Focused Ion Beam Scanning Electron Microscopy (FIB-SEM) at the AIC

Introduction #

Starting in August 2019, the Advanced Imaging Center (AIC) introduced its first pre-commercial electron microscope. This system is based on work done in Janelia’s Hess lab ¹ and makes use of a powerful technique: Focused Ion Beam Scanning Electron Microscopy (FIB-SEM). As its name suggests, FIB-SEM combines two methodologies into a single, powerful approach for imaging biological ultrastructures. This is the first in a series of blog articles that introduces this unique methodology, and the various considerations that should be made, to prospective AIC users.

Scanning electron microscopy (SEM) itself is a mature technology ², with commercial implementation going back decades. It functions via a focused electron beam that is scanned across a sample surface to form an image, either by detecting backscattered, or secondary electrons at each point along the beam’s path. Modern SEM systems can achieve electron beam spot sizes of less than 2nm, providing excellent resolution in suitably prepared samples. But while SEM is commonly used to image many types of specimens, it cannot intrinsically generate 3D information throughout the volume of a biological sample.

Focused Ion Beam (FIB) technology has existed since the 1970s and has traditionally been used as a materials characterization and preparation technique. Similar to SEM, FIB uses electromagnetic fields to focus a particle beam onto a sample. But FIB uses heavy atomic weight ions (usually Gallium), rather than electrons. So, while it can be used to form an image much like SEM, it can also be used to manipulate materials by selectively ablating (or “milling”) away thin layers from a sample surface, revealing new material underneath ³.

The combination of FIB and SEM opens the door to a uniquely powerful method of 3D electron microscopy ^{1 4}. FIB-SEM works by alternating FIB milling and SEM imaging cycles, as shown in Figure 1 below. In the FIB milling phase, a heavy-ion beam ablates away a small layer from a sample surface with exquisite delicacy, only removing ca. 2-4nm of thickness per cycle. In the SEM phase, the electron beam is scanned across the sample to form an image of the sample surface. The process repeated many times to construct an image stack of potentially many thousands of slices. Crucially, the fine-milling capability of FIB matches well with the resolving power of SEM. Thus, isotropic resolution (i.e. equal resolution in all three dimensions) becomes possible with FIB-SEM. Because of this, 3D data can be visualized with equal clarity in all directions, making FIB-SEM uniquely advantageous for discerning complex biological structures.

Advantages of FIB-SEM at the AIC #

Practically speaking, a number of issues can conspire to limit the utility of commercial FIB-SEM. To see why, we first need to consider the following. To take advantage of the isotropic resolution that FIB-SEM can offer, it’s imperative that the scanned electron beam does not penetrate too deeply into the sample. Otherwise, signal can be generated from 10s or even 100s of nanometers inside the sample – thus obviating the ultra-thin (ca. 4nm) milling capability of the FIB. So, the electron beam energy needs to be carefully controlled so only the top few nanometers of the sample are probed. That, however, results in a relatively limited amount of signal that can be detected at any given time. As a result, SEM imaging needs to proceed relatively slowly in order to accumulate sufficient signal to give good quality images. Typically, that forces a single image “slice” in FIB-SEM to take several seconds to minutes to acquire. Multiplied over many thousands of images, a moderately sized 3D volume acquisition can take many days.

Thus, a primary difficulty in using FIB-SEM for bioscience research stems from the requirement that the instrument exhibit a very high level of stability over long time periods. Simply put, commercial FIB-SEM systems cannot image continuously for long enough to capture more than a few microns of imaging depth before errors begin to accumulate. As a result, Harald Hess’ lab at Janelia has engineered a FIB-SEM system that can image continuously for weeks or months at a time. To accomplish this feat, a number of novel hardware and software improvements have been implemented.

FIB Source Re-Engineering #

Commercial FIB-SEM systems usually feature a FIB source that is angled 52-55 degrees relative to the SEM column. While that allows for some more flexible applications, it forces the imaging surface to be angled relative to the electron beam. For extended 3D imaging, this geometry is complex to deal with, and can compromise image quality. So, the AIC FIB-SEM features a FIB source that is angled 90 degrees to the SEM column, allowing a shorter and constant working distance across the image, which is much preferred for this type of application.

In addition, commercial FIB-SEM systems to not actively monitor the FIB milling rate. Knowing this rate, and being able to control it, will allow users to ensure that the z-distance between successive SEM image slices is always as close to the target value as possible. To measure (and control) the milling rate, the portion of current from the FIB source that impinges on the sample needs to be known. If, for instance, that current becomes too high, this would indicate that the milling rate is too high. Therefore, the FIB beam could be deflected away from the sample to lower it, and vice versa. To uniquely accomplish this, the AIC FIB-SEM features two custom-built Faraday cups downstream of the ion beam, as shown in Figure 2. The one furthest from the sample measures the portion of the FIB beam that is not occluded or scattered by the sample. The difference between this value and the initial FIB current should represent the amount of FIB current interacting with the sample (and thus the mill rate). However, the initial FIB current may fluctuate, causing instabilities with this approach alone. So, the second Faraday cup, closer to the sample (with an aperture in the middle) is placed in order to detect the scattered FIB beam (along with charged particles from the sample itself). This can give a more direct measure of the amount of FIB current interacting with the sample. Finally, the ion beam generates a voltage and current in the sample itself. A third detector measures this quantity. Each of these three detectors can then be used in feedback loops of varying configurations to control the FIB milling rate with exquisite sensitivity, thus making the z-distance between SEM images considerably more repeatable than commercially possible.

Image Optimization and System Monitoring, Pausing, and Restarting #

In addition to FIB instability, commercial FIB-SEM systems can experience other significant drifts over the many tens or hundreds of imaging hours throughout an experiment. Even with well-controlled FIB milling, the SEM image quality will often drop precipitously over time due to both electron beam and sample drift. To correct for this with minimal interruption in the normal milling/imaging cycle, the AIC FIB-SEM features software-controlled automatic image optimization routines. At a predefined interval, the system will run an optimization routine that corrects for drift in focus, beam astigmatism, and aperture alignment in the SEM column. Settings that result in optimal image quality are then updated to maintain high image quality throughout an experiment.

In an effort to predict instabilities, the AIC’s FIB-SEM also monitors a suite of other parameters to proactively assess its overall state. Aside from the FIB milling rate and image quality metrics, the customized control software monitors electron beam current, FIB position, shift in the x/y position of the recorded images, and column pressure among other parameters. Deviations beyond acceptable limits will trigger an email or text alert to prompt the user to investigate, or in the case of severe anomalies, the system will even automatically shut down. In this event, the custom control software can then re-start the system, and crucially, return the FIB source to its last known position. This is particularly important in the event of both planned and unplanned system maintenance. Thus, even in the event of a total system shut down, the AIC’s FIB-SEM can restart with almost no loss in data.

Practical Considerations #

The AIC’s FIB-SEM offers a uniquely robust imaging system that addresses a range of limitations in commercial instrumentation. However, total imaging time remains one of the critical considerations when considering this technology. It is necessary to weigh the time needed to satisfactorily acquire biologically useful data against the need to make this technology widely available to the scientific community. As such, users interested in applying for access to AIC FIB-SEM should anticipate approximately 1 month of imaging time to accomplish their project goals. To help guide users in defining an appropriate project scope, the table below lists expected imaging volumes (at the specified pixel resolution) that can be conservatively expected to require 2 weeks of imaging time. Note that the current resolution limit for the system is 4nm, while resolutions of above 16nm are not well justified for this technology, as other techniques adequately occupy this regime.

Staining is another major consideration that users should make. Unlike transmission electron microscopy (TEM), samples have to be stained en-bloc ⁵. So, good stain penetration throughout the cells/tissue of interest is vital. In fact, users should verify their staining well in advance of utilizing FIB-SEM by performing thin sectioning and TEM imaging on a test sample. Further details about FIB-SEM sample prep will be covered in a subsequent blog post.

Finally, it’s important to remember that once FIB-SEM data has been collected, a potentially tremendous amount of work remains in order to extract meaningful information. Unlike fluorescence microscopy, molecular specificity is limited with the FIB-SEM. Depending on the staining protocol, many structures will be visible in the images. Therefore, it becomes necessary to segment the structures of interest from the rest of the data. While this can be done manually, the typical data volume that can be acquired on the AIC FIB-SEM can quickly render this approach overwhelming. Some form of automatic segmentation is therefore almost always preferable ⁶. This subject will also be discussed further in another blog post.

Overall, prospective AIC users should always contact us to discuss their ideas. We are happy to help guide your project to ensure that your proposal (1) falls within the technical capabilities of the instruments, (2) retains an appropriate scope, and (3) the hypothesis and approach are well-formed.

Last modified Jul 16, 2020