The last time I mentioned the benefits of the PILATUS detector system. Since then, I’ve had some brainstorming sessions, and there are a few issues with the detector:

1. It looks like it is not built for use in vacuum, electronics are cooled by convection.
2. It is rather small and not square.

The first issue would mean that the detector would have to sit behind a window. Leave enough space between the detector and the window, and the air scattering would significantly reduce the benefits you had wiht the high resolution. The solution here might be to adapt the detector system, so that an external cooling system could be led through the electronics housing. In my view, it reduces complexity in a SAXS system to have as few windows in it as possible.

The second issue is of a more complex nature. The PILATUS website states that one PILATUS 100K module has an active area of 84 x 34 mm. In terms of detector systems, this is really small and non-square. Regarding the non-squaredness, a stationary single detector, positioned so that the beam falls onto one of the ends of the detector plates, would be sufficient for measuring solutions or other isotropic (randomly-oriented) systems. These systems have isotropic scattering patterns, and therefore would not require a full (square or round) detector area to be captured. For anisotropic systems, one would be very interested in capturing the entire scattering pattern.

One option to benefit from the small size is to minimize the SAXS system geometry, but this requires the use of very small pinholes. Given that the standard X-Ray generator has a source point of approx. 1mm in diameter, this would mean that you’d be cutting away most of your generated intensity, e.g. a reduction of the pinholes by a factor of 2, would cut down the beam intensity by a factor of 16. That simply is not worth the smaller size.

Secondly, the active area of the detector can be increased by tiling many 100K’s. This is indeed a good idea, but only when enough money is available. A 25×25cm detector would contain 24 100k modules, which might get rather pricey.

The third option is to move the detector, like so:

In this image, the red dot is the direct beam, and the arrows indicate the translations. Two translations are exemplified, a circular translation around the beam (A) and a lateral translation along the short axis of the detector (B).
In the case of a translation lile (A), given the fast read-out speed of the detector (100 images per second) this would mean that you could do a full circle in about 30 seconds (maximum speed), without smearing (motion blurring) over more than one pixel. Such a translation would mean that, as you go further away from the beam, the intensity would have to be corrected more and more, for the fraction of the area that the detector covers (about 1/6th in the outer regions). This method then, has an enhanced error in the outer regions. If you are only interested in analysing Guinier regions (i.e. close to the beamstop), this should not be too much of an issue.

An area with a total diameter of about 120mm could be captured with this method, using only a single detector. Measuring a half-circle would suffice, since the whole pattern can then be reflected in order to obtain the full scattering pattern. Such a motion would only take 15 seconds at maximum speed.

In case it is, option (B) has a translation that equally distributes the measurement time in each position. If a square area is to be analysed, this would mean a reduction of about a factor of three in terms of total counted intensity. The area captured by this method is equal to that of the circular translation, since the whole pattern can be reflected around the horizontal axis. This method would thus create equal errors over the entire measured area.

One can imagine many more motions like these, some highlighting the outer regions more than the inner regions and vice versa. This shows that for the price of a single detector, the options are still limitless, and the measurement times “only” increase by a factor of 3 to 6.

If intensity is that which is sought, and price would be a real limiting factor, combining a slit-collimated (Kratky-type) system with one of these detectors could be a viable solution.

A fancy new detector is being developed by the Swiss Light Source. This Pilatus detector would be of special interest in pushbutton machines, for its range in count rate would allow the use of this detector in a beamstop-less system.

Imagine a SAXS machine without a beamstop, and how convenient it would be, and it quickly becomes apparent that this would much simplify the operation and measurement requirements for samples. The advantages I can quickly think of are:
– no more beamstop alignment, 1 less set of x,y motor stages
– immediate determination of the beam center
– two-step determination of the primary and secondary beam intensity, greatly simplifying the required procedure for determining the sample attenuation/transmission factor, as well as the primary beam intensity in absolute units.

If it works as advertised, this detector could thus greatly simplify the SAXS technique.
Soon I’ll post a simple drawing of an example pushbutton machine.

Back on the topic of pushbutton machinery.

The ideal pushbutton machine could provide, for the beginning user, a control system that would not “challenge” them with a command-line interface, such as provided by TASCOM and SPEC. One option would be point-and-click interfaces on a computer screen. Another would involve a more hardware-oriented approach using midi-devices.

The first option is a no-brainer, and is probably already widespread in commercial systems. The disadvantage is that such an interface (a gui to TASCOM or SPEC), would require constant maintenance due to over-enthousiastic managers (e.g. “Ooh, wouldn’t it be cool to integrate xxx into the GUI”), and other gui-related issues.

The second option would be a nicer one, in my opinion. Recently, we’ve seen the uptake of button boxes for NMR machines, to simplify the controls of the software on the computer. Electron microscope machines have never much stepped away from dedicated controls, and for good reason. Dedicated controls provide the occasional user with a feeling of control, when button “A” is pressed, the machine performs a single specific task. Instead, pressing “A” on a keyboard could do anything, and thus requires some more warm-up time, alienating the user from the machine.

So how would we go about this? Well, the motor controls, where needed, could be controlled by an inexpensive device like the Behringer BCR2000. This midi control interface can then provide visual feedback on the approximate motor position, and provide rotary controls to change the position accordingly. The push-buttons could then initiate scripts that would control the vacuum (i.e. it could start sample_changing procedures), the generator, and the detector system (i.e. start measurements). The options are nigh endless.

I think such a control box would make the system a little more user-friendly, and thus lower the barrier to SAXS.

Next up in this series: automated processing of data.

Many consumers (dare I say Bio-people here?) would love to have a pushbutton SAXS machine. At the moment, however, there are a few issues here that have to be taken care of. One of them is the rather cumbersome pinhole alignment procedure.

After replacing a source, realigning the monochromator crystal, or after changing of the geometry of a SAXS machine, the collimation pinholes need to be realigned. Thus, the part of the beam with the highest intensity is to be selected for this. The standard method is as illustrated in this figure:

The uncollimated beam (A) produces a rather large spot on the detector. By moving in the first pinhole, most of the surrounding radiation is cut off (B). Then the second pinhole is moved into position (C), which cuts the beam into the desired shape. The parasitic scattering generated by the second pinhole is then removed by introducing the third pinhole (D) resulting in a collimated system.

There are a few issues with this method:
– If the pinhole positions are not well known, introducing pinholes two and three might become ever more tricky. This is because the only way to know where the center of the second pinhole is, is when the pinhole opening actually crosses the beam that has been reduced in size by the previous pinhole. This can get very difficult in the larger geometries, where a pinhole of say .3 mm in diameter has to be moved over a range of millimeters in two directions in order to find the beam.
– You usually find out that you moved the beam too low (for example due to misalignment of the monochromator mirror), only when you are trying to move pinhole three into position. At that point, when you can no longer move the pinhole further to center it around the beam, you have to disassemble the collimation system and realign the mirror, undoing all of the collimation work that you have done before.
– If the pinholes are aligned by the countrate over the detector, a low countrate when aligning the second pinhole may in the end cause a large error in the placement of the second pinhole. This can later, when the entire pinhole collimation system is in place, no longer be corrected (i.e. when you find out that the intensity of the collimated beam is not as much as it was once upon a time with a different collimation).

I think this could be improved using the following method:

Here, the position of each pinhole with respect to the beam center and the detector is subsequently determined (A through D). Then, a region with high beam intensity on the detector can be selected, and the pinholes can be moved there, using a spherical coordinate system to move the pinholes with the source point as center. This way, an alternative position of the beam can be selected simply by moving all the pinholes in this coordinate system to cut out the new part of the beam (F) without having to realign the pinholes.

Thus, you can quickly find out if the beam is misaligned (a region can be indicated on the detector output display that all the pinholes can reach), and an alternative part of the beam can be selected. The drawback of this method is that it would require a little more than a simple counter on the detector, i.e. it would require image processing algorithms to find circle-shaped shadows of the pinholes cutting the primary beam.

More specifically, the image processing would need to correlate the center of those pinhole shadows on the detector grid, to the motor positions of the pinhole alignment motors. Furthermore, the length of the sections between the pinholes is to be measured and input in order to be able to move the pinholes around an arbitrary center (e.g. the source point) in a spherical coordinate system.

I think, however, that this method would allow for a much more robust push-button approach to the alignment of pinholes. I hope this would make realisation of a complete pushbutton system a step closer to reality. I will attempt to implement this at the Risø system in the near future.

Leave your comments at the bottom!