Knowing your multiplication table

Share this on social media:


Greg Blackman looks at how electron multiplying technology is advancing spectroscopy

Raman scattering is well known to be a particularly weak effect, which makes it difficult to measure, especially over short time periods. Electron multiplying CCD technology has changed that to a certain extent, and is opening up areas of research that otherwise wouldn’t be possible. Now entire chemical Raman maps can be made of a sample by scanning it at a microscopic level to determine its chemical composition, aided by the ability of EMCCDs to detect very small signals. Pharmaceuticals can be probed as to their chemical makeup over the surface of a tablet, while, in the life sciences, chemical interaction in cells can be studied with the confocal Raman mapping technique.

German company WITec produces confocal Raman microscopes that capture up to 10,000 spectra to map the chemical composition of a sample. ‘In the past, people used to have to integrate over minutes or even hours to get a complete Raman spectrum,’ observes Dr Olaf Hollricher, R&D director at WITec. ‘With this Raman imaging, 10,000 spectra can be collected, so integration times of even one second per spectrum are not feasible.’ With so many spectra, extremely short integration times are required, which is where EMCCDs come in.

Electron multiplying CCD technology is designed for high sensitivity; Dr Gerald Cairns, time resolved and spectroscopy applications specialist at Andor Technology, says the technology could even be thought of as being able to measure single photons. EMCDD cameras, such as those produced by Andor, therefore offer great benefits for measuring low-level spectroscopic signals, down to 10 photons per pixel potentially, which isn’t possible with conventional CCDs.

Imaging microscopic objects like nanostructures or cells, for instance, involves intrinsically low-level spectroscopic signals and therefore benefits from EMCCDs. Also, EMCCDs are useful in areas where the excitation light source has to be kept low to preserve the lifetime of the sample, according to Dr Cairns. Lower excitation energies drives down the signal.

A third scenario is measuring on a fast timescale, for example recording a temporal profile of an event that’s changing over time. Capturing a series of spectra as quickly as possible, as in the case of confocal Raman mapping, equates to short exposure times in the orders of tens of microseconds or even shorter. Short integration times means there is not much signal per spectra and there is also more readout noise because the signal is being read out very fast. Here, the EMCCD process is useful to get a readable signal.

Minimising read noise

One of the ways to increase the efficiency of Raman spectroscopy would be to increase the laser power used to excite the sample. But there is a limit to this, explains Dr Hollricher, typically a few milliwatts for visible light excitation. ‘Focused to a 300nm spot size, this [milliwatts] equates to megawatts per square centimetre of power. You can’t increase power anymore without damaging the sample.’

High-quality optics are required to gather as much of the scattered light as possible, but the efficiency of the detector also needs to be taken into consideration, which is why the ultrasensitive EM technology is used.

Signal-to-noise ratio is the important factor with CCD detectors, not necessarily the level of signal. The main sources of noise for CCD detectors are thermal noise or dark current and readout noise, the latter being the noise from the electronics of the sensor. Shot noise, spurious noise and cosmic rays passing through the sensor are other sources of noise.

Andor uses thermoelectric cooling on its sensors to minimise dark current; EM technology comes in to minimise the readout noise. The technology multiplies the raw signal measured on the sensor before it is read out. The benefit of this is a much larger signal going into the readout electronics, which effectively overcomes the readout noise.

Amplification is through a process called impact ionisation, which accelerates the electrons as they’re transferred from one pixel to the next causing them to release secondary electrons. The gain from each transfer from pixel to pixel is small, but with a chain of several hundred pixels there is a multiplication effect moving from one element in the readout register to the next. EMCCDs will have signal gains of up to 1,000 times.

‘Reading out a standard CCD quickly can result in a noise level of 30 electrons. Therefore you need 30 photons hitting the sensor to get a detectable signal,’ explains Dr Hollricher. ‘In an EMCCD camera, five photons hitting the sensor, for instance, amplified by 100 equates to 500 electrons, which is a much better signal-to-noise ratio. With an EMCCD, the readout noise is no longer the limiting factor – the signal is virtually shot noise-limited. You want a detector that can detect single photons – in principle, an EMCCD is single-photon sensitive.’

There are some disadvantages with this – EM technology is only beneficial with low-level signals, as the amplification has a small amount of noise associated with it. ‘If you’re dealing with reasonably moderate signals in terms of hundreds of photons upwards then EM technology won’t provide really any benefit to you,’ states Dr Cairns. ‘Indeed if you’re dealing with stronger signals, a conventional CCD camera will perform every bit as well if not better [than an EMCCD].’

However, EMCCD cameras can be used in a normal mode with medium levels of signal and can then be switched to EM mode when measuring low signals, notes Dr Hollricher. It’s better to use low readout frequency when running experiments where there is plenty of time to capture Raman spectra, he says, and wait until there is enough signal to read the camera out. If time is limited though, then an EMCCD is useful.

‘With 10,000 spectra, there is only time for a millisecond per spectra integration time,’ Dr Hollricher says. ‘Every experiment that needs short integration times, in the order of a millisecond, will require an EMCCD – the camera we use can achieve 0.7ms integration time, which gives 1,300 spectra per second.’


Other areas where EMCCDs are used include hyperspectral mapping, which, again, captures large volumes of data and therefore requires short exposure times. Laser-induced breakdown spectroscopy (LIBS) is another spectroscopic technique potentially requiring EM technology. It is similar to Raman in that it uses a laser to probe the sample, and also like Raman it can require a sensitive detector to measure the signal. Raptor Photonics, based in Northern Ireland, has provided its Falcon EMCCD camera for LIBS instrumentation, among other spectroscopy methods.

Surface enhanced Raman scattering (SERS) and tip-enhanced Raman scattering (TERS) are other areas where EMCCDs are used. TERS operates by exciting the sample at the location of a probe tip, which is around 20nm in diameter. A Raman map is built up by capturing spectra on this small scale. The sample area is small and the signals are weak, which means EM technology is necessary.

Andor’s iXon EM cameras have been used in research into magneto-photoluminescence looking at silicon nanocrystals. At the nanoscale, the behaviour of silicon is different and researchers are investigating the potential to build photonic devices from silicon using nanotechnology. If these sorts of photonic LEDs or nanolasers could be fabricated from silicon, the technology could be integrated into semiconductor chips.

‘Not only were they [the researchers] looking at nanostructures, but also the fact it required thousands of actual measurements to build up the final signal – they had an average over thousands of measurements,’ explains Dr Cairns. ‘The EM gain on the camera was a key factor for them to be able to make the measurements.’

Andor has recently released a new EM sensor. One of the challenges when working with Raman spectroscopy is the signal being masked by autofluorescence from the sample. Imaging in the near-infrared reduces the autofluorescence and Andor supplies a family of EM sensors that are optimised for the NIR region (700 to 1,050nm). Most recently, Andor has incorporated fringe suppression into these sensors. Fringe suppression has been around for some time on conventional CCDs, but combining it with EM technology is a recent development, says Dr Cairns. These NIR-sensitive EMCCDs are ideal for Raman spectroscopy, where measuring in the NIR region is beneficial.

EM technology is also used in many other imaging applications, including astronomy and other scientific imaging dealing with low signal levels. In terms of spectroscopy, being able to produce a chemical map of a sample gives scientists more data to analyse all of which is aided by EM cameras.

About the author

Greg Blackman is the editor for Electro Optics, Imaging & Machine Vision Europe, and Laser Systems Europe.

You can contact him at or on +44 (0) 1223 275 472.


Find us on Twitter at @ElectroOptics, @IMVEurope, and @LaserSystemsMag.