The CCD images were bias subtracted and flatfielded using the common procedures in the IRAF software. As the object field is uncrowded, aperture photometry was found to deliver more consistent results than methods based on point-spread-function (psf) fitting. Depending on the different telescopes' fields of view, between 5 and 9 suitable field stars were used as reference "standard" stars. To perform aperture photometry with a maximum signal-to-noise, optimum aperture sizes (Howell, 1989) for each star were determined. Frame-to-frame variations in the size of the stars' psf result from changes in seeing conditions or from changes in the telescope's focusing throughout a night. To correct for this on each frame, the psf of CM Dra (which was the brightest star in the field) was fitted by a circular Gaussian. All apertures are then expressed in multiples of the FWHM of this psf. These multiples were kept constant throughout a night's data. A suite of IRAF tasks for time-series photometry with optimized apertures in uncrowded fields, named 'vaphot ', was developed for these reductions. vaphot is available upon request from H. Deeg.
The reference magnitude was based on the sum of the flux of the reference stars, against which the differential magnitude of CM Dra was calculated. Individual reference stars may have unusual brightness variations in some nights, which may result from intrinsic variability or from flatfielding residuals, as described later. To recognize these variations, the difference between each reference star's magnitude and the summed reference magnitude was checked for variability, and often the rejection of one or two reference stars led to improved light curves of CM Dra with lower noise (see Deeg et al., 1997, for an example). The resulting lightcurves were cleaned of obviously erroneous measurements, as well as of events which are most likely flares of CM Dra (see Sect. 4.3). The differential magnitude was then scaled such that CM Dra's average magnitude outside of mutual binary eclipses was zero. All the steps described in this paragraph were performed independently for each night's observations.
The large color differential between the M4 stars composing CM Dra (V-R = 1.8) and the reference stars (V-R = 0.55 to 0.7, except reference star 4: V-R = 0.33, which is a white-dwarf proper-motion companion of CM Dra at a distance of ; Lacy 1977) caused slow airmass-related changes in CM Dra's brightness from differential extinction. In lightcurves with apparent slow variations caused by differential extinction, these variations were removed by subtraction of a fit, which was either a linear or a 2nd order polynomial fit to the off-eclipse lightcurve. With some rare exceptions, the events caused by a possible planetary transit occur on the time-scale of an hour, with ingress/egress lasting on the order of 10 minutes. The removal of the extinction slopes has therefore only a small effect on the signal content of the lightcurve resulting from planetary transits. This removal will also suppress amplitude variations from star-spots, which occur on approximately the same time-scale as the extinction, since the period of CM Dra's components is very likely locked to the binary period of 1.27 days. Final lightcurves were produced in 3 versions: A 'raw' one containing spurious points and flares, a 'cleaned' one where these events have been removed, and a 'fitted' one where the nightly slopes have been removed, as described above.
At some observatories the reduction procedure to obtain the 'raw' lightcurve had to be modified: The data from the CCD at the Rochester telescope exhibited a nonlinearity (Deeg & Ninkov 1995), which required a correction step before flat fielding. The raw reduction of the data from the Skinakas telescope was performed independently, using the MIDAS software (Palaiologou, personal communication). The photometer data from Kourovka observatory only required subtraction of the reference star's magnitude, followed by removal of the nightly extinction slope.
Of special interest for the detection of planetary transits are noise and error sources which can cause deviations in the data that may appear similar to planetary transits. The two major sources for errors with time scales of transit events ( 40 min) are (i) atmospheric instabilites and (ii) flatfielding errors.
(i) Atmospheric effects: The reference stars where traced for changes in their relative brightness amongst each other. If such changes occured above the normal noise, these particular data were rejected. As mentioned, CM Dra is a much redder star than any of the reference stars in the field, whose colors are all within a relatively narrow range. The brightness ratio between CM Dra and the reference stars is therefore more sensitive to second order extinction changes (i.e. changes in the color-dependency of the extinction within the bandpass of the R-filter) as are the brightness ratios among the reference stars themselves. In rare cases, it may be feasible, that temporary effects (for example, a band of very fine cirrus, or dust) in the atmospere can cause strong second order extinction changes, without sufficient first-order extinction variations to warrant a rejection of the data.
(ii) Flatfielding: Errors from flatfielding may appear if the spatial distribution of the stellar light on the CCD undergoes positional changes between images. This happens if the telescope tracking allows the stars to appear in slightly different positions in subsequent exposures. Imperfect flatfielding corrections will then appear as brightness variations among the stars. Flatfielding effects may strongly affect the data, if a star happened to move over a bad CCD pixel or column, or over a dust-corn on an optical surface. Although care was taken to keep these `features' away from any star, and especially from CM Dra, sometimes they may not have been recognized, or dust-corns may have changed their positions. If such a flatfielding variation affected any one of the reference stars, it caused an unusual change in its brightness relative to the other reference stars, and this reference star was not used in that night's summed reference magnitude. More difficult are cases in which the program star (CM Dra) may have been affected by flatfielding variations. To avoid this, we tracked the positional changes throughout each night, and rejected data with suspicious brightness variations of CM Dra, if they correlated with positional movements.
Even in a well-tracking telescope, where the center positions of the stars do not notably change throughout a night, flatfielding effects may appear if changes in the seeing cause variations in the relative illumination among the CCD pixels. If seeing was monotonically de- or increasing throughout a night (which was the most frequent case), or correlated with the airmass, the consequent flatfielding effects will have been removed by the linear or second order fit to the data mentioned previously. Furthermore, data were rejected where brightness variations of CM Dra correlated with variations in the seeing.
In summary, care was taken to avoid error sources that could create transit-like signatures in the data. However, such error sources could not be excluded with certainty to be the source of any individual feature in the data that may appear interesting. Verification of transit features is only possible by repeated observation of transits from the same planet.
© European Southern Observatory (ESO) 1998
Online publication: September 14, 1998