Embed Size (px)
Citation preview
Department of Physics, Chemistry and Biology
Master’s Thesis
Evaluation of Flux and Timing Calibration of theXMM-Newton EPIC-MOS Cameras in Timing
Mode
John-Olov Larsson
LiTH-IFM-EX-08/1953-SE
Department of Physics, Chemistry and BiologyLinkopings universitet, SE-581 83 Linkoping, Sweden
Master’s ThesisLiTH-IFM-EX-08/1953-SE
Evaluation of Flux and Timing Calibration of theXMM-Newton EPIC-MOS Cameras in Timing
Mode
John-Olov Larsson
Adviser: Dr. Marcus G.F. KirschEuropean Space Astronomy Centre, ESAC
Examiner: Prof. Leif JohanssonDepartment of Physics, Chemistry and Biology
Linkoping, 12 May, 2008
Avdelning, InstitutionDivision, Department
Thesis DivisionDepartment of Physics, Chemistry and BiologyLinkopings universitet, SE-581 83 Linkoping, Sweden
DatumDate
2008-05-12
SprakLanguage
� Svenska/Swedish
� Engelska/English
�
�
RapporttypReport category
� Licentiatavhandling
� Examensarbete
� C-uppsats
� D-uppsats
� Ovrig rapport
�
ISBN
ISRN
Serietitel och serienummerTitle of series, numbering
ISSN
URL for elektronisk version
TitelTitle
Utvardering av flodes- och tidskalibrering av XMM-Newton EPIC-MOSKamerorna i Timing Mode
Evaluation of Flux and Timing Calibration of the XMM-Newton EPIC-MOS Cam-eras in Timing Mode
ForfattareAuthor
John-Olov Larsson
SammanfattningAbstract
XMM-Newton is a X-ray telescope launched december 1999, by the EuropeanSpace Agency, ESA. On board XMM-Newton are two EPIC-MOS X-ray detectors.The detectors are build by Chharged Coupled Devices (CCDs), of Metal OxideSemi-conductor type. The EPIC-MOS cameras have four science operatingmodes. This project aims to evaluate the calibration for one of these four modes,the timing mode.
The evaluation is divided into two parts. The first part is the evaluation of theflux calibration, performed by analysing various observation made in timing mode.The second part is the evaluation of timing properties by performing timing anal-ysis of XMM-Newton observations of the Crab nebula compared to observationsmade in the radio wavelengths.
NyckelordKeywords
XMM, Newton, Calibration, EPIC, MOS, camera, timing mode
�
http://www.ep.liu.se
—
LiTH-IFM-EX-08/1953-SE
—
Abstract
XMM-Newton is a X-ray telescope launched december 1999, by the EuropeanSpace Agency, ESA. On board XMM-Newton are two EPIC-MOS X-ray detec-tors. The detectors are build by Chharged Coupled Devices (CCDs), of MetalOxide Semi-conductor type. The EPIC-MOS cameras have four science operatingmodes. This project aims to evaluate the calibration for one of these four modes,the timing mode.
The evaluation is divided into two parts. The first part is the evaluation of theflux calibration, performed by analysing various observation made in timing mode.The second part is the evaluation of timing properties by performing timing anal-ysis of XMM-Newton observations of the Crab nebula compared to observationsmade in the radio wavelengths.
v
Acknowledgements
Throughout the work of this thesis many people have given invaluable contribu-tions. First of all, many thanks to my supervisor Dr. Marcus G.F. Kirsch whocontributed with a theoretical framework as well as practical sugestions duringthe project. Without you, this thesis would not be. Also, a great deal of thanksto Guillermo Buenadicha and Steve Sembay for contributing with valuable infor-mation and suggesting solutions to my work. I am thankful for the rewardingdiscussions with Dr. Martin Stuhlinger and Dr. Andy Pollock about theoreticaland practical astrophysics, it really helped the progress of my work.
Financial support was given through the project by the Human Resourcesdepartment at the European Space Astronomy Centre (ESAC). For the timinganalysis, radio data from the Jodrell Bank Centre of Astrophysics, operated bythe University of Manchester was used.
The trainees at ESAC owe a special mentioning for making my stay in Madridpleasant and unforgettable, and for helping whenever a problem occured.
At last, for always being there for me, and always supporting me during gooddays as well as bad days. Without you Irene, and your loving thoughts, I wouldnever have had the same determination. Thank you.
vii
Abbreviations
Abbreviation Explanation
CCD Charged Coupled DeviceCCF Current Calibration FileCTE Charge Transfer EfficiencyEPIC European Photon Imaging CameraERMS EPIC Radiation Monitor SystemESA European Space AgencyESAC European Space Astronomy CentreFF Full FrameFITS Flexible Image Transport SystemFOV Field Of ViewFPA Focal Plane AssemblyFWHM Full Width Half MaximumGTI Good Time IntervalLW Large WindowMJD Modified Julian DayMOS Metal Oxide Semi-conductorNASA National Aerounotics and Space AdministrationODF Observation Data FileOM Optical MonitorPSF Point Spread FunctionQE Quantum EfficiencyRGA Reflection Grating ArrayRGS Reflection Grating SpectrometerSAS Scientific Analysing SystemSW Small WindowTJD Truncated Julian DayTU Timing UncompressedXARV Package of scripts to analyse XMM-Newton data filesXCAL XMM-Newton Cross Calibration ArchiveXMM X-ray Multi-mirror MissionXSA XMM-Newton Science ArchiveXSPEC An X-ray Spectral fitting software developed by NASA
ix
Contents
1 Introduction 11.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 The Satellite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.3 The EPIC instrument . . . . . . . . . . . . . . . . . . . . . . . . . 31.4 Aim of the Project . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 The EPIC-MOS Cameras 52.1 The EPIC-MOS Operating Modes . . . . . . . . . . . . . . . . . . 6
3 Relevant Calibration Topics 93.1 Point Spread Function . . . . . . . . . . . . . . . . . . . . . . . . . 93.2 Pile-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103.3 Exposure Time Corrections . . . . . . . . . . . . . . . . . . . . . . 103.4 Effective Area . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113.5 Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123.6 Quantum Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . 123.7 Charge Transfer Efficiency . . . . . . . . . . . . . . . . . . . . . . . 133.8 Redistribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
4 Data Analysis 154.1 X-ray Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.1.1 Spectral Fitting . . . . . . . . . . . . . . . . . . . . . . . . . 154.1.2 Timing Analysis . . . . . . . . . . . . . . . . . . . . . . . . 16
4.2 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174.2.1 XCAL and Spectral Fitting . . . . . . . . . . . . . . . . . . 174.2.2 Timing Analysis . . . . . . . . . . . . . . . . . . . . . . . . 21
4.3 Statistical Evaluation of the Flux Measurements . . . . . . . . . . 24
5 Results 275.1 Results of Flux Calibration Evaluation . . . . . . . . . . . . . . . . 275.2 Results of Relative Timing Evaluation . . . . . . . . . . . . . . . . 285.3 Results of Absolute Timing Evaluation . . . . . . . . . . . . . . . . 31
xi
xii Contents
6 Future Work 336.1 Flux Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336.2 Timing Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Bibliography 35
Chapter 1
Introduction
1.1 Background
On December 10, 1999 the X-ray Multimirror Mission, XMM-Newton space ob-servatory was placed in orbit by an Ariane-V rocket [1] launched from Kourouin French Guyana. The orbit of XMM-Newton is highly eliptic with a perigee1
between 7000km− 22000km, an apogee2 between 100000km− 114000km and aninclination3 of ∼ −40◦ [2]. The orbit period is ∼ 48 hours. The satellite is trav-eling to almost one third of the distance to the moon [3]. The orbit was mainlychosen to allow long uninterupted observations which is improving the quality ofthe measurements, but also allow cooling of the five main X-ray cameras between−80◦C and −100◦C using only passive radiators. Many objects in the universecreate X-ray radiation. But since the earth’s atmosphere blocks all X-rays, a satel-lite is needed to help scientists to understand what happened in the past whenstars were born, or how the formation of galaxies occured. XMM-Newton can ob-serve sources such as for example hot stars, active or normal galaxies, black holes,neutron stars or supernova remnants.
1.2 The Satellite
XMM-Newton with its 3.8 tonne heavy and 10 metres long body [1] is the largestscientific observatory developed by ESA, and is dedicated to explore the universein X-ray wavelengths between 1A−120A (12keV − 0.1keV ) [4]. The satellite con-sists of three main instruments, the European Photon Imaging Camera (EPIC)[7], [8], for imaging and spectral analysis in the X-ray wavelengths, the ReflectionGrating Spectrometer (RGS) [5], for analysing X-ray spectra with high resolutionand the Optical Monitor (OM) [6], which is a camera for imaging and spectroscopy
1Perigee is the shortest distance between the earth and the satellite of the eliptical orbit2Apogee is the longest distance between the earth and the satellite of the eliptical orbit3The inclination is defined as the angle between the plane of the orbit of the satellite and
earth’s rotational plane
1
2 Introduction
Figure 1.1. Schematic view of XMM-Newton
of optical/UV light.
The satellite is built by two large modules connected by a long carbon fibretube. See figure 1.1. In the front module the OM and three telescopes are situated,of which two are equipped with Reflection Grating Arrays (RGA). The telescopesare Wolter type 1 and consist of 58 gold coated mirrors. On the front module alsotwo star trackers are situated. These are used for positioning of the spacecraft.The focal length is 7.5m and the telescope is focusing the X-rays on the rearmodule where the Focal Plane Assembly (FPA) is connected. On the FPA theRGS and EPIC are situated, but also units for data handling, power supply forthe cameras and radiators for cooling. The latter sits on the back of the FPA,pointing out from the spacecraft. In the front all EPIC cameras have identicalsections containing filter wheel, door, calibration source and radiation shielding.On board the spacecraft is also the EPIC Radiation Monitor System (ERMS)whose purpose is to measure radiation levels from earths radiation belt4 or fromsolar flares since these will pertubate the scientific instruments and cause unreliablemeasurements.
4Earth’s radiatio belt or the Van Allen radiation belt is a torus of energetic charged particles,held around the earth by its magnetic field
1.3 The EPIC instrument 3
1.3 The EPIC instrument
The EPIC consists of three cameras, designed for imaging observations and spec-troscopy. Each EPIC camera has three sections: the stand-off structure thatcontains the filter wheel, door, calibration source, interface to the spacecraft, theradiation shield and the internal vacuum bulkhead; the cryostat section that con-tains the Charged Coupled Devices (CCD) electronic interfaces and radiators tothe CCDs. The stand-off structures are identical for all the EPIC cameras.
The CCDs give a time stamp and a position for each event, i.e. when a photonhits the CCD. The cameras can run in different modes with different fixed timeresolutions, readout times and observation windows. The CCDs are sensitive forphotons in the energy range 0.15keV − 15keV . The cameras are built of twodifferent types of CCDs, the Metal Oxide Semi-conductor (MOS) camera CCDsare front illuminated with a 40µm thick layer of sensitive silicon. This makes theEPIC-MOS CCDs more sensitive to the lower energy photons, but less sensitiveto the higher energy photons. The EPIC-PN camera is back-illuminated with a380µm thick layer of silicon. This camera has a high detection efficiency overthe entire energy band. The EPIC-PN camera consists of 12 CCDs covering intotal 6cm x 6cm divided into 400 x 384 pixels. The instruments have shown tobe vulnerable to radiation damage, and therefore they are shielded with a 3 cmthick piece of aluminium all around, except in the Field Of View (FOV). In thisproject, only calibration of the EPIC-MOS cameras has been considered, thus amore detailed description of this technology will be given in chapter 2.
1.4 Aim of the Project
The aim of the work for this thesis is divided into two parts. The first part isto create software that evaluates measured fluxes in the EPIC-MOS timing modewith respect to the standard EPIC-MOS full frame mode. See section 2.1 fordetails on operating modes. The second part is to adapt existing software fortiming analysis of the EPIC-PN camera, for the EPIC-MOS cameras and evaluatethe timing properties of the latter.
4 Introduction
Chapter 2
The EPIC-MOS Cameras
In order to reach the EPIC-MOS cameras on board XMM-Newton, the X-raysmust pass through the telescopes, which are focusing the photons. Then the pho-tons travel via the RGA that diverts in total 50% of the X-rays to the EPIC-MOScameras. Structural obscuration makes ∼ 40% of the incoming photons reachingthe cameras [8]. ∼ 40% of the beam is diverted to the RGS instrument, see figure2.1. When approaching an EPIC-MOS camera, the photons first reach the stand-off structure. This contains filter wheel, which has six locations for filters andalso six apertures for the calibration source to radiate on the CCDs. The X-rayseventually reach the CCDs which are mounted on the cryostat.
There are four filters for each camera, two thin filters made of 1600A poly-imidefilms with 400A aluminium on one side, one medium filter of the same material but800A aluminium and one thick filter. The thick filter is made of 3300A polypropy-lene, 1100A aluminium evaporated on one side and also a 450A thick layer of tin.These filters are needed to avoid optical contamination, but can however also beused to lower the number of photon counts when bright sources, giving a high fluxof photons are in the FOV. The two remaining positions are the closed and openpositions. The closed position is used for protection against protons reaching thespacecraft during high proton radiation periods. The open position can be usedwhen observing weak sources with very low flux, but has so far only been used forcalibration purposes.
The cryostat is interfacing the stand-off structure and except the CCDs alsocontains the cold-finger, thermally linked to the outer radiator, heaters for thermalcontrol, preamplifiers and other electrical components. The operational temera-ture of the CCDs was changed from ∼ −1000C to ∼ −1200C in November 2002to improve camera performance.
There are seven CCDs in each MOS camera. They have a dead region of lessthan 300µm on three edges. To avoid as much dead region as possible, adjacentCCDs are overlapping 1mm, see figure 2.2 for arrangement. This still gives a slight
5
6 The EPIC-MOS Cameras
Figure 2.1. Schematic view of the lightpath from the telescope via the RGA to theEPIC-MOS camera in the prime focus, and the RGS in the secondary focus.
gap between the CCDs where it is not possible to detect photons. The central CCDis positioned in the focal point, while the outer CCDs are shifted 4.5mm towardsthe telescope to approximately follow the curvature of the focal plane. The twocameras are however situated orhtogonally in the spacecraft which make the gapsbetween the CCDs in one camera covered by the other. The gaps surroundingthe mid CCD however are not covered by the other camera. The imaging area ofeach CCD is about 2.5cm x 2.5cm so the combination of all CCDs cover the entireFOV, which is 6.2cm in diameter. The CCDs are divided into 600 x 600 pixels, insquares of 40µm x 40µm which makes one pixel cover 1.1arcsec x 1.1arcsec.
2.1 The EPIC-MOS Operating Modes
The EPIC-MOS cameras have four science-operating modes, Full Frame (FF),Large Window (LW), Small Window (SW) and Timing Uncompressed (TU) mode,also called Timing mode. See figure 2.3.
When read out, the registered energy from each pixel is shifted towards thereadout node on one side of the CCD. In the FF mode which is the basic modeof the EPIC-MOS cameras it takes 2.6s to read out all data. If a bright source isobserved, or a timing observation is performed this might not be enough. Brightsources can cause pile-up, see section 3.2 for details, and many pulsars are havingfrequencies faster than 2.6s.
2.1 The EPIC-MOS Operating Modes 7
Figure 2.2. Arrangement of the CCDs of the EPIC-MOS2 camera
Figure 2.3. View of the different scientific modes of the EPIC-MOS cameras. top left:Full Frame mode, top right: Large Window mode, bottom left: Small Sindow and bottomright: Timing mode
8 The EPIC-MOS Cameras
In LW the mid CCD only uses the 300 x 300 centered pixels giving a timeresolution of 0.9s. The outer CCDs are used as normal. SW works in the sameway but only the 100 x 100 centered pixels are used, giving a time resolution of0.3s. The difference in time resolution results from the fact that most of the timetaken comes from reading the pixel charge accurately, and since the charges frompixels outside these windows are discarded, the result is a much faster readout.
The Timing mode is however different from the others. The outer CCDs areworking as normal, but the mid CCD is only using 100 rows in the middle, parallellto the readout node. Therefore there is true spatial resolution in one axis, the x-axis, and the y-axis is a measurement of time, giving time slices of the X-ray flux.This mode cannot give images, but can be used for timing observations and forspectroscopy. Since only 100 rows are read out, this mode has a timing resolutionof 1.75ms.
Chapter 3
Relevant Calibration Topics
When X-rays enter a telescope system, the distribution of X-rays gets affected.First of all by the mirrors, that are not ideal. They absorb some of the photons,and they don’t focus perfectly. Secondly by filters that might be in use, the de-tectors are not perfect and there is gain from the electronics only to mention themost important effects. The measured data is called raw data, but what scientistswant is real data, i.e. the data the way it actually looked like before it enteredthe telescope. In order to get this we need to correct for all imperfections in thesatelite. This is called calibration. In this chapter some relevant concepts regard-ing calibration will be discussed.
Calibration observations are performed regularly, but are not allowed to takemore than 5% of the time possible for observations with XMM-Newton. Pre-launchcalibration of the MOS cameras was performed, but due to changes in propertiesover time, coming from aging or from damaging radiation, and for refinement,re-calibration is performed all along the satelites lifetime. Fitted to each camerais a calibration source, this allows the absolute energy to be calculated to anaccuracy of a few eV [8]. The on board source is using Fe55 and can shine on theentire focal plane. Calibration is also performed based on actual observation, anddifferent modes and different cameras are callibrated with respect to each other.
3.1 Point Spread Function
The Point Spread Function (PSF), is a function determining the quality of howthe X-ray mirrors are focusing the photons [9]. Each telescope on board XMM-Newton has its own PSF, see figure 3.1. The XMM-Newton mirrors have a verynarrow PSF, and vary very little in the energy range 0.1keV − 6keV . Abovethese energies the PSF becomes slightly dependent on the energy. The PSF alsodepends on the position on the detector. The middle of the mid CCD is placed inthe focal point and therefore has the smallest PSF, this is called the on-axis PSF.The other CCDs are placed to approximately follow the focal plane, but off coursehere the focusing is less accurate and the PSF is increasing. The PSF is measured
9
10 Relevant Calibration Topics
Figure 3.1. Image of the point spread function of EPIC-MOS1, EPIC-MOS2 and EPIC-PN respectively. The ”‘starlike”’ pattern comes from the spider supporting the 58 mirrorsof the telescope.
as the Full Width Half Maximum (FWHM) which is the diameter of the PSF atthe distance from the centre where only half of the intensity remains. The smallerthe PFS the better the quality of the measured data. For the MOS cameras thePSF at FWHM is 4.3arcsec and 4.4arcsec for MOS1 and MOS2 respectivlely.
3.2 Pile-up
Pile-up occurs when two or more photons hit a pixel or adjacent pixels withinone readout frame. The readout frame time is the time needed to shift the eventstowards the readout, the readout itself and the integration time. If two photonsenter one pixel during the same readout frame, their energies are merged togetherand counted as one event. This is called energy pile-up. If one photon hit closeto the border of two pixels the energy will be divided between them. This socalled double event is recognized by the software on board XMM-Newton, and theenergy is put together and counted as a single event. If however two photons hittwo neighbouring pixels, the software cannot distinguish between this case and thedouble event case. The two photons energies are merged together and consideredas a single photon. This is commonly called pattern pile-up. Since photon energiesare added, a spectrum suffering from pile-up will have too many counts for higherenergies and too few for lower energies.
3.3 Exposure Time Corrections
The X-ray spectra obtained are given as [photonstime ] as a function of energy, wherephotons is the total count of photons during the exposure and time is the exposuretime given in seconds. But the time during which data is saved can differ fromthe total exposure time.
3.4 Effective Area 11
Telemetry gaps can cause data to be lost, and the time during that data waslost must be subtracted from the exposure time. If the CCDs are measuring toomany events during a period, the data buffers will be filled before data is sent to theground. In this case the comming events will be discarded until the buffer can befilled again. The number of events discarded is however counted, hence this stateis called the counting mode. The time during solar flares or when particles fromearths radiation belt are disturbing the instruments, giving a high backgroundmust be filtered out. The remaining intervals will be the Good Time Intervals(GTI). All these three features, Telemetry gaps, counting mode and GTIs have tobe taken into account when correcting the exposure time.
3.4 Effective Area
The Effective Area (Ae), reflects the ability of the telescope to collect photonsat various energies [9]. Since the EPIC-MOS cameras have the RGAs after theirmirrors, sharing X-rays with the RGS cameras, their Ae is less than for the EPIC-PN camera. The mirrors effiecency is also dependent on the energy of the incomingphotons. With an increasing off-axis angle the mirrors let less photons throughto reach the detectors, wich makes the Ae dependent on the off-axis angle. Whenany of the filters are applied, off course this also affects how many photons reachthe detector, and thereby affects the effective area, see figure 3.2
Figure 3.2. The effective area as a function of photon energy
12 Relevant Calibration Topics
3.5 Gain
Gain is defined as the amplification when the energy of a photon is read out andconverted into units of eV . This property affects all the pixels in the same way,independent of position.
3.6 Quantum Efficiency
Quantum Efficiency (QE) is a measurement of the CCDs sensitivity to light, ormore strictly defined as the percentage of photons hitting the CCDs and produceelectron-hole pairs. If a photon is absorbed by the depleted silicon, this will reducethe QE. According to [8] photons can be absorbed in the electrode structure, inthe depleted silicon or in the field free un-depleted region. Photons absorbed inthe electrode structure will be lost, photons in the depleted silicon is detected withall its energy and photons aborbed in the un-depleted region is detected, but someenergy may be lost. These features together builds up the concept of QE. Thequantum efficiency was measured on ground using the Orsay synchrotron, but isalso measured during operation, using celestial sources. The QE is varying verymuch between different energies, as can be seen in figure 3.3.
Figure 3.3. Quantum efficiency of the EPIC-MOS detectors as a function of photonenergy
3.7 Charge Transfer Efficiency 13
3.7 Charge Transfer Efficiency
The Charge Transfer Effiecency (CTE) measures the effiecency of transferingcharges from one pixel to another. When the CCDs are read out, the chargesare shifted between the pixels towards the readout node. Therefore the chargetransfer loss from a pixel depends on the distance to the readout node. If E0 isthe energy measured at pixel 0 nearest the readout node, after shifting from apixel at position p, with energy Ep, the relation is assumed to be as follows:
E0 = Ep · CTEp
The CTE is however changing over time, and the on board Fe55 calibrationsource has to be used to determine this change. The source is illuminating uni-formly over the CCD array, and by measuring how the emission lines vary as afunction of distance from the readout node, the CTE can be determined [10].
3.8 Redistribution
When monocromatic radiation enters the detectors, the measured spectra shouldin the ideal case be a spike at the energy of the photons. In a real case scenariothe spike is expected to be a gaussian distribution with the peak at the photonenergy. However sometimes a photon penetrating the Si layer in the CCD is notfully absorbed, but some of the energy is re-emitted. This causes the CCD todetect a lower energy than the energy of the incoming photon, and part of thedistribution will be shifted to low energies. This will create a distribution with ashoulder at a lower energy.
14 Relevant Calibration Topics
Chapter 4
Data Analysis
Since most observations intended in timing determination have been using theEPIC-PN-camera due to its higher time resolution, the calibration of the EPIC-MOS timing mode has not been given as high priority as the other modes. Somespectra of observations in timing mode show difference between the modelled spec-tra and the actual spectra with respect to normalisation in comparison to standardmode or the measured EPIC-PN normalisation.
Timing analysis of the EPIC-PN camera has been performed and is well eval-uated, but for the EPIC-MOS cameras this has never been done. Timing analysisaims to determine the accuracy of timing events. There are two kinds of timinganalysis, relative timing analysis and absolute timing analysis, see section 4.1.2.In this analysis the Crab pulsar, known to produce a relatively strong periodicsignal has been used. A pulsar is a heavily magnetised rotating neutron star thatemitts a radiation beam along its magnetic axis. This beam can only be detectedwhen pointed towards earth. As long as the rotation axis is not the same as themagnetic axis, the detected beam will appear as a pulse.
4.1 X-ray Data Analysis
4.1.1 Spectral Fitting
When X-rays pass through the telescope system and reach the detectors they areaffected. This so called response of a telescope system is described by a responsematrix. The measured spectra Smeasured is the response matrix Mresponse multi-plied with the real spectra S hence:
Smeasured = Mresponse · S
But Mresponse is not invertable, so therefore the real spectra cannot easily bedetermined. To solve this problem, the method of ”forward folding” is used. Withthis method, one has to assume a model of the real spectra, ”fold” this through
15
16 Data Analysis
the response matrix and the result is a modeled measured spectrum Smodel, i.e.:
Smodel = Mresponse · Smodel
To perform spectral fitting one first needs to define a source region, D(I) anda background region B(I), where I is the energy channel. The background is thensubtracted from the source region in order to only represent the source from thesource region data. The measured spectra relates to the soure and backgroundregion as:
Smeasured(I) = D(I)aD(I)·tD
− bD(I)
bB(I)
B(I)aB(I)·tB
where tD, tB are exposure times for source and background, bD(I), bB(I) arescaling factors for the source and background areas, and aD(I), aB(I) are scalingfactors for the sensitivity, Ae, of the telescope. When a model is chosen, the sec-ond step of ”forward folding” is to validate the models accuracy by performing aχ2-test between Smeasured and Smodel as stated in [11]:
χ2 =∑I
(Smeasured(I)−Smodel(I))2
σ(I)
Where σ(I) is the error for channel I. σ(I) is estimated as√Smeasured(I).
This test is now performed for different sets of model parameters till a smallest χ2
value is reached.
4.1.2 Timing Analysis
There are two kinds of timing analysis, relative timing analysis and absolute tim-ing analysis. The relative timing analysis determines the accuracy of a measuredperiod and is defined as:
∆PP = Pr−Px
Pr
Where Pr is the period of the pulsar in radio wavelengths and Px is the periodof the pulsar in X-ray wavelengths. The radio period from the Crab pulsar is verywell established by Jodrell Bank radio ephemeris. The X-ray signal however isnot strong enough to determine by analysing the light curve directly. Instead afolded light curve has to be created. A folded light curve is created by cuttingevery period of a light curve and put them on top of each other. This will increasethe number of counts in the resulting period. But since the period is not knownfrom the begining, several periods to fold the light curve have to be tried. If theperiod is badly chosen, peaks will be folded over dips and the curve will becomediffuse. This feature can be used to determine the right period by performing χ2
test between the model and a uniform distribution. When a bad period has beenfolded, the folded light curve will be closer to a uniform distribution and give alow value. When a good period has been used, the χ2 test will give a higher value.When the χ2 values are ploted as a function of Px − Pr, a typical diagram as in
4.2 Implementation 17
figure 4.1 will appear.
Figure 4.1. χ2 fitting for a range of periods. The x-axis is Px − Pr.
Absolute timing analysis means to determine the phase difference between thethe X-ray period and the radio period. Hence:
P0,X−ray − Pradio
For the absolute timing analysis, all steps from the relative timing analysis haveto be done. Thereafter a determination of a zero phase for the X-ray light curveand the radio light curve. The difference between them is the absolute timing. Aswill be seen in the results section, this is not necessary an error, but can have aphysical origin.
4.2 Implementation
4.2.1 XCAL and Spectral Fitting
The spectral fitting during this work aimed to increase the number of observationsin the XMM-Newton cross calibration archive (XCAL). XCAL consists of datafrom XMM-Newton and other X-ray observatories such as Chandra, Swift andSuzaku. The purpose is to evaluate and improve calibration of the XMM-Newtonobservatory and to perform calibration in a standardized way. XCAL is derived
18 Data Analysis
from the XMM-Newton Science Archive (XSA) which contains all observationsmade with XMM-Newton.
XCAL is not only an archive containing spectral results but also contains scriptsto process uncalibrated data files and to perform spectral extraction. XCAL hasthree main tasks.
• Data storage
– Main data storage XMM/OBSID where data for each observation is stored,such as calibrated event files, spectra, model used for spectral fitting,source and background regions.
– Observation entries containing start and stop times for each observationlocated at lists
– Results used by the review tool, see below, such as fitted spectras andlog files located at RESULTS
• Software used for calibration evaluation
– XARV scripts located at soft/XARV
– AIO java tool used to extract event files from XSA located at AIO
• User interface
– Review tool, a web tool to allow scientists viewing spectral fits fromXCAL and browsing the data archive for spectral parameters and pho-ton fluxes.
Data analysis of data from XMM-Newton is made by the tool Science AnalysisSystem (SAS), which is a collection of different tasks specialised to analyse datafrom XMM-Newton. The data from XMM-Newton are available through Obser-vation Data Files (ODF), and contain uncalibrated raw data in Flexible ImageTransport System (FITS) format. To get calibrated data emproc is used for theMOS cameras, epproc for the PN camera and rgsproc for RGS. The tasks evselectand xmmselect can then be used to extract filtered event lists, images, light curvesand spectra. To perform spectral analysis one then needs to create response ma-trices via rmfgen/arfgen. SAS contains several more tasks. Relevant tasks will bemensioned and explained later.
4.2 Implementation 19
XARV is a package of scripts created to make the data analysis procedureinto XCAL easier and standardized. XARV lets the user specify entries about theobservations to be processed and the scripts help using the SAS tools in a properway. The following main scripts are called when processing observations:
• XARV config.csh
• XARV AIOLoader.csh
• XARV createGTI.csh
• XARV rgs.csh
• XARV extract.csh
XARV config.csh sets general SAS settings and creates a logfile for all XARVmodules. If the spectral fitting has been performed and settings saved into filesthe dothefit obs is called to fit the extracted spectra. emphXARV AIOLoader.cshobtains ODF data files from XSA and runs cifbuild and odfingest for creatingproper calibration files. emproc, epproc and rgsproc are run to get calibratedevent files. XARV createGTI.csh converts the start and stop times given by theuser into ”XMM-Newton” time and runs gtibuild, a task extracting the GTIs dur-ing the given time interval. XARV rgs.csh calls rgsproc again to produce spectra.XARV extract.csh takes the source and background regions to produce source andbackground spectra and response files. There is a master script calling all thesescripts XARV.csh which is the only script a user has to start. The XARV pack-age has one more important script, check pileup.csh used to determine the sourceregion when creating spectras.
The calibration review tool is a web interface that allows scientists to browsespectral fits for every object in XCAL and for different versions of SAS. Additionalinformation about the model used, its parameters and camera setting is provided.It is possible to download selected photon flux data for eight pre-defined energybands. In this work, such data has been used for the evaluation of the flux cali-bration.
The first step when using XARV to process data into XCAL is to define atarget file. This file should contain object name and celestical positions, given asRight Ascention (RA) and Declination (Dec) ObjectName RA-Dec and be named:target ObjectName.txt
To know what observations to process for each object XARV needs a list of ob-servations. This file ObjectName obslist.txt includes Observation ID (OBSID),start time and stop time of the observations. OBSID is a ten digit number,unique for every observation of XMM-Newton. Start and stop time is writtenin the format yyyy-mm-ddThh:mm:ss. In the file the list is written like: OB-SID starttime stoptime. This information can be found from the FITS files using
20 Data Analysis
fits view (fv), a program used to examine FITS files.
Next step is to define source characteristics in the file sourcetypes.txt. Theobject can either be a point source like a star or an extended object like somegalaxies. Here should also be stated if the observed object has a continuum orline dominated spectrum. All together the information is written: OjcectNameextended/point cont/line. This information is needed to reproduce the correctPSF in the calibration.
XARV can now be run to start the first steps of calibration. XARV usesthe Java tool AIO to extract the ODFs from the XMM-Newton Science Archive(XSA). Once the ODFs are saved in XCAL and unzipped, cifbuild and odfingestis used to determine the model CCFs and point to those. gtibuild and tabgtigencreates GTIs. emproc, epproc and rgsproc are used to calibrate the ODFs for theEPIC and RGS instruments.
When all the ODFs are calibrated it is time for the spectral extraction proce-dure. A source and background region has to be defined. The source region shouldcontain as much of the source area as possible, this gives a larger number of countsand better statistics for the spectra. For bright sources though, the centre of thesource might be affected by pile-up. The pile-up region has to be excluded fromthe source region. Instead the excluded part of the source region, or rather thePSF describing the source region will be modelled by SAS. This will give a moreaccurate spectrum than using the original data.
The choice of the regions is done using xmmselect, a graphical user interfacefor evselect, wich can create images, light curves, histograms etc. In the imagingmodes, the selection expression for the region definition is:
((X,Y) IN circle(x-coordinate,y-coordinate,radii))
For the timing mode, the background region can be selected as a circle in anyof the outer CCDs. The source region, located in the mid CCD however only hasspatial resolution along the X − axis. Instead one can use xmmselect to create ahistogram. This histogram has number of counts of photons along the y − axisand position along the x−axis. In this case a peak around the centre of the sourcewill appear, and the source region is this peak. The selection expression for theregion definition is:
(RAWX in [lower x-coordinate for peak:higher x-coordinate for peak])
As the source regions are defined, existence of pile-up must be checked. Cur-rently there is no task in SAS for doing this in the timing mode. But for imagingmodes epatplot is used. A filtered file, containing only events in the source region,is created using evselect. epatplot uses the output file to create pattern plots.The important patterns to study are the single and double events, tripple events
4.2 Implementation 21
are very rare. There is a model of the distribution of single and double eventsin the CCDs, and if the measured distributions differ from these, the observationhas pile-up. To get rid of the pile-up a circle in the middle of the source region isremoved, creating an anulus, the SAS command for this is:
((X,Y) IN annulus(x-coordinate,y-coordinate,inner radii, outer radii))
If the new pattern plots still show signs of pile-up the same procedure is re-peated. The next step is to create spectral transfer functions, or the response ma-trices via arfgen, rmfgen and rgsrmfgen. Thus XARV needs to be run again. Thismeans we have Smeasured and Mresponse. To perform the model fitting, XSPEC isused. This is a X-ray spectral fitting package created by NASA, used commonlyby X-ray scientists. XSPEC is using the given spectra files and response matrices.The energy range is divided into channels, and all photons within a certain energyrange are counted to that channel.
To do the spectral fitting a model has to be assumed. What model to usedepends on the source and was chosen by selecting a model earlier used for thatparticular source. XSPEC performs a χ2 fitting as described in the theory section4.1.1.
4.2.2 Timing Analysis
For the timing analysis of the EPIC-MOS X-ray detectors, observations of theCrab pulsar have been used, see figure 4.2. The Crab pulsar situated in the Crabnebula is commonly used in X-ray astronomy due to its brightness, stable pulseperiod and pulse profile. Different from other pulsars, the Crab pulsar has a similarpulse profile in all wavelengths from radio to gamma rays, see figure 4.3.
The Crab nebula was first observed by John Bevis in 1731 and corresponds toa nebula observed by Arab and Chinese astronomers in 1054. The name comesfrom the Earl of Rosse, who observed the nebula in the 1840s and drew it lookinglike a crab. The nebula is located 6500ly from earth. It has a diameter of 11ly,and is expanding about 1500km/s.
The centre of the Crab nebula was in 1949 discovered to be a strong source ofradio waves. In 1963 X-rays emission were discovered, and 1967 it was found to beone of the brightest celestial sources of gamma rays. One year later, it became oneof the first pulsars to be discovered when the emission was found to be radiatedin rapid pulses. A pulsar is a magnetized rotating neutron star, emitting a beamfrom the magnetic poles. When the rotating beam is pointed towards earth a pulscan be observed. The Crab pulsar rotates once every 33ms but is slowing downdue to the energy lost in the beam.
To perform the relative timing analysis, the data needs to be calibrated us-ing emproc but also barycentric1 corrected. Since the spacecraft orbits around
1The barycenter of the solar system is equivalent to the center of mass of the solar sytem
22 Data Analysis
Figure 4.2. The Crab pulsar in optical light (red) from Hubble and X-ray (blue) fromthe Chandra X-ray observatory
Figure 4.3. Pulse shapes of different pulsars for, radio, optical, X-ray and gamma rays
4.2 Implementation 23
earth and earth is orbiting around the sun with a high speed, the arrival timesof the photons have to be corrected to an arrival location at the barycenter ofthe sun-earth system. The arrival times have been corrected using the SAS taskbarycen to give arrival times in barycentric time. The barycentric time is equiva-lent to the time experienced by a clock at rest in the barycenter of the solar system.
In this analysis all available observations with the EPIC-MOS cameras havebeen used. There are 11 observations for EPIC-MOS1 and 10 for EPIC-MOS2.The source region for the Crab pulsar for all the observations, given in RAW coor-dinates is RAWX IN [300:330]. RAW coordinates are coordinates within a CCD,specifying a row and a column.
In the event files all times are given in XMM-Newton time, wich is the numberof seconds elapsed since 1998-01-01T00:00:00 UTC2. But the tasks used to furtherin the timing analysis steps, like the folding of light curves needs time to be givenin Truncated Julian Day (TJD). This is done in two steps: First calculating thestart time of XMM-Newton time expressed Modified Julian Day (MJD) and thencalculating the time in TJD.
t0,xmm = 50814 Start of the XMM-Newton time expressed in MJD
EMJD = t0,xmm + EXMM
86400
Where EMJD is the event time in MJD, EXMM is the event time in XMM-Newton time. The division by 86400 is to get XMM-Newton time expressed indays.
ETJD = EMJD − 40000
Where ETJD is the event time in TJD
efsearch is a task in XRONOS, a timing analysis software package developedby NASA, used to find the best χ2 tested period by folding over a range of periods.A start period for the folding is chosen by using the radio period at the given time.The Crab pulsar is losing energy so the period is slightly decreasing. Observationsof the Crab pulsar in the radio wavelengths are done periodically, and can bedownloaded from the Jodrell Bank Centre for Astrophysics, Manchester. Sincethe X-ray observations are not done at the same time as the radio observations,the start period has to be calculated, using a linear interpolation.
For the absolute timing analysis a zero point of the period P0 is calculated forthe radio and X-ray period by extrapolating the radio period to the XMM-Newtonepoch.
2UTC, is a high precision atomic time standard, compensating for earths slowing rotation
24 Data Analysis
4.3 Statistical Evaluation of the Flux Measure-ments
The software created, MOS tmd consists of idl and c-shell scripts. Input is datafiles containing photon flux for all observations made with XMM-Newton. The en-ergy range 0.15keV − 10keV is split into six regions covering 0.15keV − 0.33keV ,0.33keV − 0.54keV , 0.54keV − 0.85keV , 0.85keV − 1.5keV , 1.5keV − 4.0keV ,4.0keV − 10keV and two additional larger regions covering 0.5keV − 2.0keV and2.0keV − 10.0keV . The input files contain the total flux for each energy region.These files are obtained from XCAL via the Review tool. One more input is neededhowever, a list of all observations where an EPIC-MOS camera is in timing mode.This obtained from the XSA. The software is filtering out the EPIC-MOS timingmode observations that exists in XCAL and creates two plots and three tables.The first plot, see figure 4.4, shows for all the eight energy regions, the flux ratiobetween EPIC-MOS and EPIC-PN as a function of time, i.e. revolutions of XMM-Newton around earth. The EPIC-PN camera is known to be well calibrated andcan be used as a reference instrument. The ratios should therefore in the ideal casebe equal to one, if the EPIC-MOS cameras also are well calibrated. The secondplot, see figure 4.5, is showing the same data, but as a function of flux instead oftime. The tables obtained give information about flux ratios, revolution, OBSID,and for which energy band and camera these values belong. One file oot.txt onlycontains data from the 0.5keV − 2.0keV energy band, and only values deviatingmore than 10% from 1. These are the observations considered out of range, andtherefore interesting for this work.
4.3 Statistical Evaluation of the Flux Measurements 25
Figure 4.4. Plots from the diagnostics software showing EPIC-MOS/EPIC-PN fluxratios for all energy bands versus time in revolutions. Black color (crosses) correspondsto EPIC-MOS1 data and blue color (squares) to EPIC-MOS2. The numbers on the righttells how many points are out of the green colored region, and the total number of datapoints in respective plot. The green colored region shows where the MOS/PN ratio isbetween 0.9 and 1.1.
26 Data Analysis
Figure 4.5. Plots from the diagnostics software showing the EPIC-MOS/EPIC-PN fluxratios for all energy bands versus EPIC-PN flux. Black color (crosses) corresponds toEPIC-MOS1 data and blue color (squares) to EPIC-MOS2. The numbers on the righttells how many points are out of the green colored region, and the total number of datapoints in respective plot. The green colored region shows where the MOS/PN ratio isbetween 0.9 and 1.1.
Chapter 5
Results
In this chapter the result of the evaluation of flux calibration and timing calibrationwill be presented, including a brief explanation of the approach.
5.1 Results of Flux Calibration Evaluation
A first analysis using only data already in XCAL did not reveal any significantresults. Neither any pattern over time nor as a function of flux. Since the numberof observations for the EPIC-MOS1 camera was 16 and for EPIC-MOS2 only 5,more observations had to be added. In order to know what observations to add,a investigation of the existing observations out of range was made. Here one in-teresting feature appeared. For most of these the EPIC-MOS1 camera had a fullscientific buffer, see tables 5.1 and 5.2. When data is to be sent from the satelliteto the ground, it is first stored in a buffer. Full scientific buffer is reached whendata is sent to the ground slower than this buffer is filled. When it is full, thedata comming next is discarded. A counter is counting the number of events lostthough, and thus this is called the counting mode. This implies that the errorsmight have something to do with high count rates. Hence sources with a highcount rate were decided to be added to the data set.
After the addition more sources were out of range, when taken into accountthe different number of observations from each camera, the EPIC-MOS1 camera isover represented. This can be explained by a hot column in the mid CCD giving anumerous amount of counts. After discussions with Markus Kirsch, Guillermo Bue-nadicha from ESAC and Steve Sembay from University of Leicester it is clear thatall diagnostics most probably points to problems with exposure time corrections.As an example we study a observation of MKN421 with OBSID 0411080701
In this observation we can clearly see that the entire spectral fitting for EPIC-MOS1 has a flux off-set compared to the measured EPIC-PN spectrum. See figure5.1. The shape is however the same as for the other cameras, meaning the errorlies in some feature affecting the entire spectra. In figure 5.2 one can see the
27
28 Results
OBSID MOS/PN flux ratio Counting mode Remarks
0084020401 0.895 YES All cameras in counting mode0084020501 0.900 NO0106260101 1.27 NO0153950601 0.812 NO0158970101 0.887 NO EPIC-PN in counting mode0212480501 1.80 NO EPIC-PN in counting mode0303210201 4.04 YES0311590901 1.27 YES0402330301 1.93 YES0402330501 1.83 YES0404860301 0.860 YES0411080701 1.16 YES
Table 5.1. EPIC-MOS1 observations identified by MOS tmd having a MOS/PN fluxratio > 1.1 or < 0.9
OBSID MOS/PN flux ratio Counting mode Remarks
0303210201 0.730 NO EPIC-MOS1 in counting mode0404860301 0.850 NO EPIC-MOS1 in counting mode
Table 5.2. EPIC-MOS2 observations identified by MOS tmd having a MOS/PN fluxratio > 1.1 or < 0.9
measured counts over time, for each readout frame. For the EPIC-MOS1 camerathe exposure is clearly divided into intervals of data and no data at all, wherethe no data intervals corresponds to the camera entering counting mode. Onepossibility for a feature affecting the entire spectra could be gain, but since thisshifting only occurs in cases with counting mode, this cannot be the problem. IfSAS is overcompensating for the lost intervals though, i.e. the corrected time istoo small, the measured flux [photonstime ] will increase and shift the spectra upwards.
For the spectra: Normalized means counts/spectral bin
5.2 Results of Relative Timing Evaluation
For the EPIC-MOS1 camera the relative errors∣∣∆PP
∣∣ have values between 1.4·10−9
and 4.0 · 10−8, and the EPIC-MOS2 between 3.6 · 10−11 and 1.7 · 10−8. See figures5.3 and 5.4. There doesn’t exist any expected accuracy for the EPIC-MOS timingaccuracy, but compared to the EPIC-PN camera, which has a time resolution of30µs in timing mode this can be considered as good values. There exists laterobservations of the Crab nebula in timing mode, but the folded light curves fromthe timing analysis for these are very diffuse and not possible to use for furtheranalysis. Further analysis will be performed in the future with new observations.
5.2 Results of Relative Timing Evaluation 29
Figure 5.1. Spectra of MKN421
Figure 5.2. Measured counts of the EPIC-MOS cameras per readout frame
30 Results
Figure 5.3. Results of relative timing analysis of EPIC-MOS1
Figure 5.4. Results of relative timing analysis of EPIC-MOS2
5.3 Results of Absolute Timing Evaluation 31
5.3 Results of Absolute Timing Evaluation
The absolute timing for the observations of the Crab pulsar are scattering verymuch, actually in the order of ±10ms for both cameras. Compared to the EPIC-PN camera and other X-ray satellites as INTEGRAL, Chandra and RXTE wherethe phase difference of the X-ray period compared to the radio period is between−300µs to −400µs, this does not look good. See figures 5.5 and 5.6. Any furtherinvestigation why the EPIC-MOS cameras differ this much from other instrumentshas not yet been made.
Figure 5.5. Results of absolute timing analysis of EPIC-MOS1
32 Results
Figure 5.6. Results of absolute timing analysis of EPIC-MOS2
Chapter 6
Future Work
6.1 Flux Calibration
Next step in the evaluation of the flux calibration should be to investigate theexposure time correction in SAS, which is currently running by the SAS team.
6.2 Timing Analysis
For the timing analysis of the EPIC-MOS cameras several observations could notbe used in the final result. The folded light curves of these observations showed nota clear pattern of one large main peak and a secondary smaller peak, significantto the Crab pulsar, see figure 6.1, but rather look like a uniform noise, see figure6.2. This happened for the four last observations, and both EPIC-MOS cameras.The reason for this has to be investigated in the future. Measurements of theCrab pulsar for timing calibration purpose have been performed regularly for theEPIC-PN camera, but not for the EPIC-MOS cameras. This should in the futurealso be done for the EPIC-MOS cameras to make further evaluation of the timingproperties possible.
33
34 Future Work
Figure 6.1. A typical folded light curve of the Crab pulsar from a observation madewith the EPIC-MOS2 camera
Figure 6.2. A failed folded light curve from one of the later timing observations of theCrab pulsar made with the EPIC-MOS2 camera
Bibliography
[1] F. Jansen et al. 2001, XMM-Newton observatory* I. The spacecraft and oper-ations
[2] H. Barre et al. 1999, An Overview of the XMM Observatory System
[3] http://sci.esa.int/science-e/www/object/index.cfm?fobjectid=31249
As of December 2007
[4] http://sci.esa.int/science-e/www/object/index.cfm?fobjectid=31250
As of December 2007
[5] J.W. den Herder et al. 2001, The Reflection Grating Spectrometer on boardXMM-Newton
[6] K.O. Mason et al. 2000, The XMM-Newton Optical/UV Monitor Telescope
[7] L. Struder et al. 2001, The European Photon Imaging Camera on XMM-Newton: The pn-CCD camera
[8] M.J.L. Turner et al. 2000, The European Photon Imaging Camera on XMM-Newton: The MOS Cameras
[9] M. Ehle et al. 2007, XMM-Newton Users’ Handbook
[10] D.H. Lumb et al. 2000?, In-orbit calibration activities of the XMM-NewtonEPIC Cameras
[11] J.S. Bendat, A.G. Piersol 2000, Random Data Analysis and MeassurementProcedures, 3rd edition
Personal correspondance withM.G.F KirschMartin StuhlingerAntonio Martin-CarilloGuillermo BuenadichaSteve SembayAndy Pollock
35
36 Bibliography
Upphovsratt
Detta dokument halls tillgangligt pa Internet — eller dess framtida ersattare —under en langre tid fran publiceringsdatum under forutsattning att inga extra-ordinara omstandigheter uppstar. Tillgang till dokumentet innebar tillstand forvar och en att lasa, ladda ner, skriva ut enstaka kopior for enskilt bruk och attanvanda det oforandrat for ickekommersiell forskning och for undervisning. Over-foring av upphovsratten vid en senare tidpunkt kan inte upphava detta tillstand.All annan anvandning av dokumentet kraver upphovsmannens medgivande. Foratt garantera aktheten, sakerheten och tillgangligheten finns det losningar avteknisk och administrativ art. Upphovsmannens ideella ratt innefattar ratt attbli namnd som upphovsman i den omfattning som god sed kraver vid anvandningav dokumentet pa ovan beskrivna satt samt skydd mot att dokumentet andraseller presenteras i sadan form eller i sadant sammanhang som ar krankande forupphovsmannens litterara eller konstnarliga anseende eller egenart. For ytterli-gare information om Linkoping University Electronic Press se forlagets hemsidahttp://www.ep.liu.se/
Copyright
The publishers will keep this document online on the Internet — or its possi-ble replacement — for a considerable time from the date of publication barringexceptional circumstances. The online availability of the document implies a per-manent permission for anyone to read, to download, to print out single copiesfor your own use and to use it unchanged for any non-commercial research andeducational purpose. Subsequent transfers of copyright cannot revoke this per-mission. All other uses of the document are conditional on the consent of thecopyright owner. The publisher has taken technical and administrative measuresto assure authenticity, security and accessibility. According to intellectual prop-erty law the author has the right to be mentioned when his/her work is accessedas described above and to be protected against infringement. For additional in-formation about the Linkoping University Electronic Press and its procedures forpublication and for assurance of document integrity, please refer to its WWWhome page: http://www.ep.liu.se/
c© John-Olov Larsson
