Characterization of Chandra's Fe-55 External Calibration Source using Microcalorimetry
ACIS and Filters:
The ACIS consists of two different arrays of CCDs . One array is specialized for spectroscopy, while the other is specialized for imaging. Because CCDs are also sensitive to visible and UV light, Chandra contains visible/UV light blocking filters located about 2cm above the CCD focal plane. These filters are made of a 200nm substrate of free-standing polyimide and coated with 160 nm of aluminum over the imaging portion of the array and 130 nm of aluminum over the spectroscopy array.(1) The X-ray transmission through the filters was calibrated here at UW using the synchrotron light source before Chandra’s launch.
| Light-Blocking Filters || ACIS CCDs |
ACIS Calibration source:
In order to monitor the potential build up of materials on the CCDs or the light-blocking filters, Chandra contains an External Calibration Source (ECS). The ECS is composed of three separate radioactive Fe-55 sources that illuminate the entire ACIS focal pane from a distance of ~46cm. The focal plane is exposed to the calibration source only when the ACIS is out of the imaging point of the mirrors (the HRC is at the imaging point). One source emits Mn K and L photons directly illuminating the focal plane. The other two sources use Mn K-alpha and Mn K-beta photons to fluoresce either the Ti or Al target materials and excite K emission lines.(2) Any possible materials building up on the filters or CCDs would be transparent to the 5.9 keV Mn K-alpha photons. However, the materials will absorb some of the L-complex (Mn L-Fe L) emissions at .67 keV. Therefore, if the ratio of L-complex (Mn L-Fe L) flux to Mn K-alpha flux decreases this indicates that contaminants are ac cumulating on the ACIS and visible\UV light blocking filters
The first indication that there indeed were contaminants building up on either the CCDs or the filters was in May of 2000. It was then determined that out gassed hydrocarbon contaminants were permeating the outer filters, which are approximately at room temperature, and freezing on the inner filters which are cryogenically cooled. Prior to launch, Chandra planned to baked off any potential contamination using an onboard heater. Chandra has already preformed four bake offs in early 1999. However, in September of 2004, another bake off was deemed too risky for a number of reasons.
Since the heaters provided the only chance of getting rid of the contaminants, the only option that remains is to correct for the absorption in our observations. This requires a very accurate model of its distribution and chemical components. We can learn about such details by shining the on board Fe-55 calibration sources, through the contaminated filters, on to the CCDs. By comparing the expected spectrum of the source to the spectrum that is obtained when some of the photons are being absorbed, we can then learn about the contaminants that are absorbing those photons on the filter. Therefore is important to have a very accurate spectrum of the unimpeded Fe-55 source so that even subtle absorption features can be detected. In order to determine the strength of the Fe-55 emissions accurately, we must use detectors that are capable of sufficiently high energy resolution such as a microcalorimeter.
| Instead of measuring the charge liberated by an incident x-ray to determine its energy (like CCDs), microcalorimeters measure the energy by determining the rise in temperature of the absorbing material due to the thermalization of the photon. A microcalorimeter consists of three main components: an absorber, a thermometer, and a weak thermal link to a heat sink. When an x-ray strikes the absorber it will create a photoelectron, which then thermalizes its energy and heats up the absorber. The thermometer then measures the increase in temperature. The energy is proportional to the heat capacity multiplied by the change in temperature of the absorber. (E=C*∂T). The absorber then cools back down to its equilibrium temperature with a time constant Τ = C/G, where C is the heat capacity and G is the thermal conductance of the wink link.
Because microcalorimeters don’t rely on the inefficient conversion of energy to electric changes, they are able to detect the energy of the x-ray much more accurately. This means that microcalorimeters have much better energy resolution the CCDs. In addition, they also have a much higher quantum efficiency.
Absorbers must have a low heat capacity so that the x-rays will produce a measurable change in energy. Additionally, the absorbers must be kept at milli-Kelvin temperatures. In order to detect individual x-rays, the temperature of the absorber must be small relative to the energy of the incident x-ray.
Cooling down the detectors using the ADR:
The detectors are cooled down in various “stages” in order to reach such extreme temperatures. The outermost stage cools the detectors to 77K using liquid Nitrogen (the boiling point of Nitrogen at 1 atm). The second stage uses liquid Helium to reduce the temperature to 4.2K (the boiling point of He at 1 atm). The next stage involves pumping on the He which reduces the temperature to 1.4K. Lastly, an Adiabatic Demagnetization Refrigerator (ADR) is then used to cool the detectors down to 50mK.
The process of adiabatic demagnetization involves aligning the magnetic moments of molecules in a paramagnet salt (salt pill) using a strong magnetic field. By slowly ramping down the magnetic field, the magnetic moments
|of the molecules become unaligned as they absorb thermal energy and convert it in to the entropy (disorder) of the random magnetic moment alignment. The thermal energy is then dumped into a heat bath (the cold stage of the ADR) which decreases the temperature of the detectors to the desired 50mK temperature. For more information about how the ADR works check out: NASA and Wikipedia . During my research I learned how to prepare the fridge for the cool down (a very long process!) as well as how to maintain and operate it once it was cool.|
Once the detectors and the Fe-55 calibration source were sealed inside of the fridge, they were placed inside a faraday cage which shields it from external electromagnetic radiation that causes the fridge to warm up. The fridge's pre-amp box was then connected to a computer that runs specialized programs that allowed us to record each incident photon on the detector from the Chandra source. In addition to x-ray pulses, we also are interested in recording samples of noise to later compile a baseline.
The sweep program continuously takes data at a rate of about 40,000 samples/sec. However, these samples are only recorded by the computer when a trigger occurs. The trigger is set to go off when it detects a certain change in voltage during the decay of the x-ray pulse. This trigger tells the computer to record a certain amount of time before and after the trigger so that the window for the pulse is about 30ms. There are also continuous triggers that record data approximately once every second regardless of whether or not an x-ray pulse is present. In fact, it is preferable not to have an x-ray during this kind of trigger, because its purpose is to take a measurement of the noise in the absence of a pulse.
Other Data Taking:
Inside the lab ADR there were also three additional HgCdTe absorbers (sent to UW from EPIR ) and a second Fe-55 radiation source. These absorbers were sent to UW so that we could measure their heat capacity and energy resolution. I also was involved taking data for the characterization these absorbers.
Classifying the Pulses:
The sweep program takes about 4096 samples (a sample is recorded every time a trigger occurs). Ideally each of these samples would contain either an x-ray pulse or a measurement of the noise in the absence of a pulse. However, due to the inaccuracy of the trigger mechanism some of these samples may contain multiple pulses or an accidental trigger. Some of the samples may also contain pulses with bad rise times, a pre-trigger voltage that is too big/small (which usually occurs when a pulse occurred right before it and the temperature didn’t have enough time to return to equilibrium before the next pulse), or a problem with the DC level of the pre-trigger region. Using a filter program to compile all of the samples, parameters are then set to distinguish the x-ray pulses from the noise. The pulses are further classified as good pulses or pulses containing one of the above specified problems. This is necessary to ensure that only good pulses are included in the final spe ctrum.
Energy Resolution and Pulse Height Analysis:
|Because the experiment can not perfectly detect each of the Fe-55 emission lines, there is a limit on the energy resolution that can be achieved. One irreducible source of noise results from the weak thermal link in which there is a random exchange of energy between the absorber and the heat sink.(4) This noise provides a background in which the signal must be measured but it does not limit the accuracy of the measurement. Simply, subtracting the baseline from the pulse height does not provide very good energy resolution. In order to more accurately determine the pulse height (and hence the energy of the x-ray), we must look at the whole pulse in the frequency domain.|
Using a Fourier transform to look at the power spectrum of an exponential pulse and the power density spectrum of the thermodynamic noise fluctuation in the frequency domain we can see why this is so. Both the signal and noise have the same shape. This means that the signal to noise ratio in each frequency bin (of the same width) will be the same. Since the noise in different bins in the frequency domain is (generally) unrelated, each bin provides an estimate of the signal. As long as the signal to noise is constant, the resolution of the signal will improve as the square root of the bandwidth. Ideally, the uncertainty in the energy will approach zero as the number of bins averaged increases.
However, in reality there are limitations that prevent us from doing this. One limitation is the presence of an additional noise source (usually from the thermometer) that will become larger then the Thermodynamic Fluctuation Noise at some point. In addition, something that causes the signal to fall faster then the noise will reduce the signal to noise ratio. This means that there is an optimal bandwidth that can be used to estimate the signal. The process of using this optimal bandwidth to estimate the energy of the x-ray is called optimum filtering. Each pulse is optimally filtered before it can be included in the spectrum.
Expected Relative Emission Abundances:
Before compiling the x-rays in to a spectrum, we can use atomic transition and fluorescence probabilities to get an idea of the relative emission abundances we expect to see from the Fe-55 source. From the chart, one can see that for Mn with an atomic number of 25 we expect to see mostly Mn K photons and very few Mn L photons. In addition, the source is also capable of producing a very small number of Fe-L lines. About 99% of all decays will leave a K-shell vacancy due to the capture of a K-shell electron. About 1% of all decays will leave an L-shell vacancy due to the capture of an L-shell electron. These vacancies can either be filled by transitions that result in the emission of an x-ray and/or Auger electron . The probability that these vacancies will result in certain photon energies gives us an estimate of the relative line strengths.
These work out to be:
Kb = 13.6% Ka
Each L line = .1 - 2.5% Ka
| The emitted Auger electrons scatter and lose various amounts of energy and when they hit the detector, they produce events that are indistinguishable from x-rays. Onboard Chandra these electrons are blocked by the optical blocking filters, but in order to get the relative line intensities without absorption correction we are not using a filter. The blue in the ajacent figure shows the Fe-55 spectrum and the quasi-continuous background produced by Auger electrons.
The microcalorimeter detectors are mounted on a cold plate inside of the ADR and placed in front of a flight spare 55Fe calibration source. To prevent Auger electrons from reaching the detector, a “magnetic trap” was installed between the detectors and the source. Electrons are deflected by Neodymium-Iron-Boron magnets with a field of ~2000 gauss and are then trapped by four levels of BeCu baffles that are aligned parallel to the source/detector.
A close up of the L lines in the Fe-55 Spectrum
Current models of the L-complex line have been fit assuming equal contributions from the average Mn and Fe L lines. Although the data has yet to be corrected for detector efficiency and gain drifts, it seems safe to conclude the Fe L lines are actually less than 10% of the Mn L lines. Surprisingly, we have also found that F Ka (677 eV) is an important constituent of the L-complex line (about 0.38% of Mg Ka or about 115% of Mn Laß). The complex at 550eV (seen in Figure 1) had been speculated to contain bend of O Ka and Cr L emissions, but we find no evidence of the O Ka line from the source. There is a reasonably strong line consistent with Mn L1 at 552eV (at about .05% of Mn Ka) and the Cr L lines are weak (less than .016% of Mn Ka).
References:1 Advanced CCD imaging spectrometer (ACIS) instrument on the Chandra X-ray Observatory- G. P. Garmire*a, M. W. Bautz**b, P. G. Fordb, J. A. Nouseka, and G. R. Ricker Jrb aThe Pennsylvania State University; bMassachusetts Institute of Technology
4 Thermal Equilibrium Calorimeters- An Introduction –Dan McCammon
Acis and Filter: www.astro.psu.edu/xray/axaf/acis.html
Contaimination plot: http://cxc.harvard.edu/cal/Acis/Cal_prods/qeDeg/index.html
ADR: cryo.gsfc.nasa.gov/.../ADR_intro/ADR_intro.html and constellation.gsfc.nasa.gov/.../xms.html
Frequency space graph: “Thermal Equilibrium Calorimeters” - An Introduction Dan McCammon
Fluorescence Probabilities: http://xdb.lbl.gov/Section1/Sec_1-3.html