Deep Ocean Neutrino Detector Development:
Contributions by The DUMAND Project

R. Jeffrey Wilkes, J. G. Learned, and P. W.Gorham

Department of Physics, University of Washington, Seattle, WN 98101, USA,
Department of Physics and Astronomy, University of Hawaii at Manoa, Honolulu, HI 96822, USA


The DUMAND Collaboration carried out over a period of nearly twenty years, studies and tests aimed at building a high energy neutrino telescope in the deep ocean. The project saw neutrion astronomy emerge from vague notions to concrete accomplishments in science and technology in the deep ocean. In this paper we review the accomplishments of the DUMAND effort, with an eye towards making information accessable for those carrying on the drive towards high energy neutrino astronomy.


The announcement in 1996 by the US DOE of the termination of their support for the DUMAND-II project was a great disappointment to the many people who have worked so hard and so long to realize neutrino astrophysics using an underwater Cherenkov detector. The purpose of this paper is to summarize the many accomplishments of the DUMAND Collaboration in advancing our knowledge of techniques, demonstrating the feasibility of constructing a deep undersea detector, and laying the groundwork for future detector designs. The effort to begin high energy neutrino astronomy goes forward today in Lake Baikal, at the South Pole (AMANDA), and in the Mediterranean (NESTOR and ANTARES). We hope that the documentation herein may be of some aid in their success.

The DUMAND concept in its present form (more or less) has been discussed, and construction projects of various degrees of practicality have been proposed, since the mid- 70s[DUMAND 76][Roberts 92]. DUMAND-I was a ship-suspended single prototype string which was successfully operated in 1987[Babson 90]; the array of bottom-moored strings described in the 1988 proposal [DUMAND 88] is called DUMAND-II. The water Cherenkov technique was further refined in the early 80s by the successful construction and operation of large-scale nucleon-decay detectors (later used as low-energy neutrino observatories) by the IMB[IMB 92] and Kamiokande Collaborations[Kamiokande 94]. Experience gained in these projects helped make it possible for DUMAND to move from concept to reality and receive construction funding from the Department of Energy (DOE) in 1990. The cost and risks involved in deep-ocean engineering operations were still a matter of concern to agency officials.

There were also several false starts in terms of scientific and technological attempts, which are discussed below.

The complete DUMAND-II array would have consisted of 216 Optical Modules (OMs: photomultiplier tubes plus front-end electronics, encased in a standard glass oceanographic pressure sphere) deployed on nine vertical strings, moored in an octagonal pattern with 40m sides and one string in the center (Fig. 1). The instrumented portion of each string begins 100m above the ocean floor to avoid benthic boundary-layer effects. In addition to OMs, the strings include sets of hydrophones and other acoustical/environmental equipment, and calibration modules, in which a constant-output laser light source is used to excite a scintillator ball viewed by the PMTs.

In this paper we are not attempting to state every possibly useful result or conclusion to be drawn from the DUMAND labors, but to give a survey with references to guide interested people to the large amount of documentation available. A personal history of DUMAND was written by Arthur Roberts [Roberts 92], and contains many historical details of the development of the DUMAND Project, which we will not repeat here.


Stimulation of Astrophysical Consideration of Neutrinos

It is generally acknowledged that the early DUMAND Workshops ([DUMAND 1975] and [DUMAND 1976] in particular) served as the first major stimulus for astrophysicsts to seriously consider the role of neutrinos in the universe. As neutrino studies proceeded at accelerator laboratories in the 70's, and as astrophysicist learned more about strange and powerful objects throughout the universe, it slowy was realized that not only would neutrinos permit ``seeing'' into such objects, but that neutrinos might play crucial roles in these engines.

It was also noted that neutrinos provide a unique means to detect whether the source region is composed of matter or anti-matter, via the interactions of neutrinos versus anti-neutrinos (see below). This remote sensing of anti-matter regions of the universe, while not immediately going to be carried out, is the only method for doing so now known.

A historical note may be in order as well. The Baikal group has been working in Lake Baikal in Siberia for about the same time as the DUMAND group in Hawaii. Initially, in the mid-1970's, the groups worked together and indeed the 1979 DUMAND Workshop was held on the shore of Lake Baikal. However, political problems developed after the Soviet invasion of Afghanistan, and the US DUMAND group was told by its government that no funding would be available to collaborate with the Soviet groups. Hence the teams reluctantly went their seperate routes. The Baikal group, which has persisted despite enormous social upheaval, and their colleagues, deserve to share credit with the DUMAND Hawaii Collaboration, if there is such credit to be given, for the initial stimulation of neutrino astronomy.

We will not go into detail on the state of neutrino astrophysics here. The field is well launched, as may be tracked through the proceedings of the biannual International Cosmic Ray Conferences, the International Neutrino conferences, and the International Workshop on Neutrino Telescopes series held in Venice (Eight in February 1999). There have also been many specialized meetings, with two being particularly relevant to this text, [Ohashi 86] and [HENA 93]. It was at the latter meeting that the question of substantial high energy ($PeV$ range) emission from AGNs and like objects was seriously considered by a number of the world's leading astrophysicists. It was concluded that such emission was hard to avoid given the observed luminosities, time variations, and inferred source photon densities.

High Energy and Low Energy Neutrinos

At first it was not clear whether the initial neutrino telesopes should be designed for higher or lower energies. At the 1976 Workshop, groups split the hypothetical telescope designs into three major categories: low energies (such as solar neutrinos and gravitational stellar collapse neutrinos), atmospheric neutrinos, and high energies (as from various high energy cosmic sources). These categories continue to be relevant today.

The lower energy neutrinos, in the range from a few MeV to about 100 MeV seemed at first to be the most profitable avenue scientifically. Unfortunately, the calculations all reached the same conclusions: detectors with low backgrounds, few MeV levels of sensitivity, and masses of kilotons were needed for solar neutrinos (detected via Cherenkov radiation), and the size requirement increased with distance, despite the peculiar fact that the brightest astronomical objects are the most distant. The nearby brightest objects always win, witness that the sky is much brighter in photons in the daytime than by the light of the Milky Way. At this time we concluded that we would need an unaffordable detector of 100 megatons with 10 MeV sensitivity for sensing supernovae out through the Virgo cluster, which would yield an observable SN rate in neutrinos of nearly one per week. We realized that achievable detectors would likely only to be able to monitor our own galaxy. Moreover, there was little hope of a large excess of invisible galactic GSC events over visible SN, and so the observation rate might be only one per 50 years or so, very discouraging. This estimate, refined in the IMB proposal in 1979, was bourne out by the observation of SN1987A by Kamioka and IMB.

The high energy study group concluded that the most effective route to detection of multi-GeV neutrinos and beyond was through the observation of muons, which may come from interactions far from the detector. The detector effective volume is given by the array effectrive area times the mean muon range (which depends upon the neutrino spectrum). Based upon observed astrophysical object photon lumiosities one could set rough upper limits upon neutrino spectra (which could not be much greater in total neutrino luminosity in any model yet conceived). This led to the uncomfortable conclusion that to surely get into the business of high energy neutrino astronomy one needed a detector of effective volume in the range of a cubic kilometer, an estimate that remains valid 20 years later.

The study of atmospheric neutrinos was percieved as an effective avenue for approaching neutrino astronomy. It was clear that neutrino oscillations could be sought and early calculations of the effects of oscillations on the observation of muon angular distributions were made. Moreover the early workshops saw that nucleon decay seraches could be carried out with such instruments, though at that time there was not much interest in the topic in the theoretical community. This situation changed dramatically in about 1978 when the simple GUT $SU(5)$ predictions of proton decay became clear enough to justify a major experimental search. The race for nucleon decay included three large Cherenkov detectors (IMB, HPW and Kamioka). There were people from all of the major institutions engaged in these projects who were active participants in the seminal 1976 DUMAND Workshop. The designs put forward for these detectors were preceded by the ideas worked out at that workshop for low energy neutrino detectors.

Muons versus Cascades, Limits on UHE Neutrinos

A debate has continued over the years as to the relative worth and efficiency of detection of secondary muons from neutrino interactions, versus the detection of the interaction, via detection of the elementary particle cascade in water. In a large array, such cascades appear as directional point sources of Cherenkov radiation. The threshold for detection, effective volume and directional sensitivity depend upon the optics of the medium, detector density, and source spectrum, so general statements are difficult.

However, we have demonstrated with the DUMAND Prototype String [Bolesta 97] that small arrays can search for cascades in huge volumes of ocean, of the order of $0.1~km^3$ at energies in the $PeV$ range. We used a mere 17 hours of data to set the best limits (for a while) on high energy neutrinos from AGNs employing a handful of DUMAND optical modules. At the same time such an array would have an effective area for muons of hundreds of square meters. The effective voilume for muons would be perhaps megatons, for a source with a spectrum similar to the cosmic rays.

On the other hand, should high energy events be seen which are not consistent with atmospheric origin (basically any energy deposition beyond 100 TeV), a small array will have a hard time to reconstruct a direction. This is because the direction of the cascade is not well constrained by a few samples of the light intensity versus direction. There is also a strong corellation between vertex resolution and energy (misreconstructing a wavefront as more flat puts the vertex further away, and the energy estimate substantially higher). Consequently reconstruction of cascades external to the detection array is not promising. One can make explorations, but long term progress demands huge arrays enfolding the search volume. Also, as we discuss below, this tradeoff between cascade and muon detection depends upon the medium (optical scattering).

High Energy Physics With Natural Neutrinos

When we started in this business it was still interesting to measure kinematics of neutrino interactions, since done thoroughly at acclerator experiments. Hence we considered arrays where we might be able to measure, say, both the outgoing muon and the energy and angle of the target-quark induced cascade. In principle one can detect thus, in the case of muon neutrino charged current intereactions, the neutrino energy and direction, and the fraction of the energy transferred to the quark (the ``y distribution''). This is only possible for charged current interactions, and only when the charged lepton can be distinguished from the pion cascade from the target nucleon. While the measurement of the weak interaction kinematics and the nucleon quark momentum distributions are no longer interesting (in this context), the measurement of the neutrino total energy and direction, and energy transfer, remains worthwhile, particularly for oscillation studies, as discussed below.

For neutrino astronomy purposes, observing the whole interaction increases the angular resolution, potentially allowing one to beat the angular scattering, which is roughly $1.5 \deg / \sqrt{E_\nu/TeV}$. Muons in a $km^3$ array will only be contained up to an energy of $\sim 200~GeV$, so such attempts to improve angular resolution for neutrino astronomy below a $TeV$ are worth, though as usual expensive ({\it eg.} more photomultipliers).

If it is possible to measure the average energy transfer fraction ($$) of events of a given energy region from a given source, one can discriminate whether the source region is dominantly matter or anti-matter (or symmetric) for some strange kinds of sources. To this end one needs to observe both the recoil energy from the struck quark (hadronic cascade) and the outgoing charged lepton. [Learned and Stecker, 78?]. Given a volumetric array of $km$ dimensions and substantial AGN neutrino fluxes, this could become a viable measurement, and gives a unique opportunty to probe the universe.

The question of measuring the total neutrino-nucleon crossection versus energy remains an interesting subject, particularly at extremely high energies. However, at energies up to about 1 PeV, measurements at DESY can be used to reliably calculate the crossection (barring the opening with energy of any new channel unique to neutrinos). The experimental challenge is simply that one measures a rate of events versus energy, and that rate is the product of cross section and flux. When $PeV$ neutrinos are detected, one may employ the varying attenuation of the earth as it rotates relative to the source (or ensemable of sources) to extract the crossection. Here the earth attenuation is the product of cross section and column density of the earth (see below for discussion of earth tomography). This measurement will not be soon made, but remains worth watching, as no terrestrial sources of neutrinos of more than a few TeV are on the horizon.

Neutrino Oscillations

We have long dreamed of observing neutrino oscillations, in the ocean with natural neutrinos. This has been spectacularly carried out now by the Super-Kamiokande experiment [SuperK 98], employing the atmospheric neutrinos in the energy range around $1~GeV$ with a $50~kilton$ tank in a zinc mine. The characteristic oscillation length for muon neutrinos in the $1~GeV$ range appears to be about $250~km$, but at the time of this writing is not well constrained, being ucertain by perhaps as much as an orede of magnitude. Thus one may wonder if a larger neutrino detector also employing the reliable and farily well known atmospheric neutrino beam, which has built-in distance variation from $\sim 10 ~ 10,000~km$ could tighten up the situation. This would seem to be the case, particularly if the mass difference turns out to be in the range of $<10^{-3}~eV^2$, where acclerator based studies lose sensitivity (due to limited range). However, none of the detectors presently being pursued seem to be able to fill this need, because of their design being aimed towards high energies and thus sacrificing the resolution needed for this task. A possible new detector of $10^6~tons$ might suffice [Learned 98]. Another possiblity is desribed below.

Detecting Tau Neutrinos

The actual direct detection of tau neutrinos has proved elusive for 20 years now. Their existence has long been presumed, not just from symmetry, but from the kinematics of tau decay, and the neutrino flavor counting via observation of the $Z^o$ decay width. Yet no tau neutrino interactions have yet been discerned. Part of the problem is that sources of tau neutrinos are not forthcoming, so there is no known way to make a tau neutrino beam. Secondly taus themselves are not easy to pick out as they mostly decay (with lifetime $2.9 \times 10^{-13}~sec$ into many pions and there is always missing energy/momentum. Thirdly, the tau cross section comes up slowly (due to the tau mass of $1.777~GeV$), so few tau neutrino interactions are expected below about $20~GeV$.

A spectacular observation of taus could come about from seeing few $PeV$ tau production, flight over substantial distance ($\simeq 100~m$) and subsequent decay [Learned 95]. The second (decay) burst would be typically twice as bright as the first ($\sim 10^{11}$ photons), and both would be detectable at distances of hundreds of meters in the ocean. The source of such high energy neutrinos that seems most promising at the moment is from AGNs [HENA 92]. However, in all seemingly likely astrophysical circumstances the most likely neutrino flavor mix is in the range from mostly muon neutrinos to a ratio of 2/1 of muon neutrinos to electron neutrinos, with negligible tau neutrino content at the source. Thus substantial tau nuetrino content would have to arise from oscillations, which are in equilibrium for any astrophysical distance and such energy, given the parameters we see in atmospheric neutrino disappearance [Superk 98]. In the referenced paper [Learned 95], there is a discussion of a hypothetical scenario where tau neutrinos are observed as well as the resonant interactions of $\nu_e + e^- \rightarrow W^-$ at $6.4~PeV$, thus measuring the electron neutrino fraction of the putative astrophysical neutrino beam. Interesting conclusions for particle physics and astrophysics can be drawn from suh masurements. The detector required would have to be of the order of $>1~km^3$ and have lattice spacing of detection elements no less than $\simeq 100~m$. Such measurements at once observe the tau neutrino and measure the mixing angles (assuming a simple three neutrino mixing scheme).

Detection of Relic Big Bang Neutrinos

It may seem peculiar to mention the lowest energy expected for neutrinos, those left over from the Big Bang with thermal energies at present. Despite much cogitation, physicist have not come up with any viable proposed means for direct detection of these relic neutrinos. The only proposal which might work has been on the table for some years, from Tom Weiler [Weiler 82, 84, 92], involving the resonant absorption of UHE neutrinos on target relic neutrinos. One would observe a dip in the neutrino spectrum coming from distant sources. The last report on this process [Weiler 92] was not very encouraging owing to the small neutrino mass limits (and thus higher neutrino beam energies) and limits from the Fly's Eye experiment. This is another example of physics directly stimulated by DUMAND.

Earth Tomography with Neutrinos

In the early DUMAND Workshop's it was recognized [Bradner and Learned, cannot find ref] that one might employ VHE neutrinos to ``neutrino-ray'' the earth. The idea is simply that if one has an external beam of sufficient energy that the flux is measureably absorbed in passing through the earth, one can observed the time variation of the interaction rate to study the variation in column density. The expected attenuation length for muon neutrinos is one earth diameter at $2~PeV$ (?check?). Given sources at various declinations, a single detector can make a density profile assuming a spherically symmetric earth. Given three detectors, one can do earth tomography, at least in principle. The main interest in carrying these observations with neutrinos would be to examine the earth's core, about which there is little knowledge. [ref to Chaincy Kuo LBL paper here].

New Particle Searches

Over the years we have studied the ability of large Cherenkov instruments for the detection of various kinds of hypothetical particles, typically those invented to solve the ongoing mytery of the nature of Dark Matter. The Baikal group has actively worked in this area, and we simply make reference to their limits on magnetic monopoles which interact with the Rubakov process (instigating nucleon decay as it traverses the array), searches for quark nuggets and the like. The basic signature is one of two types: a particle which traverses the array at a $\beta \simeq 10^{-3 \div -4}$ and leaving a glowing trail, and a particle which makes a series of rapid flashes along a similar track. The velocities are characteristic of galactic objects. In the first case the object might be any of a number of hypothetical particles, such as quark nuggets, which heat the medium locally, radiating thermally along the path. In the second case, the particle may be inducing interactions as it passes, or sequentially interacting, but the rardiation is prompt. The difference experimentally is that in one case the array sees a light which may last for many times the light crossing time of the array, while in the other the particle would be detected as many independent events.

Atmospheric Neutrinos

It was recognized that calculations were needed for the atmospheric neutrinos over a large range of energies, from MeV's to 100 TeV or so. A great deal of work had been done in this area after the peculiar observations made by the Utah group in the Park City mine in the late 1960's. These observations, late shown to be in error, seemed to indicate evidence for the W boson, and set off a number of searches and calculations. DUMAND stimulated the revisitation of the calculations about a decade later, and advanced the work both for atmospheric neutrino fluxes which could take advantage of the newly made cosmic ray muon measurements both underground and in spectrometers at the earth's surface to calibrate neutrino flux expectations in the few to hundred GeV range. The next round of atmospheric neutrino flux calcualtions got started in the early 1980's in response to the need for reliable background claculations for nucleon decay searches.

Cosmic Ray Muons

Similarly, DUMAND pushed the calculations of muon fluxes underground. Important contributions to this were the calculations made of the various energy loss processes for muons at high energies, bremmstrahlung, pair produciton and photonuclear interactions. These have not been much improved upon over the years.

Also the DUMAND group stimulated people to make serious calculations fo the Landau-Pomeranchuk-Migdal effect, that can be thought of as the result of the high energy particle seeing the medium compressed until some processes such as Bremstrahlung are being interfered with by neighboring atoms. Multiple calculations of this (difficult to compute) phenomena were reported at the [DUMAND 1979]. It is only in recent years that the effect has actually been studied at SLAC [ref to LPM paper].


A number of things were recognized in the early DUMAND Workshops relating to the choice of location, detector and array design, which we will summarize here. In fact most of these judgements have stood the test of time and remain reasonable today.

Location Choice

One of the first questions was of course the location. People had the idea of using the almost limitless and free ocean depths since early days [ref to Reines, Markov and Greisen, dating from the 60's]. However, that assumption was reopened regularly at the workshops.

Ocean, Lakes, Ice and Mines

Lakes were heavily investigated early on as possible venues for a large nuetrino detector. It was eventually realized that the deep lakes in North America were not adequate, at least for an open array. However, the deepest lake in the world, Baikal continues to be the host to such an experiment. In the late 80's the Japanese explored making an instrument (dubbed LENA) in a lake in Japan, but the prototype was destroyed in a storm and the project abandoned. In the US, a project called GRANDE was proposed for installation in a quarry in Arkanses. This device would have attempted to do neutrino astronomy with upcoming muons. Meanwhile, being at the earth's surface it could also use uplooking phototubes to study Extensive Air Showers, in particular attempting to distinguish those initiated by gamma rays in the TeV energy range. This was largely promoted a spinoff group of DUMAND collaborators who were frustrated by the slow progress and daunting techical challenges in the ocean. That project failed to get funding, as also did several similar projects in Europe, Russia and Australia. The AMANDA Project now being brought to life near Los Alamos, for the study of gamma rays, is in fact a daughter to the GRANDE Project. There are multiple ties to the DUMAND effort, in senior researchers, students, detector and calibration techniques, and software.

The early DUMAND workshops concluded that neutrino astronomy would require detectors with a great deal of overburden for shielding (see below), and indeed took about a kilometer of water overhead as a minimum necessary overburden. The choice of venues was then restricted to deep mines or water. Mines were eliminated on the basis of cost and the impracticality of making cavities larger than sone tens of meters, which would collapse. Deep ice was not considered in these early workshops, becasue nobody thought that the ice would be clear enough and the access and infrastructure issues were too daunting. That was later shown to have been an oversite, witness the AMANDA project making great progress at the South Pole.


The depth issue has generated continuing debate until the present. It was the early conclusion of the DUMAND group that work in the ocean requried not less than 2 km depth to avoid biological activities, and that deeper was better, with the depth of the abyssal plane of 5km as initial goal. Lessor depths are clearly tolerable, but at some cost. Of course near arrays located at teh earth's surface must be enclosed to keep out ambient light. This was one of the heavy tolls on the GRANDE style array plans: the combination of cost of bag, water filtration system, and civil engineering made them very expensive (but of course much more accessable).

There is no clear break in depth in terms of background. One gains slowly with depth (as the logarithm) in being able to employ more solid angle in which neutrino generated muons dominate those from the surface. A flat overburden is superior to a mountain, and water to rock for the ability to know the overburden precisely, and thus manage backgrounds and use the solid angle most effectively.


For underearth arrays, an equatorial location is desirable since such a detector locally viewing neutrino arrival directions out to the horizon sacne the shole sky once per day. A polar location is of course interesting too, as there one sees one half the sky continuously, with sources moving around at constant elevations. With the galactic center as a prime candidate for neutrino emission, a detector in the Northern hemisphere is favored. One the other hand, many of the best studied astronomical objects are in the Horthern sky, favoring a Southern hemisphere location for a neutrino telesope. In sum we jusdged, and it seem true today that if one had only one choice for location, a near equatorial site would be optimal.

Proximity to Shore

During early DUMAND Workshops it was soon realized that power supply demanded cable connection to shore. The power requriements were too great for batteries, nuclear (SNAP) reactors were not allowable by treaty, and there was no practical means of power generetion in the deep ocean. Thus it was found, and has been confirmed repeatedly as the question has been reopened, that connection to shore for power is unavoidable. Perhaps this will change with sufficiantly low power electronics. It is obvious that fiber optics are the means of choice for data tranmission. This is an area of tremendous evolution of the technology since early DUMAND Workshops, at which time coaxial cable was the standard.

The issue of cost, reliability and complexity of long cable runs caused the DUMAND group to discount continental locations, where the continental shelf drives one out several hundred miles typically in order to find great ocean depths.

Bottom Character, Flatness

It is obvious that one would prefer a simple flat bottom in the ocean, one with sufficient bearing strength so that equipment will not sink into the bottom ooze. This turns out not to be hard to satisfy, being met by a large fration of the world's surface (the deep ocean abyssal plains). However, some deep locations considered in deep trenches should be avoided, due to steep slopes and the not infrequent ocean equivalent of an avalanche.


One hardly need mention that the availability of locally friendly territory, with available supplies, communications, transportation and kindred facilities is strongly desireable. That the Baikal group and the AMANDA groups are progressing, points out that these are desirable characteristics not mandatory.

Clarity: Absorption and Scattering

Optics of the natural waters are clearly an important issue. One happy surprise of the DUMAND adventure, was finding that the deep ocean waters were in fact about four times (absorption length $>40~m$ instead of $\simeq 10~m$) clearer than at first thought. (The amazingly long absorption lengths in the polar ice were a similar boon.) The effective scattering length in the ocean remains long compared to absorption, as had been predicted. (The surprise at the South Pole was that the scattering, with effective length around $25~m$, dominates the absorption in that environment). The relative dominance of scattering or absoprtion has a strong influsence on array deisgn and ultimately physics. For the ocean case it means that huge cascades may be observed at long distances, and with sufficient density of photomultipliers the cascade direction may be reconstructed.

Ocean Array Design Considerations

Some Design Principles

Many design principles for large neutrino detectors to be installed in the deep ocean were realized in workshops and over the years. We summarize some of those below, as they do not appear elsewhere.

Open Array for $km^3$

Calculations of anchoring requirements indicated that a closed array was impractical at the scale of a $km^3$ array. At the scales of tens os meter dimensions mines with water tanks are the location of choice. In the intermediate size scale, a closed array may be practical in the ocean, but expensive. The attraction, of course, is the avoidance of the predominantly $K40$ induced Chernkov light in the ocean, amounting to around $200~quanta/cm^2/sec$ in the open ocean. Large arrays simply have to pay the nearly factor of two penalty in photocathode area to live with the $K40$ background.

Bottom Tethered Strings

Many studies of structures for arrays were carried out, and are well documented in early DUMAND Workshops. Economics of material and construction costs have always lead back to the simplest option of fabricating the array of elements consisting of what the oceanographeers call buoys, simple anchored strings with flotation at the top. It turns out that the typical deep ocean currents (0- few $cm/sec$) do not much perturb such arrays, and the lack of small scale rotary current motions (gyres) prevent tangling. Given the studies of the DUMAND Hawaii site, it would be practical to place stroings as close as $10~m$ to each other (placement being the more difficult issue). The fact of motion with the inevitable tidal drifts, with time salces of hours in changes, permitted a plan to employ acoustic monitoring of the array position (see below).

Reliability and Connectivity

An ongoing debate has been focussed around the question of reliability and the topology of the array connections. The desire to avoid vulnerability to single point failure of significant fractions of the array leads to the idea that one should connect each module independently to shore. While this is impractical, the tremendous progress in fiber optics has made the cost of individual fiber optic links similar to the cost of optical modules. Future arrays will doubtless have multiple paths to shore, with backup paths. In any case one sould focus upon keeping the expensive and vulnerable hardware at the ends of the data connection chain, where failure will remove a minimum element of the system.

Send the Data to Shore for Processing

Another debate which has gone on over the years is over the degree of processing to carry out in the ocean. This has changed with time. At first it was assumed that triggering and filtering would have to be done at the site of the array, and that is what is done in Baikal, at least until the present. With the advent of fast links the DUMAND II group sought to send all data to shore for triggering and filtering in the laboratory instead of the ocean. Now with the decreased cost and power of electronics it is possible to empoy highly reliable and redundant electronics on site, beating the data rate down to a very small level. Still it will always be easire to repair and modify the systems on shore.

Repair, Replace and Reconfigure

It is absolutely necessary to design an array which is as simple as possible to repair and replace. Particularly in the deep ocean, where one typically works in difficult circumstances, quite often not being able to know exactly what is taking place.

This is a least partly why much of oceanographic technology is so primitive, as oceanographsers have typically launched instruments and when they did not return, as happened frequently, they knew not why. There is a long standing saying amongst oceanographers that one should `say goodbye' to an instrument when it is launched into the sea, as there is a fair chance one will not see it again. This has often lead to a strategy of making multiple instruments rather than identifiyiung and remedying existing hardware problems. A good deal of effort was spent in the DUMAND project trying to reverse this trend.

The deep ocean also offers the opportunity to reconfigure an array if one does not make premanent bottom moorings. (This is in contrast to the situation in ice where once placed, modules are irretrievable.) The use of buoyed strings of detectors makes this reconfiguration possible with fairly simple design allowances.

Use ROVs

Another long standing debate has been over the commitment to employ submarines or remotely operated vehciles (ROVs). While the NESTOR group has chosen to stick with array designs not requireing ROV's the DUMAND group chose that route. The reason is simply that if one is not going to make ocean bottom connections, then the whole array has to be installed of a piece, a frightening thought at the level of a $km^3$ array, though perhaps doable with initial arrays (as in NESTOR). ROVs are ever more available, being driven in development for the deep ocean at this time by the increasing depth of oil drilling operations.


Much effort was put into studies of triggering for large arrays. In teh end though, it seems that simple area triggers, such as the number of nearest neighbor trigger clusters in some woider array trigger window, works as well as anything and are simple to implement. Given the relatively low rate of real local physics events (muons) as compared with EAS arrays, one is not driven to the necissty of the local storeage and neighbor polling schemes as are employed in EAS arrays.

Array Shape and Spacing

Many computer studies have shown that the array shape is not very critical to performance, within large bounds. Back of the envelope calculations tell one of the maximum scale of detector spacing, being of the order of the optical attenuation length in the medium, if one wants to detect very muon travelling through a region. A figure of merit which has been multiply verified is that one DUMAND optical module ($16~inch$ diameter PMT) gets about $100~m^2$ effective are in an array in the ocean. So, one should not put the modules much closer than $10~m$ on average. With spacing much larger than that one may miss traversing muons, but spreading the detectors out gets one more leverage and better angular resolution.

Another ongoing debate has been over placing modules on the outer surface or throughout the volume. It has turned out in every study we have conducted that while a furface disposition makes sense for muons, filling in the volume costs little, and gets one the ability to detect and reconstruct cascades.

Thus one is lead to the simplest plan of placing strings of optical modules with spacings of $10-30~m$ floating up from the ocean bottom. While simplicity in calculation might tempt one to regular arrays, the breaking of degeneracies in track finding algorithms favors irregularities (which are naturally going to occur during placement at the level of a few meters).

Deployment Options

The method of deploying an array depends critically upon whether one demands preconnnection of the array or not. The DUMNAND-II group chose to employ ROVs, so the deployment was greatly simplified from earlier plans. To our view this is the only sensible option for large arrays. Placement of individual strings may proceed in various ways (we refer the interested reader to the several DUMAND deployment workshops). It seems very desireable to have strongs able to be released individually to float to the surface for recovery (which may then be accomplished by a relatively small and inexpensive vessel, without the cost of a full expedition).

Critical Items for Future Development

One of the unanticipated general difficulties with the DUMAND Project was that many areas of ocean, particularly deep ocean, engineering was (and still is) far less developed than one might wish or expect. There is nothing like the space grade of hardware and electronic qualification. There are Navy standards, but they tend to be quite out of date technologically and generally inapplicable to ocean high technology research. We found that we had to do engineering on such seemingly bread-and-butter items as mechanical cable (strength) termination methods (amongst a depressingly long list of such items). The situation is improving slowly, as oil exploration moves to greater depths and requries higher technology and remote operations. But future activities should not underestimate the difficulties which will be encountered.

We would like to highlight several items we feel should be pursued in studies for a future $KM^3$ deep ocean array. There are the obvious issues such as reliability, service, data processing, improvements in optical sensors (see below), and so on. Near the top of the `obvious' list of technology studies would go pressure tolerant connectors, the single largest source of difficulties in the DUMAND Project experience. In this regard, a requirement for local infrastructure to the assembly laboratory should be easy access to pressure and temperature test factilities (simultaneous), large enough to encompass all major pressure housings.

One area we could not afford to engage in DUMAND was the study of pressure tolerant electronics. The best way to beat the problem of leaks in housings and connectors is to employ oil filled housings. We did this successfully with power transformers, connectors and some cables (inside hoses). However, the central electronics housing, at some level almost surely a compenent of any deep ocean array is a candidate for oil filling. Solid state components are naturally pressure insentitive, and some US NAvy studies have shown that much off-the-shelf solid state electronics can be used with no problem under oil at high pressure, as long as there are no voids in the cases, and as long as pressure was applied and released slowly (minutes). Of course some items are going to present challenges, such as changing capaicatances and this is sure to make some effort for high speed circuits (clocks in particular). A modest engineering effort seems worth undertaking here.

Another area worth pursuing is the design of a deep ocean resident robot. The ability to carry out service whenever it is needed without having to mount an expedition is very attractive. (Such a device may also be useful for calibration purposes.) Oceanographic research ships cost typically \$ 10-20 K/day at present, and submersibles may double that. Commercial costs would be even greater. Hence with transit and port time, sailing to and from the site, expedition preparation costs, and so on, a single operation at the array with an ROV will cost considerably more than \$ 100 K. It will not take many trips to amortize the cost of a small ocean bottom resident robot.

Optical Sensor Studies

The DUMAND optical sensor has become somewhat of a standard, with variations. Early on we considered developing new types of photomultipliers which might be their own pressure housings. This was dropped as it was learned that such housings tend to spall internally, which could lead to the shedding to photocathode (but perhaps this is woirth revisiting with improved technology). Moreover the Benthos 16 inch housing was a well developed standard of the industry, and our initial concerns about housing dimensional creep with time were allayed. All the large Cherenkov detectors (DUMAND, Baikal, AMANDA, NESTOR and ANTARES) employ large PMTs in spherical glass pressure housings. Additional cylindrical sections to the pressure housings (which extend the housing and make for easy insertion of connectors) were explored by DUMAND but not chosen because of lack of deep ocean experience (and some discouraging test made in our pressure tank), but have been employed in AMANDA and Baikal (under less deamnding conditions). There are tricks in making the seals for glass pressure housings, the major rule being to use flat ground surfaces without any lubrication. Pressure joining between materials with differing temperature and pressure reponse can lead to disasters (spalling followed by collapse).

During the early days of DUMAND many options for optical sensors were explored, which we will not review here as they are well documented in the proceedings. We might pass on one general lesson, which is that despite substantial efforts towards developing light collectors via wavelength shifting methods, the resulting devices were impractical for the ocean. Perhaps one might employ wavelength shifters to gain light collection area on the order of a factor of two, as was done with the IMB detector (using a rectangular wavelength shifting plate in the equatorial plane of the PMT). However, in the ocean wher one would like gains for the most distant sources, which have already lost all light below $300~nm$, there is probably not much to be gained (thought the opposite may be the case for deep ice).

Various alternative configurations of photomultipliers have been considered, such as cylindrical devices. While these might have interesting properties, the development time and expense have so far proved prohibitive. The use of solid state light sensors has not been even close to practical, either economically or in terms of background noise level. Spectral response was a problem in the past, but recent improvements in astronomical detectors have made high quantum efficiency available in the UV range. The most profitable area for exploration would seem to be in the possible use of liquid photocathodes.

All that said, it continues to seem unlikely that one can beat the use of traditional photomultipliers. The new hybrid photomultipliers, which employ a solid state electron detector (with potentially many pixels) inside the vaccum envelope of an image intensifier may offer new opportunities, though they are more suited to image resolution than the simple demand for light collection with modest time resolution ($ns$) of deep ocean high energy arrays of the DUMAND style.

Acoustic Detection

Alternative to Cherenkov detection have been explored by DUMAND and other groups over the years. Radio detection has been pursued by the Russian group at eh South Pole over the years, and there is new interest. The DUMAND group made a significant atempt in the late 1970's to design an array employing acoustic detection of neutrinos. The mechanism is simply that the rapid thermal expansion of water after the heating caused by a the ionization from particle cascade leads to an outgoing tiny bipolar pressure wave. The group measured these pulses in the laboratory and in a proton beam at Brookhaven [acoustic refs]. The conclusion was that there was no unexpected new mechanism ({\it eg.} microbubble formation) which might have been more efficient at energy transfer (which is of the order of $10^{-9}$ for the thermacoustic mechanism). Calculations led to an energy threshold in the ocean for neutrinos of about $10^{16}~eV$. As there are no guaranteed sources of neutrinos (or any other particle) at those energies in the ocean it was not practical to propose such an experiment, which might have any signals at all. The DUMAND stragetgy was therefore to build a Cherenkov detector and piggy-back an acoustic detector, otherwise employed for surveying, to make a search for UHE neutrinos in multi $km^3$ volumes. The NESTOR group is keeping this approach alive.


Hawaii Site characterization

A location near the Island of Hawaii was selected as a site for DUMAND for a variety of compelling reasons: exceptional water clarity [DIR-5-86], proximity of an abyssal plain with appropriate seabed characteristics [DIR-14-83] to a suitable shore site (30 km away), presence of active particle physics and astrophysics groups at the nearby University of Hawaii in Honolulu, and pre-existing laboratory infrastructure at the shore site, due to an ocean thermal energy research project. The latter feature even provided a cost-free conduit for the DUMAND shore cable to pass through the surf zone, since the thermal energy project involved slant drilling of tunnels into the ocean. The array site is on the ocean floor at depth 4800m, 30 km due west from the Kona Coast of the Island of Hawaii (Fig. 2), connected to the shore laboratory at Keahole Point by a cable combining electrical and fiber optic elements, terminated in an underwater junction box (Fig. 3). The shore cable contains 12 fibers (including spares) and a copper layer which supplies 5 kW of electrical power at 350 VDC, using a seawater return system [Peterson 92]. The underwater site places no inherent limitation on possibilities for future expansion of the detector. With all 9 strings in place, DUMAND would have an effective detection area of 20,000 m2.

During early 1992, an array of current meters was deployed near the center of the DUMAND site for the purpose of obtaining a measurement of bottom currents over a many-month time baseline. Such studies had not been previously performed over the longer term, and the information was deemed to be useful for string deployment, since bottom currents can dominate the placement errors of the strings. The array had three vector-averaging current meters deployed at three positions along a 400 m high mooring, one near the bottom, one at the top and one near the position which corresponds to the string controller in the middle portion of the string. The meters were developed and the data collected and analyzed by Dr. Pierre Flament of the UH Oceanography Department.

The maximum current speed seen over the 5 month baseline was about 14 cm/sec, and currents were typically about 3 cm/sec or less (Fig. 4). The predominant components of the currents are tidal, with a NNE-SSW flow pattern superposed on the normal diurnal tidal flow. Occasionally this superposed flow changes to an ENE-WSW direction. The tidal behavior of the currents [Flament 92] is such that a power spectrum of the time series of the current samples projected to the N-S direction clearly shows strong components at the harmonics and subharmonics of the solar day (lunar tides tend to be suppressed relative to the solar tides in equatorial regions).

Prior to 1992, no bottom cores were successfully recovered from the DUMAND site, and a number of oceanographers expressed concern that the site might have characteristics that indicated periodic turbidity flows. Such flows, which could originate due to the proximity of the steep Hawaii Island slope, could damage or destroy the array (they have been known to break undersea phone cables); thus evidence for recent or frequent turbidity flows could impact the site selection. Turbidity flows leave sediment layers called turbidites which are usually quite apparent in a bottom core sample. The coring operations were performed under the supervision of Dr. Craig Smith of the UH Oceanography Department using a multi-coring device that his group had developed. A number of core sets were recovered from very near the nominal site center, using a ship of opportunity, the Russian research vessel Yuhzmohrgeologiya. Shipboard analysis of the core profiles indicated that, although a turbidite layer was present, the depth of the layer indicated that the turbidity flow associated with it happened of order 10,000 years ago. The bottom was also found to be relatively solid sand below a layer of soft mud about 10 cm deep. [DIR-14-92] The fact that this mud, which lies on the bottom in near colloidal suspension, was present and relatively thick was also a good indication that large transport storms were not frequent in the area.

The 1992 cruise of the R/V Moana Wave at the DUMAND site permitted a number of deployment and site-related studies to be performed [DIR-12-92]. The acoustical locating system planned for the environmental monitor system was tested, and the acoustical properties of the waters around the site were measured [DIR-13-92].

ROV operations

DUMAND deployment includes string interconnection operations, which require a manned deep submergence vehicle (DSV) or robot remotely-operated vehicle (ROV). Despite the fact that operations such as those required by DUMAND at depths of 4700 m are technically feasible and not complicated, such operations are rarely performed, since most similar operations are done within continental shelf waters at depths less than 1 km. At DUMAND depths, such vehicles are working at pressures 80% of their rated limit, compared to 15% in continental shelf operations. Thus deep operations must be tested before they can be approached with confidence.

Such a test was performed as part of a program sponsored by the US Navy Deep Submergence Unit (part of the Submarine Development Group 1 in San Diego, California) in October of 1992. The program solicited proposals for civilian research in areas close to the Hawaiian Islands, and the proposals competed for a total of 40 dive days during the September/October 1993 time frame. DUMAND was selected among these proposals on the basis of its test of forefront deep ocean technology and was awarded 4 dive days. These dives took place over the period 29 October-1 November 1992 at the DUMAND site near Hawaii Island. The operations consisted of two dives with the submarine Sea Cliff for the purpose of performing a video survey of the array site. The Sea Cliff manipulators were not functional during these dives. In addition, a test junction box was stationed on the bottom just north of the planned site, and the Advanced Tethered Vehicle (ATV), a state-of-the-art robotic vehicle, was used to perform umbilical interconnection tests with actual prototype electro-optic connectors and electro-optic umbilical cable.

The site survey and tests were entirely successful: the site was shown to be free of any debris that might complicate the placement of the strings or junction box, and the mechanical interconnection operations went smoothly and validated the basic connector design. It is worth emphasizing that even apparently simple mechanical operations can suffer unexpected failure under the ambient pressure of the deep ocean. The ambient pressure at the DUMAND site is nearly 4 tons per square inch, of order the same pressure that structural engineers must deal with in designing mechanical support systems for skyscrapers. Although the design of a simple latching connector is straightforward in principle, in practice it may fail to mechanically couple due to changes in tolerance or material deformation due to such ambient pressures. Tests of such systems are difficult and costly to perform, even in the ocean. Thus this test of the design is of great importance to the project, since it would be very difficult to recover from a problem with the connector discovered during an actual mating operation, and would damage the credibility of the system. In fact, an example of such problems was provided by the DSV Sea Cliff during this dive series, since the state-of-the-art Schilling manipulators on the submarine were found to fail at depths greater than 10,000 ft, and thus could not be used for our operations.

In addition to tests of the mechanical integrity of the connector design, the placement of the test junction box and other activities conducted at the actual site produced important input into the final design of the base platforms on which both the junction box and string ballast would rest. Thus we were able to more accurately predict the degree to which the junction box might be partially submerged in the mud, and estimate the drag force that might be expected from trailing the umbilical cables across the sea floor during the interconnect operations. The fact that no obstructions such as rocks or other debris were found at the site also relieved some of the design constraints for these systems.

In June, 1995, the US Navy's Advanced Tethered Vehicle (ATV) revisited the DUMAND junction box and successfully performed connecting operations, proving that tethered remotely operated vehicles (ROVs, which are cheaper and more readily available than manned submersibles) are also an option for DUMAND underwater maintenance activities. The ATV removed connectors which were believed to have failed, and replaced them with protective dummy plugs. This eliminateded a short circuit observed on the shore cable power conductor, but the short reappeared after a few hours. It is now believed that the remaining power line fault is due to contaminated contacts on the dummy plugs. However, due to the extremely limited time window available for ATV operations, it was not possible to complete diagnosis and repair of the junction box.

Optical Modules

DUMAND Optical Modules (OMs) come in two models, the Japanese OM (JOM) and the European OM (EOM). The JOM was based on the Hamamatsu 15” hemispherical tube designed to fit within a standard 16” Benthos Sphere (Fig. 5) [DIR-11-83][DIR-14-90][DIR- 4-91][DUMAND-2-93]. The EOM uses a Phillips PMT [HDC-14-90][DUMAND-3- 93][DIR-5-93][DIR-4-94][DIR-6-94], employing a novel design in which photoelectrons emitted from the hemispherical cathode are focussed onto a small scintillator attached to an internal structure comprising a conventional 10-stage PM lattice.

Signals from the PMTs are pre-processed within the optical modules, on a custom PC board surrounding the PMT neck (“ring electronics”)[DIR-20-90][DIR-21-90][DUMAND Optical 90], providing standard pulses which encode time of arrival (to 1.25 ns accuracy), pulse area, and time-over-threshold (TOT), a measure of pulse duration (Fig. 6). The ring boards also carry DC converters which boost 48V from the junction box to the HV levels required by the PMT, as well as housekeeping functions.

Implosion tests

A different set of experiments were conducted over the course of two cruises in 1992, aimed at determining the immunity of the array to the damage caused by catastrophic implosion of the glass pressure housings used through much of the detector, both to house optical modules and for floatation. Such implosions are rare enough that statistics cannot be reliably estimated for the planned life of the array. The kinetic energy generated in such an event is of order 1 MJ, equivalent to the detonation of a few kg of TNT. The major concerns were twofold: first, would such events sever the string riser cables, thus disabling the string; and second, could the shock wave generated destroy nearby optical modules or disable the acoustics?

With the help of Benthos Corporation, which donated pre-weakened glass pressure housings which were set to implode at depths of order 4.5 km, a total of 9 different experiments were performed. As a result of these experiments, the design of the riser cables was substantially improved with the addition of titanium "picture frame" supports around each optical module (Fig. 5). This addition was found to be adequate to protect the optical, electrical, and tensile riser cables from any significant damage during an implosion of a module adjacent to them, even though the implosions in some cases actually partially damaged the high-tensile-strength titanium frame (Fig. 8).

As described, in September 1992, a current meter string was recovered from the DUMAND site after 6 months of data logging. To our surprise, a number of its 17" float clusters showed implosion damage, and the damage had apparently destroyed the bottom current meter; this observation re-inforced the importance of the previous implosion studies, and care was taken to determine the cause of the implosions, which was traced to incorrect handling of the spheres in question. Another important lesson concerning sympathetic implosions was also learned: spheres within 1 m of another sphere that has imploded are likely to suffer induced implosions themselves. Thus we performed some further experiments to determine the minimum safe distance for adjacent spheres [DUMAND-8-93] Two meters or more was found to be safe for floats.

Optical feedthrough

Although Benthos provides electrical feedthroughs for their spherical glass housings, the stock electrical penetrators were found to be unreliable and were updated after DUMAND testing suggested a design change. No suitable optical fiber feedthrough existed. Working with engineers from the Hawaii DUMAND group, the Swiss firm Diamond SA developed a suitable feedthrough, allowing conversion of fast data from electrical to optical signals within the OMs and thus permitting noise and dispersion free transmission from the OMs to the SC.

Sea return system tests

Two major in situ tests of the sea-return electrodes have been conducted: The first took place in August 1991 aboard the RV Alpha Helix. During the period August 20-26, a series of 32 different experiments were conducted to determine the behavior of various electrode configurations, using depths of up to 4.3 km, a few hundred meters less than the actual operating depth for the DUMAND array. The various electrode configurations show some differences in intrinsic resistance, but the time variation of the resistance remains in the tens of milliohm range without any systematic trends that suggest long- term increases[DIR-8-92]. This behavior was later confirmed over a duration of approximately three months using electrodes prepared for use with the DUMAND junction box as actual return electrodes. The experiment was performed in the summer of 1993 in the Mediterranean Sea between two shore-based sites. The results demonstrated that the long-term resistance of such electrodes remains low. In fact, for the polarity being used in the DUMAND array, iron electrodes can be used in the deep ocean if the array is operated at a reasonably high duty cycle, since the deep sea electrode is effectively galvanically protected by the current that it passes into the ocean. However, the advantage of using iron is only a fraction of an ohm relative to a nobler metal such as titanium; thus the corrosion resistant material is preferable since it will not suffer during the downtime of the array.

ASIC digitizer chip (27 channel/ 1.25 nsec TDC)

Data from the 24 OMs on each string are digitized and serialized in the string controller module by a custom 27 channel (including spares and housekeeping) monolithic GaAs TDC/buffer/multiplexer chip which operates with 1.25 nsec timing precision and 2-level internal buffers. Development of this custom designed ASIC (Application Specific Integrated Circuit) represented a major achievement by the Boston University group[DIR-2- 95][DUMAND-6-93]. The data stream is sent to shore via optical fibers (one per string) at 0.5 GHz. A separate optical fiber carries environmental and acoustical ranging information which are used to measure the geometry of the array.

Chirped sonar system

To do useful astrophysics, it is necessary to point reconstructed muon tracks onto the celestial sphere with an accuracy better than 1o (the median angle between primary ? and secondary ? at 1 TeV). This means that relative OM locations must be known to the order of a few cm, and the overall geographical orientation of the array must be known to much better than 1 degree. The Global Positioning Satellite (GPS) system plus conventional oceanographic acoustical survey techniques allow us to measure the geographical coordinates of underwater fiducials (acoustical transponders) to within a few meters, satisfying the geographical orientation requirement. We were unable to find a commercial system able to reliably provide the OM positioning accuracies required, so we developed our own sonar system[HDC-10- 91],[DUMAND-10-93], which measures acoustical signal transit times with 10 ?sec precision using frequency modulated chirps and matched filtering via DSPs[Berns 93]. Other components of the environmental monitoring system (Fig.9) measure oceanographic parameters such as water currents, temperature and salinity (needed to calculate the local speed of sound)[DIR-13-92][DIR-10-94]. The acoustical positioning system components provided observations of the underwater noise environment[DIR-3-94][DIR-11-94].

Video camera system

As part of the junction box environmental module subsystem, the University of Washington group worked with a leading underwater video equipment manufacturer, Deepsea Power and Light, to develop a custom camera module integrating a high resolution camera, remotely adjustable lens and orientation sensors. These components were mounted in a compact Ti case with an acrylic hemispherical window which was designed to compensate for optical distortions. Two modules were manufactured, and mounted on motorized tilt-pan units which were controlled from the shore station via the JB and JBEM. DSP\&L also supplied high-efficiency 250W Th-I lamps, which were mounted on brackets with the camera modules. These lamps provide highly monochromatic light at a wavelength near seawater transmission peak and also near the center of efficiency of the Sony X77 CCD cameras used.

String mechanical design

A number of improvements to the construction of the tensile portion of the string mooring also resulted from the series of Benthos sphere implosion tests, including the development of a segmented riser system consisting of 10 m lengths of kevlar forming a chain with optical module frames as the joining links (Fig. 7). This system has proven to be more versatile than the originally planned continuous length of kevlar since it allows sections to be swapped in or out with much greater ease and reliability.

Calibration Modules

The Calibration Module (CM) is used to produce fast, bright UV pulses for time and pulse height calibration of the array [DIR-4-86]. It consists of two independent lasers with separate power supplies. The lasers are operated under the control of a standard OM electronics board, with some changes made to the computer operating system to accommodate commands specific to the CM. In addition, the laser intensity is varied by a mechanical optical filter assembly which is actuated by a steppeing motor under control of the same computer. The whole system is placed in a Benthos sphere of the same type used for the OMs. The lasers then shine through the glass sphere to strike a small plastic scintillator ball externally mounted 1 meter away. The scintillator ball makes the pulse more isotropic such that it can be seen by a large number of phototubes in the detector array. A small PMT monitors the scintillator flash intensity within the CM. One CM was constructed and deployed in December 1993 and was subsequently recovered with the rest of the DUMAND string 1.

Deployment and end-to-end test

One of the greatest concerns has been the difficulty of managing large and delicate string arrays during the deployment operations. Two distinct series of tests were aimed at settling these issues and these tests eventually proved to be invaluable learning exercises which paid off in a smooth operation when the actual deployment of the first string took place.

The first tests of deployment took place in September 1992 aboard the R/V Moana Wave. A short, mechanically accurate section of an actually string was deployed and recovered a number of times successfully. This test validated the new string mechanical design which incorporated the modifications suggested by the earlier implosion tests. However, fiber-optical and electrical concerns were secondary to the mechanical ones.

During the next major deployment test, one year later also aboard the Moana Wave, the deployment scheme was modified further to avoid any possibility of damaging fiber-optics. This concern arose because the original method appeared to produce sharp bends in the riser cables during the lifting and load transfer operations. A method which solved this problem and also improved the safety of the operation was found, and implemented.

The deployment tests were crucially important in that the practice loads used were comparable to the expected loads during the actual deployment; thus a number of lessons were learned about the support and transfer of these loads which would have otherwise been missed, and could have been very costly during the final operations. We were able to practice lifting and guiding the junction box over the side, as well as performing the crucial load transfers that would be required later to manage it. All of the practice acquired in this shakedown cruise proved invaluable later in the actual deployment. The shakedown of the surface navigation was also very revealing, since it was found that the radar ranging system was ineffective at that distance from shore. This led to a decision to rely on a redundant DGPS system only for the surface navigation.

In December of 1993, the DUMAND scientific team and the crew of the University of Washington oceanographic ship R/V Thomas G.Thompson successfully deployed the first major components of DUMAND, including the junction box, environmental monitoring equipment, and the shore cable, with one complete string attached to the junction box [Grieder 95].

The string was lowered in an inverted position with a sacrificial anchor and 400 m of line ahead of it. Once the anchor was set on the bottom, the string was then lowered slowly while the ship moved ahead to bring the string into an arched position, with its own buoyancy providing the tension to form the arch (Fig. 10). The junction box was then placed to complete the arch. Once the string was in this position, it could be released near the apex of the arch. The release mechanism was electrically actuated from the string controller. Unfortunately, the electrical release failed to operate when actuated. However, there was also a backup magnesium corrodible link which released the string a few days after deployment.

The junction box was placed within 90 m of the planned location, landed upright, and settled into the bottom surface a few inches, exactly as planned. Cable laying equipment was leased and mounted on the ship. The procedures for the lowering and cable laying operations had been worked out in detail by an ocean engineering contractor[MOE 92].

The deployed string was used to record backgrounds and muon events. Unfortunately, an undetected flaw in one of over 100 electrical penetrators (connectors) used for the electronics pressure vessels produced a small water leak. Seawater eventually shorted out the string controller electronics, disabling further observations after about 10 hours of operation.

Of great importance for continuing efforts is the fact that DUMAND performed a complete end-to-end system test under actual conditions in situ. This test, including the cable, junction box, and string, could not be done at ambient pressure and temperature any other way.

Cable performance

The DUMAND shore cable was designed by G. Wilkins of the University of Hawaii, and comprised 12 single-mode optical fibers in a stainless-steel tube, surrounded by a copper sheath capable of transmitting 5 kW of electrical power[DIR-15-90][DIR-16-90]. The overall resistance of the 32 km cable length laid was about 30 ohms. The cable was covered overall with two jackets of stainless steel armor with opposite helicity, to make the cable as torque-free as possible. The cable was manufactured in 8 km lenths which were spliced together during the laying operation. Despite the splices, the net attenuation of light along all fibers (except one which failed for unknown reasons during laying operations) was unmeasurably close to the diffraction limit [Rosen 94].

Shore cable lay

Following junction box and string deployment on 14 December 1993, the R/V Thomas G.Thompson successfully laid the cable from the DUMAND site to the shore station at the Natural Energy Laboratory of Hawaii (NELH) at Keahole Point. The cable path very closely adhered to the surveyed, planned route (Fig. 11). The cable end was brought ashore by a small boat, and a diver inserted it into the prepared conduit leading to the shore station. The conduit had been provided without cost to DUMAND by a separate project, a DOE-supported investigation of slant-drilling techniques for ocean thermal energy conversion (OTEC) systems. NELH operates demonstration programs in a variety of OTEC and related fields.

String Operation

The data system had been designed to cope with the background rate from radioactivity in the water (primarily from natural potassium-40) and bioluminescence and still generate minimal deadtime for recording neutrino events. The dark counting rate for a single OM was found to be on the order of 60 kHz, primarily due to trace K-40 in the huge volume of seawater each tube views. Noise due to bioluminescence is episodic and expected to be unimportant after the array has been stationary on the ocean bottom for some time, since the light-emitting microscopic creatures are stimulated by motion. K-40 and bioluminescence contribute mainly 1 photoelectron hits distributed randomly in time over the entire array. Bioluminescence caused spikes in the singles rate which reach 100 kHz for periods on the order of seconds, but their frequency of occurrence is relatively low (Fig. 12).

After about 14 hr of operation, the string controller stopped operating, due (as determined after recovery) to a leak in one connector. Seawater in the string controller effectively short- circuited the power line to the junction box, preventing operation of JB equipment such as the EM and video cameras. The short was removed when the string was subsequently released and recovered (the remotely activated release cuts the string umbilical to the junction box), but reappeared a short time later, evidently due to a similar, slower leak in the JBEM.

String Recovery

In January, 1994, the disabled string deployed the previous month was remotely released by an acoustical signal, recovered at sea, and returned to Honolulu for diagnosis and repair. A local vessel, the 120' salvage ship Noho-loa, was chartered for the recovery. The recovery was accomplished between 28-30 January 1994, about 44 days after the string was deployed. After establishing position at the site, the acoustic release was actuated, and the release operation was confirmed by simultaneous observations of the fiber continuity made using an optical time- domain reflectometer at the shore end of the cable at Keahole point, since the release caused a break in the fiber about 200 m from the end at the string controller. Release operations were performed before dawn, so advantage could be taken of the self-actuating light beacon on the string for location, with daylight for recovery operations coming shortly thereafter. The string rose to the surface faster than expected, and the VHF beacon was heard within 1 hr after the release operation. The surface strobe light was located about 3/4 hr after that and the ship and a small boat from shore approached the string cautiously. After the surface configuration of the string was noted, the recovery began, and proceeded smoothly. The recovery was completed in about 6 hours, with no damage to any of the instruments.

This rapid and economical recovery of a damaged string provided a significant test of the serviceability of the DUMAND array for long term deployment. The fact that such a complex instrument could be recovered without further damage with minimal resources gives us confidence that our original plan for possible replacement of one string in the final array every two years is both viable and cost effective.

Final Status

In mid 1996, DOE determined that further support for DUMAND-II should be terminated. At that time, the following DUMAND assets were in place at the site or at UH:

Of course, the most valuable residual asset of DUMAND-II is the experience of the physicists and skilled technical personnel who constructed and deployed the first string in 1993.


The DUMAND Project represented an extraordinarty scientific adventure for a number of people over a period of about twenty years. The lure of attempting to carry out high technology in the deep ocean using the technology of high energy physics, with the tantalizing prospect of openning an entirely new window on the universe, has drawn the energy of many talented people over the years. While the DUMAND project has ceased to exist in original organizational form, the scientific goals remain in view and being actively sought by more than five groups around the world.

We who have participated in the DUMAND Project have had a marvelous experience and we hope the project has left a legacy of scientific and technical accomplishments which can be a base for the succeeding experiments to realize high energy neutrino astronomy in the near future.


This paper was prepared with crucial assistance from many members of the DUMAND Collaboration. Portions of the text were adapted from various proposals and reports prepared over the years, and their original authorship is now impossible to determine. All errors are our responsibility.

We want to dedicate this final DUMAND summary paper to the unflagging dedication of Howard Blood, Fred Reines, and Arthur Roberts. Others desrving mention as crucial contributors in the early stages are Howard Davis, Peter Kotzer, M. A. Markov, Saburo Miyake, and George Zatsepin.


In addition to cited works, the following partial lists of DUMAND publications and internal notes will demonstrate the breadth of effort by the Collaboration. Only those notes cited in the text are listed here; complete lists can be obtained from the DUMAND World Wide Web page at Hawaii.

Internal notes were generally not restricted, but not intended for use outside the collaboration, and thus often contain DUMAND jargon without explanation. External notes were usually copies of conference papers or reports intended for wider distribution. Note that the numbering system for the external note archive was changed in 1992 from $HDC-serial-year$ to $DUMAND-serial-year$. Victor Stenger (U. Hawaii) was primarily responsible for the heroic job of maintaining these archives.

DUMAND Workshop Proceedings

External Reports

Internal Reports

Other References

Material Available on the Web

Further information on DUMAND is available via the DUMAND Home Page on the World Wide Web. The URL address is: {\it}