component and system level studies of radiation damage impact on reflective electroabsorption...

386 IEEE TRANSACTIONS ON NUCLEAR SCIENCE, VOL. 60, NO. 1, FEBRUARY 2013

Component and System Level Studies of RadiationDamage Impact on Reflective Electroabsorption

Modulators for Use in HL-LHC Data TransmissionSpyridon Papadopoulos, Student Member, IEEE, Sarah Seif El Nasr-Storey, Jan Troska, Member, IEEE,

Ioannis Papakonstantinou, Member, IEEE, François Vasey, Member, IEEE, and Izzat Darwazeh, Senior Member, IEEE

Abstract—We investigate the radiation hardness of ReflectiveElectroabsorption Modulators (REAMs) for fluence levels up to��

�� 24 GeV p/cm� to explore the possibility of utilisingREAMs for data readout at the High-Luminosity Large HadronCollider (HL-LHC). The high fluence levels used in the experimentand the online spectral measurements carried out provide signifi-cant insight into the radiation damage mechanism. The radiationlimits of an architecture based on REAMs are established andcompared to LHC and HL-LHC fluence levels.

Index Terms—Optical links, optoelectronic devices, radiation ef-fects.

I. INTRODUCTION

T HE Large Hadron Collider (LHC) [1], the largest and mostpowerful particle accelerator in the world, has been in op-

eration since 2009. Two counter-propagating beams of particlescollide at four points around its 27 km circumference. The colli-sion products are monitored using specialized detectors, called“experiments” [2]. At its design luminosity of cm s ,the LHC will generate collisions/second. The magnitudeof the collision rate leads to significant technical challenges.These include, among others, the enormous amount of data thatneed to be transmitted from the detectors for further processingand the harsh radiation environment generated around the colli-sion point. In order to handle the required data rate, LHC exper-iments make use of optical link technologies. Tens of thousandsof optical links are currently installed at the four LHC experi-ments. The optoelectronic components that are installed on-de-tector will need to withstand a significant amount of radiationover their lifetime.

Manuscript received July 12, 2012; revised October 18, 2012 and November24, 2012; accepted November 28, 2012. Date of publication January 15,2013; date of current version February 06, 2013. This work was supported byACEOLE, a Marie Curie mobility action at CERN, funded by the EuropeanCommission under the 7th Framework Programme.

S. Papadopoulos is with the PH Department, CERN—European Organizationfor Nuclear Research, CERN CH 1211, Geneva 23, Switzerland and also withthe Department of Electronic and Electrical Engineering, University CollegeLondon, London, WC1E 7JE, U.K. (e-mail: [email protected]).

S. Seif El Nasr-Storey, F. Vasey, and J. Troska are with the PH Department,CERN—European Organization for Nuclear Research, CERN CH 1211,Geneva 23, Switzerland.

I. Papakonstantinou and I. Darwazeh are with the Department of Electronicand Electrical Engineering, University College London, London, WC1E 7JE,U.K.

Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TNS.2012.2231964

The forthcoming upgrade of the LHC to High-LuminosityLHC (HL-LHC) will lead to a luminosity increase by an orderof magnitude to cm s [3] and is set to increase boththe required data rate and the amount of radiation dose that allon-detector components will be exposed to. The change in thedata rate requirement and the radiation levels necessitates an op-tical link upgrade. Furthermore, the upgrade of the optical linksis viewed as an opportunity to move to a more efficient infra-structure and to evaluate the merits of using new technologies. Inour previous work [4] and [5] we suggested more efficient waysto transmit information to and from the detectors. In [4] this isachieved through the use of a bidirectional, point-to-multipointarchitecture, while in [5] by using a bidirectional infrastructurethat is matched well to the highly asymmetric, upstream-inten-sive nature of traffic generated in particle physics experiments.In [5], we also re-visited the concept of using Reflective Elec-troabsorption Modulators (REAMs) to transmit information inthe upstream direction. Electroabsorption modulators (EAMs)are devices that modulate light through the variation of absorp-tion of semiconductor material when an external electric field isapplied. Their use in particle physics applications was first sug-gested and investigated in the 90s for the LHC [6]–[8] due totheir potential radiation hardness, as well as their low mass andpower consumption [9]. However, difficulties related to the im-maturity of the technology and to the complexity of the systemimplementation at the time led to the implementation of dif-ferent types of optical links based on directly modulated lasers[10], [11].

The EAM technology has since matured and devices are nowcommercially available. In [5] and [12] we have demonstratedthe feasibility of using EAMs in Multi-Gb/s particle physics ap-plications from a network perspective. However, there is a lackof studies on the behavior of EAMs at HL-LHC radiation levels.Although the behavior of lasers and photodiodes in such condi-tions has been recently studied [13], there is no similar researchon EAMs. We therefore consider timely a study of the radia-tion effects at HL-LHC fluence levels on EAMs and their impacton network performance. This is the topic of this paper whichaims to conclude whether EAMs can withstand HL-LHC radia-tion levels. To achieve this, we study the behavior of the deviceat higher fluence levels than have ever been studied before. Asprevious studies [6]–[8] and [14] were targeting lower fluencelevels, no device failure was observed. In this work the highfluence levels used, allow us to explore the radiation hardnesslimits of the EAMs in order to gain significant insight into the

0018-9499/$31.00 © 2013 IEEE

PAPADOPOULOS et al.: COMPONENT AND SYSTEM LEVEL STUDIES OF RADIATION DAMAGE IMPACT 387

radiation-induced performance degradation mechanism. To thisend we have also carried out online absorption spectrum mea-surements. The measurements provide us essential informationon the evolution of fundamental device properties with radia-tion. To our knowledge, only [14] reported on the impact of ir-radiation on absorption spectra. However, the authors identifiedonly small changes of the absorption spectrum with radiation,owed to the relatively low fluence levels used. Moreover pre-and post-irradiation absorption spectrum changes were exam-ined and reported only for the same and not across bias volt-ages. The authors did not explore the impact of changes in theabsorption spectrum to significant aspects of modulator perfor-mance from a system-level perspective, such as the extinctionratio and optimum wavelength of operation. In this study, wedemonstrate how the absorption spectrum of an EAM changesboth for the same and across different voltages with increasingfluence. We also utilise the significant insight gained throughthe online absorption spectrum measurements to carry out asystem-level analysis that establishes, for the first time, the max-imum operating fluence levels of a network using electroabsorp-tion modulators.

This paper is structured as follows; Section II describes theexperimental setup used to carry out the measurements as wellas the measured quantities and the measurement process. It alsoprovides information on environmental variables, such as thetemperature and radiation levels. Section III shows how the per-formance of the devices changes as they were exposed to in-creasing fluence levels. It demonstrates the radiation hardnesslimits of EAMs, highlights changes in the device behavior thatlead to degradation in the performance of the device as a mod-ulator and provides possible explanations for these changes. Fi-nally Section IV investigates the system-level aspects of the ra-diation impact and explores the fluence levels that, an architec-ture using EAMs, could withstand.

II. EXPERIMENTAL SETUP

Two unpackaged, bare-chip 10 Gb/s REAM devices, of typeR-EAM-1550-LS purchased from CIP photonics [15], wereused for the test. The two devices were irradiated at the IRRAD3irradiation zone of the CERN-PS (Proton Synchrotron) easthall irradiation facility [16]. The IRRAD3 irradiation zone usesthe PS-T7 proton beam-line that provides 24 GeV/c protonswith a flux in the region of p cm s [17].The samples were positioned on a remotely controlled “table”that could move the samples in and out of beam. The beamintensity was monitored during irradiation by a secondaryemission chamber while the beam position and profile wasmeasured by an instrument based on proton-induced secondaryelectron emission (SEE) from thin aluminium foil [18]. Themeasurements provided a 10% accuracy compared to the realfluence value, as the position of the device relative to the beamspot can not be exactly determined.

The main objectives of the test were to measure the radiationtolerance of the REAMs, as far as their capability to modulatelight is concerned and to gather information on the operation ofthe devices during irradiation in order to gain insight into theradiation damage mechanism. At the same time, our aim was touse a simple test setup utilising only commercial off-the-shelf

Fig. 1. Measurement setup.

components. As the devices tested were bare-chip, modu-lating them would have required development of submountswith impedance-controlled traces. To avoid this complication,only static measurements were carried out during the test.The selected metrics to be measured were the modulator-in-duced attenuation, the photocurrent, the leakage current andthe absorption spectrum of the REAMs. All metrics weremeasured for voltage levels covering most of the operatingvoltage range—more precisely between 0.6 and 3.6 V. Allthe measured quantities were being measured continuouslyduring irradiation.

The setup used to carry out the measurements is shown inFig. 1. The unpackaged REAM devices and a PT-100 sensor tomonitor the environmental temperature were placed into the ir-radiation environment. All the rest of the setup was placed inthe counting room that is protected from irradiation, located ata distance of 20 m from the irradiation space. Two 20 m singlemode optical fibers were used to connect the REAMs to the restof the setup. A Covega SLD-1108 1550 nm SuperluminescentLight Emitting Diode (SLED) and a Fitel FOL15DCWB 1553nm Distributed Feedback (DFB) laser diode (LD) were used toilluminate the two REAMs. It should be noted that although theREAM specifications provided by the manufacturer are quotedat 1550 nm, the performance of the devices at the LD operatingwavelength (1553 nm) in terms of insertion loss and extinctionratio was within 1 dB of their 1550 nm values; the REAMsactually performed slightly better at 1553 nm. The SLED andLD were current- and temperature-stabilized. The SLED wasused to carry out online measurement of the absorption spec-trum of the REAMs over a wide wavelength range. The LD wasmainly used to increase the reliability of the setup in case theSLED failed and to cross-check the results of the spectral mea-surements. It is also the device to be used in the architectureof our interest, where a single-wavelength LD would be usedto illuminate two REAMs (REAM A and REAM B). A JDSUSB-Series fiber-optic switch was used to connect the LD andSLED to REAM A, the “monitor” channel and to REAM Bas shown in Fig. 1 in a sequential manner, i.e., only one lightsource was connected to only one out of the three used switch


outputs at any point in time. Typically, in a network configu-ration using REAMs, circulators are preferred over splitters toconnect the REAMs to the CW source and the receivers due totheir superior performance in terms of isolation between portsand insertion loss. However, in our test, as only static measure-ments were carried out, splitters offered a cheap, simple alterna-tive with satisfying performance. Therefore 90/10 splitters wereused to connect the two outputs of the switch to the REAM de-vices. In both cases the output of the switch was connected tothe 90% input and the REAM to the “common” of the splitterto maximize the optical power injected to the REAM and mini-mize the power directly coupled from the LD/SLED to the mea-surement instruments. Indeed, good isolation of 40 dB wasmeasured between the 90% and the 10% inputs in all cases. The10% outputs of the splitters routing back the light reflected bythe REAMs and the “monitor” output of the switch were con-nected to three out of the four inputs of a 1:4 coupler. The fourthinput remained unconnected. The “common” of the coupler wasconnected, again via a 90/10 splitter to GPIB controlled op-tical power meter and Optical Spectrum Analyzer (OSA). Thelosses of all system components, as well as of the assembledsystem were measured before the irradiation test. These mea-surements, along with the measurements carried out when theLD/SLED were connected to the “monitor” channel were usedto deduce the absorption characteristics of the REAM presentedin this paper. Each of the REAMs was connected electrically toa picoammeter/voltage source to control the reverse bias voltageand measure the photocurrent or dark current (under no illumi-nation) flowing through the device under test (DUT).

The setup was controlled via LabView. In every measurementloop iteration, which was designed to last less than 20 mins, thetemperature, optical power, absorption spectrum and photocur-rent or dark current, depending on whether the measured REAMwas illuminated or not, were recorded. In each iteration the re-verse bias voltage was changed in steps of 0.2 V from 0.6 V to3.6 V. The two REAMs were irradiated for a period of 20 daysand measurements were continously taken over 27 days exceptfor short intervals of unintended interruption. Fig. 2 shows theevolution of the environmental variables, i.e., the fluence levelsthe devices were exposed to and the temperature in the irradia-tion space over this time period. As the diagram shows the flu-ence increased gradually from 0 to the level of24 GeV p/cm .

III. IMPACT OF IRRADIATION ON REAM PERFORMANCE

EAMs modulate light utilising the change in absorption ofthe semiconductor material that can be caused by one of twomechanisms; the Franz-Keldysh Effect (FKE) in the bulk activelayer or the Quantum Confined Stark Effect (QCSE) in MultipleQuantum Wells (MQWs) [19]. The modulators used in this workare based on QCSE.

Regardless of the underlying mechanism used to modulatelight, in order for an EAM to be operational, the modulator-in-duced attenuation has to follow the variations of the externallyapplied voltage. Moreover, the higher the variation of the atten-uation, the more efficient the modulation of light. Therefore an

Fig. 2. Fluence and temperature vs. time.

initial assessment of the impact of irradiation on REAM per-formance can be carried out by examining the evolution of at-tenuation at different reverse bias voltage levels as a functionof fluence. Fig. 3 and 4 show the attenuation at 0, 1, 2 and 3 Vmeasured using the LD, for REAM A and B, respectively. Forboth REAMs, the attenuation levels converge with increasingfluence, i.e., the modulation efficiency decreases. Beyond a flu-ence of 24 GeV p/cm the induced attenuation isno longer influenced by the applied reverse bias voltage for ei-ther of the modulators, thus the modulators have stopped beingoperational. This finding explains why previous studies did notobserve any significant performance degradation with irradia-tion, as the fluence levels used in [7], [8] and [14] were below

p/cm . The figures also clearly show that REAM Aperformed worse than REAM B, as its attenuation levels con-verged at fluence levels of 24 GeV p/cm . REAMA fared worse in all measured metrics; this is why from nowon the discussion will focus on REAM A that represents theworst-case scenario. A final important observation is that no an-nealing was observed after the end of irradiation.

Although Fig. 3 and 4 provide useful information related tothe radiation resistance of the REAMs, this information is notsufficient to explain the mechanism behind the radiation-in-duced performance degradation. Based on this informationwe cannot conclude whether the degradation is caused by achange in the wavelength of the excitonic absorption peak, alowering of the absorption maximum or a different cause. Thesnapshots of the evolution of the REAM absorption spectrumfor different values of reverse bias voltage as fluence levelsincrease, measured using the SLED and shown in Fig. 5 to 8,provide more insight into the degradation mechanism.

Fig. 5 uses the measured spectra of REAM A to demonstrategraphically the principle of operation of QCSE-based modula-tors. As shown in the figure, the change in the induced attenua-tion at a fixed wavelength for varying voltage levels, is causedby a move of the exciton absorption peak to longer wavelengths,its broadening and a reduction of its strength with increasing re-verse bias voltage [19]. Following this explanation of QCSE,


Fig. 3. Modulator-induced attenuation vs. fluence (REAM A).

Fig. 4. Modulator-induced attenuation vs. fluence (REAM B).

Fig. 5 to 8 show clearly why the induced attenuation levels ata specific wavelength for different reverse bias voltages—pre-sented in Fig. 3 and 4—converge. It is observed that although theshape of the spectrum remains largely unaffected by irradiationat 0 V it is obvious that it stops responding to changes in the ap-plied reverse bias voltage with increasing fluence. Thus the ex-citonic peak broadening and move to longer wavelengths is nolonger observed with increasing voltage. To our best knowledgethis effect is reported here for the first time in EAMs. A possibleexplanation for the observed behavior is the modification of thep-i-n structure in which the quantum wells are embedded due todonor and acceptor compensation as a result of the irradiation[20]. As the doping concentration within the device changes dueto displacement damage, the electric field distribution with theapplication of an external voltage also changes. Consequentlythe electric field applied to the quantum well region may be sig-nificantly weaker.

Fig. 5. REAM A Spectrum—before start of irradiation.

Fig. 6. REAM A Spectrum—Fluence � �� p/cm .

Leakage current is an important parameter as its value is di-rectly related to power consumption that should be kept min-imum for devices installed on-detector. The evolution of leakagecurrent as a function of fluence for different reverse bias voltagelevels is shown in Fig. 9. Irradiation led to a significant increasein leakage current, from values ranging in the order of tens ofnA—actually it was measured to have a value of 1 nA for 1 Vand 18 nA for 3 V at the beginning of the experiment—toseveral A by the end of irradiation for a voltage level of 3 V.More precisely leakage current at the end of irradiation reached6 A at 3 V. This value of leakage current would lead to a powerconsumption increase of 10% at 3 V assuming normal REAMoperation—i.e., ignoring the REAM responsivity decrease be-cause of irradiation and using the link budget calculations of [5]to evaluate the required incident optical power to the REAM.On the other hand, contrary to the modulator-induced attenua-tion case, annealing was observed for leakage current as shownin Fig. 10. After a radiation-induced increase of almost three


Fig. 7. REAM A Spectrum—Fluence � � � �� p/cm .

Fig. 8. REAM A Spectrum—Fluence � �� p/cm .

orders of magnitude, leakage current at 3 V decreased by 30%for REAM A after 5 days (120 hours) of annealing at roomtemperature.

IV. SYSTEM-LEVEL ANALYSIS

Following the impact of irradiation on REAM performanceanalysed in Section III, we link the above discussed parametersrelated to modulator performance, to parameters affecting net-work performance. The network architecture of interest is thearchitecture presented in [5] and [12] and reproduced in Fig. 11.As Table I—that shows the requirements of the upgraded opticallinks of the LHC—illustrates, particle physics experiments gen-erate traffic of highly asymmetric, upstream intensive nature.The architecture of Fig. 11 is designed to match this particulartraffic pattern. It allows bidirectional transmission over a singlefiber and provides the necessary network resources for upstreamtransmission, while simultaneously allowing resource sharingin the downstream direction. As Fig. 11 shows, the downstream

Fig. 9. Leakage current vs. fluence (REAM A).

Fig. 10. Leakage current annealing vs. time (REAM A).

transmitter as well as the continuous-wave (CW) light sourcethat provides the light to be modulated by the REAM for up-stream transmission, are shared by all the nodes in the detectorenvironment.

These two components are located in the counting room thatis an environment shielded from radiation. An upstream/down-stream pair use the same fiber with information being separatedusing different wavelengths as illustrated in Fig. 11. The wave-length used for upstream transmission, may be in the 1520–1560nm window, where REAMs, optical transmitters and receiversare commercially available. The downstream wavelength in thisexample is chosen to be 1490 nm mainly to explore the possi-bility of extending the work presented in [4].

To estimate the system-level impact of radiation we considerthe effect of modulator performance degradation on powerbudget [21]. Electroabsorption modulators affect the powerbudget primarily via two parameters; their insertion loss andtheir extinction ratio. The insertion loss is defined as theresidual loss of the modulator when a “1” is transmitted, whilethe extinction ratio (ER) is the ratio of the optical power at thebinary “1” to the power at the binary “0” level [22],


Fig. 11. Network architecture [5], [12].

TABLE IREQUIREMENTS FOR THE UPGRADED OPTICAL LINKS

. The impact of a non-perfect extinction ratiocan be taken into account during link budget calculations usingthe extinction ratio power penalty, defined as the additionalpower required—in terms of average transmitted power—toachieve the same performance as in the case where a perfectextinction ratio is used [21]. The extinction ratio power penaltyis calculated using the following equation [21]:

(1)

where . The impact of the insertion loss and theextinction ratio can be embedded into one performance indi-cator that we are going to term “overall power penalty” and de-fine as:

(2)

where is the REAM insertion loss and theREAM extinction ratio power penalty.

The overall power penalty as well as the extinction ratiopower penalty at 1553 nm as functions of fluence are shownin Fig. 12. However, calculating and plotting metrics directlyrelated to network performance—such as the extinction ratiopower penalty and the overall power penalty—as functions offluence is one step towards assessing the suitability of REAMsas front-end devices at the HL-LHC. In order to complete thistask, appropriate qualification fluence levels need also to bespecified. The limits set for the Versatile Link Project [23]

Fig. 12. Overall and ER power penalty vs. fluence at 1553 nm (REAM A).

could be used for that purpose, however, these are expressed in20 MeV neutrons. A conversion to 24 GeV proton equivalentfluence levels has to be carried out. Ideally, the value of sucha conversion factor would be determined by a comparativestudy, where REAMs are exposed to 20 MeV neutrons and24 GeV and the results are combined for its calculation. Forour purposes, a preliminary study can be carried out usingthe findings of [24] and [25] to calculate the 24 GeV protonequivalent fluence levels. More precisely, [25] reports that20 MeV neutrons are 1.9 times less damaging than 300 MeV/cpions, that in turn are 8.3 times more damaging than 0.8 MeVneutrons for devices fabricated in similar material systems. In[24] is also reported that 0.8 MeV neutrons are 8.4 times lessdamaging than 24 GeV protons, for the same material systems.Combining these results, we derive a 1.9 conversion factorbetween 20 MeV neutrons and 24 GeV protons—with theprotons being more damaging. The fluence level of the LHC[11] has also been calculated using the same methodology andis shown in the figures.

The results presented in Fig. 12 clearly show that althoughthe increase of the overall power penalty is low at LHC andHL-LHC Calorimeter fluence levels, it is very high for HL-LHCTracker fluence levels. More precisely, the overall power penaltyhas increased by more than 10 dB. In [12] we showed that a10 dBm CW light source can be used to serve 16 or 32 users,depending on the sensitivity of the upstream receiver. A 10 dBincrease of the power penalty would lead to a reduction in thenumber of users sharing the same CW light source by a factorof at least 8 resulting in 2 or 4 users being served. Moreoveras Fig. 12 shows most of the overall power penalty increase iscaused by an increase in the extinction ratio power penalty. Thisimplies that the extinction ratio has reached low values, thus themodulators may no longer be operational and an increase in thepower of the CW light source would not alleviate the problem.

To investigate whether an alternative solution can be found,we used the absorption spectrum measurements, carried outwith the SLED to calculate the extinction ratio evolution withincreasing fluence under two different scenarios; in the firstscenario the CW light source wavelength is assumed to be setto the optimum wavelength of operation, that minimizes the


Fig. 13. Extinction ratio vs. fluence with and without use of optimum wave-length tracking mechanism (Ream A).

overall power penalty before irradiation and then it is heldfixed. In the second scenario, the CW light source wavelengthis assumed to track the optimum wavelength of operation mini-mizing the overall power penalty. The results of the calculationare shown in Fig. 13. The fluence qualification levels for theLHC Tracker and for the HL-LHC Tracker and Calorimeter arealso displayed. We set the minimum extinction ratio thresholdat 4 dB as below this value the extinction ratio power penaltyrises rapidly, even with small variations of the extinctionratio value [21]. As the results show the extinction ratio fallsbelow the 4 dB extinction ratio threshold for fluences below

p/cm , in the first scenario. The use of an optimumwavelength tracking mechanism can deliver a performanceimprovement of 30% in terms of the maximum operatingfluence. However, in both cases the minimum extinction ratiothreshold is well below the p/cm fluence levelexpected in the inner detector layers of the HL-LHC.

Our analysis shows that a REAM-based architecture can onlybe considered for use at the outer detector layers and does notappear to be an appropriate solution for the inner detector layersof the HL-LHC. Moreover the extinction ratio values used in thispaper are based on static measurements. Although QCSE is afast process and the frequency response of commercial REAMsat the data rate of our interest is limited by the associated elec-tronics, further testing is required to investigate the impact ofradiation on the frequency response of the device. In that sense,the extinction ratio values reported here represent a best-casescenario where static measurements have been used to predictthe dynamic performance of the modulators. It is safer to usethis extrapolation of static-to-dynamic performance to establishan upper bound on the radiation resistance of REAMs.

V. SUMMARY AND CONCLUSION

In this paper we have studied the impact of irradiation onQCSE-based electroabsorption modulators at fluence levelsnever studied before. As previous studies focused on lowerfluence levels, the radiation hardness limits of REAMs were notwell established. Our study shows that the modulation efficiencyof the devices close to their designated wavelength degrades

significantly for fluence levels above 24 GeV p/cmand are no longer operational at 24 GeV p/cm .Irradiation causes convergence of modulator-induced atten-uation levels for different voltage levels and an increase inleakage current. One fundamental difference in the behaviorof these two parameters is the fact that annealing has beenobserved for the leakage current, but not for the attenuationlevels. Online spectrum measurements were carried out thatprovide insight into the degradation mechanism. The ab-sorption spectrum retains its shape even for fluences above

24 GeV p/cm , indicating that the quantumwell structure remained intact, but it becomes gradually lessresponsive to variations of the externally applied voltage withincreasing fluence. A possible explanation for this effect is themodification of doping concentrations due to irradiation thatchanges the electric field distribution within the device.

The impact of the component-level performance degradationon network performance was also investigated. We showed thatalthough the overall power penalty caused by REAM operationremains stable at LHC Tracker and HL-LHC Calorimeter flu-ence levels, it increases significantly at HL-LHC Tracker flu-ence levels. The increase in the power penalty is mainly at-tributed to the increase in the extinction ratio power penaltyand much less on the insertion loss, that shows comparativelysmall variation during irradiation. We investigated the benefitsof using an optimum wavelength tracking mechanism on extinc-tion ratio. We showed that this mechanism delivers moderate in-creases in the maximum radiation levels at which our networkcan operate. The maximum operating fluence level achieved isstill lower than the HL-LHC Tracker qualification fluence level;thus a network architecture based on REAMs cannot be usedin such an environment. The results of our test do not excludethe use of REAMs in the less hostile HL-LHC Calorimeter en-vironment, where they can possibly provide benefits in terms ofspeed and power consumption, compared to directly modulatedlasers. However, further testing on a bigger sample is requiredto identify the REAM radiation resistance variance and verifythat they can operate in such an environment in the worst-casescenario. The only possible avenue to increase REAM radiationhardness is appropriate device design and the benefits of this ap-proach remain a future direction of research.

ACKNOWLEDGMENT

The authors would like to thank Mr. Christophe Sigaud andMr Maurice Glaser of the PH Department, CERN, Geneva,Switzerland, for their help during the setup of the experiment.

REFERENCES

[1] G. Kane and A. Pierce, Perspectives on LHC Physics. Singapore:World Scientific, 2008.

[2] L. Evans, The Large Hadron Collider: A Marvel of Technology. Lau-sanne, Switzerland: EPFL Press, 2009.

[3] N. P. Hessey, Overview and Electronics Needs of ATLAS and CMSHigh Luminosity Upgrades Tech. Rep. ATL-UPGRADE-PROC-2008-001; ATL-COM-UPGRADE-2008-003, Oct. 2008.

[4] I. Papakonstantinou, C. Soos, S. Papadopoulos, S. Detraz, C. Sigaud, P.Stejskal, S. Storey, J. Troska, F. Vasey, and I. Darwazeh, “A fully bidi-rectional optical network with latency monitoring capability for the dis-tribution of timing-trigger and control signals in high-energy physicsexperiments,” IEEE Trans. Nucl. Sci., vol. 58, no. 4, pp. 1628–1640,Aug. 2011.


[5] S. Papadopoulos, I. Papakonstantinou, F. Vasey, J. Troska, and I.Darwazeh, “A network architecture for bidirectional data transfer inhigh-energy physics experiments using electroabsorption modulators,”in Proc. 16th Eur. Conf. Netw. Opt. Commun. (NOC), Jul. 2011, pp.68–71.

[6] A. Baird et al., “Analog lightwave links for detector front-ends at theLHC,” IEEE Trans. Nucl. Sci., vol. 42, no. 4, pp. 873–881, Aug. 1995.

[7] I. Dawson et al., “Irradiation tests of optoelectronic components forLHC inner-detectors,” Nucl. Instrum. Meth. Phys. Res. Sec. A: Accel-erat., Spectrom., Detect., Assoc. Equip., vol. 387, no. 3, pp. 369–376,1997.

[8] G. Cervelli, K. Gill, R. Grabit, M. Persello, G. Stefanini, and F. Vasey,“Radiation hardness of electro-optic asymmetric Fabry-Perot modula-tors,” Nucl. Instrum. Meth. Phys. Res. Sec. A: Accelerat., Spectrom.,Detect., Assoc. Equip., vol. 400, no. 2–3, pp. 435–446, 1997.

[9] C. Da Via et al., “Lightwave analogue links for LHC detector front-ends,” Nucl. Instrum. Meth. Phys. Res. Sec. A: Accelerat., Spectrom.,Detect., Assoc. Equip., vol. 344, no. 1, pp. 199–211, 1994.

[10] G. Aad et al., “The ATLAS experiment at the CERN large hadron col-lider,” J. Instrum., vol. 3, p. S08003, 2008.

[11] R. Adolphi et al., “The CMS experiment at the CERN LHC,” J. In-strum., vol. 3, p. S08004, 2008.

[12] S. Papadopoulos, I. Darwazeh, I. Papakonstantinou, J. Troska, and F.Vasey, “An electroabsorption modulator-based network architecturefor particle physics applications,” J. Instrum., vol. 6, no. 12, p. C12012,2011.

[13] J. Troska, S. Detraz, S. El Nasr-Storey, P. Stejskal, C. Sigaud, C. Soos,and F. Vasey, “Radiation damage studies of lasers and photodiodes foruse in multi-Gb/s optical data links,” IEEE Trans. Nucl. Sci., vol. 58,no. 6, pp. 3103–3110, Dec. 2011.

[14] P. Goetz, W. Rabinovich, R. Walters, S. Messenger, G. Gilbreath, R.Mahon, M. Ferraro, K. Ikossi-Anastasiou, and D. Katzer, “Effects ofproton irradiation on ingaas/algaas multiple quantum well modulators[for free-space optical communication],” in IEEE Proc. Aerosp. Conf.,2001, vol. 3, pp. 3/1523–3/1529.

[15] [Online]. Available: http://www.ciphotonics.com/[16] [Online]. Available: https://irradiation.web.cern.ch/irradiation/

irrad3.htm[17] M. Glaser, M. Huhtinen, F. Lemeilleur, C. Leroy, P. Roy, and M. Tavlet,

“Radiation test facilities in the new PS east hall at CERN,” in Proc. 5thEur. Conf. Radiat. Its Effects on Compon. Syst. (RADECS) , 1999, pp.136–141.

[18] M. Glaser, F. Ravotti, and M. Moll, “Dosimetry assessments in theirradiation facilities at the CERN-PS accelerator,” IEEE Trans. Nucl.Sci., vol. 53, no. 4, pp. 2016–2022, Aug. 2006.

[19] G. Li and P. Yu, “Optical intensity modulators for digital and analogapplications,” J. Lightw. Technol., vol. 21, no. 9, pp. 2010–2030, 2003.

[20] C. Claeys and E. Simoen, Radiation Effects in Advanced Semicon-ductor Materials and Devices. Berlin, Germany: Springer, 2002.

[21] G. P. Agrawal, Fiber-Optic Communication Systems, 3rd ed. NewYork, NY: Wiley, 2002.

[22] G. Ghione, Semiconductor Devices for High-Speed Optoelectronics.Cambridge, U.K.: Cambridge Univ. Press, 2009.

[23] Versatile Link Technical Specification, [Online]. Available:https://edms.cern.ch/file/1140665/1/VTRx_Spec_v1.8.pdf part 2.1.

[24] K. Gill, G. Cervelli, R. Grabit, F. Jensen, and F. Vasey, “Radiationdamage and annealing in 1310 nm InGaAsP/InP lasers for the CMStracker,” in Photon. Space Environ. VII, San Diego, CA, Jul. 31–Aug.01 2000, vol. 4134, pp. 176–184, Proceedings Paper.

[25] K. Gill, R. Grabit, J. Troska, and F. Vasey, “Radiation hardness qual-ification of InGaAsP/InP 1310 nm lasers for the CMS tracker opticallinks,” IEEE Trans. Nucl. Sci., vol. 49, no. 6, pp. 2923–2929, Dec. 2002.

component and system level studies of radiation damage impact on reflective electroabsorption...

Documents