Which Sensor is the Best for Confocal Imaging?

A true confocal scanning microscope focuses the light for illumination into a single spot. In order to generate a two-dimensional image, this spot must be scanned in x and y directions over the sample. Although the scanning machinery requires some efforts in design and manufacturing, the sensor can be a single device, as compared to widefield camera chips, that must provide as many sensor elements in parallel with picture elements (pixels) that are requested. As no xy-readout electronics are needed, single-spot sensors feature better signal to noise ratio as compared to array-detectors.

Photomultiplier tubes (PMT)

Probably the best-known sensor for confocal imaging so far, is the classical PMT (Fig 01), which started its career more than 80 years ago in the early 1930’s[1]. It is based on the photoelectric effect, first described by H. Hertz[2] and interpreted by A. Einstein[3]. A photo cathode, usually a layer of alkali-atoms that can release electrons readily upon absorption of a photon, is exposed to the source of radiation (here, the light is focused onto the photocathode). Best performance in terms of quantum efficiency in the visible range is achieved with Gallium arsenide phosphide cathodes (GaAsP). A voltage accelerates the photoelectron to the first dynode, where secondary electrons are released. The gain at the dynodes is typically limited to less than fivefold, due to the properties of the dynode material. The electric pulse generated by a single dynode could not be distinguished from the electrical noise. To achieve sufficient gain, the secondary electrons are accelerated to the next dynode and so forth[4]. This procedure allows gains up to circa 108 or even more, which is sufficient to pick up a measurable electrical current.

Photomultiplier tubes feature a very wide dynamic range, i.e. they can be used for very low photon densities and comparably high intensities.

Avalanche photodiodes (APD)

A semiconductor counterpart to the vacuum PMT was brought to the market with the invention of an APD, which is a derivative of the PIN diode. A PIN diode features an additional sparsely doped layer between the p-type and n-type layers. The APD is a modified PIN diode, complemented by a multiplication layer (Fig 02). If a photon is absorbed in the intrinsic layer, it will generate a charge pair (inner photoelectric effect), the charge is then accelerated by the applied voltage. The strong electrical field in the multiplication zone (p-n, in Fig 02) causes an amplification by releasing many (up to circa 1000) additional charges. The output signal is proportional to the number of photons absorbed, therefore, the device can be used as a light sensor over a certain range of intensity. At higher voltages, the absorption of a photon causes a breakdown accompanied by very high gain (up to 108, “Geiger mode”), which is typically used for single-photon detection. As a continuous breakdown would destroy the device, this mode is only suitable for sufficiently low light intensities.

APDs have been successfully employed for single photon measurements, especially in FLIM (Fluorescence Lifetime Imaging) and FCS (Fluorescence Correlation Spectroscopy) experiments. However, they are rarely used for confocal image recording.

Hybrid detector (HyD, HPD)

The PMT’s main limitation is the low gain at each multiplication step. The APD’s main limitation is its range of operation: only very low intensities (in Geiger mode). A hybrid combination of the two technologies offers a solution that fortunately shows the beneficial features of both. This combination of vacuum and semiconductor technology was originally developed for experiments in particle physics [5].

Pulse height distribution

Ideally, the output pulse for any photoelectron released from the cathode would look exactly the same. Although, practically speaking, this assumption is far from real, especially for PMTs as we will see below. Not only is the pulse width  and the time between photon absorption and output pulse peak (called “transit time”) varying, but also the height of the pulses. Pulse-height variations directly cause variations (noise) of signals that are recorded by integrating the anode current. In the case of photon counting, variations in pulse height may cause count losses if the pulse cannot be distinguished from the electronic noise of the amplification circuits (see Tab 01 and Fig 4, event #2 in the PMT case is equivalent to events #3 and #4 in the HyD case).

The height of the output pulse is essentially ruled by the gain of the first amplification step[8]. For a PMT, that is the gain at the first dynode. As mentioned above, that gain is limited by the dynode material and falls somewhere in the range of 2 to 8. The variation is governed by Poisson statistics, e.g., if we assume an average gain of 4, the variation is +/- 2, meaning there is a good chance to get from 2 to 6 electrons in the next event. The pulse height is then varying by roughly a factor of three. The gain of the later dynodes has the same statistical behavior, but as the number of electrons is growing from dynode to dynode, the variations are averaged and contribute less significantly to the noise.

Hybrid detectors (HyDs) show very low noise compared to photomultiplier tubes (PMTs), as shown in the first line of Tab 01. The dark noise level is comparable to that of an avalanche photodiode (APD). Also, the pulses are very uniform, which make them ideal devices for photon counting – similar to APDs, as seen in line two of Tab 01. The differentiator between PMTs and APDs is the very short pulse width with no delay time (line three in Tab 01). That difference allows HyDs to be used for the detection of intensities typically found with fluorescence samples, whereas an APD would be long since saturated. The HyD produces crisp fluorescence images with a truly black background, both for low intensity situations (like high speed live imaging) and standard fluorescence samples.