Next Contents Previous

2.5. The First Monitoring Programs

The early 1980s saw the first attempts to monitor the UV/optical continuum and emission-line variations in Seyfert 1 galaxies. There were two reasons this happened when it did. First, there was a realization that variability afforded a powerful probe of the structure and kinematics of AGNs on projected scales of microarcseconds. Variability was recognized as an important new tool with which to study enigmatic quasars. Second, the right technology for such investigations became widely available: it was possible to attempt such programs on account of (a) IUE, which for the first time allowed precision UV spectroscopy of low-redshift extragalactic objects, and (b) the proliferation of linear electronic detectors (first Reticons and Image Dissector Scanners, and later CCDs) on moderate-size (1-2m) ground-based telescopes.

One of the first significant monitoring programs was a multiple-year IUE-based program on NGC 4151, which was carried out by a European consortium led by M.V. Penston and M.-H. Ulrich 84. Ultraviolet spectra were obtained with a typical sampling interval of 2-3 months. The program showed that variations in the UV and optical continua were closely coupled. It also revealed that the emission-line flux variations are correlated with continuum variations, but that different lines respond in different ways, both in amplitude and in time scale. These data also showed a complicated relationship between UV and X-ray variations and led to the discovery of variable absorption lines in the ultraviolet.

The galaxy NGC 4151 was also monitored spectroscopically in the optical at Lick Observatory by Antonucci & Cohen 4. They found that the Balmer lines seemed to respond to continuum variations on a time scale less than around one month (their typical sampling interval). They also found that relative to Hbeta and Halpha, the higher-order Balmer lines and He II lambda4686 varied with higher amplitudes.

Arakelian (Akn) 120 was the first higher-luminosity Seyfert that was monitored fairly extensively in the optical 64 , 65 as a result of dramatic Balmer-line profile changes that had been detected earlier 26 , 38, 80 . Peterson et al.cite 65 found that the time scale for the response of Hbeta to continuum variations suggested a BLR size of less than 1 light month across. This was a surprising result as it suggested that there was a serious problem with existing estimates of sizes of the BLR that were based on photoionization equilibrium modeling, as these indicated the BLR should be about an order-of-magnitude larger than this. The upper limit on the BLR size was similar to that obtained by Antonucci & Cohen 4 for NGC 4151, but because Akn 120 is a higher-luminosity source, the monthly sampled data provided a more critical challenge to BLR models.

Not surprisingly, the results from these earlier monitoring programs were controversial. Several observational problems could be identified:

  1. Undersampling of the variations.   The variations tended to be undersampled because the original programs for monitoring Seyfert galaxies were designed for BLRs that were thought to be many light months in size. For example, in the case of Akn 120, Peterson et al. 65 were looking for profile structures that were expected to cross the line profile on a time scale of a year or so, and monthly observations should have been sufficient to carry out this program. There was certainly at the time no reason to believe that higher sampling rates should be required; indeed, proposals to observe AGNs as often as once per month were sometimes deemed to be "oversampled" by telescope allocation committees! There is no obvious algorithm to determine whether or not the variations that have been observed are undersampled, but it is quite obvious that if the results depend on individual data points, the light curve is almost certainly undersampled and any conclusions drawn must be eyed with suspicion. A very simple operational criterion for adequate sampling is that if the results do not change much when individual points are removed from the light curve, the light curve is probably not seriously undersampled. A nice simple test is to divide a light curve into two parts, one comprised of the even-numbered points (i.e., second, fourth, etc., in the time-ordered series) and one comprised of the odd-numbered points. If the two light curves are still very similar, i.e., the important features appear in both light curves, then the original light curve is probably adequately sampled.

  2. Low S/N of the light curves.   If the detected variations are not large compared to the signal-to-noise ratio (S/N) of typical data, then spurious results can be obtained. Stated another way, Fvar (Eq. 2) must be significantly greater than zero. This was a serious issue in the case of some of the earlier data obtained with image dissector scanners and Reticon arrays, for which uncertainties in AGN line and continuum fluxes were typically around 8-10%. To a large degree, this problem has been obviated by use of CCDs, for which typical errors in the 1-3% range are routinely achieved.

  3. Systematic errors.   There are two sometimes-related types of insidious errors that can adversely affect time series analysis: (a) correlated continuum/line errors, and (b) aperture effects.

    Correlated errors are due to systematic flux-calibration errors. Basically, if the flux calibration of a spectrum is incorrect, both the continuum and emission-line fluxes measured from it will be in error in the same sense; if the calibration is too high, both the emission-line and continuum fluxes will be too high. This introduces an artificial correlation between the line and continuum at zero lag, and can thus bias the measurement of the true lag between them to artificially small values.

    Aperture effects occur when the amount of flux entering a spectrograph is not fixed, on account of pointing or guiding errors, or variations in seeing in the case of ground-based observations. In point-like sources like stars, this affects only the overall photometric accuracy. In nearby AGNs, however, both the narrow-line region (NLR) and host-galaxy are spatially resolved, and the aperture geometry, centering and guiding, and seeing variations can lead to apparent spectral variations. Depending on their nature, aperture effects can cause either correlated, uncorrelated, or even anticorrelated errors in the line and continuum fluxes.

Next Contents Previous