HW for GMS v3
HoloWorks*
Dear customer. HoloWerk LLC is committed to its customers and so we have provided HW v5.1 as of April 2018, v5.1.1 as of December 2018 and v5.1.2 as of August 2019 for direct support of GMS 3.4.
Owners of HoloWorks v5 can upgrade for free to v5.1.2 (GMS 3.4 strongly recommended). Contact a Gatan representative or email holowerk@mac.com directly.
This web page is specifically in support of HoloWorks 5 users.
HoloWorks (HW) v5.1 features
∧ Back to top
0.0 Integrating HoloWorks v5.1 with GMS v3.3
Figure 1.0: Buttons are available for the most used functions for HoloWorks. To become accessible, they must first be enabled as shown in the short movie on the left.
As of HW v5.1.2, all buttons are integrated under 'Techniques' on the right-hand side of the GMS window (see below).
0.1 Integrating HoloWorks v5.1.2 with GMS v3.4
Figure 1.1: As of HW v5.1.2, all buttons are integrated under the 'Technique Manager/Custom/HoloWorks v5' on the right-hand side of the GMS window. This avoids having an extra floating window within the work space. For the on-line version of HoloWorks, Gatan's camera control window appears as well. HW 5.1.2 is not designed for versions of GMS earlier than v3.4.
1. Setting Up Live Phase Imaging
a. Camera and fringe spacing set-up
Figure 2: Step 1: Setup camera for preview For Live Phase Imaging, lots of electrons are needed: especially if imaging at several fps (frames per seconds). Example: for a 2k camera (i.e., 2048 by 2048 pixels) a binning factor of 4 is recommended. This allows 16x more electrons per pixels and yields good results even at >>1fps. Exposure times should be selected accordingly, e.g., 0.2s, which will yield better than 4fps (camera read-out overhead). This is sufficient to observe a decent phase image and check out the sample area for the best spot.
Fringe Spacing Setup:
Generally, for off-axis electron holography, the microscope is not in a standard optical setting. Users should carefully consider what resolution and what field-of-view is necessary to observe the desired features. Part of the considerations must be the camera performance, its MTF (modulation transfer function) and DQE (detection quantum efficiency). These are sub-optimal unless a single-electron detection camera is utilized.
Rule of Thumb:
Considering 2-3 interference fringes per resolution detail and at least 4 pixels per fringe requires typically a sampling rate s with 8 < s < 12 for the interference fringes. This is the consideration for the high-quality hologram acquisition.
For Live Phase imaging, we will break this rule.
Breaking the rules: Live Phase Imaging
Given the basic parameters (see Rule of Thumb), and a camera setup with binning of 4, the sampling rate s for the interference fringes (number of pixels per fringe) computes simply to 2 < s < 3 - or larger. The interference fringes can be fine-tuned via the voltage applied to the biprism. The location of the sideband can be easily detected in Fourier space (see movie/figure above). For a sampling rate of, e.g., s = 3.3 the position of the sidebands should coincide with the circle super-imposed on the live FFT.
Note: Assistance for selecting the correct fringe spacing is activated either via the HW menu or, e.g., by clicking the icon with the yellow circle in the HW floating window. For user friendliness, the assistance can also be added to an already existing (live) FFT. HW does distinguish between a live camera image and its FFT to optimize the use of the available screen reality.
b. Optimizing hologram acquisition parameters
Figure 3: Optimizing fringe contrast - or better: hologram quality?
It is a wide-spread believe that fringe contrast is the most important parameter to optimize. This is not quite true. There are several parameters needing attention:
- fringe contrast
- fringe drift <-> exposure time
- intensity <-> condenser setting
=> hologram quality
The best way to optimize all of them is to track them continuously as shown here.
It is a wide-spread believe that fringe contrast is the most important parameter to optimize. This is not quite true. There are several parameters needing attention:
- fringe contrast
- fringe drift <-> exposure time
- intensity <-> condenser setting
=> hologram quality
The best way to optimize all of them is to track them continuously as shown here.
Figure 4: Fringe contrast and fringe drift indicating a stable biprism. Exposure time: 0.4s. The binning factor of 4 results in a sampling rate of s = 2.53. The small amount of biprism drift (red) allows several seconds exposure time for the final holograms (drift range ~5 rad / 60 events → 1 rad per 12 events → 12x 0.4s ≈ 5s exposure time). The measured fringe contrast of ~20% provides the basis for phase image resolution of ~2π/1000 as described below.
Fringe contrast
HoloWorks measures fringe contrast very accurately and independent of the sampling rate s. By adjusting the condenser system of the microscope, fringe contrast can be optimized by tracking fringe contrast as shown in the short tutorial in Figure 3.
Word of advice: Fringe contrast significantly depends on the exposure time t in two ways.
- An increase in exposure time will decrease fringe contrast due to drift of the interference fringes (each image represents an integration over time).
- A decrease in exposure times can also decrease fringe contrast. This is caused by the deflector of the microscope.
In most cases, the deflection of the electron beam is performed by a (magnetic) deflector above the sample. While the deflection of the beam away from the sample is fast, the returning beam 'eases', so to speak, into its final position. Because the biprism is way out of focus, the interference fringes will drift until the beam has settled - thus reducing contrast when decreasing exposure time. On some microscopes, even an exposure time of 0.5s can lead to a reduction of fringe contrast and loss of signal strength in the phase image.
For the acquisition of highest-quality holograms, both the upper and lower exposure time limits need consideration.
Fringe drift
A basic rule of thumb is: fringes should drift ≤ 1 rad during the acquisition of the final image(s). Thus, the maximum exposure time computes as (see also Figure 4) :
# of exposures in the (red) line plot within 1 rad, multiplied by the exposure time per single live image.
Hologram Quality
Contrary to the widespread opinion that fringe contrast is the most important parameter for electron holograms, this isn't quite true. For example, a 100% fringe contrast and no electrons is not better than 10% fringe contrast and many electrons. Hologram quality (Q-factor) is the key parameter that considers both, number of electrons and fringe contrast. The point to remember is that this parameter is directly proportional to the noise level in the phase image!
For example, doubling the exposure time should increase the Q- parameter by sqrt(2). On the other hand, doubling the fringe contrast should increase the Q- parameter by 2.
One additional note: When adjusting the condenser of the electron microscope, the Q- parameter typically remains fairly stable over a large range of the condenser setting: as the fringe contrast increases, the number of electrons in the image decreases and vice versa. For samples indifferent to sample damage all settings with the same Q- parameter are equal. However, when working with beam sensitive samples, it is recommended to aim for best fringe contrast and best Q.
Conclusion
The ability to track all three parameters fringe contrast, fringe drift and hologram quality simultaneously and live allows determination of the hologram acquisition parameters and microscope setting that will deliver the best holograms and simplify routine phase resolutions in excess of 2π/1000.
2. Live Phase Imaging
Figure 5: Live Phase Imaging
In this video, the phase image can be observed live in relatively good quality despite low fringe contrast. This e.g., allows searching for a good spot before acquiring large stacks of full-size images while monitoring all aspects of hologram quality via a line plot.
At ≥3 or more frames per second, drifting fringes interfere little with live phase imaging. The reconstructed phase images are automatically phase-optimized thus phase offset due to lateral fringe drift between consecutive images is minimized.
In this video, the phase image can be observed live in relatively good quality despite low fringe contrast. This e.g., allows searching for a good spot before acquiring large stacks of full-size images while monitoring all aspects of hologram quality via a line plot.
At ≥3 or more frames per second, drifting fringes interfere little with live phase imaging. The reconstructed phase images are automatically phase-optimized thus phase offset due to lateral fringe drift between consecutive images is minimized.
The hologram quality in Figure 4 lacks fringe contrast and the interference fringes are moving rapidly and randomly as can be recognized from the line plot. Nonetheless, the live phase image consistently shows well. This is due to
- high binning that minimizes camera noise and provides decent electron counts per pixel and
- smart data processing that compensates for phase offsets caused by the moving interference fringes.
For strong objects especially with a phase dynamic approaching or exceeding 2π it is possible that artifacts appear in the phase image due to the low sampling rate 2< s < 3. It should be remembered that live phase images are qualitative, not quantitative data.
Another feature of HoloWorks is that live phase unwrapping can be added to the live phase image simply by selecting the menu item 'HW5.1/Unwrap phase' while holding down the 'Shift' key on the keyboard.
3. Hologram Stack Acquisition
Figure 6: Acquire Holograms into Stack
Starting point is the continuous acquisition of images. Setup parameters in this case were: exposure time: 1s and a binning factor of 2 for a 2k x 2k camera. By selecting the data cube button, the number n (here 7) of images to be stored is determined and all consecutive images are copied to the data cube until n images have been stored.
Note: this process is to be repeated for the reference holograms.
Starting point is the continuous acquisition of images. Setup parameters in this case were: exposure time: 1s and a binning factor of 2 for a 2k x 2k camera. By selecting the data cube button, the number n (here 7) of images to be stored is determined and all consecutive images are copied to the data cube until n images have been stored.
Note: this process is to be repeated for the reference holograms.
∧ Back to top
4. Hologram Stack Processing
a. Eliminating fringe drift from hologram stacks
Figure 7: Reconstruction of stacks: holograms and ref-holograms
Once the hologram stacks (hologram and reference-hologram stacks) are acquired, the reconstruction is the next step. For single pairs of hologram and reference- hologram the process is fairly simple and is the same as the first part of the movie on the left. For hologram stacks however, more steps are necessary.
Shown on the left, all holograms are reconstructed the same way. The drift of the interference fringes between succesive holograms is compensated in this step. At the end, the averaged complex wave from the reference- hologram stack is subtracted from each reconstructed object wave.
Once the hologram stacks (hologram and reference-hologram stacks) are acquired, the reconstruction is the next step. For single pairs of hologram and reference- hologram the process is fairly simple and is the same as the first part of the movie on the left. For hologram stacks however, more steps are necessary.
Shown on the left, all holograms are reconstructed the same way. The drift of the interference fringes between succesive holograms is compensated in this step. At the end, the averaged complex wave from the reference- hologram stack is subtracted from each reconstructed object wave.
The result is a stack of complex images (called image waves) and a second stack of complex images (called 'refined wave stack'). The refined wavestack is the one that has the reference data applied to. It makes sense to continue with the refined wave stack. Because of the drift of interference fringes especially over a longer time frame when many holograms are acquired, it is likely that the Fresnel fringes have not been completely removed (unless the holograms were recorded with a dual- biprism setup). Therefore, the next step must be the removal of the Fresnel fringes. This is done via the so-called line filter.
b. Suppressing Fresnel fringes
Figure 8: ''Poor man's'' Fresnel fringe removal
Starting with a single or stack of complex images, this is where Fresnel fringes are removed. The object and the interference fringes (as well as Fresnel fringes) typically drift in different directions. Thus, any leftover of Fresnel fringes in the image will interfere with sample drift correction. Therefore, any leftover of Fresnel fringes must be removed. The term ''poor man's'' derives from the fact that electron microscope systems with dual biprism setting typically are hard to find despite their advantages; they are rather expensive.
Starting with a single or stack of complex images, this is where Fresnel fringes are removed. The object and the interference fringes (as well as Fresnel fringes) typically drift in different directions. Thus, any leftover of Fresnel fringes in the image will interfere with sample drift correction. Therefore, any leftover of Fresnel fringes must be removed. The term ''poor man's'' derives from the fact that electron microscope systems with dual biprism setting typically are hard to find despite their advantages; they are rather expensive.
The movie in Figure 8 shows how to use the line filter for Fresnel fringe removal. HoloWorks typically figures out a good starting and end position for the filter, but manual adjustment is often required to remove as little as possible from the spectrum of the image wave, while removing enough of the Fresnel fringes. It should be noted that the start and end points of the filter can also be moved on a single pixel basis using the 'Control' window provided by GMS v2. If the control window does not show, it can be called from the menu item Windows/Floating Windows/Control of GMS.
Once the Fresnel fringes are successfully removed, the slices (image waves) of the resulting image stack can be aligned and averaged as shown in the next tutorial.
c. Averaging, object- drift correction & high phase resolution
Figure 9: Aligning and averaging images of type 'complex' is a bit different from averaging conventional images. Handling of the details is automated in HoloWorks and remains transparent for the user. Correcting for object drift itself can be less reliable for noisy data sets; but if the drift correction fails it is, e.g., noticeable in the absence of fine details as in this example. Low signal/noise ratio caused by low fringes contrast and low electron count should be avoided as it requires additional steps as outlined below).
The averaging part is highly automated. An important feedback is the offset of the object between slices provided in form of a line plot. The line plot is essential for gauging the quality of the drift correction. Smooth plots typically are a good sign. A ragged plot as seen in this movie is a sign for concern. An additional check for successful alignment can be done by reviewing the aligned stack (enable in Preferences, under the tab 'Stacks'). Using Gatan's slice player, a visual control check is quick and convenient.
From the complex images, also the phase images can be computed as shown here (an option to show the phase automatically can be selected in the HW5.1/Preferences menu). From here on, further processing is possible. For example, aberration correction like focusing or astigmatism correction can be performed.
The next chapter will demonstrate how to deal with situations where the automated drift correction is sub-optimal. It will be shown that even for such data drift correction can be made successful.
d. Drift determination for noisy holograms
Figure 10: Averaging noisy hologram stacks requires removal of any and all information related to holographic interference fringes - especially the interference fringes. Once the hologram stack is filtered, i.e., basically converted to a stack of conventional images, it is used to determine and plot the sample drift as a function of time, i.e., as shown in this movie. The line plot generated this way can be used for the averaging of the stack of complex images as the signal/noise ratio in the hologram can be better than the information contained in the sideband.
A successful drift plot typically is smooth. This is now an 'external' plot that can be used to correct the sample drift of the stack of complex images called 'Refined wave stack - line filtered'.
∧ Back to top
e. Averaging noisy complex images
Figure 11: Averaging noisy complex images can be challenging because there is not enough signal to correct for drift. HoloWorks provides these tools:
- Take the hologram stack and remove all the information contained
in the interference fringes.
- Run the stack alignment routine; save the line plot that shows sample drift
- Return to reconstructed stack of complex images
- Use the generated line plot as input for the stack alignment routine
When using an external data set for aligning the stack of complex images, the Fresnel fringes no longer impact the alignment and averaging process. In this way, even hologram stacks of low quality can be saved and used to obtain high quality phase images.
For more details, please check 'E. Voelkl and D. Tang, Approaching routine 2pi/1000 phase resolution for off-axis type holography, Ultramicroscopy110 (2010) 447-459'; or the manual, downloadable below.
∧ Back to top
5. Additional Details
Figure 12: Integration of HoloWorks into GMS offering buttons simplifies many operations.
'HoloWorks 5.1.0' window, left to right, top to bottom:
- Sub-image: creates sub-image from volatile ROI in any image or line-plot
- Fringe contrast measurement: tracks contrast and other parameters
- Reconstruction button: moving to 2nd line of buttons
- Collection button: drops live 2D images into a 3D data cube
- Compression button: averages and aligns images/holograms in the data cube
'HW FFT 5.1.0' window, left to right:
- Fourier transforms FFT and iFFT maintain equal scaling in Fourier space from any size image
- Peak button allows easy measurement of peaks (position, amplitude, sampling rate) in FFT
'HoloWorks 5.1.0' window, left to right, top to bottom:
- Sub-image: creates sub-image from volatile ROI in any image or line-plot
- Fringe contrast measurement: tracks contrast and other parameters
- Reconstruction button: moving to 2nd line of buttons
- Collection button: drops live 2D images into a 3D data cube
- Compression button: averages and aligns images/holograms in the data cube
'HW FFT 5.1.0' window, left to right:
- Fourier transforms FFT and iFFT maintain equal scaling in Fourier space from any size image
- Peak button allows easy measurement of peaks (position, amplitude, sampling rate) in FFT
HoloWorks v5 was designed for multi-core processing and uses independent threads where possible. Thus, the workload can be spread over different cores easily. In version 5, not only has the GUI (graphical user interface) seen a major lift, but the underlying algorithms and code have changed even more. It remains a simple but powerful interface for the novice as well as the expert while allowing a wide variety of possibilities for customizing off-axis holography technology for research purposes. Today, HoloWorks v5 remains an attractive package for off-axis holography. 50 menu items include live features as:
- View live phase, amplitude and intensity images at several frames per second
- Correct images live for isoplanatic aberrations
- Live phase unwrapping
- Live magnetic induction and magnetic contour images (color)
- Live fringe contrast, hologram quality and biprism drift measurements with line plot display for optimizing fringe contrast and hologram quality to assure optimum data acquisition
- Thickness determination (using holography's characteristics as perfect inelastic particle filter) and embedded, extensive list (>90 elements/compounds) of known mean inner potential values
- Live data cube acquisition and processing for > 2π/1000 phase resolution
- Accurate diffraction peak evaluation
- Data- and processing reproducibility through consistent tracking of processing steps and parameters under image/tags/history and the “Results” window
- Over 70 additional scripting functions supporting Fourier optics and data tracking
Note that the output of any one function of HoloWorks can be used as live input for any other HoloWorks function.
Note also that the software has been developed with a strong focus on teaching Fourier optics with hands-on use in the classroom; HoloWorks provides most features found on a modern microscope starting from changing focus, introducing astigmatism or spherical aberration, applying various numerical apertures or experimenting with dark field imaging.
New Seamless Integration
In previous versions, the simultaneous use of DigitalMicrograph® standard tools (i.e., line scan tool) with any function of HoloWorks was not possible. Now, for the first time, all HoloWorks functions have been designed as “threads.” This means, the HoloWorks tool suite and the Gatan tool suite work together seamlessly.
Menu Items Become Alive
A further important step has been added to the HoloWorks v5 software. When selecting a HoloWorks menu item while holding down the Shift key, many menu items become live menu items providing additional features. In most cases this means that the same operation continues to be applied to the image whenever one of two things happens:
• The image is being updated (or changed)
• An existing selection in the image was moved
Build Your Own Holographic Script
As electron holography continues to require additional functionality, the HoloWorks v5 software supports the user in building their own functions using the reliable core concepts of HoloWorks. Instead of rewriting hundreds or thousands of lines of code, many complex activities can now be embedded with a single function call. For example, a fully interactive reconstruction process, including image process tracking, can now be called by ~24 lines of code, representing over ten thousand lines of original code.
Improved Digital Fourier Optics
The Fourier transform in HoloWorks handles images of arbitrary size. Because of the slightly different implementation of the Fourier algorithm for non-power-of-two and non-square images, the Fourier space looks as expected from the experimentally oriented user: it looks like the back focal plane in the microscope. There is no skewing because of non-square image sizes or selections. This allows observation of symmetry features without further ado as shown here:
Figure 13: The standard FFT of non-square images looks strange in Fourier space as it does not maintain even scaling. In the example on the left, a rectangular subarea was selected as indicated by the red ROI, which is also non-power of two. It's FFT shows the effect of scaling that is different in x- and y- direction.
Right: HoloWork's FFT does maintain even scaling in Fourier space and Fourier images look as expected even from rectangular, non-power-of-two areas. Note that with HoloWorks the full rectangular area is transformed yielding a square Fourier transform.
Scale Independence
Photons, electrons, neutrons or phonons: off-axis holograms are reconstructed scale independent and aberrations can be corrected for a large scale range from meters (m) to Pico meters (pm). The only conditions being: coherent imaging conditions and Fourier optics applicability.
Continuous Data Acquisition
The continuous data acquisition attaches to live images in general. Whenever the live image updates, the new data are copied into a data cube. This allows improved data averaging via drift compensation not only for holograms.
Phase Resolution > 2π/1000
This feature utilizes continuous data acquisition into a data cube then compensates for fringe drift and specimen drift during acquisition. Both fringe and specimen movements are compensated at sub-pixel fringe accuracy with a reliability typically better than 0.1 pixel. Note that this feature can be used for drift compensation of standard TEM images.
References:
E. Voelkl and D. Tang, Approaching routine 2pi/1000 phase resolution for off-axis type holography,
Ultramicroscopy110 (2010)
447-459
E. Voelkl, Noise in off-axis type holograms including reconstruction and CCD camera parameters,
Ultramicroscopy110 (2010)
199-210.