Transmission Electron Microscopy for Imaging of Biological Molecules

Customer Case

The customer is improving their customized Transmission Electron Microscope (TEM) that is optimized for imaging unstained biological molecules with minimal radiation damage.

Diffraction-limited resolution of 0.25 nm is routinely obtained with a cold field-emission electron-gun operating at 40 keV. The sample is maintained at about 123 K (i.e. -150°C), both to eliminate contamination and to reduce mass-loss.

The scan raster is 512 x 512 points with a dwell time of 30 microseconds per pixel (over 8.5 seconds per raster-scan). For most mass measurements, the pixels are separated by 1 nm giving a scan width of 0.512 µm.

Electrons that either are elastically-scattered through large angles (i.e. 40-200 mRadians), or are elastically-scattered through small angles (i.e. 15-40 mRads), or are unscattered (i.e. direct-path, 0-12 mRads) are collected on three separate detectors (consisting of scintillators and photomultipliers). Both large-angle and small-angle signals (normalized by the total beam-current) are recorded digitally. The dose is kept low to assure that mass-loss from radiation damage is minimized.

Digital images are recorded with the TEM from the ring-shaped dark-field detectors. The background is computed in clear areas in each image, and the intensity-minus-background is summed over each particle and multiplied by a calibration factor to give a mass-value.

Data in the TEM image comes from three detectors: LargeAngle & SmallAngle (LA and SA) dark-field ring-shaped detectors (40 – 200 & 15 – 40 mRadian acceptance angles, respectively) and a bright-field detector (BF, 0-15 mRad). For thin specimens, both LA and SA signals are proportional to the number of atoms weighted by their atomic-number in the irradiated pixel. This is the basis for TEM mass measurement. The BF signal is 1 – (LA + SA), so only two signals are sufficient to give independent information. Normally, only LA & SA are recorded.

They will be getting a sync pulse (from the 60 Hz power-mains line, say, every 16.67 milliseconds) to start a series of measurements and making 50 measurements per µs (i.e. 50 MS/s). After a dwell-time of 1 to 30 µs (specified by the computer), they need to increment a scan-counter to advance the beam in their microscope to the next pixel. After 512 or 1,024 such cycles, they would stop the scanning, dump the data to the computer and wait for the next 60 Hz pulse. This would be repeated 512 or 1,024 times to scan an entire image.

The signal is the output of a scintillator/photomultiplier being struck by 40 keV electrons. The optical rise time is about 50 nanoseconds and the fall time is about 1 µs with about 1,000 photons/electron pulse.

The requirement is to capture the detected signals with the highest possible vertical-resolution (i.e. 12-bits or more). Triggering is based upon the 60 Hz of the power-mains.

Their current set-up uses one 8-bit acquisition card. An application is used to control, record, display, and save the data to a disk-file. They currently use a program written in C that runs for analysis of TEM images. TEM images are half-MB (528,384 bytes). Analysis results files are usually less than 50 KB and suitable for import into third-party programs.

The TEM image-files consist of a header followed by 512 x 512 pixels of image data consisting of two channels (8-bit each) interleaved. TEM images can be displayed and printed in PhotoShop using the “RAW” import mode and the above parameters. Some care is required to determine which image is the large angle detector signal (the channel most reliable for mass measurements). PhotoShop is the program of choice for image refinement.

They hope to discriminate single-electron events by pattern recognition in the computer.

They plan use the software development kit for Windows and program in C.

GaGe Case Solution

The 14-bit data obtained is much better than the 8-bit that they now have. GaGe’s CompuScope 14100 provides both the necessary high-resolution and the fast PCI-bus transfer-speeds.

Acquiring a single pixel-record of 1,472 points (or minimal 128 points) simultaneously on two channels at 50 MS/s would take 29.44 µs of the dwell-time (or minimal 2.56 µs). Using the PCI bus-mastered transfer-rates of over 100 MB/s, then transferring the resulting data of: 1,472 x 2 Bytes = 3 KB (or minimal 128 x 2 = 256 Bytes) from the card into the PC-RAM for later displaying/saving-to-disk would take up to about 29 µs (or minimal 2.6 µs) per channel. Rearming the cards to prepare for the next trigger takes about 30 µs per card. This means that this would take at least:

29 µs + 2 channels x 29 + 1 card x 30 = 117 µs (or 2.5 + 2 channels x 2.6 + 30 = 38 µs, minimal).

Therefore, allowing for a healthy safety-margin, a single pixel-scan will most likely take less than 300 µs. Since the rise/fall times between pixels are less than 1usec, this is obviously too long.

Alternately, instead of off-loading each record individually as described above, the Multiple Record mode of operation could be used to “stack” the pixel-records into on-board memory for a later bulk-offload. Acquiring each of the single-pixel records of 1,472 points at 50 MS/s would again take about 29.44 µs. Within 10’s of sampling clock-cycles (within 0.5 µs at 50 MS/s) the cards would automatically rearm themselves to be ready for the next pixel-trigger. Repeating this cycle for, say, up to 512 pixel-triggers would at the very least take 512 x 29.44 µs = 15 ms total. Using the PCI bus-mastered transfer-rates of over 100 MB/s, transferring the resulting bulk of 1,472 points x 512 pixels x 2 Bytes = 1.44 MB per channel (since there are two bytes per 14-bit sample-point) from the cards into the PC-RAM would take about 15 ms per channel (i.e. 30 ms total). Rearming the cards to prepare for the next Multiple Record cycle takes a negligible amount of time (30 µs per card). Thus, a single line-capture (into PC-RAM) of 512 pixels takes at least 30 ms (assuming a negligible delay between pixel-scans). If line-scans are done every, say 33.33 ms using the 60 Hz as a line-trigger source, then a 512 x 1,024 pixel frame would take at least 1,024 x 33.33 = 34 seconds (not including any GUI updates, disk-writes, data-reduction, analysis, etc).

Accordingly, GaGe recommended a combined on-board memory depth of 8 MegaSamples (i.e. 16 MegaBytes total). Between each scan, all of the data-points for one Multiple Record scan (i.e. one “line”) are offloaded across the PCI-bus into PC-RAM.

The customer was easily able to incorporate the Multiple Record sample-program from the CompuScope Software Development Kit (SDK) for C/C++ for Windows into his existing C-code ported over from their application. This sample-program is intentionally simplified for easy transfer into any given software application.

GaGe’s powerful hardware and easy-to-use software products enable the customer to quickly move from a preliminary prototype to a fully functioning imaging system.

GaGe Case Recommended Products

Medical Application Request

We encourage you to contact us and discuss your medical application in more detail with our engineering team. GaGe can provide tailored custom data acquisition hardware and software solutions to meet specific application requirements.