# Applied image sensors – EMCCDs and yeast cells

Finally, after a long break I am back!

We had exceptionally good weather here last weekend and I spent some time in the university parks with a friend who studies at the biochemistry department. Throughout our walk I had the chance to learn a bit more about the DNA replication mechanisms in eukaryote cells and the many things that can go wrong during cell replication. Our discussion gradually hopped over to commenting on the methods of imaging live organisms and eventually the imaging sensors used in fluorescence microscopy.

Anyway, I need to first say a few words about the essence of their problem before hopping over to the electronics part. DNA is a form of amino acid which is, as we all know, unique and contained in each live cell. Apparently it is also a “single molecule” in terms of quantity in each cell i.e. we can’t “suck” a bit of DNA from one cell and inject it into another. If we want to produce more DNA, or cells, we need to split the amino acid into two strings of RNA. The latter we then combine together in the new cell thanks to a comb-shaped (cell element, forgot its crazy name?) which does the magic. This process is very sensitive to external factors (as an ordinary person I imagine poison), especially sensitive towards the end of the process for which very little is known about. If we have a way to snapshot this process in extreme detail, only then we can fully understand the replication mechanism. But let’s not drift further as all above is extremely vague from a biologist’s eye and get back to the topic.

Apparently biochemists know very very little about these processes and all their experiments seem to be based solely on the “trial and error” methodology. As by today we do not have technology for 3D imaging of individual cells, biochemists are forced to use what is called fluorescence microscopy. This also forms the essence of this post – what are EMCCDs and how are they used in imaging of live organisms?

Before taking this question let me explain why fluorescence microscopy is used in biology and why other microscopy technologies are not suitable for cell imaging purposes. Imagine taking a live cell to a Scanning Electron Microscope (SEM), no, even better, imagine if you were the cell to be scanned. So you are closed inside a vacuum chamber and a gun is pointed towards you whilst it is also constantly shooting electrons with over a few keV energy. You would have a hard time getting out of the vacuum chamber alive, not to mention that the biologists want to see you replicating while they are scanning (shooting) you. It simply is not possible. Cells which have otherwise been alive die after the very first electrons piercing through the cell’s membrane. Scanning Tunneling Microscopy is also impossible, as the latter focuses on atomic scale and this time current is passed through the sample. Our cells won’t feel very good, even if it was possible to adjust the STM needle with such a precision. The passing current would kill the cells immediately. A technique which used for increasing the resolution of otherwise “standard” microscopes is the so called fluorescence microscopy technique.

Fluorescence microscopy extends our ability to image the “internals” of live cells. If a fluorescing marker is injected into our specimen, by exposing it to UV light we can observe a different wavelength emission by the former. We can therefore focus our imaging on very specific “internals” of the cells. Ordinary microscopy techniques focus on the specimen’s surface imaging and/or light absorption. The fluorescence method however, did not make any sense until recent years, as the relatively low quantum efficiency of the fluorescing markers make it difficult for the sample to be imaged. Not to mention that a naked eye is nearly not as sensitive as the sensor needed here. With the emergence of more light sensitive image sensors in the recent years, fluorescence microscopy is now a widely spread technique amongst scientists.

So what are these sensors, that have an up-to single photon sensitivity? Most fluorescence microscopes are more likely to have Electron-Multiplying Charge-Coupled-Device (EMCCD) sensors, however some systems employing Time-Delay-Integration (TDI) sensors can also be found.

EMCCDs do not differ dramatically compared to conventional CCD imagers, however what makes a difference between the two is the structure of the output register. We know that in general the readout noise from interline transfer CCDs can be very low, or less than ~3 to ~10 \$e^{-}\$. In order to decrease the relative noise floor added by the output amplifier to the signal, EMCCDs use gain registers to boost the amount of electrons (signal) to be read out. Thus the same magnitude of readout noise (3-10 \$e^{-}\$) as in regular CCDs is superimposed on a much larger charge count, the one achieved thanks to the gain register. Let’s have an overview of the charge transfer process occurring in CCDs.

CCD charge transfer principle diagram

Phase 1 and 2 are slightly overlapping and applied to two adjacent electrodes. As the electrostatic field is gradually changing from one electrode to another the trapped electrons in the wells move along the direction of the changing field. For more details I suggest reading more about CCDs. During the charge transfer process there is a slight chance that only a few electrons escape (sometimes in high-performance sensors even ~1 e-).

This is all great, but when we try to take an image at very low light level conditions, the output amplifier should practically read one, or only a few electrons. All amplifiers induce noise to the signal, thus a single electron would be hidden inside the dominating electronic noise by the amplifier. A way to solve this problem “naturally” is to use a multiplying CCD register. Here is a crude sketch of a CCD register with an emphasis on the number of charge to be read out at low light levels:

CCD readout with an emphasis of electronic noise addition from the output amplifier at low light levels

The signal is masked over the amplifier electronic noise. Here is a sketch of an EMCCD:

Principle diagram of an EMCCD

The phenomenon is self-explanatory, but how do we make a multiplying register? Here is a principle diagram:

Charge multiplying impact ionization register

Just as the regular output register the same structure is used, however now a few clock phases are introduced and the electric field applied over the electrodes is much stronger. The strong electric field causes accelerated electrons to hit other electrons in the empty register cells exciting a new hole-electron pair. The process continues and as the charge is shifted through, additional electrons are generated, and thus the name “multiplying” register. The gain of a single register is small, but with the increase of stages ionization probability is increased with $P(ioniz) = 1+P(cell)^{N}$. Where N is the number of stages and P(cell) the impact ionization probability per cell. The exact multiplication value is hard to be defined as the number of incoming electrons and the strength of the electric field cause very stochastic levels of impact ionization. Nevertheless, this amplification method is “noiseless” as compared to the regular charge-voltage conversion  and amplification in the output amplifier. The overall SNR of the system is dramatically increased.

Focusing on the microscopy technique, now having a single/few photon sensitivity sensor, we can go back to the yeast cells and image them alive. Here is a resulting image:

Live yeast cell image using fluorescence microscopy

Why yeast cells? I was told that these are one of the largest eukaryote cells and have low light absorption though their membrane. Here it is how I saw them otherwise:

Prepared cell samples

And the microscope used can be seen below. I could not detach the sensor due to obvious reasons, but it really is a regular microscope with some electronic eye “features”.

Fluorescence microscope

To wrap-up, I am extremely happy to see a real example of two sciences with a strong bond in-between. Deeply involved in our fields it is often that we forget what our research is all about.