FEI TITAN STEM 80-300 (TITAN S)

The FEI Titan 80-300 STEM is a scanning transmission electron microscope equipped with a field emission electron gun, a three-condenser lens system, a monochromator unit, and a Cs probe corrector (CEOS), a post-column energy filter system (Gatan Tridiem 865 ER) as well as a Gatan 2k slow scan CCD system. Characterised by a STEM resolution of 80 pm at 300 kV, the instrument was one of the first of a small number of sub-ångström resolution scanning transmission electron microscopes in the world when commissioned in 2006.

Users of the Titan STEM are kindly asked to quote a tecnical description of the instrument published in the Journal of large-scale research facilities 2 (2016) A42 http://dx.doi.org/10.17815/jlsrf-2-67 when referring to the use of this instrument in publications.

Microscopists planning to use this instrument should contact the Instument officers.

Instrument officer

• Penghan Lu

Microscope Manual

A short manual can be found here: File:Titan STEM Manual - english.docx

This manual is for skilled users only who have received an intense training at the microscope! This manual does NOT replace proper microscope training by a project officer or microscope officer!

Medipix3 Merlin for EM manual

[K. Müller-Caspary, August 6, 2018] File:QuickMerlin.pdf

Introduction

The Medipix3 Merlin for EM is a hybrid pixel detector that has been installed at the Titan S on August 1, 2018 as a replacement of the Gatan MultiScan camera on the right hand port of the Gatan box, located between the projection chamber and the energy filter. It is retractable to allow for the usage of the GIF still. Moreover, we mounted the Merlin camera such that it is centered under the inner shadow of the Fischione Model 3000 annular detector.

X-ray safety has been measured by Werner Pieper with the following result under parallel illumination:

Medipix state	Beam current measured on Flu-Screen	Max. dose rate measured (Pieper)
Retracted	12 nA	60 nSv/h
Inserted	5 nA (12 nA)	20 nSv/h (100 nSv/h)

These values basically reflect the noise level, so no leakage could be detected.

The camera is specified to detect electrons in an energy range of 30...300 keV. It has 256x256 pixels with a size of 55μmx55μm each. Its dynamic range is up to 24 bit, while data can be acquired in 1-bit, 6-bit, 12-bit and 24-bit depth in various modes. Although this is quite a lot, it is recommended to avoid unnecessary exposure with high beam current, especially when focused to a sharp spot. As with CCD cameras, it is good practice to keep the viewing screen down when no data is acquiring. Operation of the Merlin camera is done from a separate PC located on the right where also the acquired data is stored. It has a white Quantum Detectors sticker.

The Merlin camera can run at relatively high frame rates of up to 1.2-2.4 kHz, depending on the mode of operation. This makes it attractive for momentum-resolved STEM experiments where diffraction patterns are recorded at each raster position of the STEM beam. To allow for that, the camera frame acquisition can be synchronised with the scan engine. While the Merlin detector would be capable of being triggered with a STEM-pixel-trigger, the Titan S microscope provides only a line trigger as a kind of carriage return after each scan line. However, as the millisecond scan speed is still rather slow compared to the precision/jitter of the internal clocks (ns range), the acquisition usually doesn't suffer from synchronisation at the beginning of scan lines only significantly.

Quick notes about the working principle

The Merlin camera consists of 65536 pixels, each connected to its own electronics. This basically consists of 2 counters, each capable of 12-bit detection, that can be used in several ways, so as to:

1. Get a simple number-of-events-over-threshold counting in single pixel mode (SPM). Pixel read-out takes a gap time that adds to the actual frame time. The gap time scales with the bit depth, i.e. it is 412 μs in 6-bit mode, 822 μs in 12-bit mode etc. The maximum is the 24-bit mode where each pixel can record up to 16777216 electrons. In 12-bit mode this is 4096 electrons, in 6-bit mode we can have max. 64 electrons per pixel. Up to 12-bit, only one counter is read, for 24-bit, both are used.
2. Fill one counter while the other is read out in continuous mode (CTM). No gap time applies here, which significantly improves the speed. Maximum bit depth is 12.
3. Operate in colour mode (CM), where up to eight energy thresholds can be defined for detection. This is, however, at cost of sampling because the required additional 6 counters are taken from adjacent pixels. In other words, only each 4th pixel is used for imaging, the others are completely blind because their electronics is used for (a very coarse) energy resolution.

It is noteworthy that the detector has practically no read-out noise, which you can easily check by retracting it and close the illumination; then acquire an image. In most cases, you might operate the detector in CTM or, if necessary, in SPM mode.

Single Pixel Mode (SPM)

Whenever a certain amount of energy to be set by the user is deposited in a pixel, its counter is raised by one. For example, consider operation of the microscope at 300 keV, then the electrons have to deposit 300 keV in the Medipix3 chip. To capture all of this energy, and to protect the underlying electronics, the detector is covered by a  500μm thick silicon layer in which the beam electrons lose their energy by creating electron-hole pairs. Ideally, we would like each electron to deposit its energy within one pixel only. We could then set the detection threshold to, e.g., 295 keV and get a click whenever an electron has hit the detector at a certain pixel.

However, the electrons are scattered to travel within a potentially large volume that can penetrate several pixels, especially at high primary energies (150-300 keV). On its trajectory, an electron hence deposits its energy distributed over a certain volume which can be significantly larger than the pixel size. This means that an event where a single pixel registers an energy of, e.g., 295 keV as in the example above, is extremely rare, and the quantum efficiency drops drastically.

Hence, in practice, the threshold is set to lower values. For beam energies below ≈150 keV, it is recommended to use about 50% of the primary energy as a detection threshold. Thus, working with the TEM operated at 120 keV means that a threshold around 60 keV would be recommended.

Note that setting the threshold too low will cause multiple counting of one electron, which can distort measurements quite a lot. Putting too high thresholds will cause many electrons not to be counted at all.

At higher acceleration voltages >150 kV, things are different and not explored in too much detail at present. So far, it is recommended to stay below the 50% rule-of-thumb threshold, e.g., for 300 keV operation thresholds between 100 and 150 keV make sense. I am in contact with Quantum detectors to derive the optimum settings for us.

Continuous mode (CTM)

This is the mode you would probably use most frequently when heading for speed and when you do not need more than 12 bit depth. Because both counters in each pixel are acquiring/read-out subsequently, no gap time applies. Hence you can set the Merlin frame time equal to the STEM dwell time.

Charge Summing Mode (CSM)

To cope with multiple counting artefacts, the charge summing mode is available. When an electron hits the detector and creates signal in adjacent pixels, this is evaluated by the electronics immediately such that this split event is counted as 1 event only. It is assigned to the pixel with the highest deposited energy which, unfortunately, is usually the end of the trajectory in the chip. Note that this mode has its limitation as to the maximum local dose rate, because electrons impinging on the Si chip must be separated in time to distinguish single events. The mode can be combined with, e.g., the continuous mode, too. One can find the maximum intensity for the CSM by starting at low intensity, then increase it until the counts registered by the Merlin detector do not increase proportionally anymore.

Setting up Merlin

1. Switch on the Merlin PC (black, DELL) in the small rack behind the microscope with the Quantum Detectors sticker on it.
2. Log in as user MERLIN with password Merlin. This account has administrative rights.
3. Launch the Merlin software by double-clicking on the Desktop icon. The user interface should open.
4. In the Advanced Tab, check if HV bias is 120 V. If not, put it to this value and press Set HV Bias.
5. In the DAC Tab, set lower (threshold 0) and upper (threshold 1) threshold for electron detection.
   • In single pixel mode, threshold0 is set to 33-50% of the primary energy (i.e. 100-150 keV), and threshold1 is set to 511 keV which is the maximum.
   • When using the option Counter 0&1, two images are generated simultaneously with different lower thresholds. Then, setting threshold0 to 100 keV and threshold1 to 150 keV means that one image is recorded for a threshold setting 100...511 keV and another one with 150...511 keV (instead of an image with 100...150 keV thresholds).
   • In charge summing mode, threshold0 means the threshold at which adjacent pixels individually detect an event at all, while threshold1 sets the energy which you require the sum of 4 adjacent pixels to have in order to register their counts as one common event. Here, threshold0 could be 10 keV or more, threshold1 could be 200 keV, depending on the primary energy.}
   • Note that the noise level of the detector (thermal etc) is about 5 keV, it is hence not recommended to ever use thresholds in that range.
   • I have included the most common settings in a command file that the software reads during startup. Therefore, the parameters might be set already automatically.
6. Launch the Merlin Motion control software by double-clicking the Desktop icon. A small window opens with buttons Move In, Move Out, Stop. On the right, the current position is indicated, whereas Out means retracted. Press Move In. It takes ≈30 s for the detector to reach its calibrated position.
7. In the Image tab, set the parameters for acquisition and one of the modes mentioned above, and press Start Acquisition at the bottom, right of the window. Explanation of some parameters:
   • Number of frames. Total number of frames that you want to acquire. Set to 0 if you want to continuously run the camera. Set to 65356 if you do a momentum resolved STEM scan with 256x256 STEM scan pixels.
   • Frame time. Duration of recording of one frame. Note that this is not necessarily the speed of the camera: If you are not operating in continuous mode, you must add the gap time indicated. This is important in case you run an acquisition that is to be synchronised with the scan engine: The STEM pixel dwell time in TIA must be set manually such that it corresponds to the time the camera needs for one pixel.
     • In continuous mode, you would set both to, e.g., 1 ms. Otherwise, you must set the STEM dwell time to the Medipix frame time plus the gap time. (E.g. to 1412 μs for Frame time 1 ms and gap 412 μs in 6-bit mode.) Putting 0 in the gap time will cause the program to fill the minimum necessary value, which is what you usually want.
     • Note that in continuous mode the gap time doesn't influence the timing of your measurement, but still has a meaning in terms of accepting triggers. It should hence still be at some hundred microseconds.
     • 412 μs frame time is also the minimum at which no data loss should occur in 6-bit operation in continuous mode. If you run it faster, it may also work until one of the internal buffers spills over.
   • Number of frames per trigger. This is only important if you combine the Merlin acquisition with a STEM scan, otherwise set it to 1. For the STEM case, set it to the number of STEM scan pixels in a row, because the TIA scan engine fires a trigger whenever it starts to scan a row of your STEM image. Thus for a 256x256 STEM scan, enter 256 here.
   • Counting mode. Choose your required bit depth here. In STEM, this can often be 6 or 12, as there won't be too many counts per pixel in one frame. Note that this setting affects (detector and write) speed and determines the disk space during storage.
   • Counter. Usually this is set to Counter 0. A special thing is Counter 0&1, where you can record 2 images simultaneously for 2 different thresholds. These images are written in alternating manner to the output files.
   • Mode. Usually this should be set to Continuous mode active, the others not. If none is active, it is the single pixel mode.
   • File saving. Active or not, depending on whether you want to save acquired data.
   • Use Time Stamping. This will create a folder automatically with a generic name derived from the current date and time. It is a useful (and recommended) option to assure no overwriting of previous acquisitions happens.
   • Save all images in single file. Meaning should be clear. Be careful with the RAM available for large acquisitions.
   • Images per file. Usually I set this to 1000 or to some multiples of Frames/Trigger.
   • Filename. No explanation needed.
   • Local Save Root Directory. Browse to or enter a path where to save your acquisition.
     • Please use drive D:, because C: is much smaller and contains the Windows system.
     • Please also transfer your data to a network drive or to your computer after the session until we found a solution for mass storaging the Medipix data on-site.
8. A STEM acquisition.
   • Set the parameters in the Image tab as described above. Then, you need to select the trigger in the Triggers tab. As Start Trigger, choose Rising Edge LVDS.
   • In TIA: Assure TIA is not scanning, and enter the correct STEM frame size and dwell time for Acquire.
   • Press Start Acquisition in ther Merlin software. The Medipix recording does not yet start but waits for the STEM trigger now.
   • In TIA: Press Acquire. You should see some Medipix frames updating. If everything goes right, files are written to the folder you specified.

Some remarks

1. Under the Advanced tab, you can select Run Headless to avoid the live display in the Merlin program, which can save resources in case you approach the limits of hardware performance (RAM, write speed etc.) in large acquisitions.
2. Similarly, you can write just the raw data, but the data would need post-processing afterwards.
3. Very helpful: Under the Config Files tab, at the bottom, you can run a command file which can be selected by browsing. This is a very simple text file which contains predefined parameters to be set and commands to be executed. You can have a look in the manual, and also in one of the files I have prepared under ..\Desktop\Merlin_Run_Files\Knut.
   • These files are of great help when you want to come from a setting where you look live at images, untriggered, not saved, \dots, to a custom setting for recording a 4D STEM data set very quickly. It also assures to not forget setting parameters properly. For some settings, also the order in which parameters are set matters (e.g. triggering in continuous mode). However, please don't modify my scripts but make your personal copy, please.

End of a session

1. Shutdown the Merlin software, this takes 5-10 s.
2. In the Merlin Motion Control software, Press Move Out.
   • This assures that the next user who might not be aware of the Merlin camera, doesn't expose the camera to high beam currents accidentally. Imagine a situation where somebody wonders why, although there is super-high intensity on the screen, no signal arrives at the camera behind the GIF...
3. Copy your data.
4. Shutdown the Merlin PC.

I want to note that I have quite an extended software package ImageEval written in Matlab to analyse such data, you are welcome to contact me if you want to use it!

The same applies to Dieter's and Alex' LiberTEM software.

Network setup for remote Merlin control

The Merlin detector software can be controlled remotely by sending commands over a TCP/IP socket. TODO link details. The Merlin software performs a reverse lookup of the connecting IP address. This might lead to a delay of several seconds if the first chosen resolution method runs into a timeout. A quick fix is to add the static IP address of the computer that connects to the Merlin software to the %windir%\system32\drivers\etc\hosts file of the Merlin control PC, which usually takes precedence over any other lookup method. A proper setup would involve running a DHCP server, DNS server, and NetBIOS name server in the microscope network that ensures a coherent networking setup and quick response of name resolution.

Use FEI Scripting remotely via Python

Use Merlin computer as control PC, activate libertem Python environment and follow the setup and usage instructions (keep in mind that all necessary components are installed).

Computer assisted calculation of the convergence angle

• First two images with the MERLIN-Software need to be saved.
• One image needs to be taken of the poly-crystalline gold reference sample in parallel illumination in a way that the diffraction rings can be seen. This is called Au-rings.
• The other image needs to be taken with the same camera-length in the desired working mode of the STEM in a way that the ronchigram can be seen. This is called ronchi.
• Start Octave on the MERLIN-PC.
• Search for the directory ConvergentAngle and run the file get_conv_angle.m. Usually this file should already be open after starting Octave.
• First open the file Au-rings.
• Then open the file ronchi.
• Enter the accelerating voltage in kV.
• The Au-rings image will open.
• Enlarge it and click on three different points on the first diffraction ring.
• Do the same for the border of the ronchigram in the ronchi image.
• The program will print the convergence angle in mrad.

Model and serial number

Model No. FP 5600/30 Titan 80-300

Serial No. D 3096