IACHEC multi-mission cross-calibration procedure Matteo Guainazzi (1), Steve Sembay (2), Andy Pollock (1) (1) ESAC-ESA (2) LUX v0.1 - First draft - 14 June 2012 v1.0 - First version posted to the IACHEC web page - 02 Jul 2012 1. Scope of this document ------------------------- At the 7th IACHEC meeting in Napa (California), it was agreed that times are ripe to try a common cross-calibration exercise which may allow us to achieve progress in the mutual agreement among astrophysical observables' measurements. This document suggests a possible procedure that such an exercise may follow. It is based on a simplified version of the "Calibration Perturbation Method" (CPM) described in Lee et al. (2011). It is reckoned that an integral application of the CPM to the whole ensemble of basic calibration elements would be the best path to understand which calibration elements play the most important role in shaping the existing cross-calibration differences, as well as to estimate the systematic uncertainties on the measurement of astrophysical parameters due to calibration systematic uncertainties. However, the development of a full CPM system for most of the operating missions is deemed unpractical or unfeasible, due to the limited (and shrinking) available resources to calibration, to the loss technical expertise, and to the lack or poor maintenance of ground calibration data. For a certain number of calibration elements, any estimate of the most likely "systematic uncertainties" could be just a wild guess, implying the risk of yielding largely overestimated, and therefore useless for any practical purposes, final systematic errors. 2. Conceptual scheme -------------------- The proposed procedure can be conceptually summarised in the following basic steps: - Definition of the cross-calibration problem - Definition of the cross-calibration platform - Definition of the calibration elements to be tested - Joint fit of astrophysical and calibration parameters - Verification of the calibration results on a control sample While for easiness of presentations the steps above are described separately , they are not intended to be executed in a strict temporal sequence, neither to be carried out in isolation. 2.1 Definition of the cross-calibration problem ----------------------------------------------- At the time of writing this document, cross-calibration systematic uncertainties in the absolute flux determination are at least at the +/-10% level over the whole sensitive energy bandpass of the operational missions. While studies of galaxy clusters (Nevalainen et al. 2010) and non-thermal SuperNova Remnants (Tsujimoto et al. 2011) suggest a good agreement in spectral shape in the 2-10 keV energy band, elsewhere the cross-calibration differences are energy-dependent, yielding differences in the best-fit values of astrophysically important parameters (such as temperatures in clusters) as large as 30%. It is potentially hazardous treating these differences in "isolation", because, e.g., for any given instrument the calibration of the effective area in one band may affect the calibration of the redistribution/line spread function in another. Once again, a comprehensive approach, whereby all the calibration elements are fit simultaneously to the source astrophysics over the whole sensitive bandpass, would be the ideal approach. In what follows we will assume that it is possible to isolate a specific cross-calibration issue, and try and solve it by modifying that/those calibration element/s which most directly affects/ it. Any calibration improvements must be "anchored" to the cross-calibration status in those bands where the agreement is best, either explicitly during the common fitting procedure, or a posteriori, by verifying that the improved calibration does not spoil previously existing cross-calibration agreements. 2.2 Definition of the cross-calibration platform ------------------------------------------------ By "platform" I intend here the combination of the following elements: a) a sample of sources to be used for the common calibration exercise; b) an agreed upon astrophysical or empirical model for each source in the calibration sample; c) a set of common data reduction criteria, including: 1. a set of common Good Time Intervals (GTIs) if strict simultaneity is a requirement: 2. extraction regions for spectral products which encompass as close to 100% of the source counts as is feasible; 3. a common set of spectral analysis priors on abundances, model for photoelectric absorption, goodness-of-fit criterion, statistical confidence, common energy ranges: d) a "control sample" of sources (possibly belonging to different classes of astrophysical objects from the calibration sample sources), on which the new calibrations shall be tested. 2.3 Definition of the calibration elements to be tested ------------------------------------------------------- Each of the Instrument Teams (ITs) identify one or more the calibration element(s), that it would like to investigate during the common calibration exercise. The choice should reflect a working hypothesis on which elements may primarily affect the cross-calibration issue being tackled. Each elements should be parametrised through the smallest possible number of free parameters, without losing generalities. Limiting the number of independent calibration elements will help the convergence of the method. In summary, each IT will contribute to the common exercise with: 1) a set of spectra (extracted following the reduction procedure laid down in Sect.2.2) together with their response file 2) one/more parametrised calibration element/s 2.4 Joint fit of astrophysics and calibration --------------------------------------------- The common spectra and parametrised calibration elements will be jointly fit together. In the fit only the parameters describing the calibration elements will be left free (together with a constant overall normalisation constant, if this is not supposed to be embedded in the parametrisation of the calibration elements). It goes without saying that the parameters describing the source astrophysical parameters shall be left free and tied together across the different spectra. Example of software machinery implementing this workflow can be found in, e.g., Sembay (2010). The results of the fit will be: - joint parameters of the astrophysical model - best-fit values of the parameters describing the parametrised calibration element(s) in both cases with their statistical errors at the agreed confidence level. 2.5 Verification of the calibration results on a control sample --------------------------------------------------------------- Each IT will apply the calibration files corresponding to the parameter confidence intervals determined in Sect.2.4 to the control sample. On the bases of the quality of the astrophysical results on the control sample, the original working hypothesis will be validated, on which the choice of the calibration element to be tested was based. If the new calibration files worsen the quality of the residuals in the control sample, the IT will change its working hypothesis, another calibration element(s) (or a combination thereof) shall be proposed, and the corresponding parametrised calibration file shall be brought brought to the common exercise. The iteration cycle as outlined above will end only when ALL parametrised calibration files resulting from Step#2.4 are validated on the control sample. It goes without saying that no calibration files based on the result of the common cross-calibration exercise shall be released to the public, before the validation step is successful for all the instruments participating in the exercise. 3. Example workflow ------------------- This example aims at extending a conceptualised version of the algorithm used to determine the time- and position-dependent redistribution of the EPIC-MOS cameras as described in, e.g., Sembay (2010). **Warning**: this is to be intended as a mere example, with little, if any, reference to any existing EPIC calibration activity. Problem: improving the energy-dependent cross-calibration between EPIC-MOS and EPIC-pn in the soft energy band Constraints: - maintain the good agreement in spectral shapes between EPIC-MOS and EPIC-pn above 2 keV (Nevalainen et al. 2010, Tsujimoto et al. 2011) - maintain an average flux ratio between EPIC-MOS and EPIC-pn below 2 keV at the SASv11 level (Stuhlinger et al. 2011) Calibration sample: a set of 4 observations performed in EPIC-MOS "Epoch#12" or later. - two line-rich sources (including 1E0102-72) at different positions with respect to the boresight redistribution patch - two radio-loud AGN (w/ and w/o soft excess) Astrophysical models: - IACHEC model for 1E0102-72 - RGS-based model for the other line-rich source (e.g., active star from the XMM-Newton Routine Calibration Plan) - Exponentially curved power-law model for the AGN Reduction: - GTI . optimising the signal-to-noise ratio for the line-rich sources . in addition, strictly simultaneous for the AGN blazars - Extraction regions: . covering the same physical area of 1E0102-72 . covering the same fraction of the Point Spread Function for the point-like sources - Anders & Grevesse (1983) abundances - "tbabs" model for the photoelectric absorption - Cash statistics goodness-of-fit test - Statistical errors at the 1 sigma level for one interesting parameter - Fit in the energy range 0.2-10.0 keV (signal-to-noise permitting) Calibration elements: - EPIC-MOS: . redistribution . effective area - EPIC-pn . redistribution Calibration parametrisation: - EPIC-MOS redistribution: ratio between the main and the secondary photo-peak - EPIC-MOS effective area: depth of the C molecular contaminant - EPIC-pn redistribution: height and width of the Compton-shoulder Control sample: The 100+ observation in the XMM-Newton cross-calibration database (XCAL, Stuhlinger et al., 2011) Validation benchmark: - stacked residuals of XCAL power-law sources - distribution of "Method 2" XCAL fluxes - hard power-law index MOS vs pn distributions in XCAL blazars - stacked residuals of strong soft X-rays emission line profiles in line-rich XCAL sources - overall goodness-of-fit criterion value distributions on the whole XCAL sample - He- and H-like Oxygen/Neon emission line normalisations against the 1E0102-72 IACHEC results (Plucinsky et al. in preparation) References ---------- - Lee H., et al., 2011, ApJ, 731, 126 - Nevalainen J., et al., 2012, A&A, 523, 22 - Sembay S., 2010, presentation at the 2010 IACHEC meeting: http://web.mit.edu/iachec/meetings/2010/Presentations/Sembay_rmf.pdf - Stuhlinger M., et al., 2011, XMM-SOC-CAL-TN-0052: http://xmm2.esac.esa.int/docs/documents/CAL-TN-0052.ps.gz - Tsujimoto M., et al., 2011, A&A, 525, 25