| People involved: | Wilfred Fullagar, Dharmalingam Kurunthu, Ujjwal Mandal, Fredrik Parnefjord |
| Former members: | Sophie Canton, Jens Uhlig, Monika Walczak, Niklas Gador, Erhan Cengiz |
| Involved facilities: |
This technique has the following projects (and possibly other techniques) related to it:
Being able to detect X-rays in an informative way is as important as generating them in the first place. The requirements placed on X-ray detection depend on both the type of experiment one does, as well as the topology chosen. In any experiment it is of course necessary to register a useful number of photons in a practically viable time. For EXAFS, the energy of X-ray photons is of critical interest. However, X-ray photon energy is also of key interest in ultrafast diffraction experiments, for which we have elsewhere motivated the use of polychromatic sources with collimated or focusing X-ray optics.
The following discussion focuses is an overview of detector work associated with a laser plasma source we have built and characterised. This source emits a large number of polychromatic X-ray photons from a <30x30x30 micron source, approximately isotropically. Such bursts, which for several reasons are believed to last less than a picosecond, occur every 100 ms or 1 ms (depending on the laser used). After transmission through a sample, some energies will be attentuated more than others. We seek the energies of the remaining photons with an accuracy that permits EXAFS. Strong EXAFS oscillations of typical materials involve periods of 20-50 eV, this being superimposed on energies of several keV, depending on the element and edge.
A straightforward variation is to position the source and a line detector on the axis of a cylinder, such that the crystal itself forms the cylinder. Arranging a large and suitable curved crystal usually defines the available practical choices. This has the advantage of collecting a large solid angle and focusing it onto a detector capable of relatively high readout rates. (In an ultrafast context a wire position detector based on propagation delays cannot be used, since all the photons effectively arrive on the wire simultaneously. A linear CCD is more appropriate.)
Subsequent variations involve conically bent or asymmetrically cut crystals to allow manipulation of throughput, resolution and energy span. Some amount of disorder (or mosaicity) in the crystal can be used to advantage, however one seeks to retain the focusing ability. The crystal thickness (mosaic or otherwise) influences both throughput and resolution, and compromises must be sought to suit particular experiments. The relevant concepts are illustrated below, where we introduce the idea of using anisotropically disordered materials. Carbon fibre is a potentially interesting material in this context, alternatively it may be possible to grow low mosaicity pyrolytic graphite on suitably microcorrugated substrates. There are many tricks to play!
 Radial refocusing to the cylinder axis requires low crystalline disorder in the cylinder tangent. |
 A thicker analyser containing crystallites disordered perpendicular to the cylinder axis (here shown as a slab in cross section) will intercept more of the Rowland focusing circle, resulting in flux gains. However, increasing thickness also spoils resolution because parallel crystallites at different depths can scatter outside the Rowland circle. |
What happens if an X-ray photon arrives at the analyser crystal at an angle that does not permit Bragg diffraction of its wavelength? It will be either absorbed by the analyser crystal, or transmitted in a straight line, according to the material's attenuation of that wavelength. Either way, it is lost as far as useful detection is concerned. (Absorption is the reason for high heat loads on the first monochromator crystal in synchrotron beamlines, while the transmitted wavelengths allow development of multiple endstations using a single synchrotron beam.) The fraction of X-ray photons discarded in this way is vastly in excess of 99.9%, despite the fact that in our plasma source, all contain useful information pertaining to the sample. This is a potentially fatal drawback when implemented at a plasma source or when considering X-ray energy dispersion in an ultrafast context; X-ray photons suitably correlated to a usefully strong optical pulse are typically scarce. As long as a diffraction analyser approach is adopted using polychromatic plasma sources, there is an inescapable practical need to make the source as bright as possible to overcome this enormous loss. It is also very necessary but tedious to quantify the loss. If only we could avoid it somehow!
It is unlikely that line gratings (comparable to those used in optics to split a white beam into a rainbow) can offer a viable alternative at "hard" X-ray energies; line spacings would need to be atomically small and positioned in a correspondingly reliable manner on a macroscopic scale. Even ignoring the fundamental complication of the impractically small scattering cross section of individual atoms for X-rays, manufacture over large areas would present daunting challenges if not a result of self-assembly.
Silicon and other point detectors are directly capable of energy resolution and have a long history in this context, using many different materials. A silicon diode for example, generates one electron hole pair for every 3.65 eV of X-ray energy, so that an X-ray photon generates a current pulse whose integrated charge is proportional to the X-ray photon energy. When applied to a laser based X-ray source in this way, a point detector can usefully detect at most one X-ray photon per laser pulse, since multiple X-rays are temporally correlated well within the resolving power of the physical processes associated with readout. Bearing in mind the need to avoid pileup, this limits photon detection rates to significantly less than the repetition rate of the laser. The approach also needs a source with low shot to shot fluctuations in its X-ray yield (or sophisticated measures to account for the fluctuations), so that the amount of pileup is predictable. It would take a uncomfortably long time to build up a chemically useful spectrum (this being the histogram of the pulse integrals) using such an arrangement on a kilohertz repetition rate laser, assuming the detector's energy resolution was otherwise satisfactory.
Phosphors or scintillators provide an intermediate conversion to light. This increases the qualitative detectability of higher energy X-rays, owing to phosphors' typically higher X-ray stopping power. Unfortunately, energy resolution of individual X-ray events is largely lost. This is because generation of visible light is not a 100% efficient process (see below), also most of the optical photons are randomly scattered, absorbed and lost on their way through the phosphor to the optical detector.
An array of point detectors each operating in single photon counting mode can be read out in the time between shots of the laser. This is the manner in which we mostly use direct-detection CCDs. X-ray spectra from single shots of a laser plasma source can be obtained using this approach, with suitable measures (filters and distance) to avoid or quantify pileup. The approach is exceedingly valuable.
We have used dark images, 55Fe and 241Am sources in combination with independent point detectors to calibrate CCD dark noise, readout noise, detection efficiency and energy resolution, as well as infer device depth structure, the extent of spectral redistribution and size of thermalised electron clouds at different energies (explained below). The characterisation has also revealed electronic quirks related to the analog to digital conversion. Most of this information was not previously known for our detector and available information was at best vague.
Applying our knowledge of the CCD to the plasma source allows inference of X-ray temperatures and spectral flux from individual laser shots. This, as well as hardware binning modes that allow monitoring of shot-by-shot general flux in real time, enabled rapid characterisation of the X-ray source as functions of many experimental variables. Use of an independent point detector over a period of several hours on the plasma source quantitatively confirmed measurements relating to both to the CCD characterisation and the plasma source.
The single shot histogram image below strongly motivates applied interest, both in the established source and the development of array detectors described below. Adding the histograms arising from multiple shots is straightforward and the clarity of the edge in the noise is quickly improved. However, given the known limitations of the CCD detector, the exercise otherwise serves no purpose.
 A single shot of the laser plasma source transmitted through a thin titanium foil can easily be arranged with an acceptable level of event pileup. |
 A histogram of the event energies from the dataset at left, showing the absorption edge of titanium (4966 eV). Various corrections can be applied when processing the image to accomodate lateral and depth aspects of the spectral redistribution function (none applied in this instance). |
If one accepts pileup, and if necessary adds the effects of multiple laser shots using long exposures or software so as to best address noise considerations, X-ray imaging becomes a straightforward bonus. In this approach one can substantially overlook the complications that led to the characterisation outlined above. The energy of individual photons sought for spectroscopy is completely lost in such images. Our laser plasma source is well suited to absorption contrast imaging, perhaps also phase contrast imaging, though it should be noted that in the context of imaging, its flux typically requires a stroboscopic approach. There are other pulsed and steady lab sources that would produce nice images faster, without requiring an ultrafast laser. Such images are nevertheless useful to us in other ways, for example the three dimensional size of our plasma source is inferred from edge blurring of high contrast objects.
 Shadow image of the head of a housefly. The thickness of the tissues is comparable to extinction lengths of the polychromatic X-rays used in biological samples, resulting in good absorption contrast. |
 Shadow image of a ~250 micron nickel mesh and a 60 W incandescent lightbulb filament. Edge blurring in this and similar high contrast images allows determination of the source dimensions. |
The resolution of good silicon point detectors is generally close to ~150 eV. This is also approximately what we observe using our direct-detection CCD for ~5-15 keV emission lines from 55Fe and 241Am sources. The underlying reason for this limitation, outlined below, motivates the procedure to overcome it.
It may seem odd that the band gap of silicon is ~1.11 eV but that 3.65 eV of energy is required to produce each electron hole pair using X-rays. What happens to the difference? If we count every last electron, can we hope for the energy resolution needed by EXAFS?
The main cross section process for X-ray interaction with matter at the X-ray energies of interest is photoelectric absorption. This results in generation of photoelectrons from any and all of the energy levels in silicon, the first result being an ionised, excited atom and a photoelectron of energy less than the original X-ray energy. Within the solid, such electrons do not go far. They quickly excite electrons out of neighboring atoms, ionising and exciting them and resulting in a complex cascade involving many different processes. The number of free electrons steadily increases, but their energy decreases, until the energies are too low to cause further excitations. The result is a small cloud of free electrons. The cloud diameter is generally on the order of a micron in in silicon, though increases with the X-ray photon energy and is different in different materials. These free electrons thermally relax to the lowest point in the semiconductor's conduction band. They are trapped in this conducting state, and are ultimately read out to give the signal we measure. The thermal relaxation of the conduction electrons (by electron-phonon coupling) as well as adjacent bond length changes around ionised atoms excites phonon modes in the lattice - in other words heat. This is where the energy difference goes! Thus, the X-ray photon's energy is distributed in two channels : one which we measure (electrons in the conduction band), and one which we can't (heat). There is some variation in the division of a given quantity of energy into these channels. The limited energy resolution when reading out the thermalised conduction electrons reflects the stochastic complexity of the relaxation process. Given the situation, it is rather surprising that a resolution of ~150 eV is actually observable using semiconductor detection. More importantly in the present context, the energy resolution needed for EXAFS will clearly never be possible by measurement of liberated conduction electrons, in any material.
In phosphors, generation of light occurs when the conduction band electrons are trapped in sites that permit optical fluorescence. The generation of photoelectrons by X-rays occurs when the not-yet thermalised charge cloud happens to intersect the surface of the material (many photoelectrons of different energies can arise from a single X-ray photon). From the argument above it will be clear that phosphoresence or photoelectron measurements are also inherently unable to provide the single photon X-ray energy resolution we seek.
To measure with the required accuracy, the only foreseeable solution is to convert the entire energy of an X-ray photon into heat, and try to measure that - a bolometric approach. To avoid trapping of energy in conduction bands, we must use a metal. We want to measure minute amounts of heat from individual X-ray photons, so will need very small pieces of metal, with very low heat capacity. Other sources of thermal noise are of no interest, further indicating very low measurement temperatures. Since conduction electrons contribute to the overall heat capacity of a solid (in fact making the dominant contribution at low temperatures), we actually seek a rather poor metal. To minimise the phonon heat capacity, there is a need to operate well below the bolometers' Debye temperature. High-Z elements are better able to stop X-rays and photoelectrically generated electrons, so are generally desirable. The bolometric element must be significantly larger than the non-thermalised charge cloud in all three dimensions, and with a thickness deep enough to stop the X-ray energies of interest. A number of bolometers operating in parallel will be needed, attached to a corresponding number of sufficiently sensitive thermometers. They would typically be arranged as an array.
At the time of writing, detectors of this kind are developed for use in space-based instruments for extraterrestrial X-ray radiation detection. Energy resolutions of better than 3 eV for 6 keV photons have been achieved, sufficient for EXAFS purposes. We have initiated collaboration with a Finnish research group (Jyväskylä), for the development and adaptation of microbolometer detectors to our ultrafast terrestrial application. Our proposal for this use of microbolometric array detectors is believed to be the first of its kind, likely to givetremendous impetus to detector development.
The great advantage of this detection scheme is that all X-ray photons within the effective solid angle subtended by the detector are usefully detected. Instead of requiring plasma sources to produce ever more broadband X-ray flux, it turns out that existing and very simple laboratory plasma sources already produce very adequate flux for this detection approach, with the immediate bonus of sub-picosecond pump-probe capability. The efficiency of detection has excellent consequences in terms of the residual laser power available for sample excitation as well as general radiation safety. Microbolometric arrays are presently at a very active stage of development, so improvements in array size and spectral resolution are almost inevitable.
Can a microbolometric array approach detect photons at a rate that allows practical experiments? Compare a synchrotron EXAFS fluorescence measurement. Such facilities often use a point detector in single photon counting mode, operating at maximum count rates ~100,000 counts per second. The point detector has enough resolution to distinguish excitation energy photons from X-ray fluorescence of the element of interest (which occur at specific lower energies), as well as from other elements and emission lines. Circuits are used to discard the unwanted events and count the fluorescence events. A full spectrum of good quality is typically built up in a few hours, by scanning the monochromatic excitation energy. For comparison, in a ultrafast laser lab setting, one could anticipate transmission experiments using a 1 kHz laser and a detector of around 100 pixels, giving a generally comparable counting rate (again there is a need to avoid pileup, however no need to discard unwanted energies). The observed rate would be adjusted to optimise signal rate vs pileup, by moving the detector towards or away from the source to subtend an appropriate solid angle. To shield the detector from photons outside the energy range of interest, a material/foil of suitable thickness (low energy cutoff) and a strong absorption edge slightly above the range of interest (high energy cutoff) would be arranged in the line of sight such that fluorescence from it (line emission) does not unduly dominate the EXAFS signal.
We expect our approach will allow routine transmission EXAFS in any ultrafast laser lab equipped with a microbolometric array detector. Pump-probe EXAFS on sub-picosecond timescales is then a straightforward extension by exciting the sample using established ultrafast optical techniques. To permit high sample excitation levels and minimise temporal dispersions arising from geometrical considerations, ultrafast samples would be placed in the most intimate possible proximity to the source. This arrangement is relatively straightforward to implement on a plasma source.
General overview and motivation for X-ray based measurements
X-ray absorption spectroscopy
X-ray Diffraction and the argumentation for broad bandwidth
Overview and motivation for ultrafast X-ray measurements
Developments done on ultrafast X-ray sources in Lund
Developments on X-ray Detectors (This page)
Measurements done on ultrafast synchrotron user facilities
Steady state measurements done on synchrotrons