Sensors for True Confocal Scanning

August 23, 2012

One of the most critical elements in an imaging instrument is the light sensor. It resembles the retina, converting light into a signal that is subsequently transferred into a storable and processable ensemble of information. The sensors for single point scanning systems such as true confocal microscopes are usually photomultiplier tubes. Also, the silicon pendants of PMTs are used for particular applications, especially single-molecule measurements. A new development has led to chimeric devices called hybrid detectors (HyD) which unite the benefits of the two technologies while eliminating the disadvantages.



Fig. 1: Graphical representation of a photomultiplier tube. The photon energy is converted into accelerated electrons at the photocathode. Subsequently following dynodes multiply the initial photoelectron by secondary emission. The multiplication at each step depends on the potential difference between the dynodes, which is 1/n of the total high voltage applied when the tube contains n dynodes.
True confocal scanning (TCS) requires single point sensors. PMTs have been standard for confocal microscope systems for many years. They are still available in Leica TCS SP8 confocal systems.

PMTs are based on vacuum technology, which almost disappeared after it peaked in the 60s of the last century. However, PMTs have advantages over traditional semiconductor devices. A PMT uses two effects (which are described in detail in textbooks of physics). The first effect is the conversion of photon energy into the kinetic energy of released electrons [1]. If a photon interacts with material, the photon energy can be absorbed by the electronic system in the atoms or molecules. If the energy is high enough (shorter wavelength), the excitation can cause an electron to leave the structure. This dissociation requires a material-specific minimum energy [2]. The photocathode of a photomultiplier is made from materials that have sufficiently low dissociation energies to allow red photons down to 800 nm to cause electron dissociation. Unfortunately, red-sensitive materials also show higher thermal noise. Electrons from the photocathode are accelerated by an electric potential to a second electrode (dynode). Here the accumulated energy is used to release more electrons from the material. This is the second effect used in photomultiplier tubes. The result is a gain from the first photoelectron in up to 3 or 4 secondary electrons. These electrons are then amplified by cascaded additional dynodes (typically 6 ... 12). The entire PMT voltage is typically in the range of about 300 V to 1,200 V (tunable), thus the potential difference between dynodes is maximally ca 100 V. If the average gain for a given voltage is 3 per dynode, then the total gain in a 10-dynode PMT is 50,000-fold. After the last dynode the electrons are discharged into an anode, where the electrical current is measured. In analog mode, the current over a time interval is integrated, and the brightness of the sample is coded by that charge. If the circuitry is fast enough, single photons cause current peaks that can be resolved. The intensity then corresponds to the number of photons in a given time interval.

Multi-alkali photocathodes are standard in visible light applications [3]. These tubes exhibit 25 % quantum efficiency for blue-green light. Obviously, there is much room for improvement as regards the sensitivity of such devices [4]. The number of secondary electrons is randomly distributed. When at a certain voltage the gain is 3 electrons on average, then the signal-to-noise ratio (S/N) is 1.7 (Poisson statistics). This noise is transferred to the output signal. Since only whole electrons are released, a variation of only one electron more or less changes the signal by about 30%. A second objective for improving light sensors is therefore to increase the yield in the first amplification step.

A third important parameter is the width of the generated electrical pulses of a single photon. The pulsewidth is broadened by the temporal variation of electrons arriving at the anode, and by the circuitry. At high photon rates, the pulses must be very short in order to separate them for counting. Another discussion point is the dynamic range. Small dynamic ranges allow only limited variation in light intensity. High dynamic ranges indicate acceptance of very different illumination intensities with comparable results.

Avalanche photodiodes

Fig. 2: Upon absorption of photons in an intrinsic (i) layer, an electron-hole pair is generated. The avalanche is caused in a multiplication zone (pn+) by impact ionization through the arriving electrons.
Avalanche photodiodes are used in photon counting mode applications, especially in the single molecule detection systems (Leica TCS SP8 SMD).

Avalanche photo-diodes [5] are PIN diodes with an additional p-layer inserted between the i and n layers. The pn-region is called "multiplication zone" and has very high field strength. The photons are absorbed in the insertion layer, where it can generate an electron-hole pair (intrinsic photoelectric effect). Electrons are accelerated to the multiplication zone, where additional pairs are created by impact ionization. This process amplifies the original charge by a factor of 100 to 1,000, depending on the applied voltage. The charge multiplication occurs over a very short interval and is referred to as avalanche effect. Obviously, the sudden increase and the self-amplifying current can easily damage the device, so special precautions must be taken to avoid high currents. Much higher gain can be achieved with avalanche photodiodes, when the applied voltage exceeds the breakdown voltage, allowing amplifications of up to 108 ("Geiger mode"). The high gain is sufficient for direct detection of individual photons. Since the current is not stoppable, the diode circuitry must include specific measures to avoid demolition. Geiger-mode avalanche photodiodes are therefore very sensitive to damage from too much light and improper operation.

Avalanche photodiodes amplify by 100 ... 1,000 fold in non-Geiger mode and up to 100,000,000 fold in Geiger mode. In non-Geiger mode the APD generates a signal which is proportional to the incident light. The dynamic range is relatively low and typical light intensities in confocal microscopes are too high. Only applications with very low intensities allow the use of avalanche photodiodes. This is especially true for Geiger-mode avalanche photodiodes, which are only suitable for single-photon counting at comparatively low count rates (up to 10 MHz). For visible light, silicon-based APDs can be used at a spectral range from 300 nm to 1,000 nm. The quantum yield is better than photomultiplier tubes and can reach around 45 % – in the red zone even more. Due to the improved sensitivity in the red zone, APDs are sometimes a better solution than PMTs. Despite the red sensitivity, which increases thermal noise, APDs are less noisy than PMTs due to their small active area. 

Hybrid detectors

Fig. 3: Scheme of a hybrid detector: a photoelectron is released from the cathode upon interaction with a photon. This electron is accelerated by a high voltage. The acceleration energy is dissipated in a semiconductor target, creating many electron-hole pairs (first amplification). In a subsequent avalanche diode, these charges are multiplied (second amplification).
Hybrid Detector HyD combines vacuum and semiconductor technology. These sensors are implemented in Leica TCS SP8 confocal and multiphoton systems. They are very beneficial in single molecule applications (Leica TCS SP8 SMD).

PMTs and APDs each have their own advantages and application areas. Nevertheless, one would like to have a sensor that can accurately and quickly measure over a wide dynamic range. The solution: a combination of the two technologies. In hybrid detectors (HyD), the advantages of vacuum technology are combined with the advantages of semiconductor technology.

A hybrid detector [6] uses a photocathode to convert the photons into accelerated electrons, as in a photomultiplier tube. The best choice for the cathode material is GaAsP, which has a quantum efficiency of up to 45 % at 500 nm (corresponding to about twice the quantum efficiency of multi-alkali cathodes). Due to the technical implementation, a GaAsP-HyD is much more stable than a GaAsP-PMT, which is often the victim of light-induced damage. Hybrid detectors are much less damageable. Photoelectrons are accelerated in a single step by a very high voltage of about 8 kV. The kinetic energy is absorbed in a semiconductor material. Many charge pairs are generated by impact ionization in that single shot. The gain is about 1,500-fold (by way of comparison: approximately 3-fold per step in PMTs). However, 1,500 electrons are not easily measured. Therefore hybrid detectors contain an integrated amplifier: the charges generated by impact ionization are multiplied by an avalanche effect, as was described for the avalanche photodiodes, offering an additional gain of approximately 100-fold. In this combination, a measurable signal is produced in non-Geiger mode without noisy circuitry. The active region of a hybrid detector is smaller than that of a PMT, but larger than that of a standard avalanche photodiode. Thermal noise is therefore limited, but the light is still focusable onto the target without any loss.

The single-step acceleration creates a 500-fold higher gain in the first stage as compared to a PMT. If gain is 3-fold in a photomultiplier, there is a large probability that only 2 electrons or 4 electrons will be released. This is a variation of 33 %. The signal-to-noise ratio of peak heights is then 1.7. The first stage of amplification in a hybrid detector is 1500-fold. The signal-to-noise ratio in this case is 37 – that's a huge improvement in S/N in the first step. In PMTs, the voltage between dynodes is approximately 100 times less per step compared to HyDs, and the probability for the loss of electrons moving from cathode to anode through the dynodes comparatively high. As a result, the photon efficiency in PMTs is about 10 % less than the quantum yield. In hybrid detectors, however, the photon efficiency is very close to the quantum yield of the cathode, as there is only one step at high voltage. A second advantage of single-stage acceleration is the fact that the arrival times of the electrons show less scatter. It is only one single electron anyway. The pulse width in a PMT is strongly influenced by the different geometry of the possible movement of electrons across the dynode cascade. Consequently, the pulse width in a hybrid detector is much shorter. This allows higher frequencies for photon counting - which translates directly into higher dynamic range. The higher cut-off frequency also reduces the instrument time constant, a limiting factor for measurements of the fluorescence lifetime. Due to the relatively small active area, HyDs show a lower dark current than PMTs. Dark current causes background noise, non-zero intensity in areas that contain no signal. If the signal is very small – the usual case in biomedical imaging, background noise and signal will overlap. Lower dark current means better S/N in critical situations. 

Comparison of the sensors

All sensor types have advantages and limitations. Here, the important parameters for single point detection are compared. The figures are typical values. There is a wide variance, and specialized equipment can be very different in both directions. Our aim is not to compare individual brands or types, but to give an idea of typical performance. The figures are consistent with the view of many colleagues, but are far from being scientifically robust.

Quantum efficiency (QE) describes the ratio of incident photons and photoelectrons generated. The QE depends on the wavelength, and data for comparison refer to 500 nm (650 nm for red application). Multi-alkali cathodes have quantum efficiencies of about 25 %, GaAsP cathodes show efficiencies of nearly 50 %. Similar efficiencies are achieved by silicon avalanche photodiodes in the absorber.

Not all primary and secondary electrons will finally reach the anode. In photomultiplier tubes, the loss is primarily caused by losses of electrons between the dynodes. The quantum efficiency in PMTs is typically reduced by a further 10 % due to these losses. It is therefore important to compare the ratio between photons illuminating the active element and the signal at the anode. This is called photon detection efficiency, PDE.

The gain in the first step of a photomultiplier determines significantly the uniformity of the corresponding pulse at the anode. Pulses of uniform height are easily separated in photon-counting mode and reduce noise in analog mode.

For photon counting applications the pulse width is critical. Therefore, the circuitry must be adjusted very accurately to the required signal. Also the intrinsic variation in the arrival times of the charges at the anode will broaden the pulse. In PMTs this broadening is mainly caused by the different trajectories of the electrons. Therefore, PMTs deliver relatively wide pulses.

In order to detect a signal, sufficient contrast between signal and background is necessary. The active area size and the temperature increase the dark current. It is obvious that the increasing sensitivity in the red region also increases thermal noise. In some cases, the sensors must be cooled in order to reduce the background noise.

APDs have a very small, PMTs a relatively large active area. Hybrid detectors range in between. Although a small area is advantageous, areas that are too small are difficult to focus with sufficient long term stability. This may require additional optical elements, which reduces the efficiency

At moderate intensity, PMTs have a high dynamic range. This is the reason for their success in confocal imaging (and other areas). In photon-counting mode, PMTs allow only very small maximum count rates and are thus limited to very faint light. Geiger-mode APDs are ideal for rare single-photon events, but lack dynamics, because the sensor needs a recovery time after each pulse. Hybrid detectors combine good resolution for single-photon counting and high dynamic range with large maximum count rates, so that they can be used in the entire range of intensities that typically occur in fluorescence images.

Because of the (different) internal effect, both photomultiplier and avalanche photodiodes sometimes show a second pulse after the "original" pulse. This is not a major problem in intensity measurements, but it causes significant errors in the measurements of fluorescence correlation spectroscopy. These types of applications therefore require sensors that have little afterpulsing.

The following table summarizes the typical range of parameters discussed above for the main types of sensors used in confocal microscopy and derivative applications:

Table 1: Overview of the various parameters that control the applicability of sensors. Keep in mind that these numbers are very coarsely integrated typical values which may be very different in a given sensor type. The snowflake indicates that the value is true for cooled devices. PDE: photon detection efficiency; Dark #/s: dark counts per second; max M#/s: maximum megacounts per second.
PDE (500 nm) % 25 37 45 45
Pulse noise % 60 % 60 % 5 % 3 %
Pulse width/ns > 10 > 10 0.1 ( ) 1
Dark #/s 15,000 15,000 300 (*) 2,500
max M #/s very low very low 20 150
Area/mm2 50 50 0.05 8
Afterpulsing high high medium low

Typical applications and best fit of sensor type

Which is the best sensor for a given application? This is a relatively easy question. However, one would like to use a device that is suitable for many applications. Below, a number of applications are described together with the performance of detectors for the particular application.

  • Every-day imaging: Standard samples emit in the range between 450 nm and 600 nm. Typical examples are not too dark and relatively stable. APDs are not suitable for these samples because of their limited dynamic range. GaAsP PMTs are expensive and can be easily destroyed. A good solution is a standard multi-alkali PMT, which is quite "unbreakable" and offers a very wide dynamic range. The high dark noise may cause a problem. The best solution is a hybrid detector, which has an excellent low dark current and sufficient dynamic range, and is as robust as PMTs. Only when the light intensity from the sample is very high (reflected light mode or very strong fluorescence) do PMTs allow a higher signal and thus provide a better S/N.

  • Sunday imaging: For very weak samples or very bright and very faint structures in the same sample, the sensor of choice is clearly a HyD. The same is true for the short exposure times that are particularly important for living specimens. Such samples must be subjected to as little exposure as possible to avoid phototoxic damage.

  • NDD imaging: Since two-photon excitation creates intrinsically optical sections, emission may be detected directly behind the objective (non-descanning detection). In this case, the beam is not at rest but has a residual angular movement. Here, sensors with very small active areas require considerable optical design efforts. PMT and hybrid detectors will be the better solution. The hybrid detector is best for its low noise and high photon detection efficiency.

  • Single-molecule detection (red emission): For single-molecule detection APDs were introduced into the microscope. Due to their high quantum yield, APDs are an especially good solution for red-emitting dyes. However, hybrid detectors are also well suited here, although their QE decreases with longer wavelength. PMTs are less suitable for single-photon measurements.

  • FLIM: Fluorescence lifetime imaging using time-correlated single photon counting requires very fast sensors and small fluctuations in the transit time (low transit time spread). Therefore, hybrid detectors are most suitable. Avalanche photodiodes also provide good results, although Geiger-mode devices require long measurement times due to their low count rates. Standard photomultiplier tubes are not applicable in general, but there are special tubes that are fast enough for FLIM, if the expected lifetime is not too short. These PMTs are also offered in commercially available systems.

  • FCS: The most important prerequisite for fluorescence correlation spectroscopy is high sensitivity and low dark current. A problem for FCS is the afterpulsing phenomenon, which simulates photon correlations. For this reason, a sensor with low afterpulsing is preferable – the hybrid detector is the favorite here, too. Because of their high efficiency, APDs are the most commonly used sensors for FCS – at least so far.

  • FCCS: Fluorescence cross-correlation spectroscopy is a special type of correlation measurement. Here, two different dyes are observed, and the correlation between the two fluorescent species is calculated. Afterpulsing events are intrinsically eliminated. Avalanche photodiodes are the most appropriate choice, although hybrid detectors are also very suitable and allow  a versatile system concept that can do more than just FCCS.

The above discussion is summarized in the table below. For each type of application, each sensor type is placed in a hierarchy that describes the applicability of this sensor for this type of experiment. Higher scores mean better performance. Hybrid detectors appear to outperform other sensors in most cases, indicated by the maximum points (average 3.7). In fact, this sensor is ideal for a wide range of applications and is used for both imaging and single molecule measurements. Avalanche photodiodes are still a good solution for single-molecule and single-photon measurements – especially for far red emitting dyes. A PMT is well suited for very bright samples. Therefore, in commercial devices at least one channel is equipped with a PMT. In budget-critical situations, the PMT is the obvious choice, as other solutions are some 5 times more expensive.

Table 2: Hall of fame for four different types of sensors for single point detectors in scanning type microscope systems (including still FCS).
For a wide range of applications, the hybrid detector (HyD) is the favorite. HyDs are implemented in Leica TCS SP8 microscopes and single molecule detection systems.
 Points 1.4 2.3 2.6 3.7
Everyday imaging 3 2 1 4
Sunday imaging 1 3 2 4
NDD imaging 2 3 1 4
sm Red imaging 1 2 4 3
FLIM 1 2 3 4
FCS 1 2 4 3
FCCS 1 2 4 3


I would like to thank Lioba Kuschel and Constantin Kappel for exciting discussions and useful advice.


  1. Hertz H: Ueber einen Einfluss des ultravioletten Lichtes auf die electrische Entladung. In: Annalen der Physik und Chemie. 267: 8 (1887) 983–1000. ISSN 00033804. doi:10.1002/andp.18872670827
  2. Einstein A: Ueber einen die Erzeugung und Verwandlung des Lichtes betreffenden heuristischen Gesichtspunkt. In: Annalen der Physik 322: 6 (1905) 132–148. ISSN 00033804. doi:10.1002/andp.19053220607
  3. Photomultiplier tubes – Basics and applications Hamamatsu Photonics K.K (2006): (ca. 3 MB)
  4. Handbook of Biological Confocal Microscopy. Ed.: J. Pawley. 3rd edition. Springer US (2006)
  5. Hamamatsu Photonics K.K. Characteristics and use of Si APD (Avalanche Photodiode), 2004
  6. Suyama M and Lares M: Photomultipliers: Hybrid detector combines PMT and semiconductor-diode technologies. Laser Focus World 2008.


Help us to improve the Leica Science Lab and participate in our survey! Start Survey

Arrow Down

Arrow Up