Most electron microscopes have a fluorescent screen at the bottom of a column. And it's positioned just below a viewing screen of thick leaded glass that allows you to look through and see, in a live fashion, where the electrons are hitting. As the electrons hit this fluorescent screen, green photons are emitted and you can see the image. In addition, many are outfitted with TV cameras. So the projector deflectors deflect the image onto a TV camera that's often mounted to the side of the fluorescent screen and this can show a live image that's being produced. Nearly all electron microscopes then, have film canisters that reside below the fluorescent screen. So if you lift up the fluorescent screen, it can expose a piece of film. After the whole cassette of films is recorded, you then seal off that chamber, take the films out and develop them, and then process them. In cryo-electron microscopy, the films then need to be digitized with a scanner so that you have a digital image that you can work with in the computer. Now the advantage of photographic film, is it's, it's commonly, six, approximately six by nine centimeters. And it can be digitized as finely as, for instance, maybe, six micron pixels. In which case, you can have 10,000 pixels by 15,000 pixels in a film image. So film allows you to record very large images. And they're also high resolution because the interaction in the electrons with the grains in the film are very localized. There is also a dose regime where film has a very nice, linear response to the dose. The disadvantages of film are mostly that it's time consuming to work with. First of all, it takes time to load the films into the microscope. They always have water absorbed within the films. And some of this water escapes into the column, leading to higher contamination rates. Then, once the films are recorded, one has to develop them and then digitize them before calculations can be done. And finally, this response time is so slow that you can't record an image, do a calculation, and then adjust something on the microscope before you take the next picture. Instead, you record all of your pictures and then go digitize your film. So for that reasons, most all of cryo-EM is done today using CCD cameras. Now, the anatomy of a CCD camera or a CCD detector is that the top layer is a scintillator. So incoming electrons are coming down through the column and they're hitting the scintillator. Now in the scintillator material, the electrons are scattered, producing photons. And the photons, you know, emerge in all directions. But as the photons hit the next layer, which is a fiber optic bundle, they're captured within the fiber optics and brought down toward specific pixels of a CCD layer. And here the photons are converted into counts and stored there until the end of the image. After the given exposure time, the CCD camera is then read out through a series of out, output amplifiers. Basically meaning that the charge that was collected in each pixel is counted as it, and as the circuitry goes through the pixels one by one, it just counts how many charges are there. And you get a digital image rather quickly in a matter of seconds. Now the main disadvantage though, of CCD cameras is that, as the electron comes down through the column and hits the scintillator, it's scattered in, it can be scattered quite a distance away from the position that it first entered. This is a figure from a paper from Kenneth Downing and his co-authors in Ultramicroscopy where they calculated some example trajectories that electrons might take as they traverse a scintillator. And then pass through the scintillator into the fiber optic bundle. And as you can see, there's a chance that the electron will travel very different paths. And as it's traveling all these potential paths, photons will be produced each time the electron changes its course through a scattering event. And so you see, you get a number of photons that are going in a large area of the fiber optic bundle. And therefore one incident electron will result in counts on the CCD detector in a number of different pixels down here on the CCD detector. There, it, the spot will look large. In addition, sometimes the incoming electron is scattered not only in this vicinity but it be, can scatter and actually come back up through the scintillator in a completely unrelated position. And as it passes through the scintillator, more photons will be produced along its path there as well. And as a result, you'll get two spots of counts on your CCDD, CCD detector in two completely different positions all from one incident electron. And this contributes to noise in the images. Another major problem with CCD detectors is that some of these electrons, as they scatter in the scintillator, will produce lots and lots of photons. And as a result on the CCD chip, there will be a very bright spot. Let's say, some of them may produce 40 counts, for instance. Meanwhile, another electron might pass through pretty rapidly, with only a few scattering events. And because of that, it'll produce far fewer photons. And so it produces a spot that's much less bright. For instance, maybe it'll produce a spot with only ten counts. And as a result, we build up an image where each electron has dramatically different weight, and so this adds substantial shot noise to the image. And so, in a CCD camera the incident electrons first hit a scintillator where the electrons are converted to light. Then they pass through a fiber optic bundle to transfer those photons to a CCD camera where the light is converted to charges which are then read-out. And because of the problems we just described, there have now been developed a different class of detectors. The so called direct detectors. Which are fundamentally different, in that the incident electrons impinge directly on a CMOS detector and are directly converted into charge. Now this has a number of advantages. One is, that because the electrons are directly detected, the spread of their effect, in other words, the counts they produce within this chip are more localized. A smaller point spread function. In addition, especially if they're back thinned, there's a reduced chance of the electron being scattered through materials below it and back through the chip in some unrelated position. So this process is less likely in a direct detector. Finally, the CMOS detectors can be read-out hundreds of times a second. Now the advantage of being able to read-out the image at hundreds of times a second is illustrated well in this movie, which was recorded by the Catan Company. And what you see is the live feed from a camera, well it's slowed down significantly, as electrons are hitting the camera. And as you can see, each electron when it hits, produce a little burst of counts. And some of the bursts are bright, and some of the bursts are less bright because of a stochastic nature of the interaction of the electrons with the material of the CMOS chip. Nevertheless, because we're reading-out the camera in between electron hits, we can interpret each burst of counts as having come from the one electron. So whether it's 40 counts or whether it's 10 counts, each time we'll understand it as one electron hit. And so as we build up an image, each electron carries the same weight, and this dramatically reduces the shot noise in the images. So this is a single 2.5 millisecond frame that was recorded using a conventional CCD-style charge read-out. And as you can see, each electron hit produces a little burst of different numbers of counts. And in a traditional CCD camera, these would all be added up to form the image. A direct detector, at least in its electron counting mode, can interpret each of the bursts as a single hit. And so they all have the same weight in the final image. In addition, it's been found recently that during the typically one or two second exposure that is commonly used in electron cryomicroscopy, samples that are frozen in vitreous ice will sometimes shift position during that one second exposure. And the ability to read-out the camera at hundreds of frames per second, allows one to track that movement and it's been called deblurring, deblur the image by shifting different frames to produce a sharper net image. While being able to read-out hundreds of frames per second brings great advantages, it also significantly increases the amount of data that's recorded. And so part of the challenge of using a direct detector is establishing the required disk space and computing power to process these images.
