KM3NET-PP-SSC Report. 18 February 2012

Transcription

1 18 February 2012 KM3NET-PP-SSC Report The terms of reference of the KM3NET-PP SSC The FP7 grant agreement for the Project KM3NeT-PP Annex 1, defines as Task 3 of the Work Package WPA management to Establish a Scientific Standing Committee (SSC), composed of senior scientists, arrange periodic meetings, provide and prepare input to these meetings (mainly from WPD). The SSC was formed in February 2011 with the following composition: A. Aloisio (INFN and University of Naples), A. Bettini (INFN and University Padua; LSC Canfranc), D. Cowen (Penn State University), J. Learned (University of Hawaii), P. Lipari (INFN, Roma), D. Nygren (LBL), A. Olinto (University of Chicago), J. Pouthas (IPN Orsay), C. Spiering (DESY), G. van der Steenhoven (University of Twente). The Project Coordinator requested that the SSC produce its report before the end of the preparatory phase in February 2012, with the task to carry out a critical review of the KM3NeT science case for neutrino astroparticle physics and of aspects related to the design, construction, deployment, commissioning and operation of the detector. The SSC deliberations The 1 st meeting of the SSC was called by the Project Coordinator on March 2011 in Amsterdam. Only a fraction of the SSC members were available in those dates. In the first day, members of the KM3NET Consortium reported on various aspects of the status of the project; in the second half a day the SSC met in camera. D. Nygren was appointed as pro tempore convener. He led the following phase of exchange of opinions, via , between the SSC members on the organization and scoping of the Committee works. A. Bettini was elected SSC chair with an poll led by C. Spiering and P. Lipari on 17 April Having now been completely formed, the SSC had 10 months to complete its tasks. At that late date, it was not any more possible to arrange periodic meetings during the development of the project. The 2 nd meeting of the SSC was held in Trieste on 25 June The meeting was closed and resulted in the document KM3NET-SSC-R&Q2 (annex 1), containing a series of questions, submitted to the Consortium management on 7 July. A second document with more detailed questions, KM3NET-SSC-DQ2 (annex 2), was sent to the management on 24 September. The 3 rd meeting of the SSC was held in Amsterdam on 14 November The document Documentation for SSC Meeting was submitted by the Coordinator on 1 st November (Annex 3); it contains the answers to the SSC questions. The document An alternative strategy for KM3NeT was submitted by L. Resvanis on behalf of the Greek component of the Consortium on 7 November (Annex 4). Finally, the document Comments on the document An alternative strategy for KM3NeT was submitted by the Coordinator on 11 November (Annex 5). The SSC had a hearing of the ASC Chair in the closed session of the 3 rd meeting. In the 1

2 open session, members of the KM3NET-PP Consortium reported to the SSC on the issues raised in the two above quoted questions documents. In the closed session, which followed, the SSC drafted a set of conclusions and defined the sequence of its works for the preparation of the report. The ASC had a hearing of the Chair of the SSC via Skype on The document was elaborated via exchange between the SSC members and approved in a final meeting in EVO conference on The scenario Neutrino astronomy will transform our understanding of the high-energy universe. Experiments are beginning to reach the required sensitivity to make first observations, and/or probe upper limits in a parameter range predicted to be relevant by current theoretical models. IceCube is currently the largest neutrino observatory in the TeV-PeV energy range. The km 3 scale detector was completed at the South Pole, Antarctica in early Sensitive to upward-going neutrinos that have passed through the earth, it primarily observes the northern sky with a pointing resolution of roughly o for muons created by muonneutrino interactions. ANTARES is the largest observatory in the northern hemisphere. Deployed in the Mediterranean Sea, ANTARES has demonstrated the feasibility of operating a deep-sea neutrino observatory for an extended period of time. However, it has a fiducial volume of only 0.04 km 3. Experiments aiming to detect coherent radio emissions from neutrino interactions, such as ANITA, ARA and ARIANNA, are running or under construction and will reach 100 km 3 -scale volumes, but with energy thresholds in the EeV range, i.e., beyond the range covered by IceCube and ANTARES. These efforts will increase the sensitivity at the highest energies where diffuse sources of cosmic neutrinos are expected to contribute. A discovery of the first neutrino source may be imminent. Such a discovery will impact both astronomy and particle physics, as it will open a new window to the universe. We congratulate the KM3NeT Consortium on the progress they have made towards the construction of large neutrino detector in the Mediterranean. In particular for developing the following interesting concepts: The multi-pmt optical module, which could have advantages to the classical single or double PM structures. An elaborated all-data-to-shore concept for the DAQ design as a way to provide maximal flexibility in triggering and event reconstruction while minimizing the complexity of offshore instrumentation. The proposed vertical support structures ( towers ) that, in combination with the multi-pmt optical module, has led to a reduction of the number of wet connectors by a factor of 6 compared to ANTARES. This may substantially improve the reliability of the design, by reducing the leaking risks. However, not all the aspects of these concepts have as yet been optimized, tested and demonstrated, at an adequate scale including deployment, as detailed in the following. 2

3 Finding 1 KM3NeT has an important and unique role to play in the discovery of the first neutrino sources in the southern sky. Its goal is to become the largest neutrino telescope, observing in the southern sky in the 10 TeV 100 PeV energy range, with a pointing resolution of The current KM3NeT design of approximately 5 km 3 might give it the potential to detect theoretically favorable sources in the galactic centre with several years observing time. (IceCube has limited sensitivity to sources in the galactic centre due to its geographical location). Should IceCube discover one or more neutrino sources in the next few years, KM3NeT will be poised to take advantage of its larger volume to more closely study those sources in its field of view. Recommendation 1 The KM3NeT Consortium is urged to perform Monte Carlo optimization studies on a variety of fronts to reduce by roughly a factor of two their figure of merit. The Consortium has defined as their FoM the number of years required for a 5σ discovery of muon neutrinos from a selected group of theoretically favorable galactic point sources (see Annex 3, chapters 3.1 and 3.2). The calculated FoMs range from 7 to 15 years assuming 100% hadronic production mechanism. Reducing the FoM by a factor two will place the potential for discovery in a less model-dependent range and enhance the viability of KM3NeT as a discovery project. One or two additional FoMs may be warranted, in particular one for neutrino-induced showers, from a diffuse flux of extraterrestrial neutrinos. The chosen FoM(s) should be calculated for each site, assuming the same underlying detector hardware components, but with detector geometries (e.g., tower spacing and bar length) optimized for each site. Finding 2 The KM3NeT Consortium has taken good advantage of the Preparatory Phase to perform R&D needed to settle on designs for the geometry, optical module, and DAQ system. The proposed OM, tower, and DAQ designs demonstrate considerable ingenuity and innovation on the part of the KM3NeT Consortium. They have succeeded in reducing the price per detector unit by a significant factor with respect to ANTARES, the pioneering neutrino telescope in the Mediterranean, and in reducing the number of connectivity points substantially. However, several elements have not yet been completely tested. a. Multi-PMT. We find the case for the multi-pmt to be interesting, but its technical feasibility not yet fully demonstrated. It is designed to incorporate PMTs not yet commercially available at the time of writing this report. Consequently, no full-scale prototype unit has been produced so far. A costing exercise has been developed by the Consortium, but in view of the novelty and absence of a full-scale testing of the concept, that estimate needs to be verified. The cost comparisons were done with other less than cost effective optical modules. b. Strings of single PMTs versus bars. The SSC did not receive enough arguments and data to be fully convinced of the benefits of the use of a bar style of optical module tower. First, there has been little demonstration of the practicality of this deployment, which involves very tricky pullout of cables without tangling or sharp bending. 3

4 Second, as was visually evident with the model in the lobby at NIKHEF, the huge ratio of distance between floors and the length of the bar, it is not obvious that in the presence of rotary currents this string will not twist. c. The length of the bar. We noted that the simulations available seemed to point towards larger bars in the tens of meters scale as better from a physics standpoint. However, the constraints of deck space and deployment rigging, limit such bars (in the present design, in containers) to something of order six meters. d. Site dependence of the array design. We note that one detector design is not necessarily optimal for all sites. More shallow sites are likely to require less tall arrays and more dense packing, for example. Deeper arrays with less down going cosmic ray muon background, particularly multiple muons, can afford to be more widely spaced without admitting mis-reconstructing down going muon events as neutrinos. Less bioluminescence at greater depths also allows for lower trigger thresholds and greater reach for studying both muons and showers. These issues have not been properly considered by the Consortium. Recommendation 2 The following points should be further developed and tested as elements of a complete design of the detector a. To demonstrate the claimed superiority of the multi-pmt option the Consortium should: i. build a complete unit, ii. perform adequate tests including deep-sea deployment, iii. produce a reliable cost estimate, iv. include in the comparison the OMs already developed in previous experiments, v. evaluate the performance, including the, site dependent, detection probabilities for minimum ionizing tracks vs. distance. b. Further studies are needed to prove the superiority of the bar tower. It should be noted, in particular, that if having pairs of modules close enough to see coincidences but not so close as to see individual 40 K events, a single string with smaller vertical spacing works as well. c. We think thereby that the trade-off study ought to be with close pairs of strings to achieve the horizontal reconstruction leverage versus the bars. d. As already requested by the SSC (KM3NET-SSC-R&Q2, 4b), the optimization of the detector taking into account the characteristics of its site should be fully developed. This is a necessary process for a proper evaluation of performance and costs in different sites. Finding 3 The Consortium has considerable experience, in all the sites, especially in the Toulon one, with the challenges posed to deployment of equipment in the deep sea, and operating detectors at the bottom of the sea reliably for several years. Nevertheless, also in consideration of the points raised in Finding 2, the SSC finds that the KM3NeT Consortium is not ready to enter a Project Implementation Phase. 4

5 Recommendation 3 Since important elements of the newly developed technology are largely untested in the deep ocean environment, for the next phase of the project we recommend development and deployment of a Demonstrator detector, with enough modules to prove that the new technical concepts reliably survive deep-sea conditions for an extended period of time. Technology arguments like towers versus strings may lead to the development of more than one demonstrator. Therefore, the alternative detector architecture proposed by the Consortium, which is based on single strings without bars, could also be tested in this phase as a separate demonstrator as it may be less sensitive to complications during unfurling. Finding 4 The KM3NeT Consortium has carried out a number of measurements and studies regarding the water quality and other relevant phenomena at the various sites in the Mediterranean Sea: water depth, distance to shore, topology of the sea floor and of the cable path to shore, geological situation, water transparency, background light, sedimentation and bio-fouling, water currents, weather and sea conditions. The SSC strongly believes that intrinsic site parameters, such as depth, water quality, deep-sea surface, and bioluminescence, are of highest importance, as they correlate directly to the quality of the physics measurement and cannot be altered by human intervention. At least two of these parameters, depth and level of bioluminescence, with the corresponding lower duty cycle, rank the Toulon site lower than the other sites. Smaller differences exist between the other considered sites. Recommendation 4 The SSC recommends that the KM3NeT Consortium establish criteria and a process to clearly identify the best single site for the deployment of the full detector. A consistent set of measurements of important site properties are needed to allow for an independent and conclusive review of the results. All the relevant unpublished results should be published in refereed journals and used for comparing sites and optimizing the design. Finding 5 The KM3NeT Consortium is considering employing parts of the neutrino telescope at two or three different deep-sea sites in the Mediterranean. This idea is partly driven by the need to create a modular structure anyhow to allow for safe access by an underwater remotely operated vehicle (ROV), and because simulations have shown that the Figure of Merit of the telescope for muon tracks is not negatively affected when split in two or three parts. However, there is no clear science justification for multiple sites, and the SSC did not find the argument credible that no cost or performance penalties would be incurred. For some observations other than galactic sources, a single site might be preferred. The cost penalties have been quantified by the Consortium as 3 M per site in infrastructures and 1 M per site per year in running costs, for a total, in case of three sites, additional cost of 6 M for construction and 20 M running costs over ten years. While this cost penalty is already large, the arguments leading to those conclusions do 5

6 not appear to the SSC to be sufficiently supported. At present, there appears to be no process in place to choose a single, best site. Recommendation 5 The SSC recommends that the Consortium evaluate accurately and quantitatively the scientific (loss of FoM) and financial (more infrastructural and running costs) consequences of using more than one site. The effects of bioluminescence that are different from site to site should be carefully considered in the evaluation of the FoMs. Finding 6 The KM3NeT Consortium is not yet functioning as a sufficiently coherent collaboration to advance effectively; concern exists that the recommended Demonstrator Phase may not proceed coherently. Recommendation 6 As the various groups within KM3NeT are not cooperating in a sufficiently coherent manner, a stronger, centralized and site-neutral leadership is needed in the next phases of the project. The governance structure and the management plan should be elaborated according to the best practices in use for projects of the KM3NET scale. A Memorandum of Understanding should be defined on that basis by the parties (the ministries and funding agencies). A coherent, consensus-building path towards a well-led and wellmanaged project needs to be established. Finding 7 The establishment of the SSC, foreseen by the KM3NET-PP contract as a task of WP4, was put in place by the project management only in the last stage of the contract. Consequently the evolution of the project has not been monitored by an independent committee, contrary to the usual practices of such activities. Notice that, on purpose already in 1996, the OECD had recommended that the cubic kilometer scale detector development "should be regularly monitored by international peer-review". Recommendation 7 An independent scientific and technical advisory committee should be appointed by the agencies to follow all the phases of the project with continuity of committee membership sufficient to allow a coherent perspective over project lifetime. A project cost review should be undertaken by the committee. The purview of this panel should include a survey of OM design, deep ocean technology, DAQ, reliability, maintenance, and deployment. The panel should include reviewing the risk analysis for all of these elements. The result of such a review will bolster confidence that a Project Implementation Phase would be successful in cost, schedule, and performance. Both detector sensitivity and construction costs of any demonstrator should also be quantified and thoroughly documented. Finding 8 Procedures exist for the fabrication and the approval of the large-scale scientific projects, which however have important differences between different Countries and between 6

7 different Agencies. As an example, the collaborating Governments and Agencies have agreed upon clear procedures for the LHC experiments at CERN (which are efforts of comparable order of magnitude), with two levels of main documents, the Technical Proposal and the Technical Design Report(s), with associated reviews and decisionmaking process. A precise definition of the stages, content of the reports, and critical decisions is important for the multi-cubic-kilometer neutrino observatory project, also in consideration of its multi-agency character. The KM3NET Consortium has produced in the EU funded Design Study phase the Technical Design Report for a Deep-Sea Research Infrastructure in the Mediterranean Sea Incorporating a Very Large Volume Neutrino Telescope, which baselines the project in terms of scope, management and cost. More work has been done by the Consortium in the EU funded Preliminary Phase. The SSC appreciated this effort in the document Documentation for SSC Meeting , which is however still preliminary in its character. Recommendation 8 As already detailed in the previous points, more studies are needed to bring the project to the stage of the final construction approval, including the production of a more advanced (at the above mentioned level) Technical Design Report with all the data that are necessary for such a decision and the development of a full resource loaded schedule, in terms of labor and funds, in analogy to projects of similar scale. The process leading to that, including the reviews mentioned in point 7, should be precisely defined as soon as possible. Finding 9 The current KM3NeT Consortium has been able to bring together existing efforts in Europe aimed at the construction of a large deep-sea neutrino telescope. For the next phases of the project the Consortium is too small to carry out the required work. Recommendation 9 The collaboration will need to expand. In the nearer term, more effort should be employed in the critical effort of simulations. In the longer term, a number of strong groups should be recruited to this challenging project. Non-European collaborators should be encouraged. Finding 10 The scientific viability of the proposed design depends critically on the possible discovery or strong indication of a neutrino point source by an existing neutrino telescope such as IceCube. Similarly, an evidence of the assumed dominant hadronic emission mechanism of a few neutrino sources, if found by gamma-ray astronomers, will have critical impact on the preferred design of KM3NeT and on the science prospects of the detector. 7

8 Recommendation 10 The observation or non-observation by existing ongoing projects should be used to further optimize the design, especially in terms of required detection volume and sensitivity. Conclusions KM3NeT is the proposed second large-scale high-energy neutrino observatory, in the Northern hemisphere, after IceCube at the South Pole. IceCube is now completed but has not yet delivered an early neutrino point-source observation. Even if discoveries are possible with a longer exposure time in the years to come, this fact is not without implications for the second observatory. The deep-sea technology has made enormous progress thanks to the contributions of the KM3NeT Consortium. The phenomenology of high-energy astrophysical sources has advanced considerably. It is certainly true that the neutrino-view of the central part of the Galaxy is of great scientific interest, as is the importance of a superior angular resolution. However, the scientific risk that the current design will make no discovery is non-negligible and should be considered very carefully. Neutrino astronomy remains a great scientific challenge that should not be missed, but the scale of the necessary sensitivity is not yet sufficiently understood. Consequently, and for the technical, managerial and organization considerations discussed above, at the present time the SSC cannot recommend the final construction of the detector as proposed, rather it believes that the next phase should focus on the construction of demonstrator(s), the site selection process and - possibly most important - further consolidation of the scientific case and the related design considerations. 8

9 ANNEX 1 KM3NET-SSC-R&Q2-v3 KM3NET-PP-SSC. 2 nd meeting. 25 June Trieste Present. A. Aloisio, A. Bettini (chair), D. Cowen, P. Lipari, J. Learned, A. Olinto, G. van der Steenhoven Justified. D. Nygren, J. Pouthas, C. Spiering The KM3NET-PP SSC met at the ICTP in Grignano (Trieste), having received from the KM3NeT-PP Collaboration the following documents 1. Grant agreement for Combination of Collaborative Project and Coordination and Support Actions. N KM3NET-PP. Annex 1. Description of work 2. The contract for the Design Study for a Deep Sea Facility in the Mediterranean for Neutrino Astronomy and Associated Sciences KM3NET - Annex 1- Description of the work 3. KM3NeT. Conceptual Design for a Deep Sea Research Infrastructure Incorporating a Very Large Volume Neutrino Telescope in the Mediterranean Sea Having understood, from document 1, that its charge foresees the preparation of a report COB February 2012, the SSC decided to plan its works as follows. In the present meeting a set of basic questions, of scientific and technical character will be defined (to be finalized through e mail exchanges with the members that could not be present), based on the available documents. The KM3NeT PP Collaboration should respond, with a written document, two weeks before the next meeting of the SSC. The 3 rd meeting of the SSC (now scheduled on ) will include an open session with hearings of the management of KM3NeT followed by a closed session. The location and the details of the agenda will be defined in due time, in consultation with the KM3NeT management. The present questions to the Collaboration are of strategic character and their order below does not imply precedence in importance. Members of the SSC have raised also questions or concerns on several specific design choices, which are not included in the present document. However, a list of more detailed questions is in preparation. This level of analysis will be attached on the basis of the answers on the strategic issues, in the 3 rd SSC meeting. Should the KM3NET Collaboration need any clarification on the Committee questions, please ask them in your earlier convenience. Questions to the KM3NET Collaboration 1. Considering that the priority is the study of astrophysical neutrino sources, discuss how you rank galactic and extragalactic sources and show how the design is optimized for its scientific objectives.

10 a. Provide energy resolution, threshold and range, as well as a sensitivity plot vs. E. b. Provide example sources using a simulation based on final detector design. c. For these sources, demonstrate how a significant observation can be made in a reasonable amount of time, taking into account the recent results of IceCube, gamma ray astronomy and theoretical estimations. What is the minimum size and sensitivity detector worth building? 2. Detector design: In terms of science, cost, special conditions of deep sea deployment: a. Compare multi-pmt to single PMT (various sizes) optical module (including cost, production time, reliability, performance) b. Compare vertical structures (bar, string, etc.) c. Overall geometrical configuration and impact of blind spots d. Readout electronics and triggering 3. Highlight the relevant parameters of KM3Net compared to IceCube, with separate answers for single site and multi site options a. Compare the use of high-energy showers vs. muons. b. Size c. Angular resolution d. Energy resolution e. Energy threshold and reach 4. Site a. Quantitatively assess the three candidate sites in terms of depth, biofouling, bioluminescence, optical properties, distance to shore, b. Design the optimal detector for the best site. c. Quantify the impact on the main science goal, construction and operation costs of using one, two or three sites, using optimized designs for both single and multiple site solutions at fixed total cost, considering both neutrino-induced shower and muon signals. d. Are there other ways to use separate sites that do not require splitting the detector itself? 5. Evaluate the project risk using the formalism described in Annex 1-KM3Net-PP (B.3.2)

11 ANNEX 2 KM3NET-SSC-DQ2-v1 KM3NET-PP-SSC. Detailed questions to the KM3NET-PP collaboration for the 3 rd meeting Provide justifications of the multi-pm choice, from all the relevant aspects: scientific, technical, economical, robustness, etc. Provide information on the evolution of the perspectives of industrial production of the items. Give details on the multi-pm electronic R/O KM3Net simulations indicate that the longer scattering lengths in seawater relative to ice will permit the reconstruction of cascade directionality well enough (a few degrees) to do pointing. This is likely one of the more compelling regions of analysis parameter space where KM3Net is intrinsically better than IceCube, and it would be important to quantify a) the predicted KM3Net directionality and energy resolutions for cascades and b) the impact of multiple sites on these quantities and the subsequent impact on, e.g., searches for point sources of neutrino-induced cascades. Comparing a single site option and multi-site options and assuming a fixed total number of OMs: o how large is the effect on effective area for muons? assuming an E 2 neutrino spectrum separately for low energy µ(1tev) and high energy µ(100 TeV) o the same for 100 TeV muons (in detector) or for an E -2 neutrino spectrum applying a cut on the angular accuracy of, say, 0.3. o what is the effect on the sensitivity to an E 2 point sources without cut-off and with 30 TeV cut-off in neutrino energy? o what is the effect to diffuse fluxes (muon signature)? o the effect of three sites for cascade detection (ν e and ν τ, i.e. 2/3 of the signal) will be negative. Clear identification of isolated cascades requires veto layers from all but possibly the bottom side which shield an inner fiducial volume. The resulting question is: what is the effect to the effective volume for cascades from ν e and ν τ interaction (contained and well identified and reconstructed events), and to the sensitivity to diffuse extraterrestrial an E 2 fluxes?

12

13 Documentation for SSC meeting (2011/11/01, V3.0) 1

14 Table of Contents 1. Introduction... 4 Questions... 4 Additional questions Science Case Science Performance Optimisation Other Galactic sources Fermi bubbles Impact on other physics Cascades Effect of smaller building blocks Comparison of the KM3NeT and IceCube potential Field of view Angular Resolution Effective Area and Energy Range Technical Detector design Trigger, Readout and Data Acquisition Multi-PMT Digital Optical Module Validation of the multi-pmt DOM Cost estimate for the multi-pmt DOM Risk and reliability analysis for the multi-pmt DOM Assembly of multi-pmt DOMs Single-PMT Optical Module Cost estimate for the single-pmt OM Risk and reliability analysis of the single-pmt OM Assembly of single-pmt OMs Comparison between multi-pmt and single-pmt optical modules Towers Validation of towers Cost estimate for towers Risk and reliability analysis of towers Integration of towers Strings

15 Validation of strings Cost estimate of strings Risk and reliability analysis for strings Integration in strings Comparison between towers and strings Investment Cost and Integration Effort Site evaluations Site selection criteria Scientific and technical criteria Infrastructural and logistics criteria National and local support Site-related financial issues Current situation Answers to the questions a. Site assessment b. Optimal detector at best site c. Impact of a distributed installation d. Use of different sites without splitting the detector Project risk Programmatic threats Technical threats Bibliography Appendix A: Estimated cost investment Appendix B: List of possible threats to the project

16 1. Introduction This document aims to answer the questions posed to the KM3NeT consortium by the Scientific Standing Committee. The basis of many of the answers is contained in the technical design report [1]. This document gives an update to the present status of technology and physics insight. The questions will be summarized here with a short answer. Reference will be given to the relevant chapter in this document for more details. It should be noted that all physics analysis results are produced using three separate reconstruction and simulation packages. This allows for a considerable amount of cross checking. At present the understanding of the reference detector (towers placed at distances of 180 m in two circular footprints) is good and the three analyses agree. For the other detector configurations the agreement is not yet perfect. Investigation into the differences will still take some time, but will finally improve the confidence in the results. Where relevant the uncertainties have been highlighted in the document. The science case is presented in chapter 2, followed by the performance investigations in chapter 0. All technical and financial subjects are covered in chapter 4. The issues concerning the different sites available for the hosting of the KM3NeT detector are discussed in chapter 5. Chapter 6 finally presents the project risks. Appendices with investment cost details and risk analysis details follow at the end of the document. Questions Questions to KM3NeT Collaboration 1. Given that the priority is to find galactic neutrino sources, show how the design is optimized from that priority. Description of the optimisation procedure is given in Chapter3.1. The optimal detector is one with towers spaced at 130 m distance and a total number of 320 towers. It has a total volume of 3.5 km 3. Alternatively, 640 strings spaced at 100 m provide similar sensitivity. The optimal footprint has yet to be finalized, but first indications are that four building blocks of 80 towers or 160 strings perform identically to smaller numbers of larger blocks with the same number of detection units. Both designs utilize the multi-pmt DOM. (Chapter 3.1) a. Provide energy threshold and range, as well as a sensitivity plot vs. E. The neutrino effective area at 10 TeV is 25 m 2 for cuts optimized for maximum sensitivity to the RXJ173 flux and source size. The effective area at very high energy reaches about 1000 m 2.(Chapter 3.1) b. Provide example sources using a simulation based on final detector design. Example sources are RXJ , HESS J , and RXJ c. For these sources, demonstrate how a significant observation can be made in a reasonable amount of time, taking into account the recent results of IceCube, gamma ray astronomy and theoretical estimations. What is the minimum size detector worth building? 4

17 If the gamma ray flux from source RXJ is fully of hadronic origin the proposed detector will produce a 5 discovery after 6-8 years. For HESS J years are required while for RXJ is reached after 7 years.(chapter 3.2) 2. Detector design: In terms of science, cost, special conditions of deep sea deployment: a. Compare multipmt to single PMT optical module (including cost, production time, reliability, performance) Cost of 2 multi-pmt DOMs is similar to six single-pmt OMs with separate electronics container at 19k and 22k respectively. Production time is 5% less. (Chapter 4.4) Reliability is higher (Chapters 4.2 and 4.3). Performance is significantly better for photon counting. (Chapter 2.) An improvement of about 30% is expected on the basis of ongoing analysis (Chapter 3.1). b. Compare vertical structures (bar, string, etc.) For equal cost the tower and string structures perform equally well for E -2 spectra (Chapter 0) For Galactic sources, the bar length of 6 m seems not optimal. A detector with towers with 15 m length bars at 130 m distances performs equally well to a string detector with 100 m distances. Both give an improvement of sensitivity of about a factor 2 over the 180 m tower detector. c. Overall geometrical configuration and prevention of blind spots The overall geometrical configuration is driven by physics performance and the need for safe and efficient deployment. Building blocks approximately the size of an IceCube detector allow for reasonable flexibility in deployment. d. Readout electronics and triggering All data to shore and online software based filtering provides the most flexible triggering and readout option. Using hits from optical modules that have one PMT hit in the fit improves the angular resolution in the critical energy regime of 1 to 100 TeV by a factor of 2. Buffering of all data for transient events allows for better sensitivity. Source following for better sensitivity is also possible in this scheme. (See chapter 2) 3. Highlight the advantages of KM3Net compared to IceCube The sky coverage contains all but a few of Galactic sources, observed by HESS. IceCube sees only a few with a similar energy threshold. KM3NeT is 4 times larger than IceCube. The angular resolution is significantly better. (Chapter 3.6) a. Compare the use of high energy showers vs. muons. For Galactic sources the shower events will contribute negligibly to the sensitivity. Therefore these have not been the subject of extensive studies yet. b. Size Instrumented volume is 4.6 times as large. 5

18 c. Angular resolution The KM3NeT angular resolution versus neutrino direction is (median) for the RXJ flux and asymptotically 0.1. (Chapter 3.1) d. E threshold and reach The detector has the maximum counting rate for RXJ in the range 5 to 50 TeV. The triggered effective area at 1 PeV is slightly above 1700 m Site a. Quantitatively assess the three candidate sites in terms of depth, bio-fouling, bioluminescence, optical properties, distance to shore, See Chapter 5. b. Design the optimal detector for the best site. The detector described in the answer to questions 1 and 2 gives optimal results in all three sites. The effect of the atmospheric muon background at angles up to 15 degrees is being investigated. The required number of simulated background events is very large and being produced. c. Quantify the impact on the main science goal, construction and operation costs of using one, two or three sites, using optimized designs for both single and multiple site solutions at fixed total cost, considering both neutrino-induced shower and muon signals. There is no impact on the main science goal of maximum sensitivity to Galactic neutrino sources. (Chapter 3.5). In single a site there will already be a separation into several independent parts, including shore cables. One, two or three shore stations have to be available. Operationally, the stations must be manned with some local personnel. The impact on running costs is 1M per extra site (Chapter 5). d. Are there other ways to use separate sites that do not require splitting the detector itself? No. (Chapter 5.4) 5. Evaluate the project risk using the formalism described in Annex 1-KM3Net-PP (B.3.2) See Chapter 6. Additional questions KM3NET-PP-SSC. Detailed questions to the KM3NET-PP collaboration for the 3 rd meeting 6

19 Provide justifications of the multi-pm choice, from all the relevant aspects: scientific, technical, economical, robustness, etc. Purity for photon counting is significantly higher for the multi-pmt DOM than for the single- PMT OM. The tubes have high quantum efficiency and collection efficiency. Large phototubes suffer from large after pulses that in the high rate environment of the sea cause degradation of the trigger potential. The total cost for a detector with multi-pmt DOMs is about 10% less than a solution with single-pmt OMs. The performance of the multi-pmt DOM solution is significantly better (Chapters 4.1 and 4.8). Provide information on the evolution of the perspectives of industrial production of the items. The major items that have been investigated for industrial production are the items related to the photomultipliers: the PMTs themselves, the HV boards and the concentrator rings. These are required at the level of items a year. Of the two major PMT manufacturers Hamamatsu has indicated that the level of per year is not unrealistic, ETEL will for an up front investment provide a production line capable of the required numbers. As a reference: Photonis produced similar items per year for the medical industry in the early years of this century. Other manufacturers are available. The consortium is considering multi sourcing. Moulding techniques are being investigated for the light concentrator rings. The HV boards at the level of 2500 per week are not unrealistic. Industrial assembly of the three above items is being considered. Give details on the multi-pm electronic R/O See Chapter 0 KM3Net simulations indicate that the longer scattering lengths in seawater relative to ice will permit the reconstruction of cascade directionality well enough (a few degrees) to do pointing. This is likely one of the more compelling regions of analysis parameter space where KM3Net is intrinsically better than IceCube, and it would be important to quantify a) the predicted KM3Net directionality and energy resolutions for cascades and b) the impact of multiple sites on these quantities and the subsequent impact on, e.g., searches for point sources of neutrino-induced cascades. Not the present priority. Comparing a single site option and multi-site options and assuming a fixed total number of OMs: o how large is the effect on effective area for muons? assuming an E 2 neutrino spectrum separately for low energy µ(1tev) and high energy µ(100 TeV) See above. o the same for 100 TeV muons (in detector) or for an E -2 neutrino spectrum applying a cut on the angular accuracy of, say, 0.3. The standard cut for the reconstruction for a Galactic source used in the document is

20 o what is the effect on the sensitivity to an E 2 point sources without cut-off and with 30 TeV cut-off in neutrino energy? o what is the effect to diffuse fluxes (muon signature)? o the effect of three sites for cascade detection ( e and, i.e. 2/3 of the signal) will be negative. Clear identification of isolated cascades requires veto layers from all but possibly the bottom side which shield an inner fiducial volume. The resulting question is: what is the effect to the effective volume for cascades from e and interaction (contained and well identified and reconstructed events), and to the sensitivity to diffuse extraterrestrial an E 2 fluxes? Low energy showers contained in a denser core are not considered possible, mainly due to the technical difficulties of placing detection units very close together. Another issue is the fact that because of the optical background of the 40 K in the sea a veto is extremely difficult to implement. For cascades the emphasis will be on extremely high energies, say larger than 1 PeV, where the atmospheric backgrounds are low. These investigations are not first priority. 8

21 2. Science Case The geographical location of KM3NeT in the Northern hemisphere makes our Galaxy its prime field of operation. One could say that investigation of possible neutrino sources in the Galaxy is the raison d être of a large neutrino telescope in the Northern hemisphere. The capabilities of such a telescope will of course provide sensitivity to extragalactic sources, but at present the science priority must be the investigation of the Galaxy. In the TDR the detector was designed as somewhat of a compromise to allow for good sensitivity to both Galactic and extragalactic sources. The performance figures of the detector have shown that the Galactic sources have definitely come within reach and optimisation of the detector for the expected energy spectrum will make it possible to see for the first time direct evidence of neutrino production. In addition the IceCube collaboration is putting more and more stringent limits on extragalactic sources in their field of view. In particular the hoped for hard Waxman-Bahcall spectrum for Gamma Ray Bursts is already being excluded by the IceCube data. The models of potential galactic neutrino sources, in particular the shell type Supernova Remnants, Pulsar Wind Nebulae, Star Formations Regions and the dense molecular clouds related to them, are robustly constrained by TeV ray observations [2,3,4]. A detector of the size of the proposed KM3NeT is expected to be sensitive enough to provide the first astro-physically meaningful probes of the strongest representatives of these source populations. Among the best-bet candidates are the young shell-type supernova remnants RXJ and RXJ Estimates show that these objects, with energy flux comparable to the Crab flux at energy around 10 TeV, can be detected after several years of exposure if the major fraction of the gamma-ray flux is contributed by hadronic interactions. The main challenge here is that the gammarays from these objects can be interpreted also within leptonic (inverse Compton) models. Both, the hadronic and leptonic models have not only attractive features but also face certain difficulties. As an example many authors claim that the gamma ray spectrum from one such object, the SNR RXJ , is explained entirely by the process of inverse Compton scattering, whereas others can fully explain the spectrum by photons from the decay of pions produced in high energy collisions of accelerated protons and interstellar medium [5,6]. Measurements from the Suzaku satellite [5], have constrained strongly the synchrotron radiation spectrum from accelerated electrons. Presently, the situation is such that fits to all the multi-wavelength data can accommodate both an hadronic interpretation as well as a leptonic interpretation in terms of inverse Compton scattering [7,8]. The leptonic interpretation requires an unusually low magnetic field combined with an additional component of softer electrons that are unconstrained by the Suzaku data. The hadronic model on the other hand requires an injection spectrum for the protons with an initial spectral index of around 1.7, rather than the more conventional 2.0, combined with an exponential cut-off in the 50 TeV range. This harder power law spectrum is supported by nonlinear theories of diffusive shock acceleration. Moreover, even in the case of acceleration spectra of protons with, the effects related to the propagation of protons into the dense clumps inside the shell may lead to the significant suppression of low (GeV) energy protons. The data from the FERMI-LAT telescope [9] can be easily accommodated in both scenarios. It seems likely that at least a substantial fraction of the gamma rays has its origin in hadronic interactions. On the other hand, as long as the leptonic models cannot be robustly rejected, the predictions on neutrino signals remain model-dependent. This makes the role of neutrino observations unique for understanding of the nature of gamma-rays from SNRs, and, in a more general context, for the solution of the long-standing problem of origin of galactic cosmic rays. 9

22 Two prominent Pulsar Wind Nebulae (PWNe), the Crab Nebula and Vela X, are believed to be powered by the electron-positron pulsar winds, but one cannot exclude the large content in these nebula of protons and nuclei. This is the case for Vela X with the energy spectral distribution which peaks at 10 TeV. Remarkably, the TeV neutrino flux expected within this hadronic scenario of production of gamma-rays in this source could be detectable by KM3NeT which makes the Vela X an excellent candidate to be investigated. Finally, because of strong internal absorption of TeV -rays, detectable neutrino fluxes from (somewhat fainter) compact TeV -ray emitters like the binary systems LS 5039 and LS I , are possible, and, more speculatively, from hypothetical "hidden" or "orphan" neutrino sources. The size of the optimised detector is of the order of 4 km 3 and has a triggered neutrino effective area of about 2000 m 2 at the largest energies. The sensitivity is approximately flat as a function of declination and so the sensitivity is spread over a large area of the sky. The main effect is that for extra-galactic sources the sky coverage is larger than that of IceCube by about a factor of four, but the peak sensitivity is only marginally better. This holds for constant as well as for variable sources such as Gamma-ray Bursts. 10

23 3. Science Performance In the TDR the optimisation of the detector was done using a E energy spectrum. This resulted in a detector with an inter tower spacing of 180 m or alternatively an inter string distance of 130 m. Both these detectors perform well at large energies but compromise the lower energies. The Galactic sources produce relatively hard spectra, but are cut off at energies between 10 and 50 TeV. In order to optimize the detector for these cut off spectra the source RXJ was chosen as a test case. This source lies in a region where the visibility of the source is high (~75%) and it has a large intensity, but a relatively large size with a complex morphology. It is at present is the best measured super nova remnant in gamma ray astronomy. For the optimisation the source was simulated as a neutrino emitting disk of 0.65 extension (cone half-angle). The energy spectrum, which is suppressed significantly at energies above 10 TeV was parameterized as: Φ(E) = (E[TeV]/1[TeV]). [ ] ( e. [TeV] ) [TeV -1 cm -2 s -1 ] (1) This spectrum is the derived neutrino spectrum assuming that the gamma ray spectrum emitted by the source is fully attributable to pion production and decay. The depletion of the flux at high energies with respect to the E spectrum assumed in the TDR, has put a premium on the reconstruction of muons of lower energies, which in turn has an impact on the chosen density of light sensors in the detector. The figure of merit (FoM) chosen for the optimisation was the number of years required for a 5 discovery of the RXJ source. The 5 discovery is defined by the probability that an upward fluctuation of background is larger than the expectation value of the signal, is less than (For 3 the corresponding probability is. 013 The FoM is determined from the number of signal and background events in a search cone optimised for the detector s angular resolution. Some systematic studies, using an unbinned likelihood method and on the effect of the particular source morphology of RXJ have also been performed. The optimisation is performed using several simulation and reconstruction programmes. A great deal of time and effort is being invested in verifying the outputs of these programmes, in order to be confident in the final result. For this procedure the data provided by the ANTARES detector have been invaluable. At present the ratio of reconstructed to triggered events for the signal is not high, around 10%. This is partly due to the harsh cuts required to remove the background. A programme of tracking optimisation using directional information of the DOM and energy reconstruction algorithms is underway and is showing very promising results. From the start of the design study there have been two different design philosophies. Because of the complications and expense of making underwater connections one design was based on installing as many DOMs as possible on a single detection unit. To optimise information density and thereby efficiency the DOMs are spaced apart horizontally. The horizontal extent is then subject to technical constraints in terms of hydrodynamic behaviour and ease of deployment. The structure gives advantages in terms of torsional stability. The second design aimed at minimising cost of the single unit and thereby negating the cost of underwater connection. The cost has indeed been reduced significantly with respect to Antares for instance. Roughly a factor three reduction in price for an equivalent unit was obtained. The two designs that were adopted for optimisation were the following: 11

24 A tower structure made of 20 storeys each consisting of a 6m bar with a digital optical module at either end. The bars alternate in direction from storey to storey. The distance between storeys is 40 m. A total of 320 such units can be constructed. For a source energy spectrum behaving as E -2 the optimal distance between units is 180 m. A string structure made of 20 storeys. Each storey consists of a single digital optical module only. A total of 640 units can be constructed. For the E -2 spectrum the optimal distance between strings is 130 m. The technical designs as well as the cost considerations are described in chapter 4 of this document. Both designs are equivalent in price and initially have an instrumented volume of about 6 km 3 and have an equal sensitivity when optimised for an E -2 energy spectrum. The optimisation steps have been performed for the tower and string options separately. The performance for these detectors is very similar and is shown in Figure 1 for the tower detector. The triggered effective area reaches 2500 m 2 asymptotically. It shows a marked decrease below 10 TeV. At 10 TeV the triggered effective area is 60 m 2.The reconstructed efficiency shows a similar but more pronounced behaviour, the equivalent numbers being 1500 m 2 and 20 m 2 at 10 TeV. The reduction at low energies is amplified even more when the cuts to optimise the discovery potential for E -2 point sources are applied. Figure 1: The effective area for the reference detector of 308 towers arranged in two separate and independent blocks of 154 units. The inter tower distance is 180 m. The three different curves indicate the triggered neutrino effective area, the reconstructed neutrino effective area. Finally the effective area after cuts for optimising sensitivity for an E -2 energy spectrum produced in a pointlike source. For cuts optimised for the discovery potential of RXJ the number of signal events per year is 3.2, the number of background events is 4.1 and the FoM of this detector is years. This has been determined for three different simulation programmes and three different reconstruction programmes. Analysis (A) uses a simulation derived from the IceCube simulation package, simulation (B) uses the KM3 simulation programme presently used in Antares and analysis (C) uses a GEANT4 based simulation. For the reconstruction analysis (A) uses a general reconstruction package based on maximum likelihood with the starting direction determined from prefits to clusters of transversely 12

25 causal hits starting in several hundred predetermined directions, (B) uses a similar programme but uses the direction of the source as the direction of a prefit before a somewhat different maximum likelihood reconstruction and analysis (C) uses a 2 based Kalman filter algorithm. For the string detector with 130 m spacing the equivalent numbers of 3.2 (signal), 3.4 (background) and 10 years obtained using the method (A) and 2.7 (signal), 2.2 (background) and 10 years FoM with method (B). It should be noted that the error on the FoM is typically 1 year. Signal Background [year -1 ] [year -1 ] FoM A B C Table 1: Number of events per year in background and signal from RXJ for the three different combinations of simulation and reconstruction programmes. For the analysis (C) also an energy dependent reconstruction algorithm has been used that improves the FoM by about 25%. This method is still being perfected. A disadvantage of the, on average, lower energies from the Galactic sources is the fact that the angle between neutrino and muon in the charged current interaction becomes non negligible and the excellent angular resolution of the KM3NeT detector cannot be exploited fully. Similarly the extension of most galactic sources has a negative influence on the discovery potential. It is clear that this detector does not provide a satisfactory signal from Galactic sources. Given the flux from RXJ convoluted with the triggered effective area, around 30 events per year pass the trigger, for the background events we expect around 100 events emanating from the source disk. To optimise the signal to noise ratio quite strict cuts are necessary. Typically they reduce the number of signal events by a factor of 10 while reducing the background by a factor 25. The major difference between signal and background is concentrated at low energies. This means that to optimise the background suppression an accurate determination of the energy of the track is required. Therefore a denser and more independent sampling of the energy loss along the track is required. This can be obtained by placing the units closer together (providing more samplings per unit track length) but also in the case of towers increasing the bar length (more independent samplings). 3.1 Optimisation A systematic study is ongoing to determine the optimal detector layout for galactic sources, i.e. RXJ Presently the FoM has improved significantly from the 12.5 years obtained previously to a FoM of 7-8 years when reducing the distance between detection units to between 100 and 130 m in the case of towers and 80 to 100 m in the case of strings. The numbers of signal events vary from 3-5 events and the background is at the 2-6 event level. The angular resolution with respect to the neutrino direction for the different detectors and reconstruction algorithms ranges from 0.25 to 0.3 depending on the applied cuts (see for instance Figure 2). One issue encountered is the fact that, whereas the three different simulation and tracking programmes agreed remarkably well at 180 m distance, they begin to deviate at the 10-20% level at the shorter distances. This shows up as a significant variation in the efficiency for signal and background, although these efficiencies are correlated. This is presently being investigated. The three prong attack that is being used to investigate the programmes has already provided insights into for instance the differences in the performance of the simulation programmes and has led to a 13

26 move toward full photon tracking simulations. This work is being done in cooperation with the IceCube collaboration. Despite the differences, all analyses show a marked improvement of the figure of merit when decreasing the distance between detection units. It should be noted that the reduction of the distance between detection units has an impact on the design of detection units, the deployment strategy and the layout of the seafloor network. These issues still need to be addressed, but do not seem insurmountable. For the near future the understanding of the differences between the different reconstruction strategies is crucial and will provide the stepping stone toward more efficient algorithms. This is an ongoing process. The influence of an efficient energy estimator is being investigated and already improves the FoM by 20% in one of the reconstruction algorithms. Including the directional information of the source in the reconstruction procedure as in reconstruction (B) seems to have a significant impact on the results. The exact size of the impact is being investigated. For the 130 m detector an investigation into the length of the bar has been performed. Two longer lengths have been studied. In general the performance improves with bar length. Using the single analysis (B) the effects are summarised in Table 2. As a comparison the performance of a string detector with 100 m inter string distance using the same analysis is also given in the table. In general distributing the optical modules more evenly over the detection volume leads to a better performance. Similar results are obtained with analysis (A) be it with a FoM about 1.5 year longer. Distance Bar length Years N source N back 5s % [year] -1 [year] -1 Tower m Tower m Tower m Tower m String Table 2: Effect of bar length on the FoM. The performance improves as the bar length increases. For comparison the string detector with 100 m string distance analysed with the same simulation and reconstruction is shown. Another effect that could influence the sensitivity to Galactic sources is the position in the sky. RXJ passes above the horizon for about 5 hours a day and reaches to 15 above. Investigations are underway to determine if the background is still manageable when reconstruction is attempted above the horizon. Antares has shown that up to 5 is possible. Such an investigation requires a huge sample of simulated atmospheric muons. The production is of these is presently underway. Assuming the full 15 can be reached a further improvement of about 20% can possibly be obtained in the FoM. Performing an unbinned maximum likelihood analysis as opposed to a simple binned method also, from experience with Antares and IceCube, improves the FoM by around 20%. Combining this method with the morphology of the RXJ source yields a further 15%. 14

27 Figure 2: Median angular distance between the generated neutrino direction and reconstructed as a function of the neutrino energy. The cuts of the discovery (5 50 %) are applied. 3.2 Other Galactic sources The optimisation is taking place using the RXJ flux and morphology. A few other sources have also been considered as candidates as neutrino sources. One other source is a super nova remnant and two belong to the category of HESS sources without or with ambiguous counterparts. The FoM has been determined for these sources using reconstruction (A) and the smaller detection unit distance. The results are given in Table 3. For comparison analysis (B) yields 6.0 years for RXJ It is interesting to note that the larger source RXJ in fact gives the shortest discovery time. For the larger sources an analysis taking into account their morphology is being undertaken. Source Radius [degree] N signal [year] -1 N background [year] -1 FoM [year] HESSJ HESSJ RXJ RXJ Table 3: Details of four galactic sources that are potential neutrino candidates. 3.3 Fermi bubbles Recently the data from the Fermi satellite has revealed a peculiar structure emitting gamma rays [10]. The structure has the shape of two large bubbles one above the centre of the galactic plane and one below. The origin of these structures is subject of speculation, but one model attributes the gamma rays to a hadronic production and therefore predicts a significant neutrino flux to be emitted [11]. The expected signal in the KM3NeT detector has been simulated in the framework of this model. Figure 3 shows the 3 and 5 flux sensitivity of KM3NeT as a function of the number of years of observation for a spectrum behaving purely as E -2, and for one cut-off exponentially at 100 TeV. The predicted intensity from the model gives a 5 signal after about one year. 15

28 Figure 3: the 3 and 5 flux sensitivity for Fermi bubbles versus the number of years of observation. The estimates are shown for a pure E -2 spectrum (red) and a spectrum cut off at 100 TeV (black). The Flux predicted in [11] is at the 10-7 scale in this figure. 3.4 Impact on other physics Figure 4 shows the ratio of the effective areas of a 130 m tower detector and 180 m tower detector as a function of energy. The results are given for cuts optimised for the galactic source energy distribution and at the trigger level. The effective area at higher energies is reduced by about 20% and at lower energy especially the optimised effective area improves significantly. The sensitivity to sources with an E -2 energy spectrum is reduced by around 10% (depending on the reconstruction). For gamma ray bursts the reduction factor is larger assuming the harder Waxman- Bahcall spectrum, although this spectrum seems to be less favoured by the recent IceCube measurements. For such sources KM3NeT is complimentary to IceCube increasing the sky coverage significantly and the absolute sensitivity by more than a factor of two. Figure 4: Ratio of the effective areas of 130 m tower detector and 180 m tower detector. Circles are for triggered effective area and squares are for effective area with optimized cuts for Galactic sources. Cascades In the vast majority of neutrino reactions, a hadronic cascade of typically 5-20 m length is produced along with the final-state lepton. The charged particles in this cascade emit Cherenkov light with an intensity proportional to the cascade energy. In neutral-current reactions this is the only detectable signal, in charged-current reactions the signal of the final-state lepton is overlaid. Observing cascades in the neutrino telescope allows for detecting neutrino reactions in additional channels and 16

29 to measure flavour-dependent quantities. Initial simulation studies have been performed in the ANTARES framework, demonstrating that cascades can be detected and reconstructed with an angular resolution of roughly 5-10 degrees (median) and a rather precise determination of the cascade energy. Such studies, however, require restrictive selection cuts, in particular on the position of the interaction vertex. Given the above characteristics, the main physics objectives of cascade investigations are the measurement of diffuse neutrino fluxes at very high energies or flavour-dependent studies, mostly for oscillation analyses at low energies. Both topics are not in the core of the KM3NeT physics case and are not used in the detector optimisation process. Therefore, the use of cascades, even though being on the to-do-list, has so far not been investigated in detail. Rough estimates can however be given at present. For high energy contained showers from GZK neutrinos the variation is a factor of 1.9. For this particular source of neutrinos the predictions extracted from the Auger high energy cosmic ray spectrum under different assumptions for chemical composition and source distribution are 0.01 to 0.6 event/km 3 /year with energy above 10 PeV [12]. The reduction of the volume from 6.9 km 3 to 3.6 km 3 therefore reduces the range of event numbers in ten years form to Similar numbers are estimated for the muon signal. 3.5 Effect of smaller building blocks In the TDR the full telescope was built of two separate building blocks of 154 (320) towers (strings). The major reason for this was the realisation that the seafloor network for a full detector with twice as many units was extremely difficult to design, taking into account the required safety margins when using ROVs. Two different methods were investigated and although solutions are available, they remain challenging. This prompted an investigation into the dependence of the physics sensitivity as a function of the number of discrete sections of the telescope. This also of course gives a good indication of the impact on the physics when placing the different detector sections in different sites in the Mediterranean. Figure 5 shows the dependence of the flux sensitivity as a function of the number of towers used in the detector. The dependence is shown separately for a detector built from 1, 2 or 3 blocks. It is clear that the three section detector performs equally or better than the two section detector. Figure 5 Sensitivity flux for a source with the RXJ spectrum as a function of the number of detection units for detector made of 1 block, 2 blocks and 3 blocks. 17

30 The investigations have been done with the different analyses and the conclusion from all analyses is similar. The splitting of the detector into more than two sections has no influence on the sensitivity of the detector if the density of photocathode area per unit volume remains the same and the section does not become smaller than about one cubic kilometre. 3.6 Comparison of the KM3NeT and IceCube potential Field of view The Mediterranean location of the KM3NeT telescope at a latitude λ between 36 and 43 North, allows observation of upgoing neutrinos from most of the sky (about 3.5pisr). Declinations below 90 +λ are always visible, while those above 90 λ are never visible. Due to the rotation of the Earth, declinations between these two values are visible for part of the sidereal day. The visibility of KM3NeT (at 42 ) for the Galactic centre and RXJ1713 are 68% and 78% respectively. In contrast, IceCube at the South Pole has a more restricted sky exposure (2 sr) but sees that fraction of the sky with 100% visibility. An interesting example which illustrates the advantage of the KM3NeT location is the Fermi bubbles (large 19 spherical structures extending above and below the Galactic Centre). For KM3NeT the upper/lower bubbles are visible 72%/83% of the time whereas for IceCube only ~10% of the upper bubble is visible. The KM3NeT view of the Super-Galactic plane (75% visibility) is also enhanced compared to that provided by IceCube (55% visibility). It is worth noting that CEN-A, a well motivated potential site for cosmic ray acceleration in the Super-Galactic plane, is within the KM3NeT field of view and not that of IceCube. The KM3NeT field of view is well matched to other the major gamma (HESS) and cosmic ray (AUGER) observatories in field and thus offers excellent opportunities for multi-messenger studies. The excellent view of the KM3NeT for our local Galaxy makes likely the unambiguous discovery of the source(s) of the galactic cosmic rays, whether that be SNRs, microquasars, the Galactic Centre, Fermi Bubbles or cosmic ray interactions with molecular clouds around the galactic plane. Angular Resolution For searches of point like sources an improved angular resolution allows for an effectively reduction of the size of the search cone around the source thereby reducing the background significantly. A good angular resolution is particularly important when optimizing for a discovery rather than for setting limits. In addition, for extended sources a good angular resolution offers the potential to study the morphology of the source. The different properties of ice and seawater have important consequences on the telescope performance. Compared to ice, the seawater is more uniform and benefits from reduced light scattering, on the other hand the sea water suffers from additional random optical backgrounds due to 40 K and bioluminescence. The final analysis cuts trade off angular resolution and effective area. Optimising the selection cuts for the best upper limits typically yields an asymptotic (~1 PeV) median angular resolution of 0.1 /0.6 degrees for KM3NeT/IceCube. For energies more appropriate for galactic sources (~10 TeV) the corresponding median resolutions are 0.3 /1.0, that for KM3NeT is dominated by the muon scattering angle in the charged current interaction. Effective Area and Energy Range The performance of a neutrino telescope is significantly determined by the deployed photocathode area (PCA). In the following, this quantity is set into relation to the instrumented volume and the performance in form of the neutrino effective area for KM3NeT and IceCube. 18

31 IceCube has deployed " PMTs in a volume of about 1 km 3 resulting in a PCA density of per km 3 (as this is only an estimate, the 10" PMTs are assumed to have a flat surface). On the other hand, KM3NeT with its 308 detection units, each equipped with 20 storeys with each 2 multi PMTs, has a PCA of cm 2. With a DU distance of 180 m the instrumented volume is 6 km 3 and, hence, the DCA density amounts to about per km 3. One can therefore conclude that both detectors have about the same PCA density and therefore should also show similar performance per volume. As KM3NeT is significantly larger this should reflect in a correspondingly higher neutrino effective area. At high energies where most of the detected muons are produced outside the instrumented volume, the neutrino effective area grows approximately with the physical cross section of the detector. For IceCube this effective surface area is about 1 km 2 and for KM3NeT it varies between 3.5 km 2 and 4.5 km 2 for horizontal and vertical tracks respectively. Hence, one would expect that KM3NeT has an effective area about 4 times as large as that of IceCube. Comparing the numbers for KM3NeT and IceCube [13]one obtains: 4 (50 m 2 /12 m 2 ) at 10 TeV and 3.3 (1000 m 2 /300 m 2 ) at 1 PeV which is in good agreement with the expectation. 19

32 4. Technical Detector design The KM3NeT neutrino telescope can generally be described as a three dimensional matrix of sensors that are sensitive to the emitted Cherenkov light in the visible range. Because the attenuation length of light in the deep sea is of the order of 50-60m, at wavelengths around 470 nm, the sensor matrix can be sparse and spread out over a large volume. In designing such a detector to be placed at the bottom of an ocean there are several difficulties that must be addressed: (1) The ambient hydrostatic pressure; (2) The corrosive environment of the seawater; (3) The distance from shore for the communication; (4) The force on the structure due to the sea currents; (5) The backgrounds due to downward going muons; (6) The background dominating environment of the sea due to 40 K decay and bioluminescence. For the physical process of detecting neutrinos from sources near the Galactic centre there are additional requirements (a) optimal angular resolution of the reconstructed muon; combined with (b) a large sensitive area facing the Galactic centre. These issues led, during the KM3NeT design study and the Preparatory Phase, to an investigation of a few feasible designs, which have been studied in detail. In these, two different concepts can be recognised. One is to utilize tower structures placed at the seabed at large distance which have horizontal extents (bars) at regular vertical distances. The optical modules are distributed in clusters (storeys) along the vertical extent of the tower. To maximise the number of independent measurements the optical modules at each storey are separated by several metres horizontally using a mechanical support. The actual optical sensors inside the modules can be either one large (8- or 10-inch) photomultiplier tube or many small (3-inch) photomultipliers. This approach leads to an instrumented volume of one cubic kilometre for every 50 towers. In the other concept, slim string structures are placed on the seabed at smaller distances, while the photocathode area at each storey is concentrated in a single optical module using 31 three-inch photomultipliers. This approach with the optical modules more uniformly distributed in the detector volume, yields an instrumented volume of one cubic kilometre for every 100 strings. In January 2011, the SPB of the KM3NeT Preparatory Phase project decided, as a compromise, to give priority to validation of the multi-pmt digital optical module (DOM) and the tower structure with multi-pmt optical modules at either end of a 6 m long storey and a spacing of 40m between storeys. This decision was based on a comparison of several key performance indicators. The compromise was the result of a large effort of the KM3NeT consortium to include the return of experience of the three pilot projects ANTARES, NEMO and Nestor. Since then the technical effort in the KM3NeT consortium has been directed towards the realisation and deployment of prototypes (pre-production models) of such a DOM-tower, in The key performance indicators considered for the technical design and realisation of the KM3NeT telescope are (1) the physics performance, (2) validation of the major components, (3) validation of the assembly and integration procedures for mass production, (4) reliability estimates for a 15 year time spam and (5) estimated investment cost and assembly and integration effort. 4.1 Trigger, Readout and Data Acquisition The readout of the KM3NeT detector is based on the all-data-to-shore concept. In this, all analogue signals from the photo-multiplier tubes (PMTs) that pass a preset threshold (typically 0.3 p.e.) are digitised and all digital data are sent to shore where they are processed in real time. The physics events are filtered from the background using designated software. To maintain all available information for the offline analyses, each event will contain a snapshot of all the data in the detector during the event. Different filters can be applied to the data simultaneously. The data contain the leading edge and the time over threshold of every analogue pulse, commonly referred to as a hit. Each hit corresponds to 6-8 Bytes of data. The optical background due to decays 20

33 of 40 K and bioluminescence amounts to typically 5 khz for a 3 inch PMT. A reduction of the data rate by a factor of at least 10 6 is required to store the filtered data on disk. Various triggers have been developed which show a small contamination of random background. At a depth of 3.5 km, the event rate is then dominated by atmospheric muons and amounts to a few 100 Hz. For the detection of muons and showers, the time-position correlations, that are used to filter the data, follow from causality. In the following, the level-zero filter (L0) refers to the threshold for the analogue pulses which is applied off shore. All other filtering is applied on shore. The level-one filter (L1) refers to a coincidence of two (or more) L0 hits from different PMTs in the same optical module within a fixed time window. The scattering of light in deep-sea water is such that the time window can be very small. A typical value is T = 10 ns. The estimated L1 rate is then about 1,000 Hz of which about 500 Hz is due to genuine coincidences from 40 K decays. The remaining part arises from random coincidences which can be reduced by a factor of two by making use of the known orientations of the photomultiplier tubes. For a storey with two optical modules at either end, the relatively short distance between the optical modules can be used to define a simple higher-level filter (T1). Such a T1 filter implements a coincidence of two (or more) L1 hits on the same storey within a time window of T = 50 ns. The estimated rate of T1s is about 0.2 Hz, primarily due to contributions from (very) low energy atmospheric muons and random coincidences. Depending on the length of the storey, genuine coincidences from 40 K decays may contribute as well. A simple coincidence of two (or more) T1s can be used to trigger an event. An alternative solution to trigger an event consists of a scan of the sky combined with a directional filter. In the directional filter, the direction of the muon is assumed. For each direction, an intersection of a cylinder with the 3D array of optical modules can be considered. The diameter of this cylinder (i.e. road width) corresponds to the maximal distance travelled by the light. It can safely be set to a few times the absorption length without a significant loss of the signal. The number of photomultipliers to be considered is then reduced by a factor of 100 or more, depending on the assumed direction. Furthermore, the time window that follows from causality is reduced by a similar factor. (Only the transverse distance between PMTs need to be taken into account because the times can be corrected for the propagation of the muon.) This improves the signal-to-noise ratio (S/N) of an L1 hit by a factor of (at least) 10 4 compared to the general causality relation. With a requirement of five (or more) L1 hits, this filter shows a very small contribution of random coincidences. The field of view of the directional filter is about 10 degrees. So, a set of 200 directions is sufficient to cover the full sky. By design, this trigger can be applied to any detector configuration. Furthermore, the minimum number of L1 hits to trigger an event can be lowered for a limited number of directions. A set of astrophysical sources can thus be tracked continuously with higher detection efficiency for each source. Alternative signals with different time-position correlations, such as slow monopoles, can be searched for in parallel. It is obvious but worth noting that the number of computers and the speed of the algorithms determine the performance of the system and hence the physics output of KM3NeT. A well know feature of photomultipliers is the presence of pre-, delayed- and after-pulses. The preand delayed-pulses cause a certain degradation of the timing of the signal which normally is smaller than the characteristic transition time spread (TTS) of a PMT (about 2 ns). The after-pulses are due to ionisation feedback which produces relatively large analogue pulses. These pulses mimic a signal from a nearby muon. The probability that a photo-electron triggers an after-pulse is typically 0.5-1% for a large photomultiplier (8 or 10 inch). In the presence of optical background, each photomultiplier thus produces a rate of large pulses of about Hz. This implies that every event will be accompanied by large pulses. This severely affects the performance of the reconstruction. This was not taken into account in the studies for the TDR. It should be noted that the segmentation of the photo-cathode area by means of relatively small PMTs is not affected by 21

34 after-pulses. The number of hits, rather than total charge of an analogue pulse, yields an exact lower estimate of the number of photons that must originate from an external source. (It has been verified that the radioactivity of the glass sphere does not significantly contribute to this photon count.) As a matter of fact, the multiplicity of hits in the same optical module can be used to improve the S/N ratio of any filter. A level-two filter (L2) consisting of a coincidence of three (or more) L0 hits in a single optical module reduces the count rate by a factor of 10 compared to the L1 filter. The effect of the L2 filter on the detection efficiency of neutrinos is estimated to be less than 10%. Figure 6 The count rate as a function of the number of photons. The black dots correspond to the observed rate of a single 10 inch PMT in ANTARES. In this, bin 10 corresponds to Multiplicity 10. The green (red) area corresponds to the estimated rate of accidental (genuine) coincidences due to 40K decays in a set of 31 small PMTs. The count rate as a function of the number of photons is shown in Figure 6 for a single 10-inch PMT in ANTARES and for a set of 31 small PMTs in KM3NeT. The high rate of high-multiplicity hits observed when using a 10 inch PMT can be attributed to after-pulses. In ANTARES, there are three large PMTs on a storey. A L1 hit may be a local coincidence or a hit with a large integrated charge. In order to cope with the background, the total L1 rate should be limited to about 1 khz or so. As a consequence, the high-threshold condition for a L1 hit is set to 3 p.e. For KM3NeT, the L1 condition is simply 2 hits. This yields a similar purity (i.e. same count rate due to optical background) but significantly better efficiency (about 85% for a 2 photon signal compared to 50% for a 3 photon signal). The all-data-to-shore concept is implemented in the ANTARES telescope since The full sky is viewed with directional filters continuously. In the absence of excessive bioluminescence, a directional filter pointed to the Galactic centre is operated in parallel. This filter uses both L1 and L0 data. It has recently been shown that this filter yields a gain in the detection efficiency of neutrinos by (at least) 10%. This gain is limited by the effect of the optical background on the reconstruction which may still be improved. For KM3NeT, the estimated number of computers needed to filter the data is less than 500. The minimum number of photons to trigger an event is 8-10, depending on the filter. The chosen concept allows for a flexible, extendible, and upgradable system at a moderate fraction of the total cost (less than 10%). 22

35 Although a traditional system with a local L1 coincidence trigger at first sight may seem attractive as a tool to reduce the bandwidth to shore, it is not necessary. The currently commonly used bandwidth of optical networks provides sufficient bandwidth to sustain the expected L0-level data rate. Assuming 64 bits per recorded photon, for the envisaged photo-cathode area the total data rate amounts to about 0.2 Tb/s. This data rate to shore can be accommodated on a number of optical fibres using dense wavelength division multiplexing (DWDM) techniques. Having access to all data, the reconstruction efficiency is improved by a factor of about 1.5 at 1-10 TeV. For those events that may constitute a discovery, the availability of all data is paramount. The possibility to scrutinize the background environment, verify trigger conditions and study hit patterns in detail before claiming a discovery will be essential. Using a passive electro optical cable network, a point to point 1 Gb/s communication network will be implemented in which each optical module (or any other module) has a unique optical communication channel with the data acquisition electronics on shore. The advantage of this approach is that it minimizes the amount of active off shore electronics more prone to failure than passive electronics - while providing a dedicated wide bandwidth. It allows for staged deployment of the detection units and future upgrades of the components on shore are possible during the lifetime of the detector. The use of a passive network in the deep sea and the possibility to perform maintenance of the equipment on shore not only strongly enhances the system reliability but also reduces cost and power consumption. In Figure 8 the main functions of the readout architecture are presented. The DAQ system is based on a FPGA with an embedded processor inside the optical module (see Figure 9). The central electronics can be viewed as a hub which gathers all the information produced in the module and transmits them to shore through the assigned DWDM channel; in the opposite direction, the information received from shore (mainly slow control commands) are used to manage the functions of the optical module. The downstream data from shore consists of a Continuous Wavelength (CW) signal with a superimposed data/clock signal for detector control. The upstream signal is a modulated signal containing PMT (and other sensor) data. The electrical to optical conversion is implemented by a Reflective Electro Absorption Modulator (REAM), which is a sort of a mirror which can be turned on and off, by an electrical modulating signal. It reflects the incoming wavelength: hence, just one wavelength per optical module is needed for both directions. Since event reconstruction is based on the PMT hit time, a common timing reference must be available to front end boards, to allow for detector wide synchronization. The time offset between each acquisition channel and the fixed reference must be known in order to compare hit times. In order to facilitate the clock distribution a synchronous protocol will be used: the clock is embedded in the slow control data by an on-shore transmitter in a unique bit stream. The receiver in the optical module recovers the clock and extracts the data. The recovered clock is fed to the front end electronics which can stamp the PMT hit with the common reference. In addition, using the point-to-point connection in the network to a module, a timing marker signal can be sent forth and back to measure the propagation delay with sub-nanosecond resolution. In this way, all the receivers will be synchronised by design to the on-shore time reference, which is derived from a GPS station. The high speed transmission is timed using this clock as well: the frequency for the required serializer can be synthesized by means of a phase locked loop (PLL). The synchronous command distribution is based on the same principle. To further improve the timing resolution, the phase relation between the transmitted and received clock signals is measured continuously, thus enabling tracking of changes in propagation delay due to for example a change in temperature or a change in pressure. The data in the detection unit backbone are transported via a single cable which contains one optical fibre for each storey. The number of fibres can be reduced by a factor of 2 when the (de-)multiplexer is located approximately half way along the length of the detection unit. The power conductors reside inside the same cable. At each optical module a break out extracts the required fibre and power 23

36 SS e ~80 l/fibre < 10 Gbps P2P ~70 fibers in main cable ½ DU OM 20 APD REAM e e e APD OFM e APD 3 km 0.5 km 100 km o e PJB SJB Figure 7 Scheme of the optical communication network. SS-Shore Station, PJB-Primary Junction Box, SJB-Secondary Junction Box, DU-detection unit, OFM-Optical Fanout Module, OM-Optical Module, REAM-Reflective Electro-Absorption Modulator, APD-Avalanche Photo Diode for conversion from optical to electrical domain. wires. The multiplexing and de-multiplexing of optical signals is made inside the optical fan-out module of the detection unit. Further multiplexing of signals from different detection units is performed in the secondary junction boxes. All hardware components in the readout and data acquisition system are chosen to be standard, mass market Commercial off-the-shelf (COTS), based on server-like processors interconnected with standard 1Gbit network hardware. Exception is the reflective electro-absorption modulator (REAM). The design is 10 Gbit resistant, i.e. it is flexible enough to permit upgrade to faster processors as they become available and to migrate to a 10Gbit network should the need arise. Figure 8 Readout architecture of a detector with multi-pmt optical modules. 24

37 Figure 9 FPGA architecture of the central electronics of the multi-pmt optical module. The blocks in the green shaded area are implemented in a system on chip. 4.2 Multi-PMT Digital Optical Module The design of the multi-pmt digital optical module and its product breakdown are described in detail in the technical design report. The concept for a multi-pmt optical module has been first presented to the consortium by Esso Flykt at the first VLVnT workshop in 2003 [14]. Following his ideas, a cost effective plug-and-play sensor module has been developed for KM3NeT with the full functionality of an ANTARES storey contained in a single pressure resistant glass sphere with a diameter of 17 inch. The module is fit to be distributed in the detection volume both in a tower configuration and in a string configuration. The actual light sensors are 31 photo-multiplier tubes of 76 mm diameter, surrounded by a 102 mm diameter light concentrator ring. The total photocathode surface is 1260 cm 2, the total area of the three 10-inch photomultipliers at a storey of the ANTARES telescope. The photocathode area will be effectively enhanced by 20-40% by the use of a light concentrator ring, made of silicon gel and kept in place by an aluminium ring serving as a reflector. The photocathode has a concave shape in order to achieve appropriate timing resolution. The front end of the tube is convex with a radius matching the glass sphere. The length of the tube is less than 122 mm. It has a 10-stage dynode structure with a minimum gain of The photocathode is conventional Caesium-Potassium bialkali with quantum efficiency larger than 32% at 404 nm (larger than 22% at 470 nm). The use of Caesium-Rubidium for the photocathode is no longer considered as at 470 nm the quantum efficiency is comparable to that of Caesium-Potassium, while the dark current is substantially higher. A custom low power (<45 mw) Cockcroft-Walton base provides the high voltage for the photo-multiplier tube. It includes a chip with an amplifier and discriminator, providing a LVDS signal with a length proportional to the charge. The optical module also has instrumentation that allows for the reconstruction of its position (acoustic piezo element), determination of its orientation (compass and tilt meter) and calibration of its timing (nanobeacon). The photomultiplier tubes are supported by the light concentrator rings and a foam structure. The view of the module is made as uniform as possible. The vertical orientation of the photomultipliers varies between 50 and 180 degrees with respect to a positive axis pointing upwards in vertical direction. The acoustic piezo element and the nanobeacon are glued against the glass sphere. The photomultiplier tubes are optically coupled to the glass sphere with a thin layer of 25

38 optical gel. Since the sensitivity of the 3-inch photomultipliers to the Earth s magnetic field is very small magnetic shielding is not required. Segmentation of the photocathode area will aid distinguishing single photon hits from two-photon hits, which is important for the reduction of background hits from 40 K decay and bioluminescence in the seawater. The digitization and readout electronics is concentrated in the centre of the optical module. The central logic of the optical module is implemented using a FPGA based system. Cooling of the electronics is achieved with an aluminium mush-room shaped system that is glued to the glass sphere. Time-over-threshold values of the photomultiplier tubes are transmitted to shore via a unique optical channel using a reflective electro-absorption modulator (REAM) in the optical module and Dense Wavelength Division Multiplexing (DWDM) technologies in the optical communication network. Effectively, multiple photons in the optical module are distinguished by counting photomultiplier hits rather than determining pulse-height. Also large pulses from individual photomultipliers, e.g. due to sparking or after-pulsing, are easily recognized, as it is unlikely that this has a neutrino physics reason if only one photomultiplier will have a large pulse, while its direct neighbours have not. For timing calibration the fibre propagation delay is measured from shore using an optical pulse echo technique. Figure 10 Prototype of a multi-pmt DOM (left); the Octopus signal collection board with a Photonis PMT and PMT-bases connected (middle); technical drawing of the central logic board mounted in the cooling shield (right). Validation of the multi-pmt DOM The concept of the multi-pmt digital optical module has been validated with several reference models of the module. The performance of the cooling system, the resistance against high pressure, the optical coupling between the photomultipliers and the glass sphere and the background rates of the glass of the sphere and the glass of the photomultipliers have been measured. A long term test with 16 photomultipliers in a hemisphere in a dark box with water has shown the feasibility of the concept. During the validation tests, photomultipliers of the Photonis company have been used. With the demise of Photonis for photomultiplier production, discussions with four other providers of photomultipliers started in This has resulted in the recent delivery of the first batches of phototubes by Hamamatsu (R6233MOD) and ETEL (D783KFLA) for use in the prototype DOMs to be deployed in The delivered photomultiplier tubes comply with the specifications presented in the technical design report. The results of the acceptance tests of these tubes have been presented during the VLVnT11 workshop [15]. Together, ETEL and Hamamatsu will deliver in total 150 photomultipliers for installation in the KM3NeT tower-prototypes to be deployed in Two mini- DOMs with each five photomultipliers are in preparation for installation in the instrumentation line of ANTARES. These mini-doms will be read out using an interface to the ANTARES readout system. In addition to these two mini-doms, one full DOM will also be installed in the ANTARES instrumentation line. This DOM will be readout using a prototype of the foreseen readout system for KM3NeT. Together, the deployment of these DOMs will allow an early long-term in-situ benchmark test of the functionality of the multi-pmt DOM and the KM3NeT readout and DAQ system. Installation of these DOMs in ANTARES is foreseen for early Experience with these DOMs will be input for another in-situ test of the multi-pmt DOMs which is foreseen also in 2012 with the deployment and connection of a small version of the KM3NeT tower at the Capo Passero site. In this 26

39 tower, three DOMs will be installed again for an in-situ test of the functionality of the DOMs and for a test of the KM3NeT readout and DAQ system over a distance of 100 km. Cost estimate for the multi-pmt DOM Based on a production of components for a total of optical modules required for the full KM3NeT detector of 320 towers, the total investment cost of the digital optical module is estimated This number does not contain the mechanical interface to the storey of a detection unit, since this item depends on the choice for the storey in the detection unit. The breakdown of the costs is presented in Table 4 in section 4.4. It shows that the cost for the optical module is dominated by the cost of the photon sensor unit, i.e. the combination of the photomultiplier tube, the HV base and the light collection ring. The cost of this combination is currently estimated at 195. Options for reduction of this number are being studied. The cost of the HV base of the photomultiplier tubes will be reduced by further integration of functionalities in the ASIC on the base. A submit of the updated ASIC design is expected early In addition, it is being investigated whether using a material like Perspex-G for the convex front end of the tube to match the radius of the glass sphere is feasible. If feasible, this would replace the light collection ring and would reduce the cost for the photon sensor unit and ease the assembly of the DOM. Price-quote normalised to the price of a single PMT Trends in the price for 3 inch PMTs (2010 price-quotes) 1,2 1 0,8 0,6 0,4 0, Number of PMTs ordered Figure 11 Trends in the price for 3-inch PMTs as function of the total amount of PMTs ordered. Calculated using 2010 price-quotes. The price is normalised to the price of a single PMT. Besides ETEL and Hamamatsu, also the MELZ Company is preparing for delivery of their first prototype photomultiplier end of The newly established Chinese company Zhan Chuang Photonics, which has purchased the Photonis technology for photomultiplier production, has planned delivery of a first prototype in the summer of Of the two major PMT manufacturers Hamamatsu has indicated that the level of per year is not unrealistic, ETEL will for an up front investment provide a production line capable of the required numbers. They have presented cost estimates for the mass production of photomultipliers for KM3NeT, both for the full detector of 320 towers and for smaller batches. In Figure 11 the dependence is shown of the price for a 3-inch PMT on the total number ordered. For this figure the price-quotes made in 2010 are used. The final choice for photomultiplier companies for KM3NeT will be the result of a tendering procedure that is expected to start end The competition of the four companies allows for the choice for a multisource mass production of the photomultiplier tubes for KM3NeT, which is attractive for the pricing of the photomultiplier tubes, but also is an assurance against bankruptcy of companies or against the halting of production of producing photomultiplier tubes such as experienced with the Photonis company. The cost of the custom electronics boards inside the digital optical module is estimated at This number is dominated by the cost for the central logic board with a FPGA and the e/o 27

40 conversion board with the REAM (Reflective Electro-Absorption Modulator). About 2% of the total cost of the module is contributed to the instrumentation inside the optical module: an acoustic piezo element, a nanobeacon, a compass, a tilt-metre and a pressure gauge. Risk and reliability analysis for the multi-pmt DOM The risk analysis of the multi-pmt digital optical module is somewhat complicated as it is a custom design, but some of the components have been used extensively. For example, the glass spheres have a leakage probability below the percent level. Experience with ANTARES has shown that leakage occurs primarily on submersion and does not increase significantly with time. For KM3NeT this is acceptable only if this error does not propagate in the tower. Each optical module is therefore galvanicaly separated from the rest of the system, avoiding the propagation of any leaks by corrosion. Photomultiplier tubes are typically very reliable items. Typical FIT rates (failures in 10 9 hours) are around 10. The photomultipliers run at a gain of 10 6 and their individual photocathode area is small, therefore the integrated anode charge is small. Degradation of the performance of the photomultiplier during the foreseen lifetime of 15 year of the telescope will therefore be in the order of 1-2%. The optical modules contain many photomultiplier tubes so a single failure only causes a slight reduction in efficiency rather than a complete blind spot. The FPGAs have typical FIT rates in the range of 10 to 50 depending on their size and configuration. FIT rates of the custom electronics have been estimated from component reliability figures. The overall failure rate of the optical module has been estimated at less than 500, equivalent of 0.5% in 10 years operation. The failure is then defined as at least one photomultiplier tube becoming inoperative. Assembly of multi-pmt DOMs In compliance with a detector of 320 towers with 40 DOMs each or 640 springs with 20 DOMs each, a total of digital optical modules will have to be assembled for deployment within a period of four years. Although most components of the optical module can and will be mass produced in industry, assembly of the optical module in industry is not foreseen. Instead, the production model is to establish within the consortium dedicated DOM-assembly sites with dedicated and trained personnel hired on project basis. As a first step in the design of DOM-assembly lines the assembly procedure of the optical modules of ANTARES has been studied. This resulted in a preliminary detailed description of the foreseen assembly procedure as presented in the technical design report. The procedure has been further analysed during the preparatory phase, based on the experience with the assembly of several prototype optical modules and experience with mass production of components for the LHC detectors at the research labs in the consortium. It resulted in the recommendation for 6 separate DOM-assembly sites each with 200 m 2 space for two parallel assembly lines and 200 m 2 for storage of components and optical modules. Per assembly site, 8 fte will be required during four years, a total of 192 fte-year. The assembly and quality control procedures will be further optimised during the assembly of the prototype DOMs for the deployments in In parallel, the first prototype assembly line for mass production is being designed and will be installed in 2012 at one (possibly more) of the research labs in the consortium. The choice for six separate DOM-assembly sites has been made deliberately to allow for the various groups and institutes to locally involve personnel and make the presence of KM3NeT very visible. If this turns out not to be required, less assembly sites with more parallel assembly lines is equally feasible. This would reduce the overhead cost, e.g. in the case of a single production site it is estimated that a total of 174 fte-year would be required. 4.3 Single-PMT Optical Module Also the design of a single-pmt optical module has been described in detail in the technical design report. Moreover, in all precursor neutrino telescope detectors, the use of a single large photomultiplier tube was the common feature. Typically, the deep-sea telescopes utilize glass containers with a diameter of 17 inch and the Antarctic telescopes those with a diameter of 13 inch. The diameter of the photomultipliers used is 15 inch for the DUMAND, BAIKAL and NESTOR 28

41 experiments, 10 inch for the ANTARES and IceCube telescopes and 8 inch for the Amanda telescope and DeepCore. In IceCube a storey can consist of a single optical module, whereas in the deep sea multiple single-pmt optical modules per storey are required to allow for the reduction by local coincidences of background hits from 40 K decay and bioluminescence in the seawater. In IceCube the electronics is contained in the same glass sphere as the photomultiplier, a natural choice for use in the Antarctic ice. The deep-sea of ANTARES allowed for the choice of a separate electronics container on each storey. During the KM3NeT Design Study, the use of both 8 inch and a 10 inch photomultiplier tubes has been studied in a storey-configuration of 6 optical modules and one electronics container per storey. The diameter of the glass sphere of the optical module was chosen to be 13 inch. A nanobeacon for timing calibration is included in two of the six optical modules on a storey. With three different Hamamatsu photomultipliers the influence of the Earth s magnetic field on the performance of the tubes has been measured. These three tubes were a 10-inch R7081 tube with a standard bialkali photocathode and two 8-inch R5912 tubes, one with a standard bialkali photocathode and the other with a super-bialkali photocathode. These validation tests showed that a mu-metal shield against the influence of the Earth s magnetic field is required. Other tests showed that the increased quantum efficiency of the 8-inch super-bialkali tube almost compensates its smaller detection surface compared to the 10-inch tube [16]. In ANTARES, it has been observed that there is a presence at the level of a few percent of after-pulses that have large pulse-heights. This is due to ion-acceleration to the photocathode. The rate of these large pulses scales according to the singles rate. With a typical background rate of 50 khz in the deep-sea this translates to a rate of L1- hits (see section 0) between 250 and 500 Hz. To remove this background a significant cut must be placed on the number of L1-hits required for recognising neutrino events which for large detectors such as KM3NeT reduces the sensitivity of the detector significantly. Cost estimate for the single-pmt OM The total investment cost of a single-pmt optical module has been estimated This number includes the cost for the HV base and the connection to the electronics container but does not include as opposed to the multi-pmt optical module the cost of the readout electronics. The breakdown of the cost is presented in Table 4 in section 4.4. The cost of the PMT-unit, i.e. the photomultiplier tube, the base and the mu-metal shielding system is estimated at In Figure 12 the dependence is shown of the price for a large photomultipliers on the total number ordered. For this figure the price-quotes made in 2010 are used. The trend is shown for a standard 10-inch PMTs and a HQE 8-inch PMTs. In contrast to the multi-pmt optical module, the readout and data acquisition electronics for the single-pmt optical module is stored in a separate electronics container. This container is shared by six optical modules and requires six storey cables with copper wires for connection between the electronics and the optical modules. The total cost of the container including the electronics is estimated The breakdown of the total cost for a storey with six single-pmt optical modules is presented in Table 4 in section 4.4. The cost for the mechanical interface between the optical modules and the mechanics of the storey and that of the electronics container and the storey is not included in the table. On a tower-storey of six optical modules, two of them contain a nanobeacon. Together with an external hydrophone and a compass/tiltmeter in the electronics container about 5% of the total cost is attributed to instrumentation. 29

42 Trends in the price for 8/10 inch PMTs (2010 price-quotes) Price-quote normalised to the price of a single PMT 1,2 1 0,8 0,6 0,4 0, Number of PMTs ordered 10 inch PMTs 8 inch high QE PMTs Figure 12 Trend in the price for 10-inch and HQE 8-inch PMT as function of the total amount of PMTs ordered. Calculated using 2010 price-quotes. The price is normalised to the price of a single PMT. Risk and reliability analysis of the single-pmt OM The risk analysis of the single-pmt optical module and the electronics container can be based on the experience with the ANTARES neutrino telescope. About 900 optical modules have been deployed since Since then about 3% of the optical modules (27 out of 900) have leaked, of which about 50% due to a vacuum valve with O-ring that is no longer produced by the supplier. About 1% of the electronics containers (2 out of 300) have leaked. For KM3NeT, with regard to leakage, the same conclusions apply as formulated for the multi-pmt optical module. Also in this case, the propagation of water leaks from the electronics container to the backbone is prohibited by a galvanic separation. The FIT rate for the large photomultiplier tubes is not significantly different from those for small photomultipliers. A failure of an electronics container will eliminate all optical modules on the corresponding storey. Assembly of single-pmt OMs The assembly of the single-pmt optical module is well known from the ANTARES detector, for which the first 100 optical modules were assembled in industry and in a time spam of 17 month a total of 800 optical modules were assembled within the collaboration. For a telescope of 320 towers a total of optical modules must be assembled in a period of four year. For this, assuming a single assembly site, a total of 151 fte has been estimated, i.e. during four year of construction, about 37 fte is required for the assembly of the optical modules. In contrast to the multi-pmt optical module, assembly of the single-pmt optical module in industry is considered feasible, although an attempt to to this failed. Also for the electronics container, the experience with the assembly of the ANTARES electronics container is invaluable. In total 50 fte year is estimated for the assembly of the 6400 electronics containers in the full KM3NeT detector. I.e. during four years of construction, about 13 fte is required for the assembly of the electronics containers. Assembly in industry of the electronics container is not considered. 4.4 Comparison between multi-pmt and single-pmt optical modules The design of the multi-pmt digital optical module is the result of a careful analysis of the optical modules in other neutrino telescopes, in particular the ANTARES deep-sea detector. From this analysis it became clear that besides the need for a seriously cheaper solution two major challenges should lead the design concept: the high background and the high pressure environment of the deep-sea. 30

43 The first challenge of the high background has led to the choice for a segmentation of the photocathode area. This segmentation will aid distinguishing single photon hits from two-photon hits, which is important for the reduction of random background hits from 40 K decay and bioluminescence in the seawater. Section 0 contains the details of the readout and trigger implementation. For each hit photomultiplier the time over threshold (ToT) will be measured. This provides, for large pulses, a logarithmic dependence of the ToT on the pulse height. At small intensity, the ToT is less accurate, but the hit pattern will distinguish between 1, 2 or more photons. Two-hit separation is at the single photon pulse width for a typical threshold of 0.3 spe. Effectively, multiple photons in the optical module are distinguished by counting photomultiplier hits rather than determining pulse-height. Since photomultiplier tubes in the module also are pointing upwards up to about 45 degrees above the horizon, the background from atmospheric events will be well measured thus allowing for a better understanding of this background and as a result to a better reduction. Essential in this is the concept of all-data-to-shore for the data acquisition, which will allow for elaborated trigger studies. The second challenge of the high ambient pressure in the deep sea has led to a design with the photocathode area and the instrumentation of a full ANTARES storey in a single glass sphere together with the electronics. With only one high pressure transition through the glass and a galvanic separation from other modules, the risk of a water leak per photocathode area in the detector has been considerably reduced compared to the design of ANTARES with 6-9 high pressure transitions per storey and without galvanic separation of the optical modules and the electronics container. As already stated in the introduction, in January 2011 the SPB of the KM3NeT Preparatory Phase project decided to give priority to the validation of the multi-pmt digital optical module (DOM), in particular in a tower structure with a multi-pmt optical module at either end of a 6 m long storeys. The physics performance of such a structure is addressed in chapter 0. Here we will address the validation of the multi-pmt digital optical module, its reliability, the feasibility of mass production of the modules, both in terms of investment cost and fte required for assembly and the possibility of industrial outsourcing. We will address these issues in comparison of a two multi-pmt DOM storey with a storey containing six single-pmt optical modules and an external electronics container. In sections 4.2 and 4.3 the validation and reliability analysis of the optical modules has been described. Clearly, validation of a single-pmt optical module is delivered by the ANTARES detector. Although the lessons learned from the performance of these modules and the electronics containers have been taken into account in the design of the multi-pmt digital optical module and tests have shown the validity of the multi-pmt module in the lab, the in-situ validation of the module is still pending. The installation of two mini-doms and a full DOM in ANTARES and the deployment and connection of a prototype tower with several multi-pmt DOMs installed will be the final steps in the validation procedure. Since the multi-pmt optical modules contain many photomultiplier tubes a single failure only causes a slight reduction in efficiency rather that a complete blind spot as in the case of a single-pmt optical module. In Table 5 the design and production features of the multi-pmt optical module and the single-pmt optical module required in a full KM3NeT detector are summarized. Since the multi-pmt digital optical modules also contain calibration instrumentation and all electronics boards, also the cost and assembly effort for the electronics container and the external instrumentation is presented to allow for a proper comparison of the same functionality per tower-storey. 31

44 Multi-PMT DOM Single-PMT OM Electronics Container External instrumentation Item Cost [ ] Cost [ ] Cost [ ] Cost [ ] Glass sphere Titanium vessel 2000 Electronics Mechanics Cooling Instrumentation Penetrator PMTs+bases+light 6045 collection rings PMT+base+mu-metal 1380 shielding Connection to 650 electronics container Total Nr. of items required per storey in a tower Total cost per towerstorey Table 4 Cost breakdown for the two multi-pmt digital optical module (DOM) on a storey of a DOM-tower and of six multi-pmt optical modules, the external electronics container and the connection between them and the external instrumentation on a storey in a single-pmt OM tower. The total cost per storey is for the same functionality in both tower configurations. The cost for the mechanical interface between the modules and the mechanics of the storey is not included in this table. Investment Assembly Amount [M ] [fte year] Multi-PMT DOM including electronics Single-PMT OM including connection to the external electronics Electronics Container and an external hydrophones Table 5 Estimated total investment cost and assembly effort in fte-year for the optical modules in a full KM3NeT detector. In the case of utilization of single-pmt optical modules, the cost of the electronics container and external hydrophones need to be taken into account as well for comparison of the same functionalities. In all cases the mechanical interface with the storey mechanics is not included. For comparison, the fte-estimates are for a situation with a single assembly site for the optical modules in both cases. 201 Comparison of the numbers shows that the total investment cost for the multi-pmt DOMs in a KM3NeT detector is lower than that for the single-pmt OMs with the external electronics containers. The numbers for the total assembly effort are similar for both configurations. Clearly, the number of items to be produced for a KM3NeT detector with multi-pmt DOMs is much less than in 1 Each multi-pmt DOM contains an acoustic piezo element, a compass, a tilt meter and a nanobeacon. On a storey of 6 single-pmt optical modules, only 2 OMs contain a nanobeacon. An external hydrophone is connected to the electronics container, which also houses a compass and a tiltmeter. 32

45 the case of a detector with single-pmt optical modules. The complexity of the electronics between the two solutions will be similar; packaging of both the photomultiplier tubes and the electronics in the same vessel as in the multi-pmt DOM will be more complicated. On the other hand, testing of the plug-and-play multi-pmt DOM and its integration in the detection unit will be easier than in the case of the single-pmt optical module. The cost estimates of components are based on, in descending priority, industrial quotations, corresponding costs as occurred in the pilot projects, public catalogues, and informal or confidential statements of providers. The costs of the photomultipliers are estimated according to informal and confidential statements of the four corresponding companies. Most components will be produced in industry. For those components that have to be produced in very large quantities, such as the photomultipliers, multi-source production will be sought to reduce the risk of serious delay of the KM3NeT building project by demise or malfunctioning of companies. Outsourcing in industry of the assembly of the multi-pmt optical module with the electronics inside is not foreseen. Instead, the production model is to establish dedicated assembly lines within the consortium with dedicated and trained personnel hired on project basis and supervised by the staff of the institutes. This production model has been successfully applied many times for the mass production of e.g. detectors for the LHC experiments. Experience has shown that for detectors which require a high level of quality control, the model allows for a cost-effective high-quality production. Assembly of the single-pmt optical module in industry is considered feasible, based on the experience with the first about hundred of the ANTARES single-pmt optical modules which were manufactured in industry. As for the multi-pmt optical module, assembly of the electronics container by industrial companies has not been foreseen. 4.5 Towers During the last two years, two different tower-configurations have been considered. One is utilising single-pmt optical modules, the other one multi-pmt optical modules. Both towers consist of 20 bar shaped, 6 m long storeys with optical modules. The towers are connected to the seabed infrastructure with a wet-mateable connector. The tower-configuration with single-pmt optical modules is described in the technical design report and has been further worked out during the preparatory phase. Details can be found in a report for the European Committee as a deliverable of one of the work packages [17]. A storey in this tower contains six optical modules each connected via copper wires with an electronics container on the same storey. An electro-optical cable runs the full length of the tower connecting the electronics containers on each storey. Each storey will have a unique optical link with the data acquisition on shore. The optical modules and the electronics containers are galvanicaly isolated to prevent water leaks to propagate in the tower. The development of this tower configuration has been abandoned after the decision of the consortium in January 2011 to give priority to the validation of the digital optical module (DOM) and a tower configuration with storeys with a DOM at either end. In this tower configuration, christened DOM-tower, the multi-pmt digital optical modules are attached to breakouts in flexible electro-optical backbone cables running the full length of the tower at both sides. As opposed to the single-pmt tower, in this tower each optical module has a unique optical link with the data acquisition on shore. Each optical module is galvanicaly isolated. The choice for fibres in the backbone of the tower as opposed to copper wires is driven by several key performance indicators: (1) timing calibration; (2) signal attenuation; (3) bandwidth; (4) cost. In two independent studies during the design study the feasibility of the use of copper wires in the backbone of the tower has been investigated. In one project, the use of VDSL2 communication over twisted pair has been studied; in the other project the use of coax cable. From these studies it became clear that transitions between the copper and fibre domain complicate timing calibration 33

46 and that it is preferable to remain in the optical domain as much as possible. Signal attenuation over copper wire is more serious than over fibre; in the case of copper wires additional amplifiers in the detection unit will be required and consequently more electrical power for the unit will be unavoidable. In addition, to accommodate sufficient bandwidth in the detection unit modulation techniques must be applied which enhances the amount of electronics in the deep sea. The choice for several twisted pair copper wires would increase the diameter of the cable thus introducing a larger drag and would make the cable less flexible. Finally, the cost per metre of copper wires is higher than that for fibres, making the copper solution for the backbone cable not significantly cheaper than the fibre solution. Since the consortium has decided for the validation of the DOM-tower [17], which is not described in the technical design report, we will summarize this configuration here. A drawing of the storey is presented in Figure 13. The material used for the mechanical structure is Aluminium A system of four Dyneema ropes with a diameter of 4 mm connects the storeys in such a way that each storey in a tower is positioned perpendicular to the previous one. Flexible electro-optical backbone cables are spiralled around a rope at either side of the tower. These cables are of the type Pressure Balanced Oil-Filled (PBOF) and utilize a low density polyethylene (LDPE) tube with an outer diameter of 6 mm as a conduit for the electric wires and fibre optic lines. This option for cabling provides for a reliable and configurable cable system suitable for many subsea applications. For each optical module the cable contains a separate fibre, two copper conductors run the full length. At each storey a breakout cable with 1 fibre and 2 copper conductors connects the backbone cable with the optical module. The tower is anchored to the seabed using a dead weight of 1908 kg in sea. A separate optical fan-out module (de)multiplexing module - with DWDM (Dense Wavelength Division Multiplexing) technology is positioned about halfway the length of the cable thus allowing for a maximum of 11 fibres in the cable. A few meter above the deadweight the base structure the lowest storey in the tower without optical modules is positioned in such a way that a proper equilibrium position is assured. At the base structure, in a Titanium container the two flexible cables are connected to an interlink cable which at the other end connects to the seabed cable infrastructure with a wet-mateable connector. The two spacer storeys in the lower part of the structure are not equipped with optical modules. At the top of the structure a buoy system made of syntactic foam is installed to provide the 7000 N buoyancy required to keep the structure vertical. Buoyant material (a total of 2500 N) provides local buoyancy to each storey. The total drag causing a displacement of the top buoy of 144 m at a horizontal sea current of 0.3 m/s is sufficiently small in a detector configuration with a horizontal tower-distance of 180 m. A current of 0.3 m/s is considered a catastrophic event for which a deep-sea structure must be resistant. The full tower structure of 20 storeys with connecting cables and ropes, 2 spacer storeys and the base storey, deadweight, and two top buoys is stored as a compact package. The storeys and spacers are stored in five columns (see Figure 13). The package (about 2.5m x. 2.5m x 6 m) fits in a 20 foot ISO high cube container (6.1 x 2.9 x 2.44 m) to adapt transport directives. The compact package is deployed to the seabed. Once the deployment cable is removed the structure is released with the help of a ROV and the tower can unfurl to its full length of about 900 m. 34

47 Figure 13 Left: drawing of a single storey with local buoyancy (yellow); rope and cable management is not visible. Right: stacked tower with top buoys, anchor and top buoy (orange blocks in the middle. Validation of towers Validation of the tower configuration with six single-pmt optical modules at each storey is no longer considered by the consortium. During 2010 one in-situ test has been performed, which showed the complexity of unfurling such a tower. A mechanical prototype of the NEMO tower which has a mechanical structure similar to the DOM-tower has been deployed successfully in February 2010 (see Figure 14). Prototypes for the DOM-tower configuration are being prepared for deployment in During these deployments, the mechanical structure of the DOM-tower, the flexible backbone cable and the deployment method will be validated. Deployment and connection to shore of a small sized tower with three optical modules is also foreseen in Validation of the backbone cable in the lab is ongoing. Although not used in neutrino telescopes so far, the technology of PBOF cables is widely utilized. Figure 14 Deployment of a mechanical prototype of a NEMO tower with single-pmt optical modules at either end of the storeys (left). Unfurling of the NEMO tower from the seabed (right) Cost estimate for towers The total investment cost of the infrastructure of a tower a DOM-tower is and that for a tower with single-pmt optical modules is The breakdown of these numbers is presented in Table 6. The large difference in cost for the buoy is partly due to the fact that in the single-pmt PM tower inherently has more buoyancy delivered by the 120 optical modules and only needs six glass spheres as a top buoy. On the other hand, drag calculations for this structure has not been scrutinized to the full extent, since the consortium decided to abandon this design. The choice for two backbone cables in the DOM-tower and, as a consequence, two optical fan-out modules to allow for redundancy is another reason for the large difference in cost for the tower infrastructure. The cables run the full length of the tower, each on one side of the storeys thus providing redundancy in the tower. If the cost of the optical modules, the readout electronics and instrumentation is included, the total investment cost for the DOM-tower is and for the single-pmt OM tower The fact that the total investment cost for a DOM-tower is lower than that for the single-pmt OM tower is largely due to the fact for the DOM-tower the cost for 35

48 optical modules, electronics and instrumentation is much smaller (see also section 4.4); this more than compensates the cost for redundancy of the two backbone cables and the higher cost for the buoy system. multi-pmt DOMtower Single-PMT OM tower Tower mechanics Buoy Optical fan-out modules Backbone cable Total tower infrastructure Optical modules, electronics, instrumentation plus mechanical interface to tower infrastructure Total tower Table 6 Breakdown of estimated investment cost for a detection unit in two different tower configurations. Risk and reliability analysis of towers The risk analysis of the optical modules has been presented in sections 4.2 and 4.3. Since the modules (and in case of the single-pmt OM tower the electronics containers) are galvanicly isolated water leaks in an optical module will not propagate in the tower. The risk analysis of the vertical backbone cable is somewhat complicated, as this is a custom designed cable. Many subsea electrical and optical cable and connection systems now utilize Pressure Balanced Oil Filled (PBOF) cabling solutions. Industry claims that PBOF cable systems are both sturdy and longwearing thus providing for a reliable cable system with a life expectancy of year in the deep sea. The application of this technology in a vertical and long cable (beyond 500 m) is new. The manufacture of ANTARES interlink cables has recently changed standard dry cable to PBOF cable. This cable has been installed in ANTARES. It is an excellent opportunity to monitor its behaviour in the coming years. In case the development of the PBOF cable for KM3NeT fails, the alternative is to utilize the conventional dry cable as e.g. used in the NEMO project. In the DOM-tower two backbone cables are implemented, thus providing redundancy. At the cost of loss of redundancy, these two cables could be combined into one to reduce investment cost. It is however preferred by the consortium to let redundancy prevail over cost reduction. The unfurling method and the stability of the mechanical structure of the DOM-tower are still to be validated, although earlier deployments of prototypes of the NEMO tower have been successful. In-situ validation of the single-pmt OM tower is no longer considered. Integration of towers Assembly of optical modules has already described in sections 4.2 and 4.3. The integration of the DOM-tower will be similar to the procedure for a tower with single-pmt OMs described in the technical design report. Experience with the integration of similar towers for the NEMO project has been invaluable. Since the digital optical modules are designed as standalone modules the integration of the tower can be modularized. I.e. the optical modules can be assembled and tested separately; optical modules can be calibrated separately in small light-tight boxes; construction and testing of the electro-optical backbone cable in the tower can be possibly outsourced; integration of the tower is then limited to testing the electro-optical contact with an optical module. It is not necessary to test storeys with optical modules as a whole in a dark room as in the case of the single- PMT OM tower described in the TDR. Although it is not a strict requirement, it is preferred for towers to install the integration site in or close to the harbour for the deployment vessels. To comply with a production of 320 towers in a period of 4 years, it is estimated that the integration site for towers must have a hall of 300 m 2 for 3 integration lines and 750 m 2 space. The integration effort for the towers is presented in section 4.8 in comparison with that for strings. 36

49 4.6 Strings As described in the technical design report, the multi-pmt optical module can also be installed in a string-like mechanical structure. In that design, the optical modules are suspended by two Dyneema ropes with a diameter of 4 mm, which run parallel over the full length of the string. As in the case of the DOM-tower, a flexible backbone cable with breakouts at each optical module runs the full length of the string. Additional empty glass spheres provide buoyancy at the top buoy. The anchor is a concrete dead-weight with a volume of about 1 m 3 to which the vertical mechanical ropes are connected. The weight in air of such an anchor is 2400 kg and therefore the negative buoyancy in water is N. For the deployment of the string a custom recyclable spherical launching vehicle has been designed with a diameter of about 2.1 m (Figure 15). Three sets of cable trays run from pole to pole and are offset by 60 degrees. Between the cable trays of each set, holes in the sphere provide the space for suspending the optical modules. The vehicle is loaded top down during integration of the string. First the glass spheres of the buoy are loaded on guiding rails through the hole at the North Pole. The spherical vehicle is rotated around a winding axis perpendicular to the first cable tray. The optical modules are placed in holes and kept in place by a lever blocked by the ropes. The ropes and the backbone cable are laid in the trays. The vehicle has three tubes running through them from a spreader structure at the top to the anchor. The spreader structure is secured to the anchor with an acoustic release mechanism. When released the spreader structure floats independently to the sea surface, while the spherical launching vehicle also rise to the surface unwinding the string to its full length. The launching vehicle and the spreader structure are recovered for re-use. The total weight of the loaded launching vehicle and the anchor is about 1200 kg in air. The drag of the top of the structure is calculated 95m a sea current of 0.3 m/s, sufficiently small for a detector with an inter-string distance of 130m. Figure 15 Launching vehicle in rotator (left). Loading the launching vehicle with optical modules on a string structure in the lab (right). Validation of strings The mechanical structure and deployment method of the string using a recyclable launching vehicle has been validated with three separate deployments during two sea campaigns. During the last campaign in Jan-Feb 2011, the launching vehicle was re-loaded with mechanical optical modules on deck of the deployment vessel. The results have been reported at the VLVnT11 workshop [18]. Improvements should include the mechanical interface between the optical module and the suspension ropes and that between the optical module and the breakout box in the backbone cable. The inclusion of a real backbone cable is still to be validated. 37

50 Figure 16 Deployment of a prototype of the launching vehicle loaded with mechanical prototype optical modules (left); unwinding of the string from the seabed (right). Cost estimate of strings The investment cost for a DOM-string is estimated The breakdown of this number is presented in Table 7. multi-pmt DOM-string String mechanics 3925 Buoy 1200 Optical fan-out modules 5000 Backbone cable Total tower infrastructure Optical modules Total tower Table 7 Breakdown of investment cost for a DOM-string. Risk and reliability analysis for strings All components in the DOM-string are also present in the DOM-tower. In fact, from a technical point of view the DOM-tower can be described as two DOM-strings kept at a distance of 6 m of each other by the storeys. Except for the deployment method and the stability of the DOM-string, the risk analysis is therefore similar. String-type detection units are operational in the deep sea since 2006 in ANTARES. Experience has shown that their stability is as expected. The validation deployments have shown that the concept of a spherical launching vehicle is solid. Integration in strings The integration of a DOM-string is described in detail in the technical design report and summarized above. Experience with string-integration during the validation tests has shown that the estimate of a total of 2-3 fte-days for integration of a string is realistic. As for the DOM-tower, tests of the strings are restricted to the test of the optical and electrical connection of the optical modules, since these modules are delivered as standalone items. Since the anchor of the string is the last component to be integrated, it is attractive to choose for the solution to connect the anchor on board of the deployment vessel. This will relax the lifting requirement for a crane in the integration site and ease transport. Since the launching vehicle is at the same time a reliable transport frame, of which a few fit together in a standard transport container, it is not strictly necessary to install the integration site near the harbour of the deployment site. It is nevertheless preferred in order to avoid very strict transportation regulations over large distances. To comply with a production of 640 strings in a period of 4 year, it is estimated that one integration site for strings must have a hall of about 300 m 2 for 6 integration lines and about 750 m 2 storage space. Also a lifting crane with a capacity of about 1200 kg is required. The total integration effort for a string is presented in section 4.8 in comparison with that for towers. 38

51 4.7 Comparison between towers and strings The physics performance of a KM3NeT detector composed of DOM-towers or DOM-strings is presented in chapter 0. The difference in investment cost and integration cost for the three configurations is discussed in section 4.8. Here, we constrain ourselves to the results of a naive failure mode analysis, based on past experience. This naive analysis has been performed to achieve a first order estimate of the average loss of the photocathode area in a detection unit due to failures during a life time of 15 years. In these calculations, failures considered are those of photomultipliers, electronics components, components of the backbone cable and the connection of the detection unit to the seafloor network. Not considered, are failures inside the junction boxes or in the main cables to shore. The results of this naive analysis (Table 8) indicate that in first order all three configurations show an acceptable loss of photocathode area of about 10% over a period of 15 years without maintenance. DOMtower Single- PMT OM tower DOMstring Estimated loss of photocathode area per DU 8% 12% 8% Table 8 Estimated first order average loss of photocathode area in a detection unit due to failure of photomultipliers, electronics, backbone cable and the connection to the seafloor network. Shown are the results of a naive failure mode analysis for the three different configurations described in the text. Since the price of the photomultiplier has a large impact on the total cost of the KM3NeT detector, in particular in the case of the multi-pmt optical module, it has been investigated what the influence is when photomultipliers are ordered in smaller batches. For this study price-quotes for photomultipliers made in 2010 have been used. Price estimates were given for 3, 8 and 10 inch photomultipliers when ordered in various batch. For a KM3NeT detector of different sizes, the total investment cost for the detection units in the detector was calculated. For comparison, this number was divided by the total number of storeys in the detector. The result of this study is shown in Figure 17 for detector configurations with DOM-towers, towers with single-pmt OMs and with DOM-towers. The cost per storey is presented as a function of the number of detection units in the detector. The increase in cost per storey for a detector with a small number of DOM-strings reflects the higher price of the photomultipliers when ordered in small quantities. This effect is strongest for a small number of strings (<10). The cost per storey for towers with single-pmt OMs reaches a price-plateau at a detector with about 100 towers. 39

52 Total DU-investment cost per storey as function of the number of DUs (using PMT-price quotes in 2010) Total DU-Inestment cost per storey in k Number of detection units Configuration with DOMtowers Configuration with towers with single-pmt OMs Configuration with DOMstrings Figure 17 Total DU-investment cost per storey for three different KM3NeT configurations: DOM-towers, towers with single-pmt OMs or 640 DOM-strings. Shown is the dependence of this number on the number of detection units in the KM3NeT detector. The investment cost are calculated using price-quotes for large and small photomultipliers to be delivered in the quantities dictated by the number of detection units. Clearly, the cost per storey is lowest for a DOMstring, however this number should be multiplied by two to be compared to the cost per storey in a DOM-tower. 4.8 Investment Cost and Integration Effort In Table 10 the estimated total investment cost for the KM3NeT neutrino telescope are summarized for the three different configurations of the detector: 320 single-pmt OM towers, 320 DOM-towers or 640 DOM-strings. Given the uncertainties, the total cost of two multi-pmt options are similar at about M 225. Total cost for the 320 single-pmt OM towers is M 240. Both figures are within the targeted range of M The main cost items are listed in increasing order of the total amount required for a full KM3NeT detector. In all configurations, the PMT-unit (photomultiplier, HV base and in the case of the multi-pmt DOM the light collection ring or in the case of a single- PMT OM the mu-metal shielding) is the most frequent item. Here, cost reduction will have large impact on the total investment cost for KM3NeT, in particular in the case of the DOM-tower or the DOM-string. An option to decrease the number of relatively expensive wet-mateable connection of the detection unit to the seafloor network is to dry-connect several detection units before deployment and deploy the connected detection units as one object using two vessels instead of one. This option was considered for the DOM-string; in a preliminary contact with an off-shore company its feasibility was confirmed. The option has not been studied in detail, since the consortium has decided to give priority to the DOM-tower design. For the DOM-tower the option of multi-tower deployment is less feasible due to the larger weight and size of the towers. The estimated assembly and integration effort for a full KM3NeT detector is summarized inerrore. L'origine riferimento non è stata trovata.. Given the large uncertainties, these numbers are similar for the three configurations. With a nominal value of 100 per hour for personnel hired on project basis, the effort would add a total of 5-7 M to the project. 40

53 KM3NeT detector Amount DUs Amount (D)OMs/Electronics containers Assembly (D)OMs/ Electronic containers [fte-year] Integration DUs [fte-year] Total [fte-year] Single-PMT OM towers / / DOM-towers DOM-strings Table 9 FTE required for the assembly of DOMs and the integration of detection unit for a full KM3NeT detector. 41

54 Table 10 Break down of estimated investment cost for a KM3NeTdetector of 320 DOM-towers or 640 DOM-string 42

55 5. Site evaluations Site selection criteria The criteria relevant for the choice of one or several deployment sites for the KM3NeT neutrino telescope have been discussed intensively during the Design Study phase and thereafter. The following criteria are important: 5.1 Scientific and technical criteria Investigations of site characteristics have been intensively pursued before and during the Design Study. The site parameters impact on design, price and physics sensitivity of the neutrino telescope. The following issues are of particular importance: Water depth: On the one hand, increasing depth improves the shield against backgrounds from downgoing atmospheric muons and allows for observing an increasing fraction of the sky above horizon, thus improving the physics sensitivity; on the other hand, it tightens the requirements for the pressure resistance of optical modules, electronics containers, cables etc. and also for the deployment operations and in particular the availability and costs of Remotely Operated Vehicles (ROVs). Distance to shore: Increasing distance to shore increases the price for the main cable and its deployment, as well as ship transit tiimes and thus deployment risks. It also decreases the efficiency of electrical power transmission and the optical margin of data transmission. Also, it limits the flexibility of deployment operations in case of unstable weather conditions. Short distances may allow for designs with several main cables that would be too expensive for large distances. Topology of the sea floor and the cable path to shore: A flat sea floor without obstacles like big rocks is required for safe deployment and ROV operations. The site environment should exclude deep-sea land-slides and other catastrophic events. The cable to shore must not run across sharp edges, gaps without support or very steep slopes; also, it must have a safe and smooth landing place on shore. Geological situation: The risk of major earthquakes, volcano eruptions or submarine landslides should be taken into account. Water transparency: Absorption and scattering of light in the water is the limiting factor for the distance of Detection Units (DUs) and the vertical spacing of photo-sensors on the DUs; less absorption and scattering means higher physics sensitivity, corresponding to a reduced overall capital investment. The exact dependence of the sensitivity on the optical water parameters is complex and needs further study. A precise knowledge of the water optical properties and its temporal variation is required for physics analysis; these parameters must be monitored continuously. Background light: Whereas the amount of light from decays of K40 and other radioactive nuclei is largely siteindependent, bioluminescence exhibits strong geographical and temporal variations. Periods of high bioluminescence, as e.g. observed by ANTARES, reduce the overall data taking efficiency and thus the physics sensitivity. There is evidence that bioluminescence decreases below water depths of ~3000m (see TDR). 43

56 Sedimentation and bio-fouling: These effects can cause a layer of reduced optical transparency on the glass surfaces of the optical modules. Even though there are indications that these layers are intermittently washed off during periods of high current velocities, they reduce the overall physics sensitivity, in particular for studies that require detecting Cherenkov light coming from the upper hemisphere. Measurements by ANTARES have shown that up to angles of 45⁰ above horizon (i.e. looking upwards) there is no long-term evidence for biofouling or sedimentation effects decreasing the detector efficiency. Water currents: The KM3NeT design is safe against current velocities up to about 30cm/s; currents exceeding 45cm/s will be destructive since the anchors will start to move. Even lower current velocities could be dangerous if the current is not uniform or even turbulent. Measurements at the candidate sites indicate the 95% of the time the current velocities are below 15, 8 and 6cm/s for the Toulon, Capo Passero and Pylos sites, respectively. Weather and sea conditions: The deployment of the DUs and the deep-sea cable network will require long periods of sea operation, for which calm weather and sea conditions are required. Also, the predictability of weather changes is important to avoid operational risks. 5.2 Infrastructural and logistics criteria Preparation, deployment and operation of the KM3Net research infrastructure will require support and infrastructure at the landfall of the selected site(s). The following issues are important: Availability of shore station: There must be a suitable site (and optionally a suitable building) to house the shore station, close to a suitable landfall location of the main cable(s) and with appropriate connection to electrical power (100 kw) and a high-bandwidth data connection to the computing centre. Availability of computing centre: There must be a suitable site (and optionally a suitable building) to house the computing centre with a high-bandwidth data connection to the European data backbones. Computing centre and shore station can be combined in the same building. Harbour and vessels: A harbour to host the deployment vessels and auxiliary ships must be close. Logistics requirements: The site installations must be easily reachable for the participants, i.e. good roads (also for shock-free transport of detector elements) and the proximity of an international airport are important. Transports of standard containers by sea freight should be possible. Appropriate accommodation must be available for project members and partners over the full year. Storage capacity: There must be storage capacity for detector components and room for on-shore predeployment tests. 5.3 National and local support The host country and its scientific community should provide a series of services and guarantees: Operation teams: The core teams for the detector maintenance and servicing and for central tasks of the deployment should be provided by the host country/institute(s). 44

57 Site manager: The host country/institute(s) also have to provide a manager who interfaces the project to the local authorities and population. Planning security: The host country must make a political commitment to support the project with the necessary local resources over its full expected lifetime. It must guarantee the availability of the necessary infrastructure over this period. Scientific community: A strong national science community supporting the project and representing it in the local, regional and national science, funding and science policy bodies is desirable. Measures to guarantee this local/national scientific support over the full lifetime of the project are desirable. 5.4 Site-related financial issues It is expected that the overall funding of the project will have to be agreed upon on a multi-national, European or world-wide level. A significant fraction of the funding may be provided through European Regional Development Funds (ERDF), which very likely need to be spent locally and therefore impact on the site choice. In particular, this might imply a distributed, networked installation. The consequences of such a scenario for physics sensitivity, operation and management is currently being assessed. However, should such a scenario become reality, it will be of utmost importance to secure the following: Coherent management: All parts of the installation must be governed by a common, site-independent body making sure that they are operated as one coherent detector. Measures have to be taken to guarantee that detector operation is fully transparent, according to operation rules and modes decided by the common governing or management bodies. The host countries have to commit themselves to this policy. Current situation The scientific/technical criteria have been investigated in detail during the KM3NeT Design Study. A review of the results is contained in the TDR. No show stopper was identified for any of the sites, even though it the bioluminescence situation at the Toulon site is found to be less favourable than at the other sites and may result in reduced detection efficiency. The experience from ANTARES data taking shows that this effect is equivalent to about 15-20% of data loss. Differences in the water transparency have been observed between different sites, however these measurements are snapshots in time and the amplitude of temporal variations is not well studied at all sites. Initial simulation studies based on intermediate design configurations have been performed to assess the impact of depth and water transparency on the detector sensitivity to neutrino point sources with an E -2 spectrum. The results obtained exhibit a small depth dependence and indicate that the sensitivity is roughly proportional to the transmission length. However, these studies have been performed for upgoing neutrinos and fixed detector configurations not identical to the current design. The sensitivity gain of an enlarged angular acceptance above horizon at larger depth and from a geometry optimisation as a function of water transparency is under study but has not yet been quantified. The effect of atmospheric muon background and its dependence on depth is under study; we are still lacking sufficient Monte Carlo statistics to give precise answers. 45

58 The basic infrastructure and logistics criteria appear to be matched by all candidate sites; details, such as the choice of buildings etc., can only be negotiated after a basic commitment of the corresponding country. The same is true for the national and regional/local support. Currently, some funding is secured in the Netherlands (site-independent, but dependent on a coherent site decision of the consortium), in France and in Romania. All these together currently cover roughly 10% of the overall cost. In Greece a commitment of up to 50M (limited to 20% of the overall cost) has been announced several years ago but the consortium was never invited in writing to a common project using these resources. Significant funding requests are pursued in France and Italy, with decisions pending and expected soon. The French, Greek and Italian funding sources are mostly of regional character (in particular ERDF) and are bound to conditions on the site and on the country/region where the money has to be spent. Explicit statements to this effect have been made by INFN and in the Greek commitment letter to the EU Commission. It seems unlikely that one of these countries will take the project lead and cover a major part (50% or above) of the overall cost. Support by and access to funding through ESFRI may be possible and could depend on the one- vs. multi-site character of the project. In the current situation, most (but not all) partners of the consortium consider a distributed, networked installation at different sites the only pragmatic solution, provided the physics objectives are not compromised by such a scenario. Some of the Greek partners consider leaving KM3NeT in case a distributed installation is pursued. Answers to the questions 4a. Site assessment See TDR. 4b. Optimal detector at best site The technical detector design as pursued to date does not generically depend on site characteristics. All technical solutions are adaptable to all sites, irrespective of the distance to shore and the water depth. The geometry (i.e. footprint) optimisation is currently driven by the focus on Galactic sources with an energy cutoff in the 100 TeV regime, which as compared to E -2 sources without cutoff has a much more dramatic effect on the optimal detector setup than e.g. the water transparency is expected to have. The abovementioned simulation studies will result in quantitative statements on this subject. 4c. Impact of a distributed installation For technical reasons (complexity of sea floor network, bandwidth (i.e. number of fibres) per cable to shore, ease of deployment operations, redundancy) the full KM3NeT neutrino telescope will in any case be constructed in 2, probably several independent blocks. Simulation studies indicate that the physics performance in the search for neutrino point sources does not suffer from such segmentation. (See chapter 3.5) A distributed but networked installation therefore does not significantly compromise the priority physics objectives, provided the individual blocks are sufficiently large (the critical size if of the order of 1 km 3 of instrumented volume, i.e. IceCube-like). No results are currently available on the impact on shower analyses, which however have lower priority (see chapter 3.4) For the construction, the following additional costs arise as a consequence of multiple sites: 46

59 Additional shore infrastructure (shore station, online computing, connection to highbandwidth backbones, infrastructure for voltage supplies etc.). Since the investment cost for the online computing scales with the number of independent detector blocks, the extra costs for multiple sites are expected to be limited to the provision of housing, power and bandwidth. A generous estimate of the extra infrastructure cost per site is therefore 1 M. Additional vessels, tools and shore infrastructure for deployment. The corresponding costs depend on the time schedule, i.e. whether the deployment operations at different sites are pursued in parallel or sequentially. The latter case is equivalent to deployment in one site as far as overall time schedule and vessels are concerned (at least to the extent that vessels can easily be moved between sites). A parallel deployment would require additional vessels and crews, however for a shorter period (i.e. the same integral time); there would be overhead in training crews and equipping vessels. The shore infrastructure (building to store and possibly test/calibrate detection units) is included in the item above.the overall additional cost cannot be reliably estimated but is not expected to exceed 2 M overall. Note that no additional components are required due to the modular structure of the neutrino telescope. Also note that logistics will be more difficult but not more expensive since the integration of the detection units is planned to be done centrally in several labs of the consortium and the ready-to-deploy detection units are then transported to the deployment sites. For operation, clearly additional personnel will be necessary for servicing the online computing farm and maintaining the shore infrastructure. We assume that the actual detector operation (adjustment of online filters, run control, data quality control) will be done remotely. Nevertheless, a 24/24 maintenance service at the shore site is necessary; a crew of 6 people is deemed necessary, corresponding to 1 M per year per extra site (including overhead). The efforts for calibration and data analysis will not depend strongly on the number of sites since the marine environment in any case requires continuous monitoring and a calibration performed separately for each detector block at short intervals (typically minutes). Reconstruction and simulation software will be run independently for the individual blocks, whether or not they are at one or several sites. Some overhead can be expected from the enhanced effort to provide appropriate sets of environmental parameters at different sites for the simulations, but once the monitoring machinery is in place this effort is moderate. The above-mentioned additional effort for installation at different sites results in enhanced funding resources and, in addition, in further advantages such as: increased redundancy in the availability of deployment resources; reduced dependence on local issues (phases of high bioluminescence, power outages, etc.) reduced impact of local catastrophic events (earthquakes, landslides, etc.); increased versatility of the infrastructure for earth and sea sciences 4d. Use of different sites without splitting the detector Currently, no such scenario appears to be realistic. This could for instance be different in a situation where one of the partner countries takes a strong project lead and the remaining partner countries could use their sites for test/prototyping efforts and for earth and sea science purposes. However, as stated above, this is not the case. 47

60 6. Project risk Evaluation of the risks for the project to build the KM3NeT facility requires the identification of threats to which a probability on a scale from 1 (low) to 5 (high) will be assigned and the impact on the project will be classified as low, medium or high. As a first step threats have been generally categorized in technical and programmatic risks. Each of these can be caused by internal or external factors. Table 11 shows in more detail the categories that have been recognised. Programmatic threats Technical threats External Politics/Strategy External Context Legal/Regulatory Definition Industrial politics Internal Design Organizational R&D activities & prototypes Financial Logistics Media Realisation Environment Exploitation Internal Logical sequence Project management Performance management Organisational/Resource Budget Contractual/Legal Safety Suppliers/Manufacturers Table 11 Categories of threats for the project of building the KM3NeT facility. In each category preliminary initially conceived threats have been formulated (see the Appendix B). Further iteration inside the consortium is still to be organised to evaluate these threats and assign the probability of occurrence and the impact on the success of the project of building the KM3NeT facility. Below a preliminary inventory of the level of criticality of the major threats is given for each of the categories. Programmatic threats The main external programmatic threat to the project is a lack of funding or a funding profile which does not match the foreseen spending profile of KM3NeT. The total investment cost of the infrastructure was worked out in the TDR and is within the targeted range of at M This estimate has been confirmed in subsequent studies. It is based on the experience with ANTARES, offers from industry and prices of standard components. As such, the cost risk is limited. However, such a sizeable amount requires contributions from regional, national and European funds. The currently secured funds are insufficient for the immediate start of construction. At the time of this writing, proposals have been submitted in France, Greece and Italy to acquire funding from the structural funds for regional development of the 7th Framework Programme of the EU as well as national funds. The lack of available funds may delay the construction of the infrastructure. A phased construction may alleviate the immediate funding but it will also postpone the scientific results. Allocation of the funds with requirements of ab unrealistic spending profile will also influence the project negatively. Another threat to the project could be a too slow convergence to the creationof a collaboration, including management, for the construction phase of the project. This is related to the lack of 48

61 immediate funding and is likely to be solved when about half of the required investment budget for construction is committed. Currently, the KM3NeT community is organised in a consortium that has been formed to execute the design study and the preparatory phase of the project. For the construction phase, the choice has been made to establish an ERIC (European Research Infrastructure Consortium) to be signed at ministerial level of the participating countries. First steps have been taken in formulating a MoU for the funding of the construction of the prototypes to be deployed in This is to assure that after the end of the preparatory phase in February 2012, these projects can be completed and the tendering procedures for mass production can be initiated at the end of it. The site-issue has divided the groups in the consortium for a long time and has not been fully resolved yet. Nevertheless, it has been shown that a distributed neutrino telescope with building blocks of about the size of IceCube has no discernable effect on the science potential of the KM3NeT neutrino telescope. From a technological point of view a remotely operated distributed network of telescopes with the same technology and a common data centre is feasible with only small additional cost. However, within the consortium full consensus has not yet been reached. This lack of consensus is the clear and present danger to the successful construction and operation of the KM3NeT facility. Technical threats The design of the KM3NeT detector has been built on the experience of the pilot projects, particularly the ANTARES detector. From these projects we have recognised potential threats and have alleviated many of them, such as propagating failures and the risk of leaks. This has resulted in a design that not only is significantly cheaper, but also passes the reliability criteria of less than 10% optical modules failures in 10 years, at least on paper. However, a few potential threats remain. One of the external technical threats to the project could be the large scale of the project. In comparison to ANTARES, the enlargement is a factor 50 and compared to IceCube it is still a factor of more than 5. Although the scaling up is considerable, it is fortunate that a neutrino telescope is a relatively simple detector because of the fact that all sensors are identical. Compared to the complexity of the LHC detectors for which many of the institutes in the consortium have built components in large numbers, the sensors of KM3NeT are relatively simple. A downside to the simplicity is that some of the items are required in very large quantities from a restricted number if vendors viz PMTs. A complication particular to KM3NeT are the logistics of the sea operations. Building on the experience with ANTARES, it is expected for KM3NeT these operations will quickly become common place, in particular if dedicated vessels and ROVs with dedicated crews can be employed for the full period of construction. A possible technical threat to the project is the quality of the deep-sea components. In particular, experience in ANTARES has shown that the quality of connectors and vertical cables produced in industry can be critical. However, lessons have been learned for the design of the KM3NeT telescope and the number of connectors per photocathode area has been considerably reduced in comparison to ANTARES. Another experience from ANTARES is that industry is not used to the deployment of long vertical cables and technical input from the consortium for the realisation of such cables is indispensible. Although industry provided feasibility studies that have shown that flexible oil-filled cables are a viable solution for the KM3NeT vertical cable, they seem as yet unwilling to commit to the design and production of such an item. The consortium has therefore the design and validation into its own hands. If the cable cannot be validated in time, the alternative is to use a dry cable. Such a cable has the disadvantage that it is more expensive, is less flexible and requires more complicated handling during assembly, but it has been shown to work in the prototype NEMO tower. 49

62 Bibliography [1] KM3NeT: Technical design report. ISBN , [2] F. Aharonian et al., Rep. Prog. Phys., 71(2008) [3] F. Aharonian et al. (H.E.S.S collaboration), Astron. Astrophys., 448(2006)143. [4] F. Aharonian et al. (H.E.S.S. collaboration), Nature, 440(2006)1018. [5] T. Tanaka et al., ApJ., 685(2008)988. [6] Q. Yuan, P.-F. Yin, and X.-J. Bi, Astrop. Phys., 35(2011)33. [7] D. C. Elison, D. J. Patnaude, P. Slane, and J. Raymond, ApJ, 712(2010)287. [8] Z. H. Fan, S. M. Liu, Q. Yuan, and L. Fletcher, A&A, 517L4(2010)101051/ / [9] A.A. Abdo et al., ApJ, 734(2011)28. [10] M. Su, T. R. Slatyer, and D. P. Finkbeiner, Astrophys. J., 724(2010)1044. [11] M. Crocker and F. Aharonian, Phys. Rev. Lett., 106(2011) [12] L. A. Anchordoqui, H. Goldberg, D. Hooper, S. Sarkar, and A. M. Taylor, Phys. Rev., D76(2007) [13] F. Halzen and S. R. Klein, Rev.Scien.Inst, 81(2010) [14] E. Flykt. (2003) Workshop on the technical details of a Very Large Volume Neutrino Telescope in the Mediterranean Sea, VLVnT. [Online]. [15] Q. Dorosti. (2011) 5th International workshop on Very Large Volume Neutrino Telescopes, VLVnT11. [Online]. [16] E. Leonora. (2011) 5th International workshop on Very Large Volume Neutrino Telescopes, VLVnT11. [Online]. [17] KM3NeT consortium. (2010, Dec.) Technical description of the PPM of the KM3NeT detection unit. km3net.org/public. [18] E. de Wolf. (2011) 5th International workshop on Very Large Volume Neutrino Telescopes, VLVnT11. [Online]. 50

63 Appendix A: Estimated cost investment In the tables below, the estimated investment cost for the three configurations of a KM3NeT telescope are presented in some detail. The cost is estimated using the assumptions and procedures listed in chapter 9 of the TDR. As stated there, cost of components are taken, in descending priority, from industrial quotations, corresponding costs as occurred in the pilot projects, public catalogues and informal or confidential statements of providers. 51

64 Table 12 Breakdown estimated investment cost for KM3NeT with 320 DOM-towers. 52

65 Table 13 Breakdown estimated investment cost for KM3NeT with 640 DOM-strings. 53

66 Table 14 Breakdown estimated investment cost for KM3NeT with 320 single-pmt OM towers. 54

67 Appendix B: List of possible threats to the project Below is presented a preliminary list of threats. This is still to be improved in an iterative procedure inside the consortium. Probabilities of occurrence of the threats and their impact on the project are also still subject to evaluation inside the consortium. Technical Risks External risks in project Internal Risks in project Context Definition Design R&D activity & prototypes Logistics Emergence of new technologies calling into question previous choices Instability in needs, requirement & constraints Misreading or instability of interfaces Constraint Relations with other developments Uncertainty on feasibility, heterogeneity of system components High size of the system, instability of system architecture Specifications : incomplete, inadequately precised or too ambitious Complexity or high size of the system to be done Impossibility of realisation of Technical elements of interface Bad expression of need, Lack in supplies identification Change of need after the beginning of the project Bad interpretation of requirements (Dependability included) Complicated or innovative technology with lack of control and to be developed Lack of scenario studies (technical options) Complexity of technical solutions Out-of-date technical solutions Change in system architecture Lack of technical standards Is the specified development time correct? Difficulty to reach performances (including margins) Omission of interfaces with other systems or projects Insufficient take of exploitation contraints into account. Lack of technical data. Uncertain report of states (inventory, as build ). Lack of technical maturity of the project Reliability of components or objects (components: unreliable, identified critical by AMDEC, under wear, for which burn is impossible ) Object or element feasability (no validation or forbidden, not measurable after integration, for which process is delicate, with no guarantee on durability, limitation in manufacturing, implementation or control) Other risks on elements (long time for supply, with high impact, with short life-time, with unique supplier, vulnerable to transport, under periodic maintenance, under exploitation license) New technologies chosen in and of itself (technologies never used before, immatured or exotic). Use of "at the limit" technlogies. Obsolescence of programs Incompatibility of updates Use of proprietary softwares Potential obsolescence of components Difficulties to demonstrate/justify performances of technical choices, adequation with validation and justification methods High technological risks (innovative technical solutions but with no industrial validation) Missing data related to process or product studied High level of innovation, late development verification Late take of construction site organization, lack of storage areas, long supply time, deterioration in transport between laboratories and experiment sites Weakness of components or systems Difficulties to transport some components, equipements without possible substitution. 55

68 Realisation Exploitatio n Non-conformity Lack of qualification tests Key points badly defined Difficulties to make tests Bad definition of control plans Impossibility to validate sub-assemblies before assembly Expensive tests, Late controls Assembly and integration externalised (not done by consortium) No appropriation of exploitation data Use of complicated softwares Forgotten maintenance Programmatic Risks Lack in strategic analysis External risks in project Politic/Strategic Legal/Regulatory Industiral Politics Organizational Financial Media Instability of need Uncertainty on long-term programs Interfaces between programs (impact of changes in other programs) People or group of people with different scientifical & technical levels put neck and neck A team or a laboratory might call for or be forced tasks for which they not have needed competences. Lack of regulation Lack of standard Patent difficulties European regulation (Health ) Contradiction between countries regulations Changes in regulation standards Different understandings of standards Drift of instruction time for safety or security files Contraints related to partners Unavailability of partner's technical means Project has not prioritory for partner Architecture is not optimised Incompatibility between official regulation and some practices in industry (terms for payement) Comment: This could prevent project from some industrial support, which would respond to our needs. Customer is not well known Long time to take a decision Confusion between roles of client, project manager (general contractor), contracting owner counseling, project manager delegate Lack of representative Shortcircuit in process of decision taking Possible changes in project organisation Industrial strifes : impossibility to access to equipment in a firm, an experiment. Future user : Missing or unexperienced representative. Bad acceptation of change No experience and/or training to project management, for some big projects responsibles Lack of multi-annual agreement Financial trade-off are not in favor of the project Economical state of customer Financial constraints of supervision authorities Complexity of financal architecture Cost or time objectives are too ambitious Financial deficiency of a partner Project acceptability (public debate, publique poll, survey) Many decision-making interferences Environm ent Aggressions from earth (earthquake, mudslide, various falls, geotechnics, volcano), water (flood), air (climate, frost, wind, bad weather, lightning) Plane, road, rail traffic, and from surrounding industries 56

69 Logical sequence Man, source of danger (interactions with neighbourhood) Bad definition forgotten tasks, roles and responsibilities. Responsibilities are no defined between subsystems Mistakes of evaluation over the sequence of tasks An interface is not totally treated Project management Lack of tools, means ; insufficient scheduling and organization Lack of reporting or indicators Insufficient scheduling or work organization Insufficient margin : cost, time, performances Knowledge management / Know-How Decisional shortcircuit Bad diffusion of information Configuration management / Differences between documentation and products No performance management plan Performance management Organisational/Resource Bad identification of technical performances at the beginning of the project, bad control at the end Missing or bad definition, or bad timing of conceptual reviews No prototype Differences between physicists and engineers points of view are complementary ; but sometimes they could be at the opposite. High level of turn over (during project, departure and/or mobility of people with crucial know-how ; it's not easy to substitute them regarding human resource management in publique administration) Incompatibility between groups or persons Role and responsibilities are not well defined. Complexity of project organization. High number of implied actors Hard mobilisation of human resource Long duration of procurement Bad definition of criterions for the selection of suppliers. Too late implication of support actors (jurists, buyers, controllers, purchasing agents, project management) Lack of resources. Lack of experience of project team. Unqualified resources (to make some equipments ; due to little flexibility in human resources management ; laboratories could devote people to tasks for which they have nobody or not trained people) Internal Risks in project Budget Contractual/Legal Safet y Loss of skills (mobility, retirement) Hazards and their related provision are not taken into account Funds for travellings insufficient or not well-managed could lead to restrictions for travelling at crucial time for communication between collaboration members. Lack of reliability in financial forcasts. Lack of data to make budgets. Impossibility to attribute funds for some parts of the project. Difficult sprinkling in financial plan Difficulties to do financial realigment Bad legal management of contracts (missing legal articles) Validity of first evaluations Changes of rules in program management Clarity and completeness of contracts Long duration for procurement (CCM) Bad definition of criterions for the selection of suppliers. Cost management Reporting, "fait accompli" politics Insufficient knowledge of regulations (collaboration between several countries : differents in standards and regulations - if this is not identified and solved at the beginning of the project, issues will occured) Forgetting or bad coverage of insurances Late descovery of security requirements ; technical solution reappreasal for security ; lack of communication with security authorities ; lack of demonstration 57

70 Actors of control and safety are not implied or too late. Fire (Heat source, flammable products or materials) Insufficient means. Accidents on the experiment site. Mechanical aspects (pressurized equipments, elements under mechical contraints, elements in motion, elements with needs of handling) Source of physical explosion other than pressurised equipment, high vaccumed volum, explosive gas Source of falling, of tumbling, and other source of injuries Electrical origin (direct or alternative current, medium and high voltage, electro-statiscme, power capacitors, high frequencies) Thermal and radiation hazards (ionizing radiation, thermal sources - burn, laser, microwaves, magnetic fields) Biological hazard (Virus Bacteria - Room with controled moisture - Toxins) Man, source of danger (operator) Work station, source of danger (design of work station) Differences between physicists and engineers points of view are complementary ; but sometimes they could be at the opposite. Complete failure of a sub-contractor : bankrupt, stop of activity Suppliers/ Manufacturers "Partial" Failure of an industrial subcontractor: non-conformance of product %(revenue of project)/(global revenue) Turn over Expected contribution is not contractualized ; conflict between priorities (lack of reactivity, commitment, resources ) Borderline of suppliers (knowledge, skills, availability) Insufficient contacts Work load is under estimated Means of production / control / test Knowledge of the program team Appropriateness of industrial architecture Market situation : monopolistic situation, low competition. Change of the situation (Production cycles of scientific equipments are very long. Components, specified and validated at the beginning of the project, could not be produced anymore at the beginning of the production) 58

71 1 An alternative Strategy for KM3Net November 7, 2011 Introduction In this document we outline our position concerning recent developments with KM3NeT and the document submitted to the SSC ( Documentation for SSC meeting , 2011/11/01 V3.0; to be referred to as SSC answers ). We note that according to the KM3NeT statutes the relevant body for approving important strategic decisions is the SPB (Strategic Project Board). The SPB approval for the SSC answers has not been granted. The signatories of the present document are serving as co-coordinators in several KM3NeT-PP Work Packages (WP) (L. Resvanis-WPB/Political Convergence and WPG/Auxiliary Surface Vessels, I. Siotis-WPC1,C2,C3/Legal, Governance and Financial Issues, P. Rapidis-WPD/Strategic Issues and International Networking). The Institutions they represent account for about 24% of the KM3NeT-PP EU budget. Though we would have preferred to address only science and technology issues we feel obliged to introduce financial parameters to the extent that they affect the science issues. In this respect, we copy here the statement from thekm3net- Design Study proposal that summarizes our position. "...The selection of the optimal site for the infrastructure presents a unique challenge to our scientific community due to the intricate interplay between scientific, technological, financial and socio-political / regional considerations. It is our intention to deliver a clear prioritization of site qualities based on scientific, technological and financial aspects only. These site studies were completed during the KM3NeT-DS period but the prioritization based on the results of these studies was never concluded. Such a prioritization is necessary because it affects the telescope design significantly. The design has to incorporate features that are to a large extent site specific. The SSC answers Strategy Instead of following the approach outlined above, the SSC answers focuses entirely on financial and socio-political/regional considerations concerning the important question of the deployment site. The key argument concerning the site issues in the SSC answers is presented in p.46 as follows: Currently, some funding is secured in the Netherlands (site-independent, but dependent on a coherent site decision of the consortium), in France and in Romania. All these together currently cover roughly 10% of the overall cost. In Greece a commitment of up to 50M (limited to 20% of the overall cost) has been announced several years ago but the consortium was never invited in writing to a common project using these resources. Significant funding requests are pursued in France and Italy, with decisions pending and expected soon. The French, Greek and Italian funding

72 2 sources are mostly of regional character (in particular ERDF) and are bound to conditions on the site and on the country/region where the money has to be spent. Explicit statements to this effect have been made by INFN and in the Greek commitment letter to the EU Commission. It seems unlikely that one of these countries will take the project lead and cover a major part (50% or above) of the overall cost. Support by and access to funding through ESFRI may be possible and could depend on the one- vs. multi-site character of the project. In the current situation, most (but not all) partners of the consortium consider a distributed, networked installation at different sites the only pragmatic solution, provided the physics objectives are not compromised by such a scenario. Some of the Greek partners consider leaving KM3NeT in case a distributed installation is pursued. Comments on SSC answers and an Alternative Strategy In our opinion this statement provides the justification for the...distributed networked installation at different sites..., but is both misleading and incorrect for the following reasons: 1. The Greek Government s commitment for funding 20% of KM3NeT, up to 50 M, has been on the table since 2007 and has been communicated in writing to EU Commissioner responsible for Research Mr. Potocnik in October In December 2008 the Greek Minister informed his colleagues at the EU Competitiveness Council and extended in writing an invitation to join in a common effort. 3. In March 2010 the Greek Secretary General for Research, representing Greece at the highest possible level on the KM3NeT-PP ASC, reaffirmed the commitment and explained the reasons why it is not affected by the current financial crisis. He also stressed that all contracts must be signed by the end of 2013 and payments must be concluded by the end of 2015 and he stated that the 50 M commitment is not an upper limit. (See Appendix 1 for documentation concerning these three points) 4. Currently, some funding is secured in the Netherlands ( 8.8 M site-independent, but dependent on a coherent site decision of the consortium), 5. The French funding of about 8 M is essentially dedicated to the development of the MEUST neutrino telescope (see ) and is initially planned to double the ANTARES sensitivity. It will be scalable up to the size of one out of three KM3NeT Building Blocks (each building block is ~20 x ANTARES) but the time scale for this has not been specified. No details of a specific funding request (amount,

73 3 timescale) for this expansion has been communicated to the KM3NeT-PP Work Package C dealing with funding issues, neither has the KM3NeT-PP ASC been informed. 6. The Italian proposal for funding through ERDF (European Regional Development Fund) concerns an amount of 45 M (the max allowed in the call). Decision is expected by the end of this year and if successful, all contracts must be signed by the end of 2013, as for Greece. 7. The Romanian funding prospects are somewhat vague though a commitment of ~2 M has been mentioned by the Romanian delegate at the KM3NeT-PP ASC. 8. The statement about a possible ESFRI (European Strategy Forum for Research Infrastructures) funding is totally incorrect and misleading. ESFRI does not fund the construction of Research Infrastructures and has no plans to do so in the future. In this sense the issue of one or more sites is totally irrelevant as far as ESFRI is concerned (one of the authors of this document, I.S., is the Greek delegate to ESFRI). 9. Though true that ERDF funding imposes constraints and conditions on the site and on the country/region where the money has to be spent, such conditions do not necessarily apply to the country/region where the telescope will be deployed. An example of this is the use of ERDF funding by Estonia in establishing a beamline at the new synchrotron light source at the MAX IV Laboratory in Sweden. 10. For all of the 3 KM3NeT candidate sites the telescope will be deployed in international waters so the discussion about construction location vs. installation/deployment location, i.e. locality of investment, is irrelevant. In addition even if the shore station is in one country/region it is possible to transmit the data with a few GHz link to any other. In view of this we should be trying to get round the ERDF funding problems by exploiting this unique international aspect of KM3NeT as outlined above. Such an approach would combine scientific honesty (in the sense of choice of the best site, see later) with political inventiveness (in the sense of exploiting the broadband ICT potential) and is likely to be appreciated by all as it would show in practice how technology enables European integration. We would not be surprised if such a solution became a showcase for the EU. 11. The statement in the SSC answers that...some of the Greek partners consider leaving KM3NeT in case a distributed installation is pursued... is totally incorrect. What we have asserted is that if KM3NeT adopts a three site policy we will apply our best efforts to take advantage of the unique physical characteristics of the Pylos site and of the existing shore and sea infrastructure. This implies optimizing the telescope design which in turn may imply a different concept to that of the SSC answers. In view of these comments on the SSC answers we are very surprised by the conclusion that the...only pragmatic solution... is...a distributed, networked installation at different sites....

74 4 Our conclusion is that this so called...only pragmatic solution... is not driven either by physics or funding arguments but rather by local political considerations that have not allowed the KM3NeT consortium to focus on the important scientific and technological challenges. A striking example of this is the so called technological convergence represented by the multi PMT option combined with the 6m bars. That this is far from optimal is shown in the SSC answers document where the biggest, by far, performance difference (improvement) comes from going from 6m to 15m bars. The SSC answers shows this effect in Table 2 p.14. Other internal notes on the bar length effect, extending to 48m bar length, show that the improvement increases monotonically up to that length. Three years ago we abandoned our design efforts with extended tower structures composed of six arm stars (see KM3NeT-DS CDR p.79) as part of a compromise that would have lead, at least, to the choice of the best site. The result today is that we are faced with both a non optimal design and a non optimal multi site telescope. According to the funding scenario described above in points (1) through (8) the total amount potentially available for construction of KM3NeT over the next 5 years is ~ M or about 50% of the required budget as estimated in the SSC answers. Under these circumstances, even if all three sites were equivalent, which is not the case, we would argue that the present funding scenario should lead us to the concentration of our efforts on a single site rather than on three. The reason for this is that, for example, since the time required to achieve a 5 sigma observation of RXJ1713 is (optimistically) 6 years according to the SSC answers, it would take at least 10 years for a 5 sigma observation with the present funding! We believe that, in the absence of a discovery in the first 5 years of operation of KM3NeT it is unlikely to see the funding agencies invest another 110 M. For this reason it is extremely important to recover a 5 year discovery time horizon with the presently available budget. This might be achievable if we concentrate our efforts on the best site and the best telescope design through the following gains to be detailed in what follows. Site and Design Considerations Potential approximate gains in detector performance expressed in neutrino/ by choosing the best site and an optimized design, for the same investment, include: 1) 18% from depth, going to 4500 m vs m allows us to increase the observation angle to the horizon for the same background and reduce the atmospheric muon background. 2) 18% from better transmission length in water assuming the neutrino/ goes as the square of the transmission length. 3) 3% to 25% from lower dead time due to bioluminescence bursts that are depth and site dependent.

75 5 4) 15% lower cost due to smaller distance to shore, lower shore station overheads, and exploitation of dedicated deployment platform. 5) 20% from optimization of the telescope design, basically in the case of bar towers, going for much longer bars. The combination of all these effects leads to a gain in neutrino/ by a factor of two. In other words with 110 M we can get, with a single (best) site and a better design, the same result that we would get with 220 M at three sites. The reason for this is of course that all three sites are not equivalent and hence the present funding scenario forces to the best one in terms of neutrino/. Site Related Gains Depth, Water Quality, Bioluminescence Site properties affect and even determine to a significant extent the performance of telescopes, including that of a Neutrino Telescope. In the case of a neutrino telescope the relevant site parameters are the deployment depth, the optical properties of the sea water at the site, the bioluminescence, the sedimentation, and the biofouling. Depth is important in two ways (a) the overlying column of water serves as a shield for the down coming cosmic rays, which in our case are muons induced in cosmic ray interactions in the atmosphere and (b) in providing a larger target mass for neutrinos arriving at the detector at angles above the horizon. The background in an undersea neutrino telescope has two main components (excluding the ones due to 40 K decays and bioluminescence), one arising from atmospheric neutrinos and another from atmospheric muons. Both components are due to interactions of cosmic rays in the atmosphere. Atmospheric neutrinos are an irreducible background, i.e. there is no way to shield against it or reduce it, and any signal due to an astrophysical source can only be discriminated from this background through its directional character (for point like sources) and its energy spectrum (since atmospheric neutrinos are much softer). The atmospheric neutrino background is depth independent and due to the fact that the target mass distribution is larger in the nadir (rock under the sea bottom) results in more upwards going muons. The atmospheric muon background decreases with depth, e.g. the vertical muon flux intensity decreasing by a factor of 18.6 as one goes from 1.9 to 4.0 km water depth. Table 1 shows this decrease for various relevant sea depths, and is based on the data shown in Figure 1. The residual background caused by badly reconstructed down coming muons (i.e. reconstructed as upcoming) scales directly with the muon flux reaching a given depth. Therefore in a neutrino telescope the minimum detectable signal flux improves with depth since the contribution to background from misreconstructed down coming muons reduces with depth.

76 6 Max depth (m) Middle of DU (m) Top of DU (m) Table 1.A: Ratios of muon flux for different depths at the NEMO and NESTOR sites Depth at NEMO site Depth at NESTOR 4500m site Depth at NESTOR 5200m site Flux at NEMO/Flux at NESTOR 4500m Flux at NEMO/Flux at NESTOR 5200m Table 1.B: Ratios of muon flux for different depths at the ANTARES and NESTOR sites Max depth (m) Middle of DU (m) Top of DU (m) Depth at ANTARES site Depth at NESTOR 4500m site Depth at NESTOR 5200m site Flux at ANTARES/Flux at NESTOR 4500m Flux at ANTARES/Flux at NESTOR 5200m Table 1 shows the ratios of the vertical muon flux at different pair of depths for the NEMO site compared to the two NESTOR sites, for the maximum depth for each site, for the maximum depth minus 900m (which corresponds to the top of the KM3NeT Detection Unit (DU) ) and for maximum depth minus 500m (which corresponds to the middle of the instrumented length a KM3NeT detection unit). Table 2 shows the flux ratios for the corresponding set of depths for the ANTARES site and the NESTOR sites. The DU will have 20 storeys, spaced by about 40 m with the lowest storey at a height of about 120m above the seabed. The total length of a DU is therefore 880m with the instrumented length being 760m.

77 7 References 1 A. Tsirigotis thesis, 2004, otis/ 2 E. V. Bugaev et al, 1998, Atmospheric Muon Flux at Sea Level, Underground and Underwater, 3/files/ pdf 3 Paolo Desiati for the AMANDA Collaboration, 2003, Response of AMANDA-II to Cosmic Ray Muons, ipts/ xx _ffinal.pd f 4 V.A. Kudryavtsev et al, 2000, Energy calibration of large underwater detectors using stopping muons, f/0010/ v1.pdf 5 J.A. Aguilar et al, 2010, Zenith distribution and flux of atmospheric muons measured with the 5-line ANTARES detector, /33061/1/zenith-Antares.pdf 6 KM3NeT Technical Design Report, M3NeT.pdf Figure. 1: Plot of the vertical muon flux vs. depth. In the blown out picture the measurement obtained from the NESTOR test detector in 2003 (Page 292 from A. Tsirigotis thesis)

78 8 The basic argument of the SSC answers regarding the effects of depth on the sensitivity of KM3NeT has been developed for a source at a declination of -60 degrees, i.e. one that is always below the horizon. In this case, after making appropriate reconstruction quality cuts the atmospheric muon background can be reduced to a level below the irreducible atmospheric neutrino background. The only reference to the benefit of depth in reducing the atmospheric muon background appears indirectly in the SSC answers on p 14: Another effect that could influence the sensitivity to Galactic sources is the position in the sky. RXJ passes above the horizon for about 5 hours a day and reaches to 15 o above. Investigations are underway to determine if the background is still manageable when reconstruction is attempted above the horizon. Antares has shown that up to 5 o is possible. Such an investigation requires a huge sample of simulated atmospheric muons. The production is of these is presently underway. Assuming the full 15 o can be reached a further improvement of about 20% can possibly be obtained in the FoM. To be precise, for the latitude of the Pylos and Capo Passero sites the RX J source reaches 13.7 degrees above the horizon. It is 6.2% of the time in the angular region of 0 to 5 degrees above the horizon and 22.9% of the time in the region of 5 to 13.7 degrees above the horizon. The 5 degree above the horizon limitation of the ANTARES analysis is determined by the flux of atmospheric muons. This angular limit can be significantly increased to higher elevations for a detector in deeper water. Figure 2, which shows the dependence of the number of generated atmospheric muons (from Sapienza, Amsterdam Km3NeT general meeting, March 2011) is indicative of the achievable elevation angle for various depths. Thus a deeper detector at the Capo Passero/Pylos latitude at 5 km depth, for similar quality cuts as in the ANTARES case, is expected to be sensitive up to 14 o and therefore sensitive to RX J at all times, and thus have a gain of 14% over a detector at 3.4 km depth for the Capo Pasero/Pylos latitude which would be sensitive for up to about 9 o above the horizon (this 14% is due to 3 additional hours of observation time per day). The simulation of atmospheric muons is a daunting task, since it is very CPU intensive, especially if one considers the fact that many of the muons are in reality a bundle of many muons travelling together. To date no adequate study of this background source has been carried out, especially for angles above the horizon. In an experiment where one must take data for several years in order to establish a signal, it is imperative that the background estimation should correspond to comparable running periods. Notwithstanding the computational challenges we note that the exceptional background events will never be simulated well enough. Therefore, the only pragmatic approach is to shield the telescope as much as possible, i.e. go as deep as possible.

79 9 From : Sapienza KM3NeT general meeting, Athens March For upcoming neutrino events, i.e. for angles below the horizon, preliminary existing studies for a specific detector design show an estimated sensitivity gain of 2% for 4.5 km depth vs. 3.5 km depth and a 3.5% gain from 2.5 to 3.5 km (See Figure 3A, from Sapienza, Amsterdam Km3NeT general meeting, March 2011). These estimates refer to a point like source with an E -2 neutrino energy spectrum and at a declination of -60 degrees, which means that they are always below the horizon.

80 10 Figure 3A : Sensitivity ratios with atmospheric muons as a function of depth of deployment. From: Sapienza, KM3NeT general meeting, Amsterdam March Figure 3B: Sensitivity for various sites. At each site the depth, water absorption, and geographic latitude are different. From Cogniglione, WPD KM3NeT meeting, Paris May

81 11 Figure 3B (from Cogniglione, WPD KM3NeT meeting, Paris May 2010) shows the depth, water clarity, and site latitude effects for the three sites for a source at declination δ=-40 o (essentially that of RX J at 39 o 46 =39.78 o, thus we use the name source for now to refer either to the galactic object or to a source at -40 o ) as well as for a source at a declination δ=-60 o. The red colored lines include the effect of the atmospheric muon background. Tracks 6 o above the horizon and higher are not used because of the increase of the muon background at larger angles. We observe the following a. For δ=-60 o and no atmospheric muons (black lines) the change in sensitivity is entirely due to the variation of the light absorption in the water from site to site. b. For δ=-40 o no corresponding change in sensitivity is seen in going from Toulon to Capo Passero. This is due to the fact that the improvement in water quality (see the next section on water quality) is compensated by the latitude difference between the two sites. At the Toulon site the source reaches a maximum of 7.4 o above the horizon and is visible using the 6 o for almost all of the time whereas in Capo Passero it reaches 13.7 o and spends a substantial amount of time at high horizon angles (above 6 o, i.e. when the detector is blinded by the muons). c. For δ=-40 o including the muon background leads to a loss of sensitivity of 9% at Toulon, 4% at Capo Passero and no loss at Pylos in spite of the rejection of track with angles above the horizon larger than 6 o. This is clearly a depth related effect and would become more important if were to incl tracks up to 13.7 o which is the highest angle for the source at Capo Passero and Pylos. By extrapolating the observed losses for tracks up to 6 o we estimate that if we go up to an angle of 13.7 o the overall gain in sensitivity in going from Capo Passero to Pylos is about 18%, due to the better shielding against atmospheric muons. The physics advantage achieved by the fact that by going deeper one can see above the horizon is directly applicable for any diffuse neutrino flux of cosmic origin as well. In the case of very high energy neutrinos (e.g. E>100 TeV) this effect becomes quite significant because the Earth becomes opaque to neutrinos, thus looking sideways (near the horizon) and above is the only way to observe. Estimates for the increase in neutrino effective area for E>100 TeV are 26% (from 2.4 km to 3.4 km) and 33% (from 2.4 km to 5 km) (Sapienza, Athens KM3NeT general meeting, March 2009). A similar result was obtained for GRB s where the expected event rate is larger by 12% for 4500 m depth as compared to 3500 m (Tsirigotis et al., Nucl. Instr. and Meth. A, 639 (2011) 79). Since the SSC answers addresses only galactic point like source searches we have not included such considerations in our neutrino/ estimates. Water quality is a major determinant of the detector performance. The transmission length of the water plays a dominant role in selecting the separation distance between optical modules. Clearer water leads to a larger effective area for a given detector design. In addition, clearer water allows for building a sparser detector since light travels further. Thus the dependence of the effective neutrino area for an optimized detector will be in some power of the increase of the transmission length. A naïve expectation would be a cubic dependence on the transmission length, but such a

82 12 behavior has to be determined by proper MC simulation. Water quality is an environmental parameter that may show temporal (seasonal) variation. Long term studies at the Pylos and Capo Passero sites demonstrate the constancy of the water quality and show very little (if any) seasonal variation. Comparison between sites is not easy; differences in instrumentation and methodology make them so. To date three series of measurements have been carried out with the same instrument and at the same time period (measurements taken within a few days of each other). The first series was in August 2002 and utilized the AC9, a commercial instrument with a short optical length (25 cm) which does not allow for a high accuracy (see Riccobene, Pylos KM3NeT general meeting, April 2007). Measurements were taken at the Toulon and Capo Passero sites and show that the absorption length at Capo Passero is much better than at Toulon (50 m vs 70 m). The second series of measurements were taken with the AC9 at the Pylos and Capo Passero sites in July These measurements were presented only in preliminary form (raw data) in 2007 (Riccobene, KM3NeT WP5 meeting, Rome, October 2007). The third set of measurements was made in Capo Pasero and in Pylos with a specially built instrument with a long optical path (>10 m), to address the inherent accuracy problem of short optical path devices in clear water. These measurements were carried out in a five day cruise in May 2009, and we refer to the published description for details (Anassontzis et al., Astroparticle Phys. 34 (2010) 187). The situation can be summarized by stating that at the Pylos area the water has 10% larger transmission length than the one in Capo Passero. Figure 4: Sensitivity as a function of water absorption length From: Sapienza Amsterdam general meeting March

83 13 Detailed MC studies of the dependence of the effective area (or the sensitivity) on water quality have not yet been carried out. The only study where the water absorption length was varied (Figure 3, Sapienza, KM3NeT general meeting, Amsterdam March 2011), but otherwise this specific detector s parameters were not changed, shows that the change in sensitivity is roughly linear in the absorption length, with dφ/φ= x dλ/λ. The obvious study of increasing the optical module separation for clearer water has not been done. Thus for 10% better absorption length (longer λ) we get an 8% better limiting sensitivity flux (smaller φ) without utilizing an optimized detector. We can expect at least a quadratic dependence for an optimized detector leading to a 16% better sensitivity for the Pylos site. Bioluminescence can be a significant problem. The ANTARES site has suffered episodes of long term increases in bioluminescence that have led to a complete stop of data taking for 13% of the time and compromised data taking for 21% of the time (for the period of ) (Payet, Amsterdam KM3NeT general meeting, 2011). It is the case that a higher bandwidth detector design, more sophisticated reconstruction schemes, and running with decreased PMT gain will help, but the basic problem will still remain. Bioluminescence in the deep Ionian Sea is not as severe. Operation of the NESTOR prototype has shown that it suffered only 1% dead time due to bioluminescence (i.e. bioluminescence burst leading to singles rates larger than 200 khz). Studies of bioluminescence at the two Ionian sites as a function of depth and of the season are shown in the TDR (Fig 5-16 and related text). Based on this data we expect a 2-3 times more dead time due to this effect at the Capo Passero site than at the Pylos site. Sedimentation and biofouling lead to a film formation on the surfaces of the glass optical modules and therefore to a decrease in light collection. Sedimentation affects mostly the upper part of the optical module; while biological growth is probably independent of direction. Long term studies of sediments at the Pylos site show very low sedimentation and the sediments on the sea bottom indicate long term stability of the sedimentation. The sedimentation at Capo Passero is notably larger than at Pylos (TDR fig. 5-8). There is evidence of significant sedimentation at the Toulon site, quite likely a reflection of significant regional geological activity (see e.g. ). An evaluation of the effect of sedimentation (i.e. of diminished light collection efficiency for upwards pointing PMTs, which may be significant in the high sedimentation sites) has not been done. Biological growth has been observed on at least one optical module of the ANTARES device but it did not lead to any measurable effect on the OM performance. Devices at the Pylos site that have been in the water for periods of a few years do not show any growths. There seems to be insufficient data at this time to evaluate any potential problems due to this source for very long times.

84 14 Site Related Gains Distance to Shore, Use of a Dedicated Deployment Platform, Shore Station Overheads The total cost of deployment for all three detector configurations, 320 multi-pmt DOM towers, 640 multi-pmt DOM strings, or 320 single PMT OM towers is estimated in Table 10 on p.42 of the SSC answers to be 30 M. This figure is dominated by ship time cost for both for the deployment vessel and the ROV mother ship. If conventional deployment vessels and methods are to be used then this figure does not depend significantly on the distance to shore. On the other hand for a short distance to shore it is possible to reduce considerably this cost by using a dedicated deployment vessel, built especially for this purpose. Such a vessel, aiming to take advantage of the short distance to shore at the Pylos site has been designed and built by the Greek teams in KM3NeT. The Delta-Berenike vessel (Figure 5) was launched in 2007 and at present is being equipped with all the necessary fittings. The final funding installment of 1.2 M for completion, provided by the Greek ERDF, was activated in 2010 and the first tranche of this final amount has been spent. The Delta-Berenike is a self propelled special purpose vessel to be used as a stable platform for the deployment at a depth of down to 5200 m of the components of the Cubic kilometer Neutrino Telescope. This is a novel instrument which will facilitate the deployment and reduce the required time and costs. She is a ballasted vessel with a central well able to keep station to ±1m and operate up to wind/sea state force of 5 on the Beaufort scale. Figure 5. : The Delta-Berenike being fitted in the shipyard.

85 15 The cost of deployment for 320 towers with the use of this dedicated vessel combined with a dedicated landing craft vessel for bringing DUs from the shore is considerably reduced as this combination also satisfies the requirements for deployment and ROV mothership. The cost difference between deploying at 100 km from shore with conventional vessels and deploying at 30 km with such a combination of dedicated vessels is about 22 M. To this one should add the cost difference for the procurement and deployment of the electro-optical cable which we estimate to be about 1 M (for a 10 /meter cost) thus leading to a total cost saving of about 23M. If more than one cable runs are requires, which is quite likely the case, the savings will scale accordingly. An additional cost saving of about 2 M would result from having one rather than three shore stations as detailed in SSC answers, p.47. The overall investment cost saving thus amounts to 25 M of the total KM3NeT budget without taking into account the annual savings in operational costs that are estimated in SSC answers, p.47 to be 1 M /year per extra site. The discounted present day value from the savings in annual operational cost of 2 M s (for two additional sites) with a discount rate of 5% per annum is approximately 7 M s. Therefore the overall savings amount to 32 M s, some 15% of a 220 M total cost. Detector Design Gains Systematic optimization studies of the detector layout, both in terms of the distance between OMs (e.g. the bar length in the design of the SSC answers, or the distance between storeys) as well as of the distances between detection units have not been carried out. It is very telling that the SSC answers the optimum bar length is found to be much longer than the one that was being considered until recently. As shown in Figure 6 (from Kopper, WPD meeting, 12 Sept 2011, Amsterdam) increasing the bar length from 6 meters to 48 meters for a design similar to the SSC answers one, but with 180 m intertower separation, decreases the required running time for a 5 sigma discovery for the RX J source from 12.1 to 8.2 years, a gain of almost 32%. This is by far the most striking evidence of the importance of an optimized layout. A similar effect for the same source is seen in Table 2 (p. 14) of the SSC answers for intertower separation of 130 meters when the bar length is increased from 6 to 15 meters, leading to a decrease in running time for 5 sigma discovery from 8 to 6.2 years, a 23% gain. Clearly a lot more work needs to be done on these issues; we estimate that a further 20% gain from such an optimization.

86 16 Figure 6. Event rates per year for RX J using a detector similar to the SSC answers DOMbar detector, but with 180m intertower spacing, as a function of bar length (6, 12, and 48 m lengths). The time to reach a 5 sigma discovery for this source is also given (from Kopper as indicated in the text). Note: Atmospheric muon background is not included in the above figure. From : Kopper, Amsterdam WPD meeting Sept 12, A systematic study of various footprints for a detector were carried out in S Kuch s thesis (U of Erlangen 2007, Report no FAU-PI1 DISS Dissertation.pdf ), but that study was for smaller sized detectors and was done mostly at the trigger level simulation and thus is not directly applicable to the current designs. Designs that include variable density regions (e.g. a central dense core and a sparser outer region) or non-conventional geometries (e.g. a thick walled cylinder arrangement) have been discussed but not fully studied. The physics objective determines to a large extent the detector design. If one focuses on galactic sources, with a lower energy spectrum, then denser detectors are more appropriate. A focus on GRB s or GZK neutrinos, i.e. neutrinos of larger energies, favors sparser layouts. As a result a first generation undersea neutrino telescope should aim at addressing both objectives, and not compromise any of them. The studies of the intricate interplay between intertower distance and bar length mentioned above may allow for such an optimal detector.

87 17 Conclusions In this short document we have tried to show that the funding conjecture underlying the SSC answers document should lead to the choice of the best site as well as a detector optimized for that site. In case that arguments as to the importance of depth are not deemed crucial or convincing by the SSC, we bring to the attention of the SSC and the Consortium that the Pylos area has the possibility to host a telescope at shallower depths (at 3000 m depth at 13 km from nearest shore, and at 3750 m depth at 14 km from nearest shore) as detailed in the TDR and in the provided booklet on the Pylos site characteristics. These shallower sites still benefit from all the advantages shown in this document, excluding obviously depth. This subset of the Consortium believes that a strong concerted effort by the Consortium in such a direction is more likely to be accepted by the scientific community and the funding agencies as serious and supportable.

88 vx 18

89 APPENDIX

90

91 HELLENIC REPUBLIC MINISTRY OF DEVELOPMENT GENERAL SECRETARIAT FOR RESEARCH & TECHNOLOGY SECRETARY GENARAL Athens, 22 th October 2007 R.N: SG 1500 To: Mr. Janez Potocnik Commissioner for Science and Research European Commission B-1049 Brusselles Belgium c.c Minister of Development Mr. C.Folias Dear Commissioner Potocnik During your visit to Athens on May 31 and June 1, we officially informed you of our Government's decision to host the Cubic Kilometer Neutrino Telescope (KM3NeT), a European Research Infrastructure on the ESFRI Roadmap, and fund it from 2009 on with 20% of the total cost with a ceiling of 50 Million euro if it were located in Pylos. We hereby reconfirm our offer and we solicit the ECs support considering that 1. The deep sea near Pylos is the best possible site because a. It is the deepest in the Mediterranean down to 5200m very near the shore and at the same time there are numerous plateaus literally next to the coast, at depths of 3000m, 4500m etc.. b. Proximity to shore of the various deep sea plateaus minimizes the cost and time of the various deployment/recovery/maintenance operations as well as the cost of power and data transmission to/from the telescope. Further, it makes all sea operation safer because there are 4 harbours within miles where one can find shelter in bad weather. c. The optical properties of the deep water are truly excellent, underwater currents are either minimal or essentially non existent and sedimentation is minimal Pylos is located on the bay of Navarino, a very safe and large natural harbour, where we already have a fully instrumented test station that can be used continuously throughout the year.