Novel photosensors for neutrino detectors and telescopes

Abstract

The volume and the photosensitive area of next generation detectors of the numerous rarely occurring phenomena will greatly exceed the sizes of the current experiments. These phenomena include cosmic neutrinos, atmospheric neutrinos, long-baseline neutrino beams from accelerators, geo-neutrinos, geo-reactor neutrinos, and hypothetic proton decays. Similar requirements hold for a new type of a large scanning device for homeland security and nuclear proliferation control, and for the future widely accessible medical imaging devices. Photon detectors are the most important component of such detectors. Existing photosensors are based on vacuum tubes and dynode electron multipliers that are essentially hand-made, expensive and nearly impossible to produce in large enough quantities. Silicon detectors are too small for experiments requiring a very large photosensitive area. Our laboratory is developing novel detectors with a large photosensitive area that can be mass-produced, similar to large flat panel TV displays.

1 Introduction

Neutrinos present a unique probe of physical processes in space, in particular of those processes that take place deeply within various celestial bodies including supernovae, the sun, and the earth (geo-neutrinos and geo-reactor neutrinos). As the only exclusively weakly interacting particles, neutrinos have a virtue of easily penetrating through matter and the extragalactic background light.¹ However, for the same reason their detection is notoriously difficult—only detectors of a very large mass may stop a significant number of neutrinos. This requirement holds for the detection of neutrinos of any origin, including of those that originate in particle showers created by cosmic rays in the atmosphere, neutrinos emitted from reactors, and long-range neutrino beams from accelerators. Detectors of a similar size are also necessary for the detection of the extremely rare (hypothetic) nucleon decay process.

In the construction of detectors on such a large scale, no other option remains than to use inexpensive natural media—deep packs of ice, water, the atmosphere, or liquid scintillators. The interactions of neutrinos in these transparent media create charged particles, which in turn radiate Cherenkov or scintillation photons—the finally detected signals of a captured neutrino. That is why photon detectors have been the most important component of a neutrino telescope. In addition, large neutrino detectors are under construction, based on the detection of Cherenkov ‘light’ in radio spectrum emitted by the excess negative charge in particle showers, and the detection of acoustic waves associated with the shower development. The radio and acoustic detection methods are only useful for the observation of the highest energy neutrinos, where most of the primary energy is converted into visible particles (it is currently excluded that one could detect muon tracks). Although the photon-based neutrino telescopes ultimately detect visible light, just like the optical telescopes do, there is a significant difference—neutrino telescopes may not provide light concentration to a small area, because the angular distribution of the initial light is already virtually isotropic.² That is because the Cherenkov light cone has a large intrinsic angle (>41° in water and ice), the scintillation light is isotropic, and because a neutrino telescope should have a wide, ideally isotropic acceptance. The photosensitive surface of an ideal neutrino detector should therefore virtually enclose the large detector volume.³

The volume and the photosensitive area of the next-generation neutrino detectors will greatly exceed the sizes of already very large current experiments (e.g. the proposed water Cherenkov projects UNO, Hyper Kamiokande, or Memphis would each be 20–50 times larger than the Super Kamiokande detector). Existing photosensors, the photomultiplier tubes (PMT), are based on vacuum tubes and dynode electron multipliers that are essentially hand-made, expensive and nearly impossible to produce in large enough quantities. Silicon detectors, like those used in digital cameras, are too small for experiments requiring a very large area. To solve this problem, our laboratory is developing novel detectors with a large photosensitive area that can be mass-produced, similar to large flat panel TV displays. The large photosensitive area in our panels is a cheap photocathode layer. When photons hit that layer, electrons are emitted inside the device, and thanks to some key inventions, concentrated into a very small (and therefore inexpensive) detection and amplification area on the other side of the panel. This technology has been developed for a new type of large scanning device for homeland security and nuclear proliferation control, and it may as well lead to wide accessibility of new medical imaging devices in the future.

We will discuss two complementary flat-panel solutions, named ReFerence (Ferenc US Patent, 2001, 2003; Ferenc et al. 2003, 2005, 2006a) and ArcaLux (D Ferenc submitted). In addition, we developed an intermediate solution that combines some of the novel elements with the old hemispherical vacuum enclosure, the Hemispherical Light Amplifier, discussed in Sect. 3.

2 From flat-panel TV displays to flat-panel photosensors

For various reasons, neither the current vacuum photosensor technology, nor the various modern semiconductor photosensor technologies may be able to meet the high demands of the future large-area projects, regarding quality, cost, and most important, producible quantity. In contrast to several modern semiconductor photosensor technologies that have rapidly evolved during the last few decades, but towards small-pixel devices that are unsuitable for large-area and large-pixel applications, the vacuum photomultiplier tube (PMT) technology did not make any significant progress since the late 1960s. The complex and bulky construction, and the labor-intensive manufacture are inherent to the PMT concept—mass production on the required scale is therefore virtually inconceivable, irrespective of the cost. Some important drawbacks in PMT performance are also intrinsic to the PMT concept:

• (i) low photoelectron collection efficiency (at most ∼70% in large area PMTs);

• (ii) low quantum efficiency (20–25%), limited in addition only to a narrow spectral region;

• (iii) PMT arrays are composed of 100s of individual PMT tubes, held individually by complex and expensive 3-dimensional structures, which also leaves a lot of dead area between the tubes.

• (iv) complicated and expensive electrical installation; each PMT is individually supplied with 10–12 different levels of high voltage, which requires a dense network of high voltage and readout cables, with a separate high voltage power supply, and a separate voltage divider for each PMT. The cost of connector sockets is also high;

• (v) intrinsic fragility, and a very large stored energy due to the vacuum in the large PMT volume (as dramatically experienced in the massive Super Kamiokande phototube implosion);

• (vi) high buoyancy force in liquid media;

• (vii) high sensitivity to magnetic fields; PMTs are sensitive already to the geomagnetic field, and their detection efficiency depends on the orientation in space. Special mu-metal shielding of every individual PMT is necessary, but impossible in many applications in neutrino physics when PMTs must be completely exposed, and

• (viii) almost complete lack of single-photon resolution (i.e. of the ability to resolve the number of photons in a photosensor pixel).

Nearly half of the PMT cost corresponds to the glass bulb, and another half to the dynode column. Large part of the bulb complexity comes because it needs to host the long dynode column. Both components are essentially handmade. In addition, a PMT has a closed topology, and the photocathode formation process actually takes place within an assembled PMT—every PMT is virtually ‘its own factory.’ This method is far from being economical, and it is in stark contrast to modern continuous production-line technologies.

We will discuss below the proposed new photosensor concepts, and the specific advantages. The following list presents some of the goals that have guided us in our search for next-generation photosensors for neutrino telescopes and comprehensive nuclear proliferation control:

• 1. Open topology—necessary for continuous production lines, both in the production of the components, and in the assembly,

• 2. Industrial mass production, with large and stable markets (other than physics research)

• 3. Low cost per unit of area—ideally two orders of magnitude cheaper than PMTs of a similar pixel size,

• 4. Assembled from high-purity (low radioactivity) materials,

• 5. Assembled from high-tolerance industrially prefabricated components (no hand-made parts),

• 6. Completely safe in case of accidental overexposure to strong light,

• 7. Mechanically robust,

• 8. Simple means of mounting (like a ‘picture on the wall’),

• 9. Simple means of electrical connection (single high voltage line for the entire device),

• 10. Neutrally buoyant in water,

• 11. Simple means of readout and high-voltage connection (without multi-pin sockets for each pixel like in PMTs),

• 12. Applicable in very high pressure environments, like in deep-sea neutrino telescopes, possibly without any additional protection,

• 13. The ability to discriminate weak optical light flashes from the ubiquitous background from ⁴⁰K decays and thermionic noise—by high granulation (small pixel size), and fast time resolution,

• 14. Full area coverage (no dead area),

• 15. Single-photon resolution, at least up to five photons,

• 16. High detection efficiency,

• 17. Possibility of photocathode cooling to reduce thermionic noise,

• 18. No need for expensive preamplifiers,

• 19. No need for expensive shielded cables,

• 20. Insensitive to the geomagnetic field, possibly also to stronger fields.

The recent phenomenal success of hi-tech vacuum flat-panel TV technologies, which truly exemplify modern industrial mass production, indicates that an equivalent breakthrough might be possible also in the field of vacuum photosensors. The revolution in the TV industry came because of the profound conceptual shift—from the bulky cathode ray tube (similar to a PMT); to the lightweight and attractive flat-panel configuration, that perfectly suits continuous production lines.

The ultimate solution is therefore a flat-panel technology, inspired by the production of modern flat panel TV displays, see Fig. 1. This technology may lead to inexpensive industrial mass production that could satisfy the needs of the largest future markets, like the nuclear proliferation control, and the medical imaging. Small-scale prototype development is under way at UC Davis, supported by a grant from the US Department of Energy’s National Nuclear Security Administration.

Fig. 1

The flat panel TV displays exemplify modern mass-production technology. Our goal is to make an equivalent step in the vacuum photosensor world, where the vacuum photon detectors have not significantly changed since 1960s

The proposed technology essentially combines three established and well-understood technologies:

• 1. For the vacuum enclosure and high voltage management—mass-production of the modern flat-panel plasma and field-emission TV-screens,

• 2. For the photon conversion into photoelectrons—vacuum deposition of the classical (bialkali or multialkali) photocathodes, and

• 3. For the detection and amplification of photoelectrons—Mass-production of small semiconductor sensors (in particular, CMOS technology).

How would such a flat panel device detect light? In simple terms, the ‘vacuum part’ of the detector transforms incoming photons into electrons in the inexpensive photocathode layer, and compresses the signal to such a small area that a very small and inexpensive semiconductor sensor may be used for the electron detection (Ferenc US Patent, 2001, 2003; Ferenc et al. 2006b; Lorenz and Ferenc 2007; Bisello et al. 1995; Golovin and Saveliev 2004; Buzhan et al. 2003; Dolgoshein 2006). This sensor may detect photoelectrons either directly, or indirectly, through the scintillation light that the photoelectron creates in a small scintillator (Light Amplifier, discussed in Sects. 2.3 and 3). This strong, more than 1,000-fold concentration of information from the irreducibly large photocathode area to the readout device, effectively bypasses Liouville’s theorem by replacing a photon by a photoelectron, i.e. a charged particle that can be acted upon by an electrostatic field. Photoelectrons are focused to the small electron sensor by a special electron lens, while at the same time the components of the lens are integral parts of the flat panel structure. To put the benefit from the concentration into perspective, let us assume that the total light-receiving area of a hypothetic one-megaton water Cherenkov detector (for simplicity, a 100 × 100 × 100 m³ cube) is 60,000 m². With a 1,500-fold concentration, like in the first ReFerence photosensor prototype, that area would effectively concentrate to the photoelectron readout area of only 40 m² (expected to cost not more than several million of US Dollars, since the cost of industrially made Geiger-mode avalanche photo diodes is expected to become lower than $50 cm^-2 in the future).

Our research has been strongly motivated by R&D in the field of field-emission displays, particularly by the pioneering work of the Candescent company from San Jose, California. Soon after successfully completing the development of small 5 inch displays, Candescent has dissolved, selling its intellectual property assets to the Canon-Toshiba joint venture, and a significant fraction of their research equipment to our laboratory. That equipment has served as a basis for the construction of our UHV Transfer System (Ferenc et al. 2006a), designed for the assembly of 5 inch panel prototypes.

For a long time LCD and plasma screens have dominated the market, while field-emission displays have spent nearly 30 years in research laboratories, without reaching the market. However, driven by the introduction of a new generation of ink-jet printed and self-assembling field emitting surfaces, the field-emission technology seems to be making its definite breakthrough. Several major manufacturers have recently announced new field-emission displays, and the early expectations rise even to the point that LCD and plasma screens will be soon completely replaced. The Canon-Toshiba venture that inherited Candescent’s technology, recently commercialized a field-emission device, the Surface Emission Display (SED), which outperformed all plasma and LCD screens. At the same time, its production costs are allegedly significantly lower. This square-meter size device is very similar to our photosensor panels, both in the size and in its construction; it is based on electron acceleration in a high potential (∼10 kV), over a very narrow gap of only few mm, in an ultrahigh vacuum. The success of this device demonstrates the feasibility of the mass production of large inexpensive panels operating with vacuum (plasma screens are also vacuum-processed, but then filled with gas for operation).

Field-emission displays focus electrons from field emitters within each image pixel, to a luminous phosphor area of appropriate color. Our task is different, as we need to convert photons entering through the front face of the panel, now acting as a window, into electrons in the photocathode, and then accelerate and focus those electrons in a vacuum to a point-like detection area on the opposite side of the panel. The real challenge has been to find a solution for efficient electron focusing within large pixels (several centimeters in diameter), while simultaneously assuring continuous sensitivity to light over the entire panel surface, with a negligible dead area between the pixels. As discussed in the following two sections, two different inventions have allowed us to solve this problem, in two very different ways.

ReFerence and ArcaLux panels comprise significantly lower constructional complexity than any flat-panel TV screen. In particular, pixels are very large (few cm in diameter), and the tiniest feature within a pixel still looks comfortably large from the manufacturing point of view (the photoelectron sensor of a ∼1 mm diameter). Based on these and other differences in constructional complexity, we have derived a very rough cost estimate for ReFerence and ArcaLux panels—less than one third of the market price of a plasma TV screen, i.e. currently less than $1,000 per square meter (note that the current production cost of TV screens is already below $1,000). This should be contrasted with $30,000, the cost per square meter of popular 8-inch PMTs, or yet with much higher figures of ∼$200,000 for PMTs of comparable (pixel) sizes, like the ones currently used in the imaging atmospheric Cherenkov gamma-ray telescopes. However, much more important than the dramatic cost reduction will be the high production capacity that would not only satisfy all the projected needs, but would change the way we used to design experiments, and would encourage us to readily go for much large and better.

2.1 The ReFerence flat-panel photosensor

The ReFerence photosensor (Ferenc US Patent, 2001, 2003) is based on a fortunate coincidence: the very same mirror/electrode shape may act simultaneously as an optical concentrator, and as an electron lens. The compact parabolic light concentrator, see Fig. 2a, concentrates light from the entrance aperture (on the left) to the photocathode (on the right), and focuses photoelectrons to a very small photoelectron sensor; a p-on-n PIN diode, a p-on-n avalanche diode, a fiber-coupled scintillator in the Light Amplifier (see Sect. 2.3), or other, placed in the middle of the entrance aperture. The concentration factor is nearly optimal in both directions, which leads to an optimal overall signal concentration of about factor 1000. Thus, a very small electron sensor in the focus may detect virtually all photoelectrons (∼100% collection efficiency). Note that this small sensor effectively replaces the entire dynode column–the most expensive, bulky, hand-made, rear-end component in a PMT.

Fig. 2

ReFerence photosensor. The functionality of an elementary ReFerence cell, (a). Result of a prototype test, (b); the bright spot in the middle of the image presents the electron focal point as seen on an inserted phosphor screen. This prototype is hexagonally shaped. Schematic front view of a ReFerence panel, (c). Each hexagonal pixel presents a ReFerence cell. Note the low fraction of dead area. Schematic cross-section through the Light Amplifier panel, (d), see Sect. 2.3. The APD array is symbolically located behind the panel and serves as readout for the scintillators coupled to the APDs by means of optical fibers (fibers not shown)

As we demonstrated in a series of generic 3-inch prototype tests (Ferenc et al. 2003, 2005), the point-like photoelectron focusing persists in a hexagonally formed ReFerence pixel, see Fig. 2b. This feature is essential, because it makes the honeycomb multi-pixel flat-panel configuration possible (Fig. 2c, d). Since the front section of each pixel has a larger diameter than the photocathode, it may be hexagonally reshaped while at the same time the photocathode and the entire rear section maintain axial symmetry. When the electrons reach the hexagonally modified section, they are already fast and driven close to the axis, where the electric field created by the hexagonally shaped electrode looks practically axially symmetric. Electrons are anyway too fast to feel minor residual field deviations, and they are properly focused, just like in a cylindrically symmetric pixel in Fig. 2a.

In our experiments, a 2 mm diameter back-illuminated photocathode was used in order to strictly localize the position of the photoelectron source. This spot-photocathode was then precisely positioned in small increments across the photocathode plane, from the center to the periphery, and the electron focusing was checked by imaging of the electron spot on a phosphor screen in the focal plane. With a correct potential setting, both the position of the focal spot and its size were found to be invariant of the photoelectron source position in the photocathode plane. From the area of the spot size we derived a concentration factor of ∼1500, which is extremely high for such a simple electron lens configuration.

In addition to its suitability for mass-production, as discussed above, the ReFerence flat-panel concept offers other important advantages (Ferenc US Patent, 2001): nearly 100% electron collection efficiency (in contrast to typical 70% efficiency in PMTs); high photon detection efficiency that extends over a much wider spectral range, thanks to the reflection-mode photocathode; small dead area; efficient magnetic shielding; and a very low level of thermionic noise, if the photocathode would be cooled in vacuum. In addition, the exotic TransReFerence configuration (Ferenc 2003) could provide non-destructive color sensitivity, and a very wide spectral sensitivity.

After verifying the basic functionality of a single-pixel ReFerence prototype, in a series of vacuum-unsealed prototype tests, we started with the development of fully functional vacuum-sealed flat-panel ReFerence prototypes (Ferenc et al. 2005, 2006a). The goal of that effort is to demonstrate the feasibility of an industrial mass-production technology, and to find the optimal production and assembly techniques. At our UC Davis lab, we have developed a new oxide-free, low-temperature glass-to-glass sealing technique (Ferenc et al. 2006a), based on a vacuum-evaporated multi-layer structure consisting of Chromium, Gold, and Indium. The first flat-panel prototype comprises 7 pixels hosted by a 5-inch diameter vacuum enclosure. The prototype diameter is limited by the throughput of our UHV Transfer System (Ferenc et al. 2006a), but otherwise the prototype shares all the important features of a large-scale panel. The enclosure of the first panel prototype is made of glass, while the honeycomb plates were made of Aluminum, see Fig. 3. Recently we started with the development of a series of gradually improving all-glass ReFerence panel prototypes, fabricated with the ultimate industrial methods (a variation of precision molding in double-sided molds). We have engaged specialized industries to design the tools and produce the prototype components.

Fig. 3

Components of a ReFerence panel prototype; design and reality (Ferenc et al. 2005, 2006a)

ReFerence panels would be best suited for detection of light from high energy particle showers in the atmosphere, for coupling to relatively thin dense media, or in any application that requires sharply limited angular acceptance.

2.2 The ArcaLux flat-panel photosensor

Prompted by applications that simultaneously require:

• (i) perfect optical coupling to large volumes of dense media,

• (ii) full surface coverage,

• (iii) and full angular acceptance,

we searched for an optimal solution. Recently we arrived at a new concept that combines the virtues of a spherical electron lens with flat-panel topology.

The ArcaLux concept (D Ferenc submitted) merges the flat-panel topology with the seemingly incompatible hemispherical dome configuration. The latter is appealing for applications in thick dense medial, or highly pressurized media, thanks to:

• i. Excellent optical coupling of a hemispherical tube to dense media, like water, ice, or liquid scintillator,

• ii. Full angular acceptance, and

• iii. Optimal mechanical stability in high-pressure environments.

The way we merged a flat panel with a sphere is through a synergy of several constructional elements, as illustrated in Figs. 4–6:

• A matrix comprising small hemispherical cavities (2) submerged into a flat window plate (1); the cavity diameter may assume any value, but for practical panels it should be on the order of few cm.

• An integrated optical concentration system (3). The optical reflectors (3) in the form of cavities integrated in the window plate reflect a large fraction of otherwise lost photons into the photocathodes. This component is the first of the two key innovative elements.

• An integrated high-voltage distribution system (4), where the central plate distributes both the anode and the cathode potential to every pixel, and provides a continuous voltage divider within each pixel. This component is the second of the two key innovative elements.

• An integrated and distributed getter vacuum pump (5), which allows permanent vacuum pumping throughout the panel, and

• In the examples presented in Figs. 4–6, we have assumed the Light Amplifier readout (see Sect. 2.3); consequently, the readout network (6) consists of integrated large-diameter light guides, covered with a thin scintillator film. The alternative solution is, again, to use APDs for direct photoelectron detection.

• The quantum efficiency will be significantly enhanced, since a photon trajectory will practically always pass the active photocathode layer twice, either directly, or after a reflection from the central plate, to be coated with a dielectric reflector.

Fig. 4

A 4 × 4 pixel segment of an ArcaLux flat panel detector, seen at an angle from above (D Ferenc submitted). The top flat surface of the glass plate is assumed transparent (not outlined in this sketch)

Fig. 5

Cross section through a segment of a single-sided ArcaLux panel. The incoming photons (ph) hit the hemispherical photocathodes (2) either directly, or after reflection from the reflector cavities (3). A photoelectron (e) released by the photon (ph) from the photocathode is focused and accelerated to a small electron detector (in the presented case, a fiber-coupled scintillator). The amplified secondary light signal from the scintillator travels via an integrated light guide (6) out of the vacuum enclosure, to a Geiger-mode APD (not shown), where it is read-out and strongly amplified. The presented sketch is highly schematic, and it misses details. The real mass-produced panel will be assembled of the few prefabricated, highly integrated components, one of which will include the light guides (D Ferenc submitted)

Fig. 6

Cross section through a double-sided ArcaLux panel, shown as immersed in water. The double-sided panel comprises a fully symmetric multi-dome structure, ideal for applications in high-pressure environments. It may be sensitive to light on both sides, which is potentially very useful in many applications. If only one side of the panel should be sensitive, the other side may be passive, but still shaped in the same way in order to play its important pressure-supporting role

All features, in all components are open and accessible, which allows both: (i) the application of industrial glass-forming techniques in the component production, and (ii) continuous production-line technology for the panel assembly.

Each hemispherical cavity (2) forms a Roman Dome structure with strong ‘pillars’ that give the panel a high mechanical stability. One of the reasons why we opted for a rectangular, rather than a hexagonal pixel pattern, is that the former offers significantly stronger pillars.

The list of expected advantages of the ArcaLux panel over classical PMTs would be similar to the wish list in the Introduction. The advantages over the ReFerence concept are, a better optical coupling to dense media, and the full angular acceptance. The latter may be also viewed as a disadvantage for those applications that require limited acceptance angle for the incident light. A disadvantage is that a transmission photocathode is less efficient and has a significantly narrower spectral sensitivity than the reflection mode photocathode in ReFerence. On the constructional side, ArcaLux bears more complexity in the windows, while ReFerence incorporates a complex conical structure.

The ArcaLux concept is completely new, and we expect funding for the experimental verification. However, the hemispherical electron lens is so simple that there is hardly any need to prove that feature, provided that the voltage distribution plate (4) is indeed feasible. There is no a priori reason to suspect that, or any other assumptions we made.

2.3 The Light Amplifier concept

Light Amplifier is a general concept (Ferenc US Patent; Ferenc et al. 2005, 2006a, b; D Ferenc submitted) based on a robust and inexpensive readout solution for hybrid photon detectors (HPDs) with strong photoelectron focusing, such as ReFerence panels (Sect. 2.1), ArcaLux panels (Sect. 2.2), and large hemispherical HPDs (Sect. 3). The readout is based on new Geiger-mode avalanche photodiodes (G-APDs) (Bisello et al. 1995; Golovin and Saveliev 2004; Buzhan et al. 2003; Dolgoshein 2006). In conjunction with optically coupled scintillators, G-APDs provide a uniquely robust and inexpensive readout solution (Ferenc Us Patent; Ferenc et al. 2005, 2006a, b; D Ferenc submitted).

In general, Light Amplifier is a vacuum device that comprises:

• (i) Photocathode that converts a photon to an electron,

• (ii) Electron lens that accelerates and focuses all the photoelectrons to a very small focal area covered with

• (iii) A thin scintillator film that emits photons (100s to 1000s) upon the electron impact. The scintillator film is over-coated by a very thin aluminum film that acts both as a shield against light feedback to the photocathode, and as an electrical conductor, to set up the large potential between the cathode and the scintillator; the application of these thin films is suitable for industrial mass-production in continuous production lines, and may be applied (in different processing steps) on top of the window, or on the receiving area of an integrated light guide, depending on the actual detector configuration (flat panel, or hemisphere);

• (iv) These photons are transported via integrated light guides, or simply through a window (see below), from the scintillator to

• (v) The Geiger-mode avalanche photodiode (G-APD), placed outside the vacuum enclosure, and optically coupled to the light guides, using some standard coupling methods (like e.g. in medical imaging PET scanners based on APDs). The G-APDs may also be integrated in the panel structure, together with the light guides, in the early component processing stages (D Ferenc submitted).

Thanks to the strong electron focusing, the entire photosensitive area effectively maps to a very small active G-APD readout area (∼1,000 times smaller than the panel area). The signal concentration is therefore the most important function of the Light Amplifier. Another important function is the amplification of the light signal. Its goal is merely to provide a light pulse that is significantly stronger than the intrinsic noise of the G-APD, i.e. at least 30 photons, when measured outside of the vacuum enclosure. This may be achieved by using a high light yield scintillator, and at least 15 kV of electron acceleration potential. The G-APD will then itself provide the real signal amplification, typically of a ∼10⁶ gain.

Note that Light Amplifier presents an ideal match between its vacuum part and the G-APD. The two main disadvantages of G-APDs—the high level of intrinsic noise, and the small G-APD pixel size, are fully compensated in the Light Amplifier by light amplification, and electron concentration, respectively. In return, the G-APD readout efficiently replaces what used to be a bulky dynode column in classical photomultiplier tubes (PMT), and in addition, it provides a number of unique benefits discussed in more detail below.

The placement of the G-APDs outside the vacuum enclosure leads to a simple and robust construction without any electronic components enclosed within the vacuum panel. That also prevents chemical cross-contamination between the photocathode and the silicon sensor. Since G-APDs are not an integral part of the flat panel, they could be upgraded at any time. More important, that would allow one to modify the position resolution of the panel, merely through remapping of the G-APD pixels, to a different number of primary pixels in the Light Amplifier. But, probably most important, that would allow the production of a standard panel for all applications that require different resolutions.

The following is a summary of the specific advantages of the Light Amplifier (Ferenc et al. 2006b):

• a. Robustness,

• b. Safety against a catastrophic burnout at accidental exposures to strong light, since G-APDs may be safely exposed to ambient light while fully powered,

• c. Very strong signal amplification (several times 10⁶), and consequently,

  • No need for expensive electromagnetic shielding,

  • No need for a preamplifier,

• d. Very small G-APD size (low cost),

• e. Optimization of the cathode formation process by avoiding the need to take into account also the conditions for dynode surface formation,

• f. Nearly 100% photoelectron collection efficiency which is normally a big problem in large area PMTs with secondary dynodes that results in lower and non-uniform photoelectron collection efficiency (at most ∼70% in large area PMTs),

• g. Minimized timing jitter due to nearly perfect electron focusing, as previously demonstrated with virtually all designs of single-pixel Hybrid Photon Detectors, which also use high acceleration potentials of 8–15 kV; this cannot be achieved in large area PMTs with classical dynodes and a low acceleration potential of only several 100 Volts in the first stage.

• h. A device that is virtually insensitive to the earth magnetic field, both because the primary device, the vacuum Light Amplifier, is nearly unaffected by the earth magnetic field due to its high acceleration voltage, and G-APDs are completely insensitive to (even extremely large) magnetic fields,

• i. G-APDs require only a low voltage supply (∼50 V), and

• j. Excellent intrinsic position resolution (which may be modified by remapping of the primary pixels to the G-APDs).

The following section summarizes our tests of a hemispherical Light Amplifier device.

3 Hemispherical light amplifier

The Hemispherical Light Amplifier (Ferenc et al. 2005, 2006b; Lorenz and Ferenc 2007) is an application of the general Light Amplifier concept (Ferenc US Patent; Ferenc et al. 2005) (reviewed in Sect. 2.3) to a large hemispherical vacuum enclosure, rather than to a flat-panel, see Figs. 7, 8. Since hemispherical devices may be manufactured with relatively simple modifications to the old PMT technology, this idea may present a fast intermediate solution, useful before industrial production of flat panel devices has fully developed.

Fig. 7

Hemispherical Light Amplifier (Ferenc et al. 2006b). Photoelectrons originating from the hemispherical photocathode are accelerated within the hemispherical vacuum tube (left) to an energy of ∼25 keV, and focused to a small scintillator surface. The slightly amplified light signal from the scintillator is detected in a Geiger-mode APD array, placed outside the vacuum enclosure, on the other side of the fiber plate window that preserves the image configuration (middle). Our first prototype, based on a QUASAR tube, did not have a fiber plate coupling (right)

Fig. 8

Two QUASAR tubes (Ferenc 2001) are shown, the one we tested, enclosed in a dark box (on the left), and a vented tube without a photocathode (on the right), in which the interior is nicely visible, including the dark focal area in the middle that hosts the scintillator. Photoelectrons from the entire photocathode are mapped onto that small surface (Ferenc et al. 2006b)

The first results of the experiment with a Hemispherical Light Amplifier prototype derived from an old QUASAR phototube (Fig. 8) are presented in (Ferenc et al. 2006b). Figure 7 shows a functional sketch of the Hemispherical Light Amplifier. QUASAR phototubes (a variation of the original concept developed by Philips, the so-called Smart-PMT) have been used in the Baikal Lake neutrino telescope (Lubsandorzhiev 1999, 2005). The original QUASAR tube is a spherical vacuum tube made of glass, 370 mm in diameter, with a bialkali photocathode deposited on the inside surface, and a flat 18 mm diameter focal area covered with a thin scintillator film (blue light emitting scintillator Y₂SiO₅(Ce)), which is over-covered with a thin layer of Aluminum to prevent light feedback to the photocathode, and to establish the anode electrical potential. Unlike in the original QUASAR tube (Dolgoshein 2006; Lubsandorzhiev 1999), which uses a small PMT for the secondary light readout, our Light Amplifier prototype uses Geiger-mode APDs (Bisello et al. 1995; Golovin and Saveliev 2004; Buzhan et al. 2003; Dolgoshein 2006).

As illustrated in the middle sketch in Fig. 7, the ultimate tube should include a fiber plate to couple the scintillator to the G-APD array, both for an efficient coupling and for precise imaging. The implementation of a fiber plate should be rather simple at the time the glass envelope has been blown up, although the typical fiber plates available on the market have a way too good resolution for this application, and they also cost significantly more than a thin, ∼1 mm resolution plate that would satisfy all our needs. However, the existing QUASAR tubes do not comprise a fiber plate, but rather an ordinary 5 mm thick window, and therefore our first prototype (Ferenc et al. 2006b) has offered a rather poor coupling efficiency, but still sufficiently strong for our proof-of-concept tests. In summary, despite the weak amplification, and the strong photon dispersion, we detected pulses corresponding to single photoelectrons, we resolved individual photoelectrons within multi-photon events and groups of pulse amplitudes within well-defined multi-photon peaks (Ferenc et al. 2006b; Lorenz and Ferenc 2007). We also verified the intrinsic imaging capabilities of the QUASAR tube, with a remark that a new electron lens design would grant a much better result (Ferenc 1999) (note that QUASAR was designed as a single-pixel device).

Apart from the improved position resolution, there would be another important benefit from the application of a position-sensitive G-APD matrix in the readout. The active photosensitive area of a large hemispherical tube virtually splits into a large number of independent pixels, acting like an array of individual photosensors. While a single-channel readout collects together signal photons, and all the photons from the environmental background (from the thermionic emission of the entire photocathode, and the radioactive decays), the multi-channel readout will localize, and effectively ‘dilute’ the noise contributions in the individual pixels, and thus improve the signal-to-noise figure. This effect is similar to the more familiar benefit in the signal-to-noise ratio that comes from the improvement in the time resolution. In particular, this will allow for a major improvement (depending on the granulation) in deep-sea neutrino telescopes—the ability to better discriminate weak optical light flashes from the ubiquitous background from ⁴⁰K decays, and from the strong intrinsic PMT noise.

A very promising solution is a fully Spherical Light Amplifier (Lorenz and Ferenc 2007), outlined in Fig. 9. This single-pixel device would provide almost full angular acceptance, a very large photosensitive area, perfect timing resolution, and 100% photoelectron collection efficiency. Its shape would be ideally fitted for deep-sea or ice neutrino telescopes.

Fig. 9

Spherical Light Amplifier (Lorenz and Ferenc 2007). Sketch of a possible configuration with a single spherical scintillator for perfect spherical symmetry

Based on our test results (Ferenc et al. 2006b; Lorenz and Ferenc 2007), the Hemispherical Light Amplifier concept should be considered as an excellent solution for high quality photon detection in the near future. We have also implicitly verified that G-APDs present a superior solution for the readout of any Light Amplifier, either spherical or flat-panel.

Let us point out that the quantum efficiency of a spherical Light Amplifier will be significantly enhanced, since the photon trajectory will almost always pass the active photocathode area twice. In a hemispherical device, a dielectric reflective coating on the internal passive tube surfaces will send the photons back to the photocathode.

Some new developments promise significant progress: new scintillators, such as BrilLanCe 380 from Saint-Gobain (16 ns decay time, about twice the light yield compared to the currently used scintillator), will offer a higher light yield, and shorter decay times; new G-APDs with enhanced blue sensitivity and photon detection efficiency approaching 40% have emerged recently; and the first generation of G-APDs with larger area (3 × 3 mm²) have been already used in our latest tests (Lorenz and Ferenc 2007). First prototypes of multi-pixel G-APD arrays are also expected soon.

4 Summary

Next-generation large neutrino detectors will study neutrinos originating from various sources: (i) celestial bodies and processes in space, (ii) particle showers created by cosmic rays in the atmosphere, (iii) radioactive days in the earth, (iv) nuclear geo-reactors, (iv) man-made nuclear reactors, and (v) particle accelerators. Detectors of similar characteristics are needed also for the detection of the hypothetic nucleon decay process, and for a comprehensive nuclear proliferation control.

Our aim is to find a replacement for the more than 50-years old PMT photosensors that are essentially hand-made, expensive, and impossible to produce in large enough quantities. We propose novel flat-panel detectors with a large photosensitive area that can be mass-produced, similar to large flat panel field-emission TV displays. The proposed technology merely combines three existing and proven mass-production technologies. In addition, we propose an intermediate solution that combines some of the novel elements with the old hemispherical vacuum enclosure—the Hemispherical Light Amplifier. While the flat-panel technology promises inexpensive mass production, but requires a relatively high upfront investment, the Hemispherical Light Amplifier may present a quick solution for some neutrino telescopes, since small quantities of large spherical vacuum tubes may still be manufactured.

We are currently making a small-scale (5 inch) all-glass ReFerence panel prototype. We have designed and constructed a state-of-the-art facility for prototype assembly at UC Davis, the UHV Transfer System. We developed a new vacuum sealing method specifically for the assembly of all-glass flat-panel detectors (Ferenc et al. 2006a). We also confirmed the feasibility of the Light Amplifier concept in a series of experiments using a prototype made of an old hemispherical Quasar tube, and a series of ever improving Geiger-mode APDs (Ferenc et al. 2006b; Lorenz and Ferenc 2007).