Single Photon Detectors
Side-On or Head-On PMT, Dynodes or Channeltrons ?
Single Photon Detection Specialities
Usually, light is measured using a device either modulating it’s resistance with being
illuminated or a device creating a current which, within certain limits, is proportional to the light intensity on the device. Unfortunately, neither of these devices is capable of measuring extremely small light amount, such as a single photon, for example. While, sticking to photonic description of light, every, or at least a good amount of these, create electron / hole pairs in a photo-diode for example, a current as small as the one generated by the photon is well below the thermic noise
of the diode and thus not measurable at all. Many photons per second must hit the photosensitive area of such a device to ensure a current generated large enough to become measurable. While the actual amount of required current is very dependent on the bandwidth the device is being used at (that is, it’s capability of correctly following changes in the light intensity at a certain frequency), the general trend is easy to see - the higher the bandwidth, the larger the required current. Typical
values for a not too large bandwidth are in the order of 1 pA, which corresponds to approx. 6.5 Million photons per second at 633 nm wavelength - quite far from single photon resolution. Workarounds for such (analog) measurements are the use of diode with internal gain mechanism (APDs), which helps for high speed applications (thus increasing the bandwidth), but not much in terms of minimum detectable light intensity, unless a very small bandwidth is used along with pulsed illumination and
lock-in techniques to reduce the 1/f noise semiconductors always show.
Quite obviously, a single photon detector requires a certain amplification mechanism, as noiseless as possible, which per primary
generated electron or electron / hole pair, generates a secondary pulse of electrons consiting of at least 100.000 electrons per pulse. If the pulse itself is short in time, the resulting instantaneous current can be quite large and the pulse becomes easy detectable by additional electronics. A typical member of the family of internal gain photodetectors is the PMT (photomultiplier tube). It is a vacuum device which has a photo-sensitivy area, the photocathode, and a certain number of “dynode
stages” which multiply the number of electrons the photon generated pulse consists of. While a photon is capable of releasing an electron from the photocathode, the electron gets accelarated by a rather strong electric field, collids with the first dynode and by doing so releases on average 3...6 secondary electrons, mainly depending on the dynode material and electric field strength used. These will do the same on the next dynode stage etc. It does not need much algebra to prove, that after
8 ... 14 stages, the average amplification is in the order of 106 easily. Since the flight time within the tube is very short, as is the electron release time from the photocathode, which can be viewed as being instantaneous, a pulse in the order of a few 10 ns width is generated. Since around 1 Million electrons are “moved” within approx. 10...20 ns, the instantaneous current is relatively high (a few 10 µA) and thus easy to post-amplify. The drawback of the PMT is the comparably small quantum efficiency - since the electron must be physically released from the photocathode, the photocathode can not be made arbitrarily thick - however, a thin layer does not allow high photon capture probabilities - the photocathode is semi-transparent. Usually the quantum efficiency is not better than 20% in the UV/blue wavelength and below 5% in the red wavelength regime. Although other photocathode techniques exist, such as “opaque” photocathodes which operate in reflection mode and (using GaAs as material) allow up to 30% Q.E. in the red wavelength regime, both, the rather severe count rate linearity limitations and the cost of these devices (difficult to process photocathode, photocathode requires massive cooling) prevented the use for photon correlation experiment so far. There seems to be little hope that this will change in future. Additional problems with all PMTs arise by two effects, recharge effects of the last dynode chains, which are the ones moving a quite high number of electrons in a quite short time and afterpulsing, “phantom” pulses generated some time after the primary photon pulse was generated. Usually these are caused by residual gas atoms which get ionised if travelling through the electron “cloud” between the last dynode chain and the anode of the PMT, recombine and release photons doings so. This afterpulsing can be minimised by special tube design (fast focused and head-on design), but never really prevented.
Side-On or Head-On PMT, Dynodes or Channeltrons ?
Although offering somewhat enhanced quantum efficiency, the significantly larger afterpulse contribution of side-on PMTs (which usually use a circular cage type dynode design) make them at best second choice for correlation experiments and
such tubes should be avoided unless special selection procedures were applied to ensure low afterpulsing. An annoying effect of side-on PMTs, which can not be suppressed by selection at all, is the rather limited count rate linearity which is a plain result of the dynode design used - up to 10 x worse than a good head-on PMT, no matter of the photocathode material used (however, even worse using GaAs). ALV thus uses head-on PMTs only due to their much better linearity and afterpulse
performance. A head-on PMT with a special photocathode ensuring near to 10% quantum efficiency in the red wavelength regime and < 200 cps dark count was part of a development project between ALV and a PMT manufacturer which resulted in the ALV-SIPC-II.
of using dynodes as secondary electron amplifier, other technologies can be used, such as channeltrons, for example. The overall performance can be better or worse compared to traditional PMTs, however the semiconductor doted channeltron seems to generally have some advantages over dynode chains in terms of pulse height distribution, specially if compared at the same equivalent overall gain.
The other member of the family is the APD in geiger mode. Unlike the analo mode, where the internal gain of an APD is in the order of 100 (and thus too little to ensure single photon counting capabilities), the Geiger mode is reached if the diode is operated well above it’s break-down voltage. This condition is best visualised by a field of mouse
traps. All the mouse traps are equipped with a (well known) ping-pong ball. As long as there is no free electron (ball) nothing happens. However, the first free electron (ball) can quickly generate an avalanche of free electrons (balls) - which is nothing else but a large current pulse, even larger than the dynode chain generated electron pulse of a PMT. The difference to the classical mouse trap picture - best known from the nuclear chain reaction - is that the mouse traps are continuously
reloaded - this causes the following situation : once the avalanche was generated, the APD stays in a conducting mode forever - without further electronics “quenching” the diode back to a non-conducting mode, the APD would literally be a single photon detector - a single photon would be sufficient to have it conducting and any further photon could not be detected anymore. “Quenching” is a relatively easy task - simply ensure that shortly after the avalanche was generated the number of mouse
traps still being reloaded is small enough to stop the chain reaction. As long as the average multiplication factor of balls (electrons) is below 1, the avalanche will stop and so will the conducting mode. “Quenching” the APD can be performed in a passive or active mode, the passive mode limits the current that can flow through the diode to a value small enough to ensure the avalanche is quenched, the active mode in fact reduces the reverse voltage over the diode to be significantly below
break-down for a certain period of time. While the passive quenching usually suffers from rather slow time response, since the resistors required to stop the avalanche are quite large and the R x C time constant is as well, active quenching requires special fast electronics and FETs capable to switch the required several 10 V within a few ns. Still active quenched APDs show overall performance quite comparable to standard photon counting PMTs, as long as fast quench times are guaranteed -
otherwise the count rate linearity can quickly suffer.
In contrast to the PMT, the quantum efficiency of an APD is mainly depending on the electron multiplication behaviour, since the Q.E. of the light absorption in the intrinsic area of the diode itself usually is in the order of 80 ... 85% peak and much less wavelength dependent than that of a typical photocathode. Overall quantum efficiencies of 70% at 633 nm are reachable for modern APDs - and they still show more than 20 ... 30% in the near IR and in the blue wavelength regime. Afterpulsing is visible with APDs as well - however the underlying physics is a bit different. While ionisation was the major effect in PMTs, APDs suffer from “electron capture” - electrons are captured within the semiconductor and will be released after a certain time triggering another pulse - the afterpulse. The average capture time seems to be depending on the actual temperature (thus kT) and is prolonged at strong cooling. Strong cooling on the other hand is important to reduce the dark current / dark counts and typically APDs annoyingly show afterpulses in a much longer lag time regime than PMTs. While the author does not know whether special design mechanisms exist to reduce afterpulsing in an APD, the ultimo ratio, as usual, is strict selection for lowest afterpulsing on APD level.
The todays APDs and PMTs show very low dark count contribution - that is the unavoidable contribution of thermal noise in terms of generated pulses. For a good photon correlation setup, the dark counts should be smaller than a few 100 cps for DLS and below 250 cps for FCS experiments.
Single Photon Detection Specialities
MCP - Multi Channel Plate. These are principally equivalent to PMTs, but have as secondary electron multiplier tiny channels of a few 10 µm diameter in a glass substrate. They are most comparable to channeltrons in how they work, but usually require much higher voltage (8 .. 10 kV in contrast to 1 kV for a PMT) and still do not show the same amplification, even if used in multiple stages. Their advantage is that they can easily act as 2D-detector and allow images to be created. Not surprisingly, they major use is for military purposes as image intensifiers. In this case, instead of multiple anodes, phosphorescent material is used as image screen.
Hybrid Detector - a mixture of APD and PMT. Using a semi-transparent photocathode, the Hybrid Detector at first strongly accelerates the free electron released from the photocathode by a photon. This electron, in a second stage, bombards an APD structure which acts as secondary electron amplifier. Since the APD can
not be operated in Geiger Mode due to the large diode area required (typically several square mm), the overall amplification is not better than 10.000 usually (for 8 ... 10 kV acceleration voltage), just on the edge for reasonable single photon detection. Since, besides shorter pulse rise time, shorter pulse duration and better double/triple photon resolution, the detector offers no advantage over a standard PMT, specially the quantum efficiency is as good as that of the PMT, its use
for photon correlation experiment is rather questionable.
Photon Counting CCD - charge coupled devices offer excellent quantum efficiencies comparable to that of good photo-diodes. Recently, a CCD with additional ionisation based charge amplifier at the output stage was release which was claimed to allow photon counting, at least if strongly cooled. Still, two major problems appear, the size of the single CCD spot, which is a few µm diameter only (larger areas would not allow photon counting anymore) and the very low readout time. The same spot on a CCD usually is read out not faster than 30 ...60 times a second, which makes this detector unusable for 99.95% of all photon correlation experiments. Unfortunately, there is little hope, the frame readout time will ever be as fast as 10 .. 20 ns, not even 1 µs.
Quantum Dots - a more or less academic device so far. Transistors with extremely small sizes were created which showed strong quantum effects. These were used as photon detectors. They have been shown to allow single photon detection, however at very low Q.E. of 1% only (which was addressed to the poor incoupling of light to the device, the researchers mentioned a maximum Q.E. of 10%) and rather complicated read-out electronics which did not allow very fast pulse detection. However, the interesting point about quantum dots is, that they do not require strong fields and thus should be very little susceptible to afterpulsing, if at all. The disadvantage seems to be that the “quantum well” must be reset after a certain time and/or accumulated charge, just like a CCD and the naturlly small size, which makes efficient coupling of light to the device extremely difficult.