Clusters as calibration Clusters as calibration toolstools
S. Molendi (IASF-MI)S. Molendi (IASF-MI)
MotivationMotivation
• To what level can we trust the calibration of our instruments?
• How far can we push spectral modelling before we end up fitting instrument systematics rather than astrophysically relevant features?
The final frontier?The final frontier?
This is a frontier that is worth exploring, major future X-ray missions are not around the corner, in many respects XMM-Newton and Chandra will be the best experiments for many years to come.
Cross Cal with ClustersCross Cal with Clusters
• Has enjoyed some success• Lots of photons and no pile-up!• No need for simultaneous
observations• Spectrum is not a simple power-
law however at the level of precision we are dealing with are there any pure pl spectra in the X-ray sky?
• Background can be a limitation
The SampleThe Sample
only EPIC to be extended to Chandra16 observations of 13 objectsspectra selected to be:• Observed in thin or medium filters• high SB• 1T • 2 < kT < 8• 0.015 < z < 0.09• long exposures
WORK IN
PROGRESS
The SampleThe Sample source ring T SB Z arcmin keV erg cm2s-1amin2 Solar units ------------------------------------------------------1. 2A0335 2 5 3.2 5.0e-13 0.46 2. A85 3 5 6.0 4.5e-13 0.39 3. A262 2 4 2.2 3.0e-13 0.31 4. A478 2 5 6.5 5.7e-13 0.31 5. A496 4 7 4.5 2.7e-13 0.31 6. A1060 2 5 3.0 4.2e-13 0.46 7. A1650 1 2 5.8 4.0e-13 0.35 8. A1795 3 5 6.0 3.6e-13 0.3 9. A2029 2 4 7.6 7.5e-13 0.37 10. A2199 2 5 4.2 7.5e-13 0.44 11. A2597 2 4 3.7 8.9e-14 0.28 12. AWM7 2 4 3.7 8.7e-13 0.28 13. MKW3s 2 4 3.6 3.2e-13 0.30
BackgroundBackground
BackgroundBackground
AnalysisAnalysis
Fit spectra with 1T and 2T models In most cases no substantial
improvement with 2T, even when improvement is substantial, typically for objects with better statystics, 2T modeling is un-physical, possibly associated to systematics
CalibrationCalibration
Consider data from each camera individualy
X-ray vs radio NX-ray vs radio NHH
● MOS1● MOS2● pn
X-ray vs radio NX-ray vs radio NHH
Investigating residuals Investigating residuals
• Explore residuals in the form of ratio data/model
• Heavily regroup data (beyond resolution limit) to achieve few % errors
2A 0335 Ratio data/model2A 0335 Ratio data/model
Investigating residuals Investigating residuals
• Compute mean residuals averaged over all 16 observations
• Statistical errors are reduced, systematics should show up
Deviations from a simple Deviations from a simple thermal model thermal model
MOS 1
Deviations from a simple Deviations from a simple thermal model thermal model
MOS 2
Deviations from a simple Deviations from a simple thermal model thermal model
pn
Stability of residuals Stability of residuals
Do different objects show similar residuals?
ColdCold vsvs Hot Hot
MOS2
ColdCold vsvs Hot Hot
pn
SampleSample vsvs Perseus Perseus
MOS1
SampleSample vsvs Perseus Perseus
MOS2
SampleSample vsvs Perseus Perseus
pn
Investigating residuals Investigating residuals
• Residuals at least in some instances appear to be similar in different objects
• Spectral model 1T, hot cold, 4T for Perseus -- unlikely
• Instrument systematics
• Redistribution -- rmf
• Effective Area -- arf
pn soft band residuapn soft band residua
Centaurus
pn soft band residuapn soft band residua
Perseus
pn soft band residuapn soft band residua
Mkn 421
mos soft band residuamos soft band residua
Centaurus
mos soft band residuamos soft band residua
Perseus
Implications Implications
• Systematics likely related to rmf, particularly true for pn
• MOS less certain
De Grandi+SM 09
Cross-CalibrationCross-Calibration
Compare different instruments
MOS1 & pn NH vs MOS2 MOS1 & pn NH vs MOS2 NHNH
● MOS1● pn
MOS1 & pn kT vs MOS2 MOS1 & pn kT vs MOS2 kTkT
● MOS1● pn
MOS1,MOS1, MOS2 MOS2 && pnpn vsvs pn pn modelmodel
MOS1,MOS1, MOS2 MOS2 && pnpn vsvs pn pn modelmodel
MOS1,MOS1, MOS2 MOS2 && pnpn vsvs pn pn modelmodel
MOS1 MOS1 && MOS2 MOS2 vsvs pn modelpn model
SummarySummary
Calibration• Redistribution problem on pn• MOS, if there is one it’s smaller
Cross-calibration
• Good to 3% in 0.7-3.5 keV band
• problem with Effective areas at energy
ACIS S3 vs EPIC pnACIS S3 vs EPIC pn
• We compare Chandra ACIS S3 with EPIC pnWe compare Chandra ACIS S3 with EPIC pn• About 3×10About 3×1066 events for each spectrum events for each spectrum
extracted from annulus with bounding radii of 1' and extracted from annulus with bounding radii of 1' and 2 2 ‘‘
• Used old and new Chandra calibrations Used old and new Chandra calibrations (CALDB 4.1.1 with hrmaD1996-12-20axeffaN0008.fits)(CALDB 4.1.1 with hrmaD1996-12-20axeffaN0008.fits)
• Multi T spectral model (Molendi & Gastaldello Multi T spectral model (Molendi & Gastaldello
09)09)
ACIS S3 vs EPIC pnACIS S3 vs EPIC pn• We start from the EPIC pn spectrum We start from the EPIC pn spectrum
ACIS S3 vs EPIC pnACIS S3 vs EPIC pn• We start from the EPIC pn spectrumWe start from the EPIC pn spectrum• Perform fit with multi T modelPerform fit with multi T model
ACIS S3 vs EPIC pnACIS S3 vs EPIC pn• We start from the EPIC pn spectrumWe start from the EPIC pn spectrum• Perform fit with multi T modelPerform fit with multi T model• Fold best fitting model with ACIS Fold best fitting model with ACIS
response and compare it with ACIS response and compare it with ACIS spectrumspectrum
ACIS S3 vs EPIC pnACIS S3 vs EPIC pnPlot residuals in the form of ratio data/modelPlot residuals in the form of ratio data/model
Renorm applied to match spectra at 1.5 keV Renorm applied to match spectra at 1.5 keV
pnpn
ACIS S3ACIS S3
ACIS S3 vs EPIC pnACIS S3 vs EPIC pnPlot residuals in the form of ratio data/modelPlot residuals in the form of ratio data/model
Renorm applied to match spectra at 1.5 keV Renorm applied to match spectra at 1.5 keV
pnpn
ACIS S3ACIS S3
ACIS S3 vs EPIC pnACIS S3 vs EPIC pnSimilar result when using a different region Similar result when using a different region
Annulus with Annulus with bounding radii of 2' and 3 ‘bounding radii of 2' and 3 ‘Showing only plot with new ACIS calibrations
pnpn
ACIS S3ACIS S3
Cross CalCross Cal
pn and ACIS S3 spectral shapes are pn and ACIS S3 spectral shapes are now in much better agreement! now in much better agreement!
Differences are almost everywhere Differences are almost everywhere less than ~5%less than ~5%
Major discrepancy in 0.7-1.0 keV Major discrepancy in 0.7-1.0 keV rangerange
ACIS S3 vs EPIC pnACIS S3 vs EPIC pnResiduals in the form of ratio data/modelResiduals in the form of ratio data/model
Renorm applied to match spectra at 1.5 keV Renorm applied to match spectra at 1.5 keV
pnpn
ACIS S3ACIS S3
ACIS S3 vs EPIC MOSACIS S3 vs EPIC MOSMOS2 appears to be more similar to ACIS in the MOS2 appears to be more similar to ACIS in the
0.7-1 keV band. 0.7-1 keV band.
MOS1 appears to be somewhere btwn. MOS2 MOS1 appears to be somewhere btwn. MOS2 and pn.and pn.
MOS2MOS2ACIS S3ACIS S3
Flux crossFlux cross-cal pn vs -cal pn vs ACISACIS
• Both figures have renorm factors: 5% for Both figures have renorm factors: 5% for the first; 15% for the second: let’s take the first; 15% for the second: let’s take them out.them out.
Flux crossFlux cross-cal pn vs ACIS-cal pn vs ACIS• Both figures have renorm factors: 5% for Both figures have renorm factors: 5% for
the first; 15% for the second: let’s take the first; 15% for the second: let’s take them out.them out.
Flux crossFlux cross-cal-cal
• The new HRMA calibration impacts on the The new HRMA calibration impacts on the ACIS/EPIC flux cross calibration.ACIS/EPIC flux cross calibration.
• Our analysis indicates that the flux cross Our analysis indicates that the flux cross calibration below calibration below ~ 2 keV ~ 2 keV will be shifted will be shifted by about 10%. by about 10%.
• Although comparing spectra extracted Although comparing spectra extracted from a given region of a cluster may not from a given region of a cluster may not be the best way to go, our data indicates be the best way to go, our data indicates that the flux cross calibration change is not that the flux cross calibration change is not for the better.for the better.
SummarySummary• The The newnew HRMA effective area reduces HRMA effective area reduces
ACIS S3 vs pn residual calibration errors ACIS S3 vs pn residual calibration errors to less than 5%to less than 5%
this is no small achievement!this is no small achievement!• Major remaining discrepancy in 0.7-1 Major remaining discrepancy in 0.7-1
keV bandkeV band• The new spectral calibration modifies by The new spectral calibration modifies by
about 10% Chandra fluxes below 2 keV, about 10% Chandra fluxes below 2 keV, ACIS vs EPIC flux cross calibration will ACIS vs EPIC flux cross calibration will be affectedbe affected