Photomultiplier tubes stay the course

When researchers require both fast response time and high sensitivity in photonic detection, even to the degree of counting individual photons, photomultiplier tubes (PMTs) may still be the best solution. Application areas for these detectors remain diverse, including biomedicine, astronomy, medical diagnostics, environmental monitoring, industrial quality control, and aerospace.

Among the oldest detector technologies, PMTs use vacuum, rather than solid-state techniques to multiply electrons generated by photons—but this doesn't mean the technology isn't advancing, albeit at a slow pace. Current research needs are pushing designers to develop PMTs with even larger sensitive areas. Other issues on their wish list include greater immunity to magnetic fields and smaller pixels for higher position resolution.¹

The basics

All PMTs operate basically the same way. Light passes through an input window and strikes the photocathode, causing emission of photoelectrons into the evacuated tube. The electrons, which have been accelerated and focused by an accelerating electrode, then strike a metal dynode upon which multiplication occurs. Each primary electron leads to the emission of several, usually around 10, secondary electrons. These are then accelerated again toward another dynode, and the process repeats to produce overall multiplication factors into the millions range.

The electrons from the last dynode, which collect at the anode, then leave the detector for external measurement circuits. Generally, the total potential between the photocathode and anode is near 1 to 2 kV. With 10 dynodes, the drop between dynodes is 100 to 200 V. This yields both good multiplication levels and nanosecond response times. In some cases, response occurs in as little as 300 ps.

Depending on the photocathode material, these detectors will be sensitive to a range of spectral windows. Cesium iodide, for instance, provides for an ultraviolet (UV) range of 0.115 to 0.2 µm. Antimony cesium and bialkali (antimony rubidium cesium, antimony potassium cesium) are available for the visible range. Multialkali types and gallium arsenide extend sensitivity from 0.3 to 0.85 µm. Silver oxygen cesium and indium gallium arsenide allow moving further into the infrared (IR) range, are available, but can also extend sensitivity down to 0.3 µm.

Even the window material can affect the range of spectral sensitivity. Borosilicate glass, the most common window material, has a UV cutoff of 0.3 µm. For shorter wavelengths, magnesium fluoride, sapphire, and synthetic silica produce UV cutoffs of 0.115, 0.15, and 0.16 µm respectively.

For still shorter wavelengths in the x-ray range, researchers at the European Nuclear Research Center (CERN, Geneva, Switzerland) have demonstrated the use of yttrium aluminum peroxide doped with cerium (YAlO₃:Ce or YAP) for windows.² With the window producing scintillations in the visible range, the resulting PMT was able to produce x-ray spectra ranging from 2 to 400 keV.

There also are several different dynode arrangements. The most common involve the box and grid, in which the dynodes are a series of semicircles set in a line, and the circular cage, in which electrons bounce around in a circular pattern. Another design is the venetian blind, in which each dynode consists of a series of metal slats. More recent developments include a fine-mesh type for use in position-sensitive PMTs, in which position information is preserved as the electrons go from one grid-like dynode to the next.

One particular dynode arrangement that has gained in importance in the past decade is the multichannel plate (MCP). Here, instead of having separate dynodes in a series, each small metal channel serves as a multiple dynode, with the electrons bouncing back and forth off the channel walls, causing multiplication at each bounce. A detector array can include thousands of channels, each with an internal diameter of 6 to 20 µm. Since MCPs considerably shorten the path traveled by each electron, response times reach tens of picoseconds.

This dynode arrangement also reduces sensitivity to magnetic fields. In conventional PMTs, such fields can distort electron paths and reduce gain. In MCPs, though, the electron path detail does not matter since the electrons are confined to the channels. This produces a degree of insensitivity to magnetic fields. Up to 40-kG fields parallel to the channels can be tolerated, but only about 700 G is allowed across the axis.

Photomultipliers operate in two distinct modes, depending on the intensity of illumination. In extremely low light levels, the time between individual photons is greater than the device response time, so each photon counts as a single pulse. As illumination increases, the pulses merge into a time-varying analog signal. These two modes naturally require different measurement circuitry. The analog case involves the measurement and averaging of current values, while photon counting relies on pulse-height analyzers to count all pulses over a given threshold.

For some applications, it is useful to vary the gain of the PMT, which can be done by varying the accelerating potential. A Hamamatsu (Bridgewater, NJ) PMT, for example, has a gain of only 200 with a 300-V potential, but this rises to 105,000 with a 700-V potential.³

Scintillation and position sensitivity

By themselves, PMTs can detect radiation into the UV range. Linking them to scintillators allows radiation detection far into the gamma ray range. When radiation enters a scintillate, it produces a brief flash of fluorescence detectable by a PMT. Common scintillator crystals include barium fluoride, bismuth germania oxide (BGO), and cadmium tungstate.

In some applications, improvements in spatial resolution are limited by the physical size of the PMTs. As a result, there has been an intense development effort to create ways to detect and measure the position of photons on the face of the PMT, which boosts resolution. In positron emission tomography (PET) scanning, in particular, the use of animals in research has placed a premium on such advances in resolution.

There are now a number of such approaches to position-sensitive PMTs, including a detector system based on multiple anode MCPs. Here an anode array, such as a 10 × 10 matrix, is located behind an MCP array. Since the MCP preserves position information, the anode array can pick up the information with a resolution of around 1.5 mm.

More interest, though, has focused on two other position-sensitive detector types that use fine-mesh dynodes. One of the designs uses a similar multiple-anode array to that of the MCP, but with somewhat coarser resolution of a few millimeters. Here, the fine mesh, which is in fact a series of meshes, keeps the electrons localized as they are multiplied, since each electron travels only a short distance before hitting the next layer of mesh.

The second, and most recent design, uses a coarser mesh or grid dynodes, but adds a focusing mesh between each grid dynode. The magnetic fields created by the wires in the focusing mesh act as an electron lens, causing the electrons to curve back toward their original positions. This further limits the spread of the secondary electrons from the position of the primary electron. The main innovation in this approach is in the cross-wire anode, which consists of two sets of wires perpendicular to each other, with each wire in a set separated from the next by a constant resistance. The charge is read out from each end of both the x and y sets.

The beauty of the cross-wire grid is that it allows the accurate determination of the peak or center of gravity of the broad spot created by an electron shower. While the shower may be 4 mm across at the base, the peak of the shower and thus the location of the primary electrons is measured to an accuracy of a few tenths of a millimeter. The actual accuracy depends on how many photons are generated by a given scintillation at a given location in the crystal. With this process, the x coordinate of the original spot is found by dividing the charge from one end of the x set by the sum of the charge from both ends. The y coordinate is found the same way.

Using such cross-wire PMTs with arrays of BGO scintillators, researchers at Hamamatsu Photonics, which developed these position-sensitive PMTs, have shown that PET scanners can be built with resolutions of 2 to 3 mm, good enough for detailed studies in even the smallest lab animals.

Comparing detectors

Photomultipliers have a variety of strengths and weaknesses compared with other photodetectors. The greatest strength is speed—with response times as fast as 300 ps in the case of conventional designs and 30 ps for MCPs. In contrast, photodiodes take at least 1 ns, which can be too slow for many research applications. Charge coupled devices (CCDs) are even slower, with 10 ms being a typical response time.

While PMTs remain the most sensitive photodetectors, this advantage has narrowed somewhat, as some CCDs and avalanche photodiodes now have sensitivities within a factor of two of PMTs. On the other hand, solid-state detectors clearly have advantages in spatial resolution. While resolution of an MCP may range near 1 mm, an avalanche photodiode could provide a resolution as low as 0.1 mm. A CCD could offer resolution below 10 µm.

REFERENCES