300nm ranges showed the EO effect that qualitatively and quantitatively corresponds to the numerical calculations. Electro-optic effect was investigated in more details in the frame of Task 3.3. Task 3.3: design and fabrication of 1.3 µm GaAs based wavelength-stabilized edge-emitting TWL Design of TWL emitting at 1300 nm is based on the theoretical and experimental results obtained for 980 nm TWLs. Laser wafers emitting at 1300 nm were grown by molecular-beam (MBE) epitaxy on n-GaAs (100) substrate. The active region contained ten layers of self-organized InAs quantum dots (QDs) separated by 35 nm thick GaAs spacers. It was inserted into the middle of 0.95 μm thick GaAs waveguide layer. Active region composition and growth conditions are similar to ones described earlier. To obtain 1.3 µm emission the QDs were directly formed on the GaAs matrix by deposition of 2.5 monolayers of InAs and then covered with a 5-nm thick In0.15Ga0.85As layer. The structure had 1.5 μm thick p-Al0.35Ga0.65As cladding layer and 200 nm p-GaAs contact layer. The lower part of the wafer was composed of 10.5 μm thick n-GaAs coupled waveguide and 2 μm thick n-Al0.35Ga0.65As reflecting layer. Two waveguides were coupled via 300 nm thick n-Al0.35Ga0.65As layer. Figure 3.3.3 shows temperature dependence of the lasing spectra of the TWL. In the range of 20-400C (∆T=20K) temperature-induced shift of the lasing wavelength is -9 nm, which gives temperature coefficient dλ/dT of -0.5 nm/K. It is important to notice that the temperature coefficient is negative which is unusual for any edge-emitting laser. This trend has been observed in numerical simulations though it was not so pronounced. Due to the negative temperature coefficient TWL operating temperature range is limited since lasing in TWL takes place only if the mode wavelength lies within the gain spectrum. This experimental result confirms the fact that temperature behavior of the lasing wavelength in TWL is defined by the laser cavity parameters and can be engineered as it was predicted.

19

EOM properties of grown TWLs were investigated in the lasers with two-section configuration (Fig. 3.3.4). The lasers have directly biased sections (gain sections) and reversely biased sections. The sections were electrically isolated. Maximum applied reverse bias before breakdown was as high as 20 V. To avoid excessive optical losses the gain sections were much longer than reversely biased ones (section length ratio > 10). EOM parameters were extracted from analysis of evolution of far-field patterns and light-current characteristics properties. The measurements have been carried out at room temperature under CW and pulsed pumping current. In Fig. 3.3.5 far-field pattern dependence of the reverse bias voltage is presented. The curves are normalized. In this figure redistribution between outer lobes and central lobes is clearly seen.

Figure 3.3.4 Two-section construction of EOM-TWL. Figure 3.3.5 Far-field patterns vs. reverse bias in EOM TWL.

The waveguide coupling strength controls intensities of the major and minor lobes: the stronger the coupling, the more intensive the two outer lobes. In TWL combined waveguide the coupling mainly depends on the thicknesses and refractive indexes of thin cladding layer and thin waveguide layer. Electric field caused by the reverse bias exists in i-type thin waveguide surrounded by n and p claddings. This electric field induced refractive index changes and the far-field redistribution therefore. Numerical fitting of the experimental far-field patterns allows calculating changes of the refractive index of the thin waveguide. Linear EO effect is

described by the following equation , where is the refractive index change, n is the

refractive index, is the electro-optic coefficient, E is the average electric field. We have calculated electro-

optic coefficient for quantum dot medium (InAs QD). It is as high as 2∙10-11 m/V. This value corresponds

to the experimental data obtained earlier.

Figure 3.3.3 TWL lasing spectra temperature dependence.

20

A family of the light-current curves with different reverse bias voltages is presented in Fig. 3.3.6. Total cavity length L of the laser was 2 mm. The gain section was pumped by pulsed current while dc voltage was applied to the modulator section. Increasing of reverse bias aroused small changes in the threshold current and pronounced decreasing of the intensity. Thus, light from two-section TWL can be modulated by applying reverse bias.

Task 3.4: Design, fabrication and high-speed testing of 1.3 µm GaAs based TWL with integrated electro-optical modulator (EOM) Potential for using TWL wafer for high-speed operation has been investigated. The modulator dynamic response depends on the carrier escape time. In TWLs total thickness of the waveguide exceeds 10 μm. However, only thin waveguide is undoped so broad coupled waveguide does not affect dynamic parameters of the modulator section. In real optoelectronic devices modulation frequencies are much lower than theoretical maximal frequency due to a number of limiting factors. Dynamic performance of the modulators integrated into tilted wave lasers can be estimated from mode-locking (ML) regime of operation when lasers emit a train of pulses with a repetition frequency approximately equal to an integer multiple (a harmonic) of the cavity round-trip frequency. The reverse biased saturable absorber can double as an optical modulator, through the electro-absorption (quantum-confined Stark) effect, if an AC voltage is applied to it. We have processed TWL wafer into lasers with integrated modulator section. Laser cavity length L was 3.0±0.3 mm, length of modulator section was ~0.1L. Stripe width was 7 μm, which ensures single-mode lasing.

We have investigated EOM TWL of three types: 1) Type A: absorber section is located nearby the laser facet. This construction produces pulses with fundamental frequency; 2) Type B: absorber section is located half-way along the laser cavity (Fig. 3.3.7a). ML is achieved at the second harmonic of the fundamental frequency; and 3) Type C: absorber section is located at the distance of 1/3L from either of the facets (Fig. 4.3.7b). In this construction ML is achieved at the third harmonic of the fundamental frequency.

Figure 3.3.6 Light-current curves vs. reverse bias in EOM TWL.

21

Gain section was pumped by cw current, modulator was reversely biased. The largest reverse bias was 10V. Voltage Vr applied to the modulator sections increased threshold current by 25%. Passive mode locking has been obtained for all three types of TWLs. Autocorrelation functions for three types of lasers are presented in Fig. 3.4.8. Parameters of mode-locking operation are summarized in Table 3.3.1. The highest repetition frequency of 46.1 GHz has been obtained for the lasers of type C operating at the third harmonic of the fundamental repetition frequency. This type of lasers has shown the shortest pulse duration as low as 3.2 ps (Gaussian approximation). Time-bandwidth product for laser type C was 0.6 which is very close to the

theoretical limit of 0.44.

Figure 3.3.8 Autocorrelation functions for mode-locked TWLs. 1 – type A (fundamental frequency); 2 – type B (second harmonic); 3 – type C (third harmonic). Table 3.3.1 Parameters of EOM TWL mode-locking operation.

Laser type

Pulse duration,

ps

Frequency, GHz

Reverse bias, V

Time-bandwidth

product Type

A 8.7 12.5 -3.0 1.1

Type B

5.1 25.7 -6.0 0.7

Type C

3.2 46.1 -8.5 0.6

(a) (b)

Figure 3.3.7 Images of EOM TWL: a – type B, b – type C.

22

Reversely biased modulator section of mode-locked TWL lasers can be additionally modulated by RF signal. In this case ML frequency is a carrier frequency. Thus, RF signal should have a smaller frequency. In the best scenario RF signal has a subharmonic frequency. This concept is represented in the idea of Radio over Fiber (RoF) referring to a technology whereby light is modulated by a radio signal and transmitted over an optical fiber link to facilitate wireless access. RF spectrum corresponding modulated mode-locked pulse train (modulation frequency was 650 MHz.) is presented in Fig. 3.3.9. It shows fundamental mode-locked

frequency of 12.5 GHz (carrier wave) and two sidebands shifted for . Frequency of the carrier wave can

be dramatically increased if lasers with harmonic ML frequencies are used. TWLs of type C with ML frequency of 46.1 GHz can be modulated by a signal with frequency of (n - integer number).

Thus, specific optical mode formation and corresponding output parameters (increased wavelength temperature stability, narrower far-field) of tilted-wave lasers with integrated modulator may allow considering it as a light source for RoF applications.

Figure 3.3.9 RF-spectrum of optical signal matched to low oscillographic trace. Central peak: ML signal at 12.5 GHz, left and right peak shift on 650 MHz of external modulation signal

3.4 WP4: Development of advanced directly modulated VCSELs Work Package 4 of the VISIT project has aimed at the development of high speed, directly modulated 850 nm VCSELs for high capacity, short reach optical links and interconnects based on high speed multimode (OM) fiber. The work has followed the plan in the Description of Work. In the first phase of the work (Task 4.1, M1-M28), a first generation (Gen I) VCSELs operating at speeds of 16-32 Gbps up to 85°C was developed. In the second phase (Task 4.2, M10-M41), a second generation (Gen II) VCSELs operating at 40 Gbps was developed. Summary of most important achievements: Task 4.1:

• A new VCSEL design (Gen I) with improved differential gain, reduced gain compression, reduced capacitance and improved thermal management was developed.

• A modulation bandwidth in excess of 20 (15) GHz at 25 (85) °C was achieved. The bandwidth was found to be limited by thermal effects and electrical parasitics.

• With the VCSEL biased at a current density as low as ~ 10 kA/cm2, error-free transmission over 50 (100) m OM3 fiber was demonstrated at 32 (25) Gbps up to 25 (85) °C.

23

• The VCSELs were found to have a dynamic range sufficient for next generation links employing multi-level modulation formats for increased capacity.

• A set of 12 VCSELs was delivered to VI Systems for mounting on high speed carriers and packaging. Task 4.2:

• A second VCSEL design (Gen II) with lower capacitance and reduced damping of the modulation response was developed. The latter was achieved by reducing the photon lifetime.

• A record modulation bandwidth of 23 GHz was achieved, with the bandwidth being limited primarily by thermal effects.

• A comprehensive analysis of the effects of the cavity photon lifetime on VCSEL dynamics was performed, revealing a trade-off between damping and resonance frequency and providing design rules for high speed.

• A new method for the analysis of VCSEL thermal performance was developed and implemented and used to identify major sources of heat and reasons for thermal saturation.

• Error-free transmission (back-to-back) at 40 Gbps using a directly modulated 850 nm VCSEL was demonstrated for the first time. 35 Gbps error-free transmission over 100 m OM3 fiber was also demonstrated.

• A record dynamic range at high frequencies was achieved (>90 dB·Hz2/3 up to 19 GHz). • A set of 8 Gen II VCSELs was delivered to VI Systems for mounting on high speed carriers and

packaging. The work represents a collaborative effort between partners Chalmers University of Technology (WP leader; VCSEL design, fabrication, and evaluation), Tyndall National Institute (active region design), IQE Europe (epitaxial growth), TU Berlin (transmission measurements) and University of Cambridge (noise and linearity measurements). Task 4.1: Design, fabrication and evaluation of VCSELs operating at 16-32 Gbit/s (Gen I) A first generation high speed, oxide confined 850 nm VCSELs, aiming at modulation speeds up to 32 Gbit/s at temperatures up to 85°C, was developed. The design (Fig.3.4.1) includes a number of features for improving the intrinsic and extrinsic modulation bandwidths. Strained InGaAs quantum wells (QWs) are used in the active region to improve differential gain. To reduce gain compression, the active region is surrounded by graded composition separate confinement layers and thin transport layers for fast carrier transport and capture. Two oxide layers, a thick layer of BCB under the bond pad and an undoped substrate are used to reduce capacitance. Modulation doping and graded interfaces are used in the DBRs to reduce resistance and binary compounds are used in the major part of the lower DBR to reduce the thermal impedance.

double oxide aperture

BCB

p-contact

n-contact

undoped substrate

high thermal conductance n-DBR

low resistance p-DBR

strained InGaAs/AlGaAsQWs

Figure 3.4.1 Cross-sectional view of the Gen I high speed directly modulated 850 nm VCSEL. Left: top image showing the ground-signal-ground contact configuration. Right: SEM image showing part of the two oxide layers. The design of the strained InGaAs/AlGaAs QW active region was optimized using an 8-band k·p model for the band structure. With optimum concentrations of In and Al, the calculations predicted a doubling of the differential gain and a 30% reduction of the threshold carrier density compared to unstrained GaAs/AlGaAs QWs, due to both strain and quantum size effects. The final optimum design has five 4 nm thick In0.10Ga0.90As QWs with Al0.37Ga0.63As barriers. Additional numerical tools for optical, electrical and thermal simulations were used to optimize the longitudinal and transverse designs with respect to e.g. optical confinement, injection uniformity and thermal transport.

24

Figure 3.4.2 shows the output power and voltage as a function of current for a 9 µm aperture (multimode) VCSEL. Low threshold current (0.6 mA), high slope efficiency (0.8 W/A) and high maximum power (9 mW) were achieved at 25°C. The maximum power efficiency is 33% and the differential resistance at a typical bias current under modulation is 80Ω. The thermal impedance is as low as 1.9 K/mW.

Figure 3.4.2 Output power (at different temperatures) and voltage (25ºC) as a function of current for a 9 µm aperture VCSEL. Results from small signal modulation response measurements are shown in Fig. 3.4.3. A maximum modulation bandwidth of 20 GHz was achieved at 25°C, being reduced to 15 GHz at 85°C. Clear improvements with strained InGaAs QWs are observed. Fits of a transfer function to the measured modulation response and impedance measurements together with equivalent circuit modelling showed that the bandwidth is limited by a combination of capacitance and thermal effects. The predicted doubling of the differential gain was also confirmed. The performance under large signal modulation was investigated by transmission measurements using an OM3 multimode fiber (2 GHz·km bandwidth) and a 25 GHz optical receiver. Results are shown in Fig. 3.4.4. Error-free (BER < 10-12) transmission over 100 m fiber at 25 Gbps up to 85°C, and over 50 m fiber at 32 Gbit/s and 25°C was demonstrated. At the highest data rate, the VCSEL did not provide enough bandwidth to transmit at 85°C while the fiber did not provide enough bandwidth for transmission over more than 50 m. In these experiments, the VCSEL was biased at a current density as low as 11-14 kA/cm2. The spectral width (rms) under modulation was measured to be in the range 0.39-0.45 nm.

Figure 3.4.3 Left: small signal modulation response at 25°C and different bias currents. Middle and right: 3dB bandwidth and resonance frequency vs. square root of current above threshold at 25 and 85°C, also comparing strained InGaAs and unstrained GaAs QWs.

-6 -5 -4 -3 -2 -1 0 1-14-12-10

-8

-6

-4

-2

Received optical power (dBm)

log(

BER

)

BTB

50 m

100 m

-6 -5 -4 -3 -2 -1 0 1-14-12-10

-8

-6

-4

-2

Received optical power (dBm)

log(

BER

)

BTB

50 m

100 m

-6 -5 -4 -3 -2 -1 0 1

-14-12-10

-8

-6

-4

-2

Received optical power (dBm)

log(

BER

)

BTB

50 m

100 m

-6 -5 -4 -3 -2 -1 0 1-14-12-10

-8

-6

-4

-2

Received optical power (dBm)

log(

BER

)

BTB

50 m

100 m

-14-12

Received optical power (dBm)

log(

BER

)

BTB

50 m

-4 -3 -2 -1 0 1 2

-10

-8

-6

-4

-14-12

Received optical power (dBm)

log(

BER

)

BTB

50 m

-4 -3 -2 -1 0 1 2

-10

-8

-6

-4

Figure 3.4.4 BER vs. received optical power back-to-back and over 50 and 100 m of OM3 fiber using a PRBS7 bit sequence. Left: 25 Gbps, 25°C, middle: 25 Gbps, 85°C and right: 32 Gbps, 25°C. Received eyes are shown as insets.

25

The VCSELs were also evaluated with respect to noise and linearity, properties that are of utmost importance for more advanced multi-level modulation formats (PAM, PSK, QAM) which can make better use of the bandwidth provided by the VCSEL compared to on-off keying, thereby increasing the link capacity. A peak spurious free dynamic range (SFDR) of 98 dB·Hz2/3 was measured at 2 GHz. At the highest frequency (19 GHz) the SFDR was reduced to 87 dB·Hz2/3. These values prove that the VCSELs are suitable for next generation links employing multi-level modulation formats. Finally, a set of 12 VCSEL chips was delivered to VI Systems for mounting on high speed carriers and packaging. Task 4.2: Design, fabrication and evaluation of VCSELs operating at 40 Gbit/s (Gen II) The design of the Gen II VCSELs is shown in Fig. 3.4.5. To further improve speed, the number of oxide layers was increased from two to six. The first two oxide layers (also present in Gen I), placed just above the active region, define the aperture for current and optical confinement. The purpose of the additional four oxide layers, with a larger aperture, is to reduce capacitance without affecting the optical or transport properties. Impedance measurements and equivalent circuit modeling confirmed a reduction of the mesa capacitance by 30-40% (bias current dependent), in good agreement with estimates from device geometries and dielectric constants.

2 deep oxide layersBCB

p-contact

n-contact

undoped substrate

high thermal conductance n-DBR

low resistance p-DBR

strained InGaAs/AlGaAs QWs

4 shallow oxide layers 96% Al

98% Al

phase of reflection adjusted for optimum photon lifetime

Figure 3.4.5 Cross-sectional view of the Gen II design. The SEM image (right) shows part of the six oxide layers. The optical microscope image (left) shows the ground-signal-ground contact configuration. It was also found that the cavity photon lifetime has a significant impact on VCSEL performance, including the modulation response and bandwidth. Therefore, the impact of photon lifetime was studied by varying the reflectivity of the top DBR. This was accomplished by varying the thickness of the top layer using precise dry etching, thereby controlling the phase of the reflection at the semiconductor/air interface, Fig. 3.4.6.

in-phase reflection anti-phase reflection

Top

DB

R re

flect

ivity

Figure 3.4.6 Left: Dependence of top DBR loss rate and reflectivity on etch depth. Right: Corresponding dependence of cavity photon lifetime. The four etch depths used in the experiments (0, 25, 40 and 55 nm) are indicated. Figure 3.4.7 shows the output power as a function of current for VCSELs with an 11 µm oxide aperture and different cavity photon lifetimes (different etch depths). Performance measures are listed in Table 3.4.1.

6.4 ps

5.3 ps

3.3 ps

1.2 ps

26

Figure 3.4.7 Output power vs. current for an 11 µm oxide Table 3.4.1 Static performance parameters as a function aperture VCSEL with different etch depths (25°C). of photon lifetime. As expected, the threshold current, the slope efficiency and the maximum output power increase with reduced reflectivity of the top DBR. In addition, the thermal roll-over current increases since less photons are lost in the cavity through free-carrier absorption, thus reducing self-heating through internal optical absorption. Thermal effects, caused by current induced self-heating, have large impact on all VCSEL characteristics. Thermal effects lead to a premature saturation of the resonance frequency of the modulation response with increasing current and can thus limit the modulation bandwidth and speed. Therefore, a detailed thermal analysis, based on measurements and empirical modeling of dissipated power and heat generation as a function of current, was performed, aiming at identifying major sources of heat and providing guidelines for design improvements. The analysis showed that, for all VCSELs tested, thermal roll-over occurs at a current corresponding to an internal VCSEL temperature in the range 110-130ºC. Furthermore, the analysis showed that resistive loss and internal optical absorption are the major contributors to an increase of temperature with current (Fig. 3.4.8) and that thermal roll-over occurs primarily due to a rapid reduction of the internal quantum efficiency as the temperature approaches 100ºC. It was also confirmed that a reduction of top DBR reflectivity results in a significant reduction of power dissipation due to internal optical absorption (Fig. 3.4.8, right).

Figure 3.4.8 Contributions from the different heat sources to the increase of internal VCSEL temperature with current at an ambient temperature of 25°C. Left: oxide aperture = 9 µm, photon lifetime = 6.4 ps. Right: oxide aperture = 9 µm, photon lifetime = 1.2 ps. Results from small-signal modulation response measurements on 7 µm aperture VCSELs with different photon lifetimes are shown in Fig.4.4.9. A significant reduction of the damping of the modulation response is observed with decreasing photon lifetime, enabling an increase of the 3dB modulation bandwidth from 15 to 23 GHz when the photon lifetime is reduced from 6.4 to 3.3 ps. This is the highest modulation bandwidth ever reported for an 850 nm VCSEL. However, a reduction of the resonance frequency is also observed, resulting in a reduced bandwidth for photon lifetimes smaller than 3.3 ps. The net effect is an optimum photon lifetime for maximum bandwidth.

Photon lifetime (ps) 6.4 5.3 3.3 1.2

Threshold current (mA) 0.50 0.51 0.70 1.20

Slope efficiency (W/A) 0.56 0.71 1.00 1.07

Max power (mW) 5.4 7.1 12.7 14.2

Roll-over current (mA) 14.5 15.5 19.2 20.1

27

6.4 ps 3.3 ps5.3 ps 1.2 ps

Figure 3.4.9 Small-signal modulation response at different bias currents for VCSELs with a 7 µm oxide aperture and different photon lifetimes (at 25ºC). Figure 3.4.10 (left) shows a reduction of the D-factor with decreasing photon lifetime, resulting in a reduction of the resonance frequency and a lower maximum resonance frequency at thermal roll-over. This is due to a reduction of the differential gain, Fig. 3.4.10 (right).

25 °C

6.5 GHz/mA1/2

4.7 GHz/mA1/2

Figure 3.4.10 Left: Resonance frequency vs. square root of current above threshold for an 11 µm aperture VCSEL with different etch depths (25ºC). Right: Differential gain vs. etch depth at three different temperatures.

25 °C

0.35 ns

0.14 ns

Figure 3.4.11 Left: Damping rate vs. resonance frequency squared for an 11 µm aperture VCSEL with different etch depths (25ºC). Right: Gain compression vs. etch depth at three different temperatures. Figure 3.4.11 (left) shows the reduction of the K-factor with decreasing photon lifetime. This is responsible for the reduced damping seen in the modulation response (Fig. 3.4.9). The reduction of the K-factor is a direct result of the reduced photon lifetime. However, eventually the K-factor saturates and damping is no longer reduced. The extracted variation of gain compression with photon lifetime is also shown in Fig. 3.4.11 (right). The relatively small values of gain compression are a result of an active region designed for fast carrier transport and capture. We conclude that, with a reduction of the photon lifetime there is initially a large reduction of damping which increases the modulation bandwidth. With a further reduction of the photon lifetime, damping saturates and the bandwidth is reduced due to the reduction of the resonance frequency. This results in an optimum photon lifetime for maximum modulation bandwidth.

28

The large signal modulation characteristics of the Gen II VCSELs were investigated by transmission measurements using an OM3+ multimode fiber (bandwidth-distance product = 4.7 GHz·km) and a 30 GHz detector (VI Systems). Fig. 3.4.12 shows results from measurements using the 7 µm aperture VCSEL with a photon lifetime of 3.3 ps. In a back-to-back configuration, 40 Gbit/s error-free transmission was achieved with less than 0 dBm of received optical power. This is the highest data rate at error-free transmission ever demonstrated with an 850 nm VCSEL. With 100 m of fiber, data could be transmitted error-free at 35 Gbit/s.

Figure 3.4.12 Measured BER vs. received optical power. Left: back-to-back transmission with the VCSEL biased at 6.5 mA (17 kA/cm2) for 25 Gbps operation and 8.0 mA (21 kA/cm2) for 40 Gbit/s operation. Right: transmission over 100 m fiber with the VCSEL biased at 7.5 mA at both bit rates. The SFDR was found to consistently remain above 90 dB·Hz2/3 throughout the frequency range 1-19 GHz, which represents an improvement over the Gen I VCSELs. A successful 64-QAM OFDM test again proved that the VCSELs are suitable for next generation links employing multi-level modulation formats. Finally, a set of 8 Gen II VCSELs was delivered to VI Systems for mounting on high speed carriers and packaging. 3.5 WP5 Packaging, System-level Evaluation, Standardization, and Exploitation WP5 covers a wide range of activities, from laser diode packaging to optical systems specifications and validation to active participation in IEEE and other Standards groups. In WP5 we assembled high speed (up to 40 Gb/s) test beds and characterized the three types of lasers developed in the VISIT project: 1) the EOM BR VCSEL (WP2); 2) the EOM TWLs (WP3); and 3) the directly current modulated VCSEL (WP4), as the prototype laser devices become available. Also, WP5 includes the active participation in, and reporting of IEEE Ethernet and other industry Standards activities in order to help infuse the VISIT-developed technology into the evolving Standards as soon as is possible, and to help keep the VISIT work focused on the needs of the various evolving optical link markets. Summary of most important achievements:

• Assembled 40G optical test beds at UCAM and TNI and performed bit error ratio (BER) tests, optical eye diagram measurements, fiber link tests, and other high speed measurements on unpackaged and packaged devices.

• Performed high order modulation format tests on VCSELs and EOM BR VCSELs including the following formats: RZ, NRZ, PAM4, QPSK, and QAM16.

• Generated VCSEL device performance data and models and systems specifications based on measurements from prototype DM VCSELs. Used this data to develop system specifications for a DM VCSEL and EOM BR VCSEL driver ICs and the electrical and optical specifications for related short reach 40G optical link systems.

29

• Designed and produced prototype, high frequency (up to 40 Gb/s) VCSEL and EOM BR VCSEL sub-assembly packages and characterized the performance of these packaged components individually and in systems and link tests.

• Actively participated (attended, gave presentations) in numerous Standards and other related industry meetings and workshops including: IEEE P802.3ba 40G and 100G Ethernet Standards meetings, Fibre Channel (T11) 16G and 32G Standards and FCIA workshops, OIF CEI workshops, an Ethernet Alliance Forum, workshops at the Optical Fiber Conference (OFC), OIDA workshops, and more.

For the laser diode packaging work VIS focused on the development of a low-cost, easy-to-manufacture, easy-to-use, high frequency VCSEL module concept that includes both an industry standard connectorized multi-mode fiber pigtail and a standard high-frequency electrical connector for direct plug-and-measure operation. This included 3D high frequency simulation, bonding and assembly studies, and the development of ceramic testboards. In Figure 3.5.1 we show a first prototype VCSEL module consisting of a VCSEL die wire bonded to a high frequency V-connector. The V-connector provides the electrical contact to the VCSEL and is secured to a small metal housing. A multimode fiber pigtail with a lensed tip is positioned over the VCSEL’s emitting aperture and secured in the housing using a rapidly drying epoxy. The measured characteristics of a first VCSEL module are given in Figure 3.5.2. In Figure 3.5.3 we show a packaged CUT VCSEL. To validate the package for DM VCSEL, CUT provided twelve (12) “generation 1” VCSEL chips to VIS. Four (4) of these 12 chips were successfully packaged and characterized by VIS for basic static LIV and high-frequency performance. In general the coupling efficiency is about 90% for a lensed fiber and 50% (- 3 dB) for a cleaved fiber. Figure 3.5.4 shows BER tests from one of the packaged CUT DM VCSELs over up to 100m of OM3+ MMF. One can see that the 100 m optical link introduces a power penalty of ~3dB. Note that these results are only preliminary and issues with a faulty PIN photo-detector limited the performance of the test bed at the time. The optical eye diagrams indicate a capability for “error-free” operation at ≥ 30 Gbit/s data bit transfer rates. In Figure 3.5.5 we demonstrate how error-free transmission can be achieved at both 28 Gb/s and 40 Gb/s over 103m of OM3+ fibre. The latter result has a 4 dB power penalty compared to the 28 Gb/s case. Figure 3.5.1 Prototype DM VCSEL sub-assemblies concepts (left and right), both with an OM3 multi-mode fiber pigtail and a high frequency electrical V-connector, designed for operation at or above 40G. Figure 3.5.2 Measured CW: (a) emission spectra at bias currents of 1, 4, and 8 mA; and (b) measure L-I-V curves for a first prototype packaged DM VCSEL as shown in Fig. 3.5.2. Inset: optical eye diagram at 20 Gb/s.

846 848 850 852 854 856 858 860 862-70

-60

-50

-40

-30

-20

-10

0

VIS-VCSEL-module

Inte

nsity

, dB

Wavelength, nm

CW, RT 1mA 4mA 8mA

0 2 4 6 8 10 12 140

1

2

3

4

5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5VIS-VCSEL-moduleRTIth=0.7 mA

ηdiff

~0.5 W/A

Pow

er, m

W

Current, mA

Volta

ge, V

(a) (b)

20 Gb/s optical eye

30

Figure 3.5.3 White light microscope images of a CUT Gen I VCSEL (left) and a packaged Gen I VCSEL (center and right).

Figure 3.5.4 BERT results of DM VCSEL Package Module # 4 at room temperature. These measurements were performed at TUB on a prototype CUT VCSEL module assembled by VIS.

Figure 3.5.5 LIV characteristics of a packaged VCSEL (left) and record bit-error ratios at 28 Gb/s and 40 Gb/s over 103 m OM4 fibre for a packaged VIS VCSEL. These measurements were performed at TUB on a prototype VIS VCSEL module assembled by VIS. In Figure 3.5.6 we show a packaged EOM BR VCSEL. VIS delivered several of these packaged devices for subsequent high frequency testing including extensive studies of modulation formats by UCAM.

Figure 3.5.6 Packaged EOM BR VCSEL and an EOM BR VCSEL driver integrated circuit (both left and right images) both mounted with a thermally conducting adhesive and wire bonded onto a high frequency ceramic printed circuit test board.

EOM BR VCSEL die

31

To facilitate the addition of a VCSEL driver IC to the packaged VCSEL we developed a VCSEL ceramic printed circuit board concept as shown in Figure 3.5.7a, where a VCSEL and a VCSEL driver IC are glued and wire-bonded onto a test board. We also developed a second a second test board for flip-chip bonding as shown in Figure 3.5.7b.

(a) (b)

Figure 3.5.7 Digital images of a prototype: a) VCSEL and VCSEL driver IC glued and bonded onto a HF ceramic test board; and b) a top down view of a computer generated TOSA test board designed for a flip-bonded VCSEL and an optional VCSEL driver IC. Next we report on our series of high frequency tests performed at TNI. A schematic of the high frequency test set up at TNI is shown in Figure 3.5.8. The results of bit error ratio (BER) tests on packaged VCSELs at 20 Gb/s is shown in Figure 3.5.9. Next as shown in Figure 3.5.10 error-free transmission (BER = 1e-9) was measured for 40 Gb/s B2B, 36 Gb/s over 100 m, and 32 Gb/s over 300 m OM4 with corresponding bias currents of 11.3 mA, 10.6 mA, 8.2 mA. The pseudorandom binary sequence (PRBS) pattern length was 27-1 and the peak-to-peak modulating voltage was 0.52 Vpp. The corresponding VCSEL emission spectra are presented in Figure 3.5.11 and 3.5.12. For comparison, the results of BER tests on a CUT Gen 2A VCSEL module performed by TUB are shown in Figure 3.4.12 (in the previous section).

Figure 3.5.8 The 40 Gb/s test bed configuration used at TNI.

VCSEL die

VCSEL driver IC

site for a flip-bonded VCSEL

32

Figure 3.5.9 20 Gb/s link measurements using 100 m OM4 MMF and one of the Gen 1 DM VCSEL supplied by CUT. These measurements were performed at TNI.

Figure 3.5.10 BER tests and corresponding eye diagrams for the B2B and transmission cases up to 300 m for packaged CUT Gen 1 VCSELs. These measurements were performed at TNI.

Figure 3.5.11 Unmodulated and modulated spectra at (a) 25 Gb/s and (b) 40 Gb/s for packaged CUT Gen 1 VCSELs.

(a) (b)

33

Figure 3.5.12 Modulated spectra as a function of transmission via different fibre lengths (B2B, 100 m and 300 m) at (a) 25 Gb/s and (b) 35 Gb/s for packaged CUT Gen 1 VCSELs.

Figure 8a shows “back-to-back” BER curves for 25 Gb/s and 40Gb/s using the Generation 2A VCSELs. Error-free operation is achieved even at 40Gb/s, although with a 3.5dB power penalty. Figure 8b shows transmission results over 100m of OM3+ fibre. At 25Gb/s there is no evidence of a transmission penalty. At 40Gb/s there is a small penalty and some evidence of error flooring at 10−12. Error! Reference source not found. 3.5.13 shows the BER measurements for the Chalmers Gen 2A (SN: 113-10-63-S1) packaged device for each of the worst case tributaries. B2B performance has improved slightly over Gen 1 device (this is due to the improved photon lifetime) but the transmission penalty has increased for these measurements due to an increase in spectral width of the devices. Modal filters will be included in next generation devices to reduce the modulated spectral width and improve overall transmission performance.

Figure 3.5.13 Results of BER measurements for the CUT Gen 2A devices. These measurements were performed at TNI. We also characterized EOM BR VCSELs. These structures include a multimode oxide confined VCSEL operating at a wavelength of 850 nm. The VCSEL mesa diameter is 35 μm and the active region consists of multiple GaAs quantum wells and an 11 μm oxide aperture. The LI and VI characteristics of the EOM BR VCSEL are shown in Fig. 3.5.14. The threshold current is found to be 1.6 mA and is independent of the bias applied to the EOM section demonstrating negligible absorption of light in the EOM section. From the LI characteristic, the slope efficiency is calculated to be 0.18 W/A with 0 V applied to the EOM section and reduces to 0.11 W/A as the EOM voltage rises to 10 V. The dynamic resistance at 6 mA is found to be 58 Ω which undergoes very little change as the EOM voltage is varied.

34

The spectra of the EOM BR VCSEL are shown in Fig. 3.5.15 under two different EOM bias conditions. Clear multimode operation is seen with a peak lasing wavelength around 842 nm. The spectra remain the same regardless of the EOM voltage demonstrating that the EOM section has little effect on the VCSEL lasing wavelength as desired. Figure 3.5.14 The LI and VI characteristics of an EOM VCSEL. These measurements were performed at UCAM.

Figure 3.5.15 Emission spectra of the EOM BR VCSEL.

A network analyzer is used to measure the frequency response of the EOM BRVCSEL. The highest bandwidth of the EOM VCSEL is 18 GHz when biased at 8 mA with a strong resonance seen at 5 GHz and does not change with varying the EOM voltage. The response curves are shown in Fig. 3.5.16.

Figure 3.5.16 Frequency response of the EOM BR VCSEL.

Frequency, GHz

10 V

0 V 1 V 2 V 3 V 4 V 5 V 6 V 7 V 8 V 9 V 3dB

0 5 10 15 20 25 30 -10

-5

0

5

10

15

Mod

ulat

ion

Res

pons

e, d

Bo

0 2 4 6 8 10 12 140

0.5

1

1.5

2

2.5

3

Current, mA

Volta

ge, V

0-10 V0-10V

Current, mA

Volta

ge, V

0 2 4 6 8 10 12 140

100

200

300

400

500

600

700

800

Current, mA

Opt

ical

Pow

er, µ

W

0-10 V0-10V

Current, mA

Opt

ical

Pow

er, µ

W

35

NRZ modulation is a very common modulation format in widespread use and is one of the formats chosen to modulate the EOM VCSEL with. However, photoluminescence studies performed at TUB have identified an unexpected absorption region very close to the VCSEL gain region resulting in intensity spikes followed by damped oscillations which negatively impact the NRZ eye quality. Due to this issue, NRZ transmission is unsuccessful. RZ modulation is another common modulation format which is not as adversely affected by the absorption feature present in the device. This format is therefore used to modulate the VCSEL. The eye diagram produced using this format is shown in Fig. 3.5.17. The VCSEL is biased at 8 mA with 4 V applied to the EOM section. A 10 Gb/s RZ signal with a 25% duty cycle and 0.5 Vp-p amplitude is used to modulate the device. The eye quality is sufficient for bit error rate measurements. The BER test results are shown in Fig. 3.5.18. The lowest BER measured is 4 x 10-11 under back to back optical conditions where a short length of 50 µm fiber patch cord is present. This reduces to 2 x 10-10 when 100 m of OM3 fiber is used indicating successful error-free transmission with a 0.6 dB power penalty compared to the back-to-back measurement.

Figure 3.5.17 Results RZ eye diagrams. These measurements were performed at UCAM.

Figure 3.5.18 BER plot using 10 Gb/s RZ modulation. These measurements were performed at UCAM.

Received Optical Power, dBm

-2

-4

-6

-8

-10

-12

-14

-20

log 1

0(B

ER)

36

Similar BER tests were performed using the EOM BR VCSELs and QPSK modulation. For this modulation scheme the lowest recorded bit error rate after transmission over a 100 m OM3 optical fiber link at 7 Gb/s and 8 Gb/s per channel is 5.5 x 10-12 and 9.4 x 10-12 respectively with a power penalty of 0.9 dBm compared to their respective back-to-back cases. The BER plot is shown in Fig. 3.5.19.

Figure 3.5.19 Eye BER measurement of EOM VCSEL under QPSK modulation. These measurements were performed at UCAM.

Table 3.5.1 Selected Standards, application areas, data transfer bit rates, and expected implementation timelines that could be directly impacted by the high speed optical data link technology under development in the VISIT project. The term OIF is the Optical Interconnect Forum; USB is Universal Serial Bus; PCI is the Peripheral Component Interface; and HDMI is High Definition Multimedia Interface. Standard applications generation Gb/s timeline Fibre Channel SANs 16G FC 17 2009 32G FC 34 2012 Infiniband high-end computing per lane 20 2011 OIF chip-to-chip CEI-28G-SR 25-28 2010 USB consumer USB 3.0 4.8 2009 products USB 4.0 ~20-60 ~2014-15 PCI Express optical links Gen3 8 2009 Gen4 17 ~2011-12 HDMI home/TV/audio 1.3 3x3.4 2006

-2

-4

-6

-8

-10

-12

-14

-20

Received Optical Power, dBm

log 1

0(B

ER)

37

We completed a first reliability test on a sample of 1x4 DM VSCEL arrays – 4 units giving a total of 16 VCSELs. The use of arrays in ageing tests reflects the fact that many higher bandwidth applications will rely on arrays of multiple VSCELs to provide the predicted higher data rates from 100 Gb/s and beyond. The VCSELs were previously proven suitable for data rates of between 25 to 40 Gb/s and were selected with aperture diameters of 9-10 m, i.e. a relatively large aperture size which was prev