The calibration of read-out-streak photometry in the XMM-Newton Optical Monitor and the construction of a bright-source catalogue
M.J. Page, N. Chan, A.A. Breeveld, A. Talavera, V. Yershov, T. Kennedy, N.P.M. Kuin, B. Hancock, P.J. Smith, M. Carter
aa r X i v : . [ a s t r o - ph . I M ] D ec Mon. Not. R. Astron. Soc. , 1–11 (2013) Printed 22 February 2018 (MN L A TEX style file v2.2)
The calibration of read-out-streak photometry in theXMM-Newton Optical Monitor and the construction of abright-source catalogue
M.J. Page , N. Chan , A.A. Breeveld , A. Talavera , V. Yershov , T. Kennedy ,N.P.M. Kuin , B. Hancock , P.J. Smith , M. Carter Mullard Space Science Laboratory, University College London, Holmbury St Mary, Dorking, Surrey, RH5 6NT, UK XMM-Newton Science Operations Centre, ESA, Villafranca del Castillo, Apartado 78, 28691, Villanueva de la Ca˜nada, Spain
Accepted —-. Received —-; in original form —-
ABSTRACT
The dynamic range of the
XMM-Newton
Optical Monitor (XMM-OM) is limited at thebright end by coincidence loss, the superposition of multiple photons in the individualframes recorded from its micro-channel-plate (MCP) intensified charge-coupled device(CCD) detector. One way to overcome this limitation is to use photons that arriveduring the frame transfer of the CCD, forming vertical read-out streaks for brightsources. We calibrate these read-out streaks for photometry of bright sources observedwith XMM-OM. The bright source limit for read-out streak photometry is set by therecharge time of the MCPs. For XMM-OM we find that the MCP recharge time is 5 . × − s. We determine that the effective bright limits for read-out streak photometrywith XMM-OM are approximately 1.5 magnitudes brighter than the bright sourcelimits for normal aperture photometry in full-frame images. This translates into bright-source limits in Vega magnitudes of UVW2=7.1, UVM2=8.0, UVW1=9.4, U=10.5,B=11.5, V=10.2 and White=12.5 for data taken early in the mission. The limitsbrighten by up to 0.2 magnitudes, depending on filter, over the course of the missionas the detector ages. The method is demonstrated by deriving UVW1 photometry forthe symbiotic nova RR Telescopii, and the new photometry is used to constrain the e-folding time of its decaying UV emission. Using the read-out streak method, we obtainphotometry for 50 per cent of the missing UV source measurements in version 2.1 ofthe XMM-Newton Serendipitous UV Source Survey (XMM-SUSS 2.1) catalogue. Key words: techniques: photometric – space vehicles: instruments – ultraviolet:general – stars: individual: RR Tel.
The
XMM-Newton
Optical Monitor (XMM-OM;Mason et al., 2001) is a 30 cm Ritchey Chr´etien tele-scope working at ultraviolet and optical wavelengths. Itis co-aligned with the X-ray mirrors on the EuropeanSpace Agency’s
XMM-Newton observatory. A similarinstrument, the Ultraviolet/Optical Telescope (UVOT;Roming et al., 2005) is carried on NASA’s
Swift gamma-ray-burst observatory.The XMM-OM detector is a micro-channel plate(MCP) intensified charge coupled device (CCD) (MIC;Fordham et al., 1989). Incoming photons first encounter amulti-alkali S20 photo-cathode, where they are convertedinto one or two electrons which are then proximity focused onto a series of MCPs. The MCPs multiply the number ofincoming electrons by a factor of 10 . The resulting electroncloud then encounters a phosphor screen, which converts thesignal back into photons. A tapered block of optical fibresthen transmits the photons to a frame-transfer CCD with ascience area of 256 ×
256 pixels. A single photon incident onthe photo-cathode results in a large splash of photons overseveral pixels of the CCD. Photon splashes are detected inindividual CCD frames and centroided by the onboard elec-tronics to a precision of one eighth of a CCD pixel, equivalentto 0.5 arcsec on the sky.When two or more incident photons give rise tooverlapping photon splashes on the CCD during a sin-gle CCD frame, they are recorded as a single incident c (cid:13) M.J. Page et al. photon. This phenomenon is known as coincidence loss(Fordham, Moorhead & Galbraith, 2000). When the arrivalrate of photons from an astronomical source is small com-pared to the frame rate of the CCD, coincidence loss isunimportant and the response of the instrument is linear.Coincidence loss leads to an increasingly non-linear responsefrom the detector as the incident photon rate approaches theframe rate of the CCD. For XMM-OM, the non-linearityis calibrated, and photometry can be corrected in groundprocessing for incident photon rates up to 3.6 times theCCD frame rate. Photometric uncertainty becomes large asthis limit is approached (Kuin & Rosen, 2008). Beyond thislimit, the detected photon rate saturates at approximatelythe frame rate of the CCD and photometry is not recover-able via the usual aperture photometry method.One way to overcome this limitation to the dynamicrange of the instrument is to use the photons that arriveduring the frame transfer of the CCD. At the end of eachCCD frame, the CCD image is shifted to the frame storearea for read out. Photons that arrive during the frametransfer (i.e. while the image is being moved downwards to-wards the frame store) are displaced in the vertical direction.Bright sources therefore show vertical streaks of displacedphotons which we will refer to hereafter as read-out streaks.The frame transfer is sufficiently fast that coincidence-lossis greatly reduced in the read-out streak compared to thedirect image. Page et al. (2013) showed that the read-outstreaks in
Swift
UVOT images can be used for photometricmeasurements, and demonstrated a photometric precision of0.1 mag. They found that the bright limit for UVOT pho-tometry could be decreased by 2.4 magnitudes for full-frameimages using this method, with coincidence-loss within theMCPs rather than the CCD dictating the bright limit forread-out-streak photometry.In this paper we test and calibrate read-out-streak pho-tometry for the XMM-OM, before applying the method toconstruct a catalogue of photometry for bright sources ob-served with XMM-OM. In Section 2 we describe the ba-sic principles of photometry using read-out streaks. In Sec-tion 3 we describe the data and methods used to calibratethe recharge time of the MCPs and verify read-out streakphotometry against sources of known brightness. The resultsare presented in Section 4. In Section 5 we demonstrate thetechnique by deriving XMM-OM UVW1 photometry for thesymbiotic nova RR Telescopii and examining its long termphotometric evolution. In Section 6 we describe the con-struction of a catalogue containing photometry of sourceswhich exceed the brightness limit of the XMM-SUSS 2.1catalogue. Our conclusions are presented in Section 7.Unless otherwise stated, magnitudes are given in theVega system.
The XMM-OM and
Swift
UVOT detectors were built toidentical specifications, so the principles of read-out streakphotometry are much the same in the two instruments andthe description in Section 2 of Page et al. (2013) is validfor XMM-OM as well. However, the two instruments areoperated rather differently. Whereas UVOT has only three hardware-window modes and three corresponding frametimes, XMM-OM routinely uses a variety of hardware win-dows to image different parts of the field of view (see Fig. 1for an example), and so a variety of frame times. The math-ematical description that follows therefore differs from thatgiven in Page et al. (2013) in that it is valid for any frametime, rather than for specific values.During frame transfer, charge is shifted 290 rows down-wards from the imaging portion of the CCD to the framestore area at a rate of 1 row per 6 × − s, for a to-tal transfer time of 1 . × − s. Photons arriving duringthis frame-transfer time form the read-out streak. FollowingPage et al. (2013), we base our read-out streak photometryon a measurement in a 16 ×
16 unbinned image pixel aper-ture, which corresponds to 2 × . × − s. For a frame time t F , in whichan image is accumulated and then transferred, the ratio S ofthe exposure time in the static image to the read-out streakis therefore S = t F − . × − . × − (1)The zeropoints for read-out streak photometry are re-lated to the normal imaging zeropoints by the ratio of equiv-alent exposure time S , such that Z s = Z i − . ( S ) (2)where Z s is the zeropoint for read-out-streak photometryand Z i is the imaging zeropoint.A complication arises in that imaging zeropoints forXMM-OM, given in the Current Calibration File (CCF) are given for 6 or 17.5 arcsec radius apertures, dependingon filter (Talavera, 2011), which have somewhat larger foot-prints than the 16 ×
16 aperture used for the read-out streakphotometry. Therefore we have converted the published ze-ropoints to those which would be appropriate for a 5 arc-second radius aperture using the encircled energy fractions for low-count-rate sources in the CCF to match better theread-out-streak aperture, and to maintain consistency withthe approach in Page et al. (2013). As the CCF does notcontain a suitable encircled energy fraction calibration forthe White filter, we have used the encircled energy fractionfor the U filter for this purpose. The adjusted imaging zero-points, appropriate for use as Z i in Equation 2 are given inTable 1.The dead time of the final-stage MCP is likely to limitthe read-out streak photometry at the bright end througha second stage of coincidence loss that is independent ofthe CCD (Fordham et al., 2000a; Page et al., 2013). The ex-pected count rate after coincidence loss is given by Equation3 of Page et al. (2013) which is reproduced here: R o = 1 − e − ( S R i τ MCP ) S τ
MCP (3)where R o is the count rate observed from the read-out streakin a 16 ×
16 pixel aperture, R i is incident count rate that calibration file OM COLOURTRANS 0010.CCF calibration file OM PSF1DRB 0010.CCFc (cid:13) , 1–11 ead-out streaks in the XMM-OM C oun t s p e r p i x e l Column number 200 400 600 800 C oun t s p e r p i x e l Column number300 400 500 600 700 C oun t s p e r p i x e l Column number200 400 600 800 1000 C oun t s p e r p i x e l Column number C oun t s p e r p i x e l
800 950Column number C oun t s p e r p i x e l
50 250Column number a b
Figure 1.
Illustration of the read-out-streak and measurement process. Panel (a) shows a full-frame image before (top) and after (middle)masking of sources. In the graph at the bottom of panel (a), the black line shows the mean counts per pixel for each column of themasked image, while the grey line shows the background estimate (see Section 3.3). The read-out streaks appear as positive spikes abovethe background estimate. Panel (b) shows the same field imaged in the default imaging mode (also known as ‘Rudi-5’ mode) in which 5separate images are taken to cover the field of view, together with the counts per column and background levels for each of the 5 imagesafter masking. would be observed if there were no coincidence loss in theMCP, and τ MCP is the MCP recharge timescale.Rearranging Equation 3, we obtain the following equa-tion for R i : R i = − log e (1 . − ( S R o τ MCP )) S τ
MCP (4)The gradual decline in XMM-OM sensitivity with time willaffect read-out streak photometry in the same way as itaffects normal aperture photometry. Therefore, R i shouldbe corrected by the time-dependent sensitivity correction which is contained within the CCF. If C is the time-dependent sensitivity correction factor and R c is the read-out-streak count rate after correction, then calibration file OM PHOTTONAT 0004.CCF R c = C × R i (5)and R c can be used with the appropriate zeropoint Z s toobtain a magnitude m : m = − . ( R c ) + Z s (6) In order to calibrate and qualify read-out-streak photome-try for XMM-OM, we must measure the read-out streaksfor sources of known (or predictable) brightness in one ormore XMM-OM photometric passbands. The most reliableand widely available photometry for stars in the magnituderange relevant for read-out-streak photometry comes from c (cid:13) , 1–11 M.J. Page et al.
Table 1.
XMM-OM imaging Vega and AB zeropoints scaled toa 5 arcsec radius aperture, suitable for use as Z i in Equation 2.For the White filter we do not provide an AB zeropoint becausethe CCF does not contain an AB zeropoint for this filter.Filter Vega zeropoint AB zeropoint(mag) (mag)V 17.9501 17.9098B 19.2481 19.0629U 18.2408 19.1705UVW1 17.1247 18.4871UVM2 15.7004 17.3400UVW2 14.8200 16.5252White 20.2370 - the Tycho-2 catalogue (ESA, 1997). Tycho-2 is a catalogueof stars for which photometry was recorded by the star map-per on ESA’s Hipparcos satellite. Tycho-2 contains photom-etry in two bandpasses, V T and B T , which cover similar butnot identical wavelength ranges as the V and B bands ofXMM-OM.In order to obtain a suitable dataset, we correlatedthe XMM-SUSS 2.1 catalogue (Page et al., 2012) with theTycho-2 catalogue using a matching radius of 5 arcsec. Inorder to avoid crowded fields we excluded matches within20 degrees of the Galactic plane. Matches were restricted toTycho-2 sources with V T
12 mag which have the mostreliable Tycho-2 photometry. Transformation from Tycho-2 photometry to photometric systems similar to that ofthe XMM-OM is known to be less reliable for stars ofspectral type M than for stars of earlier spectral type(Page et al., 2013; ESA, 1997). It is difficult to differentiateM stars from earlier types using Tycho-2 B T − V T colour, sowe obtained near-infrared photometry by cross correlatingour sample with the Two Micron All Sky Survey (2MASS;Skrutskie et al., 2006). Stars with V T − J > . XMM-Newton , foreach filter we used data from only a single
XMM-Newton pointing. Hence each star is represented by a single XMM-OM measurement in any filter, and the level of systematicerror derived from the analysis will be representative of thesystematics occurring in individual
XMM-Newton pointings.
To derive the transformation between Tycho-2 and XMM-OM photometric systems, we generated synthetic photom-etry for the Pickles (1998) atlas of stellar spectra in bothsystems. Transformations involving the Tycho-2 system areknown to have a much larger scatter for M-type stars than − . . . . V T − V −0.5 0 0.5 1 1.5 2 . . B T − B B T − V T Figure 2.
Relations between XMM-OM V and Tycho-2 V T mag-nitudes and between XMM-OM B and Tycho-2 B T magnitudeswith respect to Tycho-2 B T − V T colour. The black dots cor-respond to synthetic photometry for stars in the Pickles (1998)stellar library. The grey lines correspond to the linear relationsadopted in this work for transformations between Tycho-2 andXMM-OM systems. for earlier types (ESA, 1997; Page et al., 2013), and henceM star templates were not used for the synthetic photome-try. The colours relating V and B in the XMM-OM systemto V T and B T in the Tycho-2 system from the syntheticphotometry are shown in Fig. 2.The transformations between Tycho and XMM-OM areapproximately linear over a significant range in B T − V T . Thefollowing transformations were obtained from least squaredfits to the synthetic photometry. V = V T − . − . B T − V T ) (7)for 0 < ( B T − V T ) < B = B T + 0 . − . B T − V T ) (8)for 0 . < ( B T − V T ) <
2. These relations are shown as solidlines in Fig. 2. The rms scatter in V T − v and B T − b aboutthese relations are 0.006 mag and 0.009 mag respectively. c (cid:13) , 1–11 ead-out streaks in the XMM-OM Read-out streaks were measured from raw XMM-OM imagesafter correction for modulo-8 noise by the standard
XMM-Newton pipeline processing.Prior to measuring the streaks, sources and bad pixelswere masked from the images. The first step was to maskbad pixels identified in the standard XMM-OM calibrationfile within the CCF. Next, bright sources with count rates >
40 counts s − were identified via a sliding-box search;bright sources are surrounded by dark regions produced bycoincidence loss, so a 24 arcsec radius region was maskedaround each bright source. Masking of other sources wascarried out on a column by column basis to prevent theread-out streaks themselves being detected as sources. First,for each column, the median brightness was calculated andany pixels brighter than the median by more than 3 σ , orby more than 3 counts if the median of the column is σ , or 3 counts if the median of the column is
1, were flagged. All pixels flagged in either step were thenmasked.Next, for each column the mean value for all non-masked pixels was computed, and used to populate a sin-gle row of column brightness values. For each pixel in thisone-dimensional row, the median of the pixel values withina 128 unbinned-pixel-wide box was used to define a back-ground, and read-out streaks were then detected using asliding 16 unbinned-pixel cell. Examples showing the col-umn brightness and corresponding background for imagestaken in full-frame mode and the default imaging mode areshown in Fig. 1. For streaks detected at > σ , the countrate was summed over the 16-pixel interval and backgroundsubtracted. Next, the count rates were multiplied by a fac-tor of 16 to scale them from the equivalent of a single-pixelslice of the aperture to the full 16 ×
16 pixel aperture. Fi-nally, the count rates of the read-out streaks were correctedfor the time-dependent sensitivity degradation of the XMM-OM (Talavera, 2011).The procedure just described differs from that describedin Section 3.2 of Page et al. (2013) for UVOT read-outstreaks only in two aspects. First, no bad-pixel mask was em-ployed in Page et al. (2013), but in the present work we havefound that masking the bad pixels improves the read-outstreak detection and photometry towards the edges of theimages. Second, whereas corrections for large scale sensitiv-ity variations of UVOT were applied in Page et al. (2013),there are no such corrections for XMM-OM.Where multiple full-frame exposures have been takenthrough the same filter in the same
XMM-Newton observa-tion, the read-out streak count rates were averaged. To en-sure good quality photometry for the calibration, only mea-surements with a signal to noise ratio >
20 were used. Thereare suitable measurements for 54 stars in the V band and25 stars in the B band. calibration file OM BADPIX 0005.CCF . . . . O b s e r v e d c oun t r a t e i n x p i x e l a p e r t u r e ( c oun t s s − ) V (mag)
Figure 3.
Datapoints show the observed read-out-streak countrates in 16 × R o in Equations 3and 4) of Tycho-2 stars against XMM-OM V magnitude, derivedfrom Tycho-2 photometry (equation 7). The dashed line showsthe expected relationship if there were no coincidence-loss in theMCPs, (i.e. if R i = R o ). The solid line shows the expected rela-tionship for an MCP recharge time τ MCP = 5 . × − s, asderived in Section 4. Figure 3 shows the count rates measured in the read-outstreaks against the predicted V magnitudes for the Tycho-2stars. The dashed line shows the predicted relationship ifthere were no coincidence loss within the MCPs. While themeasurements approach the dashed line at faint magnitudes,the difference is large at bright magnitudes, implying that,as expected, coincidence loss in the XMM-OM MCPs has asignificant effect on the read-out streaks of bright sources.For sources brighter than v = 10 mag the read-out streakcount rate shows only a weak dependence on source mag-nitude. To determine the MCP recharge timescale, τ MCP ,we performed a χ fit in which the observed count rate isrelated to the incident count rate according to Equation 3to the sources fainter than V = 10 mag in Fig. 3. We obtaina best-fitting τ MCP = 5 . ± . × − s.The solid line in Fig. 3 shows the expected relation be-tween V magnitude and count rate for the best fitting valueof τ MCP . The model reproduces the data well up to the V = 10 mag bright limit of the fit, but for V <
10 magthe model and data diverge systematically towards brightermagnitudes, with the count rates falling below the expecta-tions of the model for all sources with
V < . c (cid:13) , 1–11 M.J. Page et al. . . . . C o I − c o rr ec t e d c oun t r a t e i n x p i x e l a p e r t u r e ( c oun t s s − ) V (mag)
Figure 4.
Datapoints show the read-out-streak count rates in16 × R i in Equations 3 and 4) against XMM-OM V magnitude, derived from Tycho-2 photometry (Equation 7). Thedashed line shows the expected relationship. The solid line showsa count rate of 0.14 counts s − , below which saturation effectsare not important. thermore, at the brightest magnitudes, the statistical un-certainties in the measurements are small, and therefore wewould expect the scatter in the measured count rates todiminish towards bright magnitudes as the count rates con-verge towards the saturation limit. Instead, the small num-ber of measurements at V < . Swift
UVOT, who suggested that charge bleeding inthe CCD, or positional dependencies in electron mobility inthe MCPs, might be responsible. It is interesting to notethat τ MCP = 5 . × − s is significantly longer than thatfound by Page et al. (2013) for Swift
UVOT, such that co-incidence loss in the MCPs becomes significant at a lowercount rate in XMM-OM than in UVOT. The bright limit be-yond which the data deviate from the coincidence-loss modelis also shifted to fainter magnitudes in XMM-OM compared . . . . C o I − c o rr ec t e d c oun t r a t e i n x p i x e l a p e r t u r e ( c oun t s s − ) B (mag)
Figure 5.
Datapoints show the read-out-streak count rates in16 × R i in Equations 3 and 4) against XMM-OM B magnitude, derived from Tycho-2 photometry (Equation 8). Thedashed line shows the expected relationship. The solid line showsa count rate of 0.14 counts s − , below which saturation effectsare not important. to UVOT. This finding suggests that it is the behaviour ofthe MCPs, rather than charge bleeding in the CCD, whichis responsible for the break down of the model at brightmagnitudes.Correcting the observed read-out-streak count rates forcoincidence loss in the MCPs using Equation 4 and τ MCP =5 . × − s, we obtain Fig. 4. A close correlation is seenbetween count rate and magnitude for coincidence-loss cor-rected count rates R i < .
14 counts s − . Fig. 5 shows thecoincidence-loss-corrected count rates for the B band com-pared to B magnitudes derived from Tycho-2 photometry.There are fewer B measurements, and the Tycho-2-derived B magnitudes have larger uncertainties than those for the V band. In addition, read-out streaks typically have largerstatistical uncertainties in B than in V because XMM-OM B images typically have higher background levels than V images. Consistent with the findings from V -band photom-etry, saturation effects associated with the MCPs are onlyseen for R i > .
14 counts s − . c (cid:13) , 1–11 ead-out streaks in the XMM-OM − . . ∆ m ( m a g ) − . . < ∆ m > ( m a g )
10 10.5 11 11.5 12 . . . σ ∆ m ( m a g ) V (mag)
Figure 6.
Top panel: differences ∆ m between the V magnitudesobtained from Tycho-2 and the V magnitudes obtained from thecoincidence-loss corrected count-rates of the XMM-OM read-outstreaks. Uncertainties are the quadrature sums of the errors onthe count-rates and the Tycho-2 magnitudes. Middle and bottompanels: mean and dispersion of ∆ m respectively in 0.5 magni-tude bins. The dashed line in the bottom panel shows the best-fitdispersion, σ ∆ m = 0 .
10, over the full 10 < V <
12 magnitudeinterval shown.
We now examine the scatter of the read-out-streak photom-etry measurements with respect to the Tycho-2 photom-etry to estimate the level of photometric accuracy whichcan be achieved for XMM-OM using read-out streaks. ForUVOT, Page et al. (2013) found that systematic errors limitthe photometric accuracy of read-out-streak photometry to0.1 mag.The top panels of Figs 6 and 7 show the differences be-tween the read-out-streak and Tycho-2 derived magnitudes(∆ m ) for the individual stars in V and B . The uncertain-ties on ∆ m have been computed by adding in quadraturethe statistical uncertainty on the read-out-streak photom-etry with the magnitude uncertainty given in the Tycho-2catalogue. The V data in Fig. 6 are of higher quality thanthe B data in Fig.7, both in terms of the precision of theindividual measurements, and in the number of measure-ments. We have used the method of Maccacaro et al. (1988)to derive from these measurements in 0.5 magnitude bins,maximum-likelihood estimates of the mean and intrinsicstandard deviation of the distribution of ∆ m , which wehave assumed to be Gaussian. The middle panels of Figs 6 − . . ∆ m ( m a g ) − . . < ∆ m > ( m a g ) . . . σ ∆ m ( m a g ) B (mag)
Figure 7.
Top panel: differences ∆ m between the B magnitudesobtained from Tycho-2 and the B magnitudes obtained from thecoincidence-loss corrected count-rates of the XMM-OM read-outstreaks. Uncertainties are the quadrature sums of the errors onthe count-rates and the Tycho-2 magnitudes. Middle and bottompanels: mean and dispersion of ∆ m respectively in 0.5 magni-tude bins. The dashed line in the bottom panel corresponds to adispersion of 0.1 magnitudes. and 7 show the means, h ∆ m i , and the bottom panelsshow the intrinsic standard deviations σ ∆ m . The valuesof h ∆ m i are all consistent with 0, and no systematictrends are evident with magnitude. The individual valuesof σ ∆ m are less well determined. For the V band, apply-ing the Maccacaro et al. (1988) method over the full rangeof magnitudes we obtain h ∆ m i = − . ± .
025 and σ ∆ m = 0 . ± .
02. The individual values of σ ∆ m ob-tained in 0.5 magnitude bins in both V and B bands are allconsistent with σ ∆ m = 0 .
10, which represents the level ofsystematic uncertainty which must be added in quadratureto the statistical uncertainties to reproduce the distributionof ∆ m . Thus photometry obtained from read-out streaksshould be considered to have a systematic uncertainty of0.1 mag, in addition to the statistical uncertainty. It is in-teresting to note that this level of systematic error, 0.1 mag,is identical to that derived for UVOT read-out-streak pho-tometry by Page et al. (2013). c (cid:13) , 1–11 M.J. Page et al.
From Figs 4 and 5 we identified a bright limit of R i =0 .
14 counts s − for read-out streaks in full-frame XMM-OMimaging, beyond which saturation of the MCPs degrades thephotometry. Different window configurations correspond todifferent frame times, and so different exposure-time ratios, S , according to Equation 1. For full-frame imaging the frametime is 11.0388 ms, corresponding to S = 9054. The max-imum acceptable countrate will scale inversely with S . sofor an arbitrary frame time the bright-limit count rate forread-out streak photometry is R max = 1268 S (9) The Nova RR Telescopii is a symbiotic binarywhich underwent a nova outburst in 1944. In opti-cal radiation it has been slowly dimming ever since(Kotnik-Karuza et al., 2006). Between 1978 and 1995 itsultraviolet flux declined by a factor of 2–3, dependingon wavelength (Nussbaumber & Dumm, 1997). Short-ward of 1500 ˚A, the UV flux of RR Tel is thoughtto be dominated by the central star, while longwardof 1800 ˚A the UV flux is likely to be dominated bynebular emission (Nussbaumber & Dumm, 1997). In astudy of the X-ray and ultraviolet emission from RR Tel,based on an
XMM-Newton observation taken in 2009,Gonz´alez-Riestra, Selvelli & Cassatella (2013) show thatthe 2009 XMM-OM photometry through the UVW2 andUVM2 filters are consistent with an extrapolation ofthe exponential decay seen in International UltravioletExplorer (IUE) data between 1978 and 1995. However,Gonz´alez-Riestra, Selvelli & Cassatella (2013) describe theXMM-OM UVW1 photometry as uncertain, because itis subject to a large degree of coincidence loss. Indeed,in the XMM-SUSS 2.1 catalogue there is no UVW1magnitude for this source because it is flagged as beingtoo bright for a reliable photometric measurement inUVW1. This is unfortunate, because it can be seen inGonz´alez-Riestra, Selvelli & Cassatella (2013) that theIUE-based photometry corresponding to the UVW1 filteris of a higher statistical quality than the photometrycorresponding to UVW2 or UVM2. Using the read-outstreak technique, we are now able to derive valid UVW1photometry from the 2009
XMM-Newton observation ofRR Tel, and so can examine the UV photometric evolutionover a 31-year time base.Photometry was derived from the 2009 XMM-OMUVW1 observation using the read-out-streak technique de-scribed here. The read-out streak is present in three of thefive sub-exposures used to image the field around RR Tel,and is detected in all three sub-exposure with signal to noiseratios of between 46 and 48. Thus the uncertainty on thephotometry derived from the read-out streak is dominatedby the 0.1 mag systematic term described in Section 4.1. . . . UV W m a g Date IUEXMM−OM
Figure 8.
Lightcurve of RR Tel in the UVW1 bandpass, us-ing read-out streak photometry from XMM-OM and synthesizedUVW1 photometry from IUE. The dashed line shows the best-fitexponential dimming model, which corresponds to an e-foldingtime of 19.4 years.
IUE large aperture, low-resolution spectra, reduced us-ing the New Spectral Image Processing System (NEWSIPS;Nichols & Linsky, 1996), were retrieved from the MikulskiArchive for Space Telescopes. UVW1 photometry was syn-thesized from the IUE spectra by integrating the productof IUE flux and the the XMM-OM UVW1 response curve.The UVW1 passband extends further to the red than thespectral coverage of IUE, which ends at 3350 ˚A. In orderto correct the IUE synthesized photometry for the missingred flux, we made use of a Hubble Space Telescope (HST)Space Telescope Imaging Spectrometer (STIS) observationof RR Tel from 2000, which provides full spectral coveragethroughout the UVW1 passband. From the STIS data, wedetermine that the photometry of RR Tel synthesized onlyto 3350 ˚A is 0.12 mag fainter than the photometry synthe-sized over the full bandpass.The statistical uncertainties on the UVW1 photometrysynthesized from IUE spectra were initially propagated fromthe statistical uncertainties contained within the NEWSIPSspectral files. Examination of the scatter of the photometricdata suggests that these uncertainties are too small. In par-ticular, the rms of the UVW1 photometry derived from fourIUE observations taken within a single day, 7 May 1994, isobserved to be 0.13 mag, whereas the statistical uncertain-ties derived from the spectra are only 0.01 mag. Thereforewe have adopted 0.13 mag as the statistical uncertainty onthe IUE-derived photometry.Fig. 8 shows the XMM-OM UVW1 photometry to-gether with the corrected, synthesized UVW1 lightcurvefrom IUE. We have not included the photometry derivedfrom the HST STIS observation because this was obtainedfrom a much smaller aperture (0.2 arcsec) than the IUEor XMM-OM photometry, and the spatial extent of thenebula is not known. The dashed line shows the best-fit exponential decay model to the IUE and XMM-OMUVW1 photometry. From the fit we derive an e-foldingtime of 19 . ± . ± c (cid:13) , 1–11 ead-out streaks in the XMM-OM wavelength UVW2 and UVM2 bands, which were 23 ± ± The XMM-SUSS is a catalogue of sources detected inXMM-OM images and consists of astrometric, photomet-ric, and morphological information together with qualityflags to appraise the user of the validity of the measure-ments (Page et al., 2012). The latest release of the cata-logue, XMM-SUSS 2.1 , contains more than 4.3 millionunique optical and UV sources. Sources which exceed athreshold of 0.97 counts per frame in a given band aredeemed too bright for reliable photometry in that band. Insuch cases, magnitudes are not provided in the catalogueand quality flags indicate the bands for which the source istoo bright. The read-out-streak photometry method whichhas been described in this paper permits photometry ofsources up to 1.5 magnitudes brighter than this threshold,and therefore offers the possibility to recover some of themissing photometry for these bright sources. We have there-fore used read-out-streak photometry to produce a catalogueof photometric measurements for bright sources to form asupplement to the main XMM-SUSS 2.1 catalogue. In thefollowing subsections we describe how we constructed thesupplementary catalogue and describe its basic properties. A pipeline was constructed to measure read-out streaks inXMM-OM images and match them to stars in the XMM-SUSS which were too bright for their photometry to bemeasured from the direct image. Such sources can be identi-fied in the XMM-SUSS 2.1 catalogue through having qualityflag number 11 (with numerical value of 2048) set in theQUALITY FLAG column corresponding to the affectedfilter(s). As for the XMM-SUSS, each individual XMM-Newton pointing, corresponding to a unique observationidentification number (OBSID) is treated independently inthe supplementary catalogue. For each
XMM-Newton obser-vation containing one or more saturated stars, we retrievedthe standard XMM-OM sourcelist and modulo-8-correctedraw-image products from the ESA
XMM-Newton
ScienceArchive (XSA). Read-out streaks were measured from theimages using the software developed by Page et al. (2013),with bad pixels masked using the standard XMM-OM cali-bration file from the CCF. The columns containing read-out streaks were then matched to the columns containingsaturated stars in the sourcelists. For XMM-OM observa-tions carried out in the default imaging mode (also known as‘Rudi-5’ mode) in which 5 separate images are taken in a mo-saic pattern to form a full-field image, the read-out streaks calibration file OM BADPIX 0005.CCF often cross multiple images, but the saturated star is typi-cally found in only one. Therefore the pipeline merges theindividual sourcelists prior to matching the read-out streaksto stars.It is possible for more than one bright source to land onthe same column within the XMM-OM image, particularlyin crowded fields near to the Galactic plane, and in suchcases multiple stars will contribute to the read-out streak,invalidating the photometry. We have discarded read-out-streak photometry for saturated stars in which one or moreadditional stars with count rates of >
20 counts s − arefound in the source list within eight raw columns of thesaturated star. The threshold of 20 counts s − correspondsto a maximum contaminating contribution of 8 per cent tothe read-out-streak of a saturated star.We also identified cases in which two read-out streaksare detected in close proximity (within 24 raw columns),and discarded any corresponding read-out-streak photom-etry. Although this condition occasionally identifies casesof bright sources lying on close-enough columns that theirread-out streak photometry is affected, its main purpose inthe pipeline is to identify stars which are far too bright forread-out streak photometry but can not be identified as suchfrom the count-rates of their read-out streaks (see Section5.1 of Page et al., 2013). Visual screening was used to re-ject other examples of such stars which are too bright forread-out-streak photometry, but which were not identifiedby their count rates or by the apparent proximity of multi-ple read-out streaks.During the construction of the catalogue it became ev-ident that in extremely crowded star fields, background de-termination becomes very challenging. In extreme cases, thecolumn-by-column background as determined in the pipelineby the algorithm described in Section 3.3 shows large sys-tematic deviations that compromise the read-out streakmeasurement. From visual examination of the backgroundcurves for fields containing different densities of sources wehave chosen a threshold of 2000 sources per full-frame XMM-OM image; where the field is imaged through a sequenceof windowed exposures as in the default imaging mode,the sources are summed over all the exposures. We havediscarded read-out-streak photometry derived from imageswhich exceed this threshold. Basic statistics for the supplementary catalogue are pre-sented in Table 2. Overall, valid photometry is recoveredfor 50 per cent of the UV measurements and 30 per centof the optical measurements, which are saturated in XMM-SUSS 2.1, using read-out streaks. The magnitude distribu-tions for the supplementary catalogue entries in the six pass-bands are shown in Fig. 9. The dashed lines correspond tothe nominal brightness limits for read-out streak photome-try and aperture photometry for full-frame images early inthe mission, so we would expect the vast majority of sourcesto lie within the 1.5 mag interval between the two lines.A small number of sources are found to be brighter thanthe left dashed line because the bright limit has increasedslightly (by up to 0.2 mag depending on filter), over thecourse of the
XMM-Newton mission as the XMM-OM de-tector has aged. Some spill-over of sources beyond the right c (cid:13)000
XMM-Newton mission as the XMM-OM de-tector has aged. Some spill-over of sources beyond the right c (cid:13)000 , 1–11 M.J. Page et al.
10 11 12 13 N u m b e r V mag 11 12 13 14 N u m b e r B mag 11 12 13 N u m b e r U mag9 10 11 12 N u m b e r UVW1 mag 8 9 10 11 N u m b e r UVM2 mag 7 8 9 10 N u m b e r UVW2 mag
Figure 9.
Magnitude distributions of the supplementary catalogue entries for the six passbands used in the XMM-SUSS. The two dashedlines in each panel indicate the bright limits for read-out-streak photometry and aperture photometry. dashed line is expected because these are the faintest read-out streak measurements and hence have the largest phot-metric errors. However, a tail of sources is seen in severalpassbands between 0.5 and 2 magnitudes fainter than theright dashed line. Stars with these magnitudes should bewell below the 0.97 counts/frame saturation limit of XMM-SUSS 2.1 by which they were selected for read-out-streakmeasurement. Therefore we have visually inspected each ofthe sources which are more than 0.5 mag fainter than thedashed lines in Fig. 9, of which there are 23 in total. Wefound that in 12 cases the bright sources were on edge ofthe image, such that the count rate measurement in XMM-SUSS 2.1 is flawed. The read-out-streak photometry appearsto be correct in 11 of the 12 cases. In a further 5 cases, thesources are bright galaxies, with the XMM-SUSS 2.1 mea-surement including counts from a much larger region thanthe standard point-source aperture. In these cases the read-out-streak photometry provides reasonable estimates for thenuclear regions of these galaxies. The remaining 6 cases in-clude 2 spurious scattered light features which have beentreated as extended sources in XMM-SUSS 2.1, 2 sourcesin which the presence of bright sources in the neighbouringcolumns has caused problems with the local background es-timate for the read-out streak photometry, and 2 sources inwhich Flag number 11 appears to have been set incorrectly;in both these latter cases the aperture photometry from theimage agrees with the read-out streak photometry. Overallthe great majority of read-out streak measurements appearto be correct among these apparently outlying points, sothe tail of faint read-out-streak magnitudes is not cause forconcern.
Table 2.
Source statistics for the Supplementary Catalogue. Themean magnitudes for the sources in each filter are Vega magni-tudes in the XMM-OM photometric system.Filter Number of Number of Mean magnitudesaturated objects in in Supplementaryobjects in Supplementary CatalogueXMM-SUSS 2.1 CatalogueV 978 258 11.06B 1670 460 12.34U 1252 440 11.35UVW1 675 329 10.28UVM2 88 53 8.80UVW2 9 4 7.65
The supplementary catalogue is available as a fits file viathe MSSL XMM-SUSS2 web pages and will be availablefrom the ESA XMM-Newton
Science Archive (XSA) . We have shown that read-out streaks in XMM-OM imagescan be used for photometric measurements of stars that arebrighter than the coincidence-loss limit for normal aperturephotometry. The study is based on XMM-OM V - and B -band measurements of stars in the Tycho-2 catalogue. Wefind that the recharge timescale for the pores of the mi-crochannel plates in XMM-OM is 5.5 ± . × − s, whichsets the bright-source limit for read-out streak photometryto be 1.5 magnitudes brighter than the limit imposed bycoincidence-loss in the CCD in full-frame images. We findthat systematics and unaccounted errors limit the precision (cid:13) , 1–11 ead-out streaks in the XMM-OM of the read-out streak photometry to 0.1 mag. As a demon-stration, we have derived UVW1 photometry of the sym-biotic nova RR Tel using read-out streaks, and comparedit to photometry derived from earlier IUE observations. Wehave used the read-out streak method to construct a supple-mentary catalogue of photometry for sources which were toobright for photometric measurement in the XMM-SUSS2.1catalogue. Using this method, photometry is recovered for50 per cent of the UV measurements which exceeded theXMM-SUSS2.1 bright limit. ACKNOWLEDGMENTS
Based on observations obtained with XMM-Newton, an ESAscience mission with instruments and contributions directlyfunded by ESA Member States and NASA.
REFERENCES
Kotnik-Karuza D., Friedjung M., Whitelock P.A., MarangF., Exter K., Keenan F.P., Pollacco D.L., 2006, A&A, 452,503ESA, 1997, The Hipparcos and Tycho Catalogues, ESASP-1200Fordham J. L. A., Bone D. A., Read P. D., Norton T. J.,Charles P. A., Carter D., Cannon R. D., Pickles A. J.,1989, MNRAS, 237, 513Fordham J. L. A., Kawakami H., Michel R. M., Much R.,Robinson J. R., 2000a, MNRAS, 319, 414Fordham J. L. A., Moorhead C. F. & Galbraith R. F.,2000b, MNRAS, 312, 83Gonz´alez-Riestra R., Selvelli P., & Cassatella A., 2013,A&A, 556, A85Høg E., et al., 2000, A&A, 355, L27Kuin N. P. M. & Rosen S. R., 2008, MNRAS, 383, 383Maccacaro T., Gioia I.M., Wolter A., Zamorani G., StockeJ.T., 1988, ApJ, 326, 680Mason K. O., et al., 2001, A&A, 365, L36Nichols J.S., & Linsky J.L., 1996, AJ, 111, 517Nussbaumer H. & Dunn T., 1997, A&A, 323, 387Page M.J., et al., 2012, MNRAS, 416, 2792Page M.J., et al., 2013, MNRAS, 436, 1684Pickles A. J., 1998, PASP, 110, 863Roming P. W. A., et al., 2005, Space Science Reviews, 120,95Skrutskie M. F., et al., 2006, AJ, 131, 1163Talavera A., 2011, Technical Report XMM-SOC-CAL-TN-0019 issue 6.0,
XMM-Newton
Opticaland UV monitor (OM) Calibration Status, ESA;http://xmm2.esac.esa.int/docs/documents/CAL-TN-0019.pdf c (cid:13)000