2 Observations

We report here two data sets: 48 hr of time series photometry acquired in August 1996, with the journal of observations presented in Table 1. The second data set consists of 323 hours acquired in May-June 2000 (see Tables 2 and 3 for the observing log). Both of these data sets were obtained simultaneously with time resolved spectroscopy with the Hubble Space Telescope, which will be reported elsewhere.

In 1996, the observations were obtained with three channel time series photometry using bi-alkali photocathodes (Kleinman et al. 1996), in Texas, China, and Poland. During 23 May to 23 June, 2000, we observed GD 358 mainly with two and three channel time series photometers using bi-alkali photocathodes and a time resolution of 5 s. The May-June 2000 run used 13 telescopes composing the Whole Earth Telescope. The telescopes, ranging from 60 cm to 256 cm in diameter, were located in Texas, Arizona, Hawaii, New Zealand, China, Lithuania, Poland, South Africa, France, Spain, Canary Islands, and Brazil. As the pulsations in white dwarf stars are in phase at different wavelengths (Robinson et al. 1982), we used no filters, to maximize the detected signal.

Each run was reduced and analyzed as described by Nather et al. (1990) and Kepler (1993), correcting for extinction through an estimated local coefficient, and sky variations measured continuously on three channel and CCD observations, or sampled frequently on two channel photometers. The second channel of the photometer monitored a nearby star to assure photometric conditions or correct for small non-photometric conditions. The CCD measurements were obtained with different cameras which are not described in detail here. At least two comparison stars were in each frame and allowed for differential weighted aperture photometry. The consecutive data points were 10 to 30 s apart, depending if the CCDs were frame transfer or not.

After this preliminary reduction, we brought the data to the same fractional amplitude scale and converted the middle of integration times to Barycentric Coordinated Time TCB (Standish 1998). We then computed a Discrete Fourier Transform (DFT) for the combined 2000 data, shown in Fig. 1. Due to poor weather conditions during the run, our coverage is not continuous, causing gaps in the observed light curve; these gaps produce aliases in the Fourier transform. At the bottom of Fig. 1 we present the spectral window, the Fourier transform of a single sinusoid sampled exactly as the real data. It shows the pattern of peaks each individual frequency in the data introduces in the DFT.

Figure 2: Fourier transform of the GD 358 data, year by year.

Figure 3: The amplitude modulation of GD 358 423 s mode observed in the optical in August, 1996. The timescale of this change is surprisingly short. We have never observed such a fast amplitude modulation in any of the pulsating white dwarf stars.

The Fourier spectra displayed in Fig. 1 looks similar to the ones obtained in 1990 and 1994 (see Fig. 2), but the amplitude of all the modes changed significantly. As we describe in more detail later, the most striking feature of the 2000 data is the absence of triplets, except for k=9. The 1996 data are even more unusual than the WET runs due to the observation of amplitude changes over an unprecedented short time. We describe these observations in more detail in Sect. 3. We then describe the 2000 observations and our interpretations of them in Sects. 4 through 7.