TESS first look at evolved compact pulsators: Known ZZ Ceti stars of the southern ecliptic hemisphere as seen by TESS

Context. We present our findings on 18 formerly known ZZ Ceti stars observed by the TESS space telescope in 120s cadence mode during the survey observation of the southern ecliptic hemisphere. Aims. We focus on the frequency analysis of the space-based observations, comparing the results with the findings of the previous ground-based measurements. The frequencies detected by the TESS observations can serve as inputs for future asteroseismic analyses. Methods. We performed standard pre-whitening of the data sets to derive the possible pulsation frequencies of the different targets. In some cases, we fitted Lorentzians to the frequency groups that emerged as the results of short-term amplitude/phase variations that occurred during the TESS observations. Results. We detected more than 40 pulsation frequencies in seven ZZ Ceti stars observed in the 120s cadence by TESS, with better than 0.1 microHz precision. We found that HE 0532-5605 may be a new outbursting ZZ Ceti. Ten targets do not show any significant pulsation frequencies in their Fourier transforms, due to a combination of their intrinsic faintness and/or crowding on the large TESS pixels. We also detected possible amplitude/phase variations during the TESS observations in some cases. Such behaviour in these targets was not previously identified from ground-based observations.

A known characteristic of the ZZ Ceti pulsations is that we detect different pulsational behaviour within the temperature range that defines their instability strip. While variables closer to the blue edge (higher T eff ) are more likely to show pulsational frequencies with stable amplitudes and phases, short-term (days-weeks) amplitude and phase changes are more common closer to the red edge (lower T eff ). Longer periods and larger amplitudes are also detected at the cooler objects. The short-term amplitude variations can be results of the insufficient frequency resolution of the data sets, possible physical explanations are interaction of pulsation and convection (e.g. Montgomery et al. 2010), and resonant mode coupling (e.g. Zong et al. 2016).
The Kepler space-telescope (Koch et al. 2010) observations revealed another unusual behaviour of these objects recently, the so-called outburst events: recurring increases in the stellar flux (up to 15 per cent) in cool ZZ Ceti stars (see e.g. Bell et al. 2017a). This phenomenon might be in connection with the cessation of pulsations at the empirical red edge of the ZZ Ceti instability strip .
For comprehensive reviews of the observational and theoretical aspects of pulsating white dwarf studies, see the papers of Winget & Kepler (2008), Fontaine & Brassard (2008), Althaus et al. (2010), and Córsico et al. (2019). For a study on the pulsational properties of ZZ Ceti stars based on mainly space-based observations, see the paper of Hermes et al. (2017).
This paper is part of a series presenting the first results on compact pulsators based on TESS measurements. The other papers of this series focus on different types of compact variables: an sdBV (hot B subdwarf pulsator) star (Charpinet et al. 2019), another three sdBV stars (Sahoo et al., in prep.), a DBV (pulsating helium-atmosphere white dwarf) (Bell et al. 2019) and a DOV (hot hydrogen-deficient post-Asymptotic Giant Branch) variable (Sowicka et al., in prep).
This manuscript presents the study of 18 formerly known ZZ Ceti stars bright enough to be targeted with 2-min cadence observations with TESS, observed during the survey of the southern ecliptic hemisphere. We present the 120 s cadencemode TESS measurements collected on these stars in Sect. 2, give details on the frequency analysis of the data in Sect. 3 both for the stars with detected light variations and those not observed to vary this time, and we discuss our findings in Sect. 4.

TESS observations
We downloaded the light curves from the Mikulski Archive for Space Telescopes (MAST), and extracted the PDCSAP fluxes provided by the Pre-search Data Conditioning Pipeline (Jenkins et al. 2016) from the fits files. We removed all data points with quality warning flags, and finally we corrected the light curves for long-period systematics. We normalised the light curves by fitting a fourth-order Savitzky-Golay filter with a three-day window length computed with the Python package lightkurve (Barentsen et al. 2019). This correction does not affect the frequency domain of the short-period white dwarf pulsations. The panels of Figs. A.1 and B.1 show the resulting light curves. Note that the Pre-search Data Conditioning pipeline corrects the flux for each target to account for crowding from other stars. MCT 0145-2211, and L 19-2 were observed in one sector, BPM 31594 was measured in two, while HE 0532-5605, located in the TESS Continuous Viewing Zone (CVZ), was observed in 13 consecutive sectors. Tables 1 and 2 show the journal of observations of our targets observed and not observed to vary by TESS, respectively.

Light-curve analysis
We performed standard Fourier analysis and pre-whitening of the data sets with the photometry modules of the Frequency Analysis and Mode Identification for Asteroseismology (famias) software package (Zima 2008) 1 . We also used in-house developed software for the least-squares fitting, used for the analysis of Whole Earth Telescope (Nather et al. 1990), Kepler, and K2 data. We chose the detection limit to be at 0.1 per cent false alarm probability (FAP), that is, in this case, there is 99.9 per cent chance that a peak reaching this limit is not just result of noise fluctuations.
We calculated the 0.1% FAP threshold following the method described in Zong et al. (2016): we generated 10 000 synthetic light curves of Gaussian random noise for the times of the real observations, calculated their Fourier transforms up to the Nyquist limit (≈ 4167 µHz), and then determined the probability for the different signal-to-noise (S/N) ratios that a single peak emerges at that S/N value only due to noise fluctuations. The noise level was chosen to be the mean amplitude level of the given synthetic light curve. This way we could calculate the S/N values corresponding to the 0.1% FAP thresholds, and we accepted a frequency peak only above this S/N limit as significant.
However, in some cases, we were unable to remove all of the observed power by simple pre-whitening, suggesting the presence of amplitude/phase variations in the signal. We fitted Lorentzian envelopes to these peaks as described in Bell et al. (2015) and Hermes et al. (2017). We fitted the peaks in the power spectrum with the function presented in eq. 1 of Bell et al. (2015) and eq. 1 of Hermes et al. (2017). The Fourier transforms were oversampled by a factor of six.

Ground-based observations
Ross 548 (TIC 29854433, G = 14.23 mag 2 , α 2000 = 01 h 36 m 14 s , δ 2000 = −11 d 20 m 33 s ) is also known as ZZ Ceti, and is the namesake of the class of hydrogen-atmosphere pulsating white dwarfs. It was one of the first pulsating white dwarfs discovered by Lasker & Hesser (1971). The star has shown a remarkably stable pulsation spectrum for over 48 years, with two dominant periodicities (each now known to be triplet (ℓ = 1) modes). Lasker & Hesser (1971) isolated two modes with periods of 212.864 s (4697.8 µHz) and 273.0 s (3663 µHz). These relatively short periods are consistent with its position near the blue edge of the ZZ Ceti instability strip, with an effective temperature of 12 200 K (Giammichele et al. 2015, and references therein).
The stability of the pulsation amplitudes and frequencies has been studied in depth, in part to determine the secular period 1 famias is a package of software tools for the analysis of photometric and spectroscopic time-series data. It enables us to search for periodicities in these data sets using the method of Fourier analysis and nonlinear least-squares fitting techniques. famias is also capable of performing mode identification of the detected periodicities utilising different photometric and spectroscopic methods. 2 Gaia Data Release 2 G magnitude, Gaia Collaboration (2018) Zs. Bognár et al.: Known ZZ Ceti stars of the southern ecliptic hemisphere as seen by TESS Table 1. Journal of observations of the eight targets showing light variations in their TESS data sets. All data were collected in 120 s-cadence mode. N is the number of data points after the removal of all those with quality warning flags, δT is the length of the data sets including gaps, and Sect. is the serial number of the sector(s) in which the star was observed. The start time in BJD is the time of the first data point in the reduced data set. The CROWDSAP keyword represents the ratio of the target flux to the total flux in the TESS aperture. change rates driven by the cooling of the star. This decades-long effort is summarised in, for example, Mukadam et al. (2013). The two main modes identified by Lasker & Hesser (1971) have been resolved as triplets, with the amplitude of the central m = 0 peak about 3-5 times smaller than the flanking m = ±1 peaks. The splitting is 4.0 µHz for the 213 s mode, and 3.5 µHz for the 275 s mode. Mukadam et al. (2013) report that the amplitudes of the m = ±1 modes in Ross 548 have been very stable -a fact that has enabled measurement of the long-term secular change in the pulsation phase, and therefore the rate of period change. This decadeslong stability in principle allows us to use Ross 548 to help calibrate the pulsation amplitudes of white dwarfs observed by TESS without having simultaneous ground-based observations.
In Table 3, we present the weighted average amplitudes of the largest-amplitude modes in Ross 548. Amplitudes we used are those reported by Giammichele et al. (2015), Mukadam et al. (2013), and references therein.
The bandpass of TESS is much redder than that of ground based photomultiplier and CCD observations, but we can compute the expected amplitude in the TESS bandpass quite readily. Since the g−mode pulsations reflect temperature changes across the white dwarf surface, we can (assuming a blackbody with T ≈ 12 200 K) compute the amplitude ratio in the TESS bandbpass (centred at approximately 825 nm) to that in the optical (assumed to be 425 nm), following Kawaler et al. (1994): For T eff =12 200 K, this implies that the amplitudes observed by TESS should be about 63% of those observed in the optical. The last column in Table 3 are therefore the expected amplitudes for TESS observations. We compiled the published observed amplitudes of Ross 548 to obtain an estimate of the expected amplitude for our observations with TESS.

TESS observations
TESS observed Ross 548 during Sector 3. The light curve is shown in Fig. A.1. Note that towards the end of the first segment, the noise level in the light curve increased (a common issue for Sector 3 data). We chose to retain those data to improve the duty cycle after examining the periodogram with and without the problematic segment to ensure that the noise did not adversely impact the frequency analysis. Our 0.1% FAP level is at a S/N of 4.59 (corresponding to an amplitude of 1.24 mma). In the periodogram, the only peaks that rose above that threshold were precisely at the frequencies corresponding to the known pulsation modes (or their Nyquist aliases) listed in Table 3. While the 275 s modes (at 3646.3 and 3639.4 µHz) were below the Nyquist frequency of 4167 µHz, the 213 s modes (at 4692.6 and 4699.2 µHz) are above the Nyquist frequency. They are "reflected" back across the Nyquist frequency to 3640.7 and 3634.1 µHz. As we show in Fig. 1, the two aliases of the 213 s modes intermingle with the two 275 s peaks. Beyond the Nyquist frequency, the same pattern repeats, with the alias of the 275 s modes intermingling with the "true" 213 s peaks. We performed a least-squares fit to the time series of Ross 548 using the prior values of the four frequencies as a starting point. The formal fit to the light curve is given in Table 4. Pre-whitening the data using these four frequencies completely removes all peaks in this region, and their mirror aliases, at or near the 0.1% FAP. Table 3. Periods and frequencies derived by ground-based observations on Ross 548. The next-to-last column lists the corresponding weighted mean amplitudes (near V magnitude), while the last column is the expected amplitude in the TESS bandpass assuming a blackbody spectral energy distribution as per Eq.   Table 3 (red lines), and the aliases (reflected across the Nyquist frequency) of f 1 and f 2 (blue vertical lines). The right-hand panels show the transform above the Nyquist frequency, where the real periodicity -alias situation is reversed. The two bottom panels show the residuals in the spectrum after pre-whitening by the least-squares fit sinusoids f 1 , f 2 , f 3 , f 4 listed in Table 4.
For more details considering the super-Nyquist analyses of compact pulsators, see the papers of Baran et al. (2012) and Bell et al. (2017b).
Finally, for comparison with the ground-based measurements of the amplitudes of the modes of Ross 548, we need to correct the fit values of the amplitudes for the fact that the frequencies lie close to (and beyond) the Nyquist frequency. Thus the signals are attenuated by the smoothing/averaging over the 120 s TESS integrations. Assuming a sinusoidal waveform, the functional form of the attenuation is a sinc function in frequency, with the first zero at twice the Nyquist frequency. The corrected amplitudes are in the fifth column of Table 4. The corrections were applied to the uncertainties, and the ground-based optical amplitudes were reduced by the bandpass and integration near the Nyquist limit to compare with the TESS observed values.
Considering the 11-frequency solution presented by Giammichele et al. (2015), we found that only the four highestamplitude modes can be seen in the TESS data set. The rest, with ground-based amplitudes of 0.3-1.4 mma, all fall below the detection limit.
Comparing the (corrected) observed amplitudes from TESS to the expected values from ground-based data over the past 48 years, we see that there is close agreement for all 4 modes resolved by TESS. Therefore, we expect that amplitudes from TESS can be directly compared with archival and future groundbased measurement of the amplitude of white dwarf pulsations (after correcting for different bandpasses and exposure times). . They also detected the second harmonics of several frequencies, and additional closely spaced frequency components at the 1010.86 µHz peak. Apart from the discovery paper, the results of only one observation have been published up to now: Thompson et al. (2005) presented a two-frequency solution with periods at 878.8 and 508.1 s, based on time-series spectroscopy.

Results of the TESS observations
If we pre-whiten the data set with the standard procedure down to the significance threshold, we can identify 17 frequencies above the 0.1% FAP limit (S/N = 4.74). However, these are not all independent pulsation modes, but they form groups of peaks in some cases, suggesting amplitude/phase variations occurring during the TESS observations. Actually, we found that the high frequency peaks (those with periods shorter than 800 s) are stable enough in amplitude and phase that pre-whitening leaves no residuals above the detection threshold. However, significant residuals remain after removing the lower frequency peaks. This feature is consistent with fig. 5 of Hermes et al. (2017), that shows the stochastic nature appearing at periods longer than about 800 s. Therefore, we fit a Lorentzian envelope to the regions at 1144 and 1206 µHz (see Fig. 3). The dominant frequency at 1367 µHz (∼ 730 s) shows intermediate behaviour -removing the main peak leaves behind some additional peaks, but that one peak's dominance makes fitting a Lorentzian challenging. The damping times calculated for the 1144 and 1206 µHz regions are 0.183 and 2.252 days with Lorentzian widths of 25.5 ± 1.0 µHz and 1.59 ± 0.11 µHz, respectively. Table 5 lists the results of the 10-frequency fit, including the parameters of the Lorentzian fits, while Fig. 2 shows the Fourier transform of the whole light curve. Red dashed lines mark the frequencies listed in Table 5.
Note that the dominant peak and the peaks at 1372.116 and 1376.720 µHz form a triplet with 5.034 and 4.604 µHz frequency separations, respectively, while we see similar frequency difference of 4.085 µHz between the peaks at 1143.895 and 1139.810 µHz. This suggests that we see the results of rotational frequency splitting at least in the first case. Considering the 4.085 µHz separation, the approximate rotation period of the star could be 1.42 or 2.36 d, assuming the rotational splitting of ℓ = 1 or ℓ = 2 modes, respectively.
We also mark in Table 5 the frequencies obtained by groundbased observations. Apparently, Stobie et al. (1993) detected a set of modes completely different from those we determined by the TESS observations, as if we observed two different stars. Besides the detected complex frequency structure with closely spaced peaks in the Fourier transforms, this also suggests that some of the mode amplitudes are not stable.  McGraw (1976). The authors presented the results of four nights of observations. They reported an interesting change in the pulsational behaviour of BPM 31594. On the discovery night it showed light variations with 314 s, but on the subsequent nights, its periodicity changed to 617 s. However, it turned out later, that this change was not real, but only result of erroneous data reduction of the first night's measurements (Odonoghue 1986). McGraw (1976) also detected the two harmonics of the 617 s mode ( f 0 ), significant power in the spectra near (3/2) f 0 , and combinations of frequencies.
O' Donoghue et al. (1992) presented the results of the longterm investigations of BPM 31594 carried out between 1975 and 1989. They found the dominant frequency to be at f 0 = 1.62 mHz (617 s) again, and detected significant peaks at multiples of 1.54 f 0 . In some cases, further frequencies appeared at multiples of 1.48 f 0 . They also identified, that at f 0 there is actually a rotationally split triplet with 12.7 µHz frequency separations. Besides these, they detected the harmonics of f 0 , an additional frequency at 1.8778 mHz (533 s), and several combination terms. Considering the stability of frequencies, they found variability in the period of f 0 , and also variable amplitudes at the other frequencies from season to season. The authors summarised their findings as the observed structure of the frequencies of BPM 31594 can be described by only a few independently excited modes, which drive others through non-linear coupling by direct resonance.

TESS observations
The dominant peak is at 1618.405 µHz, in agreement with the results of the previous observations. This peak is also the central component of a triplet with 13.42 µHz frequency separations. This is also in good agreement with the 12.7 µHz finding of O'Donoghue et al. (1992) by ground-based observations. Note that there are also residual closely spaced peaks at the triplet components. These are partly caused by the (apparent) changing noise level between the two sectors and sub-sector parts.
We also detected the second harmonic of the dominant frequency, significant peaks at 1.54 and 2.54 times the dominant frequency, two independent frequencies at 1461.119 and 1548.455 µHz, a combination term at 3166.876 µHz, and a low-A&A proofs: manuscript no. 18DAVs  Table 5, while the blue dashed line represents the 0.1% FAP level (S/N=4.74) calculated by this Fourier transform. Table 5. This work: results of the fit of the EC 23487-2424 data set with 10 frequencies. The errors are formal uncertainties. The results of Lorentzian fits are marked with star symbols ( * ). In the case of the supposed rotationally split frequencies, we mark the possible azimuthal order values. For completeness, we also list the periods detected by Stobie et al. (1993) and Thompson et al. (2005  Besides these, four additional peaks can be determined by the data set, close to, but well-separated from the peaks at 1.54, 2, and 2.54 times the dominant peak, respectively. These are at f w = 2483.202, f x = 3245.652, f y = 3252.593, and f z = 4115.083 µHz. It is remarkable, that the spacing between f w and 1.54 f 1 is 6.952 µHz, which is nearly half the spacing within the triplet around the dominant peak. This is also true for the spacings between f x and f y (δ f = 6.941 µHz), and f z and 2.54 f 1 (δ f = 6.50 µHz). The origin of these additional peaks is in question. Table 6 lists the 14-frequency solution without the closely spaced residual peaks (the 0.1% FAP limit is S/N=4.85). These frequencies are also marked in Fig. 4, showing the Fourier transform of the whole light curve.  Hesser et al. (1976). They identified several periodicities between 1190 and 1640 µHz in 24 observing runs spanning several months. The power spectrum was variable, but most of the peaks in the Fourier transforms of individual runs were situated in the 1000 -1900 µHz domain (see Table 7). In retrospect, we recognise this as typical behaviour of cool DAV stars, with complex pulsation spectra of modes with periods longer than 800 seconds (Hermes et al. 2017). To our knowledge, no time-series photometric observations of BPM 30551 have been performed since the discovery observations in 1974. Given this, the lowresolution Fourier transforms available in Hesser et al. (1976) make it difficult to compare with the TESS results.

TESS observations
BPM 30551 was observed for 25.4 days during the 27.5 day observing sector 2, so therefore with a duty cycle of 92.5%. The light curve of BPM 30551 (see Fig. A.1) shows no evidence for bursts, which occur in some DA white dwarfs that pulsate in similar periods (Bell et al. 2016). There were no apparent lowfrequency variations (1 day or longer) following examination of the un-flattened light curve.
The Fourier transform of the light curve of BPM 30551the first observations since the mid 1970s -confirms that it is indeed a DAV pulsator, with the largest-amplitude variations at frequencies between 1100 and 1300 µHz. Fig. 5 shows the FT over the range of frequencies reported by Hesser et al. (1976), with red lines showing the approximate positions of their identifications. In the TESS data, the only significant signals (i.e. S/N above 4.69) are in a narrow frequency range between 1195 and 1215 µHz, with a second region around 1260 µHz that lies just above the significance threshold.
An expanded view of the FT of the BPM 30551 data is shown in Fig. 6. The signal in the range from 1200-1215 µHz is extremely complex, indicating that the mode or modes of pulsation are not coherent over the 27 day duration of the run. In fact, this is very similar to what is seen in the longer-period pulsations of DAV stars near the cool end of the instability strip, which is reminiscent of stochastically driven oscillations seen in solar-type asterosesimic data. Attempts to identify and remove coherent pe-riodicities in this region were unsuccessful -pre-whitening was unable to remove all of the observed power, and as individual coherent periodicities were removed, new nearby peaks appeared. Therefore, we fit a Lorentzian envelope to this region, and the region around 1260 µHz. The central frequencies of the Lorentzian fits to the frequency groups are as follows: 1204.17 ± 0.02 µHz, and 1264.49 ± 0.16 µHz. Note that as Fig. 6 shows, we can fit a third frequency group at ∼ 1140 µHz with the central frequency at 1144.75 ± 0.15 µHz, even though the corresponding peaks are below our detection threshold.
We find a damping time for the strongest modes of about 0.77 days; thus the 27 day observing span dilutes the true power of the oscillations over a broad range of frequencies. Shorter observations, such as a typical 1 night terrestrial observation which is shorter than the typical damping time for these modes, would be expected to see much large amplitude pulsations. Indeed, simulations show that modes which conform to the fitted Lorentzian power envelope widths and heights can produce peaks in excess of 20 mma within that frequency range. Thus the results from Hesser et al. (1976) are consistent with what TESS observed. However, the modes seen by Hesser et al. (1976) at other frequencies are (apparently) either too low in amplitude to be detected, or are no longer present.  Fontaine et al. (2003). It was observed over two consecutive nights for 3740 s on the first and 5780 s on the second, each with sampling times of 10 s. They identified three periodicities at 1215, 1374, and 2164 µHz with amplitudes ranging from 17 to 25 mma; they noted that they expected that more periods could likely be resolved with longer observations. They determined the temperature of the star to be around 11 550 K, making this another example of a DAV being in the middle of the instability strip.
These observations were followed up by Kilkenny et al. (2014). They observed for 50 min with a sampling time of 0.1 s. The frequency resolution of their data was approximately 8 µHz, estimated using their data using Equation 5.52 of Aerts et al. (2010). They identified three periodicities in this short data set at 1370, 2200, and 2290 µHz with amplitudes between 6 and 10 mma which, given the brevity of their observations, are consistent (in frequency) with two of the three identified by Fontaine et al. (2003).

TESS observations
From the TESS light curve, we see that MCT 0145-2211 is still pulsating, with periodicities that are in the same range as reported earlier. However, some differences with the earlier data are evident. The highest peak in the FT in Fig. 7 is a single, fully resolved peak at 2407.20 µHz, with an amplitude of 11 mma, that was not seen in the earlier work. In this context, a frequency is considered resolved when it is a single peak that can be pre-whitened with a sinusoid. The second-highest peak(s) lie between 2212 and 2216 µHz, consistent with what was reported earlier. There are at least three peaks within this frequency range, but the 20.3 day span of Sector 3 data on this star was not long enough to fully resolve them. Least-squares fits and A&A proofs: manuscript no. 18DAVs Table 6. 14-frequency solution of the BPM 31594 data set. The errors are formal uncertainties calculated by famias. In the case of the rotationally split frequency, we mark the azimuthal order values.      pre-whitening in this frequency range revealed at least 3 closely spaced but unresolved modes.
At lower frequencies, the TESS data show several groups of significant peaks that fall within roughly Lorentzian envelopes, at 956 µHz, 1257 µHz and 1306 µHz (periods of 1046 s, 796 s, and 766 s). As with the other cool DAVs in this study, we were unable to completely pre-whiten peaks within these three regions, indicating that the modes may be amplitude/phase unstable over the span of the observations. In Fig. 8, we expand the view around several peaks, and show the residuals after removing the largest peaks in each range. This illustrates that the 2407.20 µHz mode is a single coherent periodicity, while the modes with frequencies below 1400 µHz are complex. The 2215 µHz mode seems under-resolved -we can find 4 frequencies using least-squares sinusoidal fitting that removes most of the peak, but the frequencies are separated by small amounts that are nearly at the expected resolution for a run of this length.
Summarising, the only resolved, coherent oscillation can be fit with a sinusoid with a frequency of 2407.200 ± 0.014 µHz, with an amplitude of 10.9 ±0.6 mma. Considering the mode near 2215 µHz, there is insufficient time to resolve this into individual peaks, and therefore we can not yet say if this is a stochastically excited mode or if it is a closely spaced multiplet. If it is a closely spaced multiplet, and rotation is what causes the splitting, then it would be one of the most slowly rotation white dwarfs with asteroseismic results as per Hermes et al. (2017).
There seems to be a 'triplet' at 1305.215, 1306.980, and 1308.724 µHz (all about ±0.04 µHz) though all three are close to the false-alarm limit. A 1.7 µHz splitting corresponds to a rotation period of 3.3 days if that 'triplet' is indeed associated to an ℓ = 1 mode. The peaks on either side of that 'triplet' at 1297.55 and 1317.98 are nearly equally spaced around that 1307 µHz clump. However, the three peaks close to 1307 µHz are not a triplet but a single stochastically excited mode, and the 1298 and 1318 µHz to either side make up the rest of the (more widely spaced) triplet that would be present if the rotation period was 0.57 days. We were unable to get a satisfactory fit using Lorentzians for that wider-triplet case.
As we mentioned above, there are two additional frequency regions with peaks at 956 and 1257 µHz, respectively. Note that they are also not resolved, and all close enough to the significance threshold.
3.6. L 19-2 3.6.1. Ground-based observations  be a relatively simple ZZ Ceti with a pulsation spectrum dominated by two short-period modes with periods of 193 s and 114 s. This star was the target of a Whole Earth Telescope (Nather et al. 1990) campaign in 1995, which revealed L 19-2 to pulsate in at least 10 detectable eigenmodes (preliminary analyses are presented in Sullivan 1998;Yeates et al. 2005;Sullivan & Chote 2015). Continued monitoring from Mt John University Observatory was used to measure the secular rate of period change from the cooling (and proper motion) of this white dwarf of dP/dt = (4.0 ± 0.6) −15 s s −1 (Sullivan & Chote 2015), the third such measurement for a DAV. This value is slightly higher than predicted by asteroseismic models, causing Córsico et al. (2016) to argue that the cooling could be accelerated by the emission of axions. Bradley (2001) also presented asteroseismological constraints on the structure of L 19-2.
Article number, page 10 of 21 Zs. Bognár et al.: Known ZZ Ceti stars of the southern ecliptic hemisphere as seen by TESS

TESS observations
The periodogram of the TESS Sector 12 data on L 19-2 reveals a single significant signal at 3141.461 ± 0.013 µHz with an observed amplitude of 1.92 ± 0.11 mma. We recognise this as the alias of the dominant mode observed in previous studies of this star, reflected across the Nyquist frequency. We identify the intrinsic signal from this mode in the periodogram computed out to twice the Nyquist frequency, as displayed in Figure 9. Refining the fit of a sinusoid for this signal to the light curve, we obtain a frequency measurement of 5191.853 ± 0.013µHz (192.6095 ± 0.0005 s period). The 2-minute exposures smooth the light curve, reducing the observed amplitude of this signal to 47% of its intrinsic amplitude. The amplitude measured in shorter exposures would be close to 4 mma, which is comparable to values measured from high speed photometry in the literature, allowing for the difference in observational filter. The second highest amplitude signal previously observed from L 19-2 has a period of 114 s; this is very close to the TESS exposure time, suppressing this signal to well below the periodogram noise floor at 5% its intrinsic amplitude.  Castanheira & Kepler (2009) found this star to be near the red edge of the ZZ Ceti instability strip at T eff = 11 560 K, thus it is expected to have complex and long periodic oscillations generally exceeding 600 s. They identified additional periodicities in the Fontaine et al. (2003) data, ranging from 1100-1950 µHz. The most recent observations of HE 0532-5605 came from Kilkenny et al. (2014), who used the Berkeley Visual Imaging Tube detector to observe HE 0532-5605 for roughly 3000 s with a 0.1 s integration time. They detected periods of 438, 707, and 1380 s (720, 1410, and 2280 µHz respectively) and indicated that they believed that the periodicity around 1410 µHz may be unstable and related to the peak originally detected by Fontaine et al. (2003).

TESS observations
HE 0532-5605 is located near the southern ecliptic pole and as a result can be observed by TESS for nearly a year. Here, we report on data from all of the 13 sectors available. Each sector was reduced independently, as described earlier, after which the sectors were combined into a single light curve. The Fourier transform of the entire data set is shown in Fig. 10 -the 0.1% FAP level (S/N = 5.24) lies at 1.93 ppt, which is remarkable given the faintness of the star and the limited aperture of the TESS camera.
There are no significant periodicities detected in HE 0532-5605 in the complete data set. The amplitudes of the modes reported from ground-based studies ranged from 2-8 mma, and were seen in much shorter data sets than the full TESS light curve. Given that the reported periods are in excess of 600 s, and that this is a relatively cool DAV, it is likely that the modes observed from the ground have short lifetimes resulting from changes in phase and amplitude. As we saw in other targets in this study, and described by Bell et al. (2015), this lack of coherence can reduce the signal amplitude in the Fourier transform if the duration of the observations is much longer than the damping time. Scaling down the expected amplitudes to account for the redder bandwidth of the TESS data can also help explain the failure of these modes to appear in Fig. 10.
To look for shorter-lifetime modes, we examined each sector of the TESS data individually to see if significant periodicities were present. However, since each sector is 1/13 as long as the total run, the noise in the Fourier transform in the individual sectors is nearly three times higher, putting our 0.1% FAP threshold at about 6 mma. No significant periodicities were detected, however, in Sectors 1, 3, and 9, some peaks came close to reaching the 0.1% FAP threshold. We show Fourier transforms of these three sectors in Fig. 11. In Sectors 3 and 9, the highest peaks were close to frequencies reported by Fontaine et al. (2003) and Castanheira & Kepler (2009).
The peak near 2500 µHz in Sector 1 is broad and reminiscent of those seen in the shorter runs on cool DAV stars here (and in Hermes et al. 2017) and in K2 observations of the DAV PG 1149+057 . In the latter case, the star in question is a "burster", which shows semiregular increases in brightness coincident with large amplitude pulsations. Close examination of the Sector 1 light curve of HE 0532-5605 suggests that such a burst may have occurred at about BJD 2 458 340. In Fig. 12 we expand the light curve around this interval. For this figure, we did not flatten the light curve, as the flattening process reduced the amplitude of the apparent burst. The possible burst reaches an amplitude of nearly 20%, and the duration is approximately 24 hours.
The feature at 2530 µHz in the top panel of Fig. 11 appears to be associated (in time) with the feature at 1341 d (BJD -2 457 000) in Fig. 12. Figure 13 shows the Fourier transform of the first and second halves of the Sector 1 data on HE 0532-5605. The burst appears soon after the start of the second half of the observation. The feature at 2530 µHz is only present in the second half of the sector as well.
Since this star only showed mildly coherent pulsations during a portion of Sector 1, determining the frequency or frequen-cies present is a very uncertain task. The FT rose above the significance threshold between approximately 2520 and 2550 µHz. We fit a Lorentzian envelope to the strongest part of this range and so estimate the frequency for this mode at approximately 2527 µHz.
It appears, therefore, that HE 0532-5605 may be a bursting, pulsating DA white dwarf. The duration and amplitude of the burst, and the increased pulsation amplitude during the burst, has been seen in other stars. However, most of the bursting white dwarfs recur on an irregular basis, but roughly every 8 days or so. No other clear bursts have been seen in the TESS data over 13 sectors, though it is possible that the nearly-significant features in Sectors 3 and 9 may reflect smaller bursts that aren't apparent in the relatively low S/N data from this faint star. Also, the periodicity present during the burst is at a relatively short period (395 s, or 2530 µHz). HE 0532-5605 merits monitoring with larger instruments to confirm its possible membership in the rare class of bursting, pulsating DA white dwarfs.

HS 0507+0434B
3.8.1. Ground-based observations HS 0507+0434B (TIC 455094688, G = 15.40 mag, α 2000 = 05 h 10 m 14 s , δ 2000 = +04 d 38 m 55 s ) was discovered to be a member of the ZZ Ceti class by Jordan et al. (1998). Their observations showed that the star's light variations can be described with three or four independent modes at 278.4, 355.1, 445.2, and 558.7 s, and a large number of combination terms, with 18 frequencies altogether. Further observations revealed that at some frequencies there are actually rotationally split triplets. Handler et al. (2002) detected 10 independent frequencies consisting of three triplets and a singlet frequency, and also a large number of combination frequencies, while Kotak et al. (2002) detected two doublets and three additional modes besides the combination terms. Castanheira & Kepler (2009) listed four modes for their asteroseismic fit, while the latest time-series photometric observations on HS 0507+0434B were presented by Fu et al. (2013). They derived 18 independent pulsation frequencies consisting of five triplets, a doublet, and a singlet frequency besides the combination terms.

TESS observations
In contrary to the rich frequency spectrum observed e.g. by Fu et al. (2013), we only detected five peaks in the TESS data set of HS 0507+0434B above the significance level. Table 11 lists the frequencies, periods and amplitudes of these peaks. The 1680.072 µHz frequency represents a new mode not observed before, while the other four peaks are side components of formerly detected triplets. No combination frequencies could be derived by the TESS data, however, we have to take into account the amplitude suppression factor of the 2-minute exposures, especially in the case of peaks expected to emerge with low amplitudes. Note that we listed an additional peak in Table 11 in parentheses. This represents the central component of the ∼ 2810 µHz triplet, however, this peak is under our significance threshold (S/N=3.7).  Fontaine et al. (2003), dashed lines are additional frequencies from that data reported by Castanheira & Kepler (2009), and blue vertical lines (offset upwards) mark those from Kilkenny et al. (2014).  (72) The f 2 and f 3 frequencies are found to be stable during the time-span of the TESS observations, as these frequency domains can be pre-whitened with the doublet frequencies alone, and no further closely spaced frequencies emerged as in e.g. EC 23487-2424. Figure 14 shows the original and the pre-whitened Fourier transforms, respectively.
Considering the influence of the companion HS 0507+0434A (TIC 455094685) on the frequency determination of HS 0507+0434B, we found that TESS cannot separate individual signals from these two objects since the separation between the two stars (∼ 15 arcsec) is smaller than the plate scale of the TESS detector (21 arcsec px −1 ). However, investigating the amplitudes of the signals originate from the two sources, we found that at the position of HS 0507+0434B the measured amplitudes are a factor of 4.02 ± 0.04 (error of the mean) larger than at the position of HS 0507+0434A, so they must come from component B. Another argument that the frequencies detected belong to HS 0507+0434B, is that there are ground-based studies which clearly resolve both components, including the paper of Handler et al. (2002), where the authors studied the variability of HS 0507+0434B, using HS 0507+0434A as well as other stars for comparison.
3.9. ZZ Ceti stars not observed to vary in the TESS data TESS targeted additional previously known, bright ZZ Ceti variables in the southern hemisphere that did not exhibit significant pulsation signals in the TESS data, namely MCT 2148−2911, HE 0031−5525, EC 00497−4723, MCT 0016−2553, WD 0108−001, HS 0235+0655, KUV 03442+0719, WD J0925+0509, HS 1013+0321, and EC 11266−2217. For completeness, we report on these ten stars in Appendix B, comparing the upper-limit detection thresholds of the TESS data to pulsation amplitudes reported in the literature.

Summary and conclusions
We presented the frequency analysis of 18 formerly known ZZ Ceti stars observed by the TESS space telescope during the survey observation of the southern ecliptic hemisphere (sectors 1-13). We compared our results to those from previous groundbased observing campaigns. Eight out of 18 of our targets show at least marginal evidence of pulsations that we were able to characterise, with at least 40 statistically significant (FAP < 0.1%) pulsation detections in total. From ∼month-long TESS light curves with very high duty cycles, we have measured pulsation frequencies to 0.1 µHz precision. These are eigenfrequencies of these white dwarf stars and they can be compared to stellar models to infer the interior conditions of these compact pulsators. A follow-up paper from TASC WG#8 is planned (Romero et al. in prep.) that will use different sets of models to interpret the precision measurements reported here. Figure 15 shows the ZZ Ceti instability strip with known ZZ Ceti stars (atmospheric parameters are from Bognár & Sódor 2016), together with all the 18 objects studied in this paper. The corresponding atmospheric parameters from the literature are listed in Table 12 for the 18 targets (Fuchs 2017). The T eff and log g values were determined by the use of the ML2/α = 0.8 version of the mixing-length theory, and corrected according to the findings of Tremblay et al. (2013) based on radiationhydrodynamics three-dimensional simulations of convective DA stellar atmospheres.
We demonstrated the difference between the TESS amplitudes and those detected in the ground-based measurements  through the case of Ross 548. We have to take several different effects into account. First, TESS observes in a redder bandpass than typical ground-based observations, which results in lower pulsation amplitudes, see Eq. 1. Second, the relatively large pixel size of TESS often causes contamination, and hence signal-to-noise suppression. Third, the 120 s exposure times of the TESS observations smear out the short-period signals of the ZZ Ceti stars. And finally, phase and/or amplitude variations on timescales shorter than the month-long TESS light curves reduce the pulsation amplitudes by distributing power across multiple frequency bins. These effects explain why we did not detect pulsations in 10 low-amplitude pulsators, and why there are missing frequencies compared to the ground-based results in some of the TESS variables.
The detection threshold corresponding to an uncontaminated target in a single sector is defined as 0.1%FAP × √ CROWDSAP × √ number of sectors and is shown in Fig. 16. The plot demonstrates the pulsation-detection performance of TESS. This threshold obviously depends on the magnitude of the ZZ Ceti star. The eight stars with confirmed pulsations have G < 15.5 mag (except for HE 0532-5605). The brightest of the stars with unconfirmed pulsations is G = 15.3 mag and most are fainter than G = 15.8 mag. There is, however, significant scatter in the calculated threshold at a given brightness, especially at targets fainter than 15.5 mag, possibly due to instrumental effects.
In spite of the difficulties, we were able to detect new frequencies for five stars (EC 23487-2424, BPM 31594, BPM 30551, MCT 0145-2211, HS 0507+0434B). We found that HE 0532-5605 may be a new outbursting ZZ Ceti star. We also found possible amplitude/phase variations during the TESS observations, which resulted in the emergence of groups of peaks in the data set of EC 23487-2424, BPM 30551, and MCT 0145-   Fig. 13. Fourier transform of Sector 1 data on HE 0532-5605. The black curve is for the first half of Sector 1, and the red curve is for the second half, which contains the candidate burst 2211. We fitted Lorentzians to these frequency groups to approximate pulsation frequency values. Such behaviour in these stars was not identified from the ground before.
Another thing we can learn from the TESS measurements is that we have to be cautious with the interpretation of the frequencies considering the Nyquist-limit of the 120 s exposures (≈ 4167 µHz). In the case of the peaks close to the Nyquist frequency, we have to take into account the possibility that these are intrinsically above the Nyquist-limit. The barycentric correction applied to the observation timings makes the data deviate from strict equidistancy that, in turn, lifts Nyquist-degeneracy, that is, the pseudo-Nyquist alias will have lower amplitude in the Fourier spectrum than the real frequency peak (Murphy et al. 2013). However, in the one-month-long TESS observations, this effect is far too weak for safe discrimination. We found that the amplitudes across n×Nyquist behave differently depending on a star. In some stars, the amplitudes are the highest in the sub-Nyquist region, getting lower with increasing frequency, however, sometimes we see that the amplitudes of the signals in question are changing significantly up and down, while being higher in the super-Nyquist region across many times the Nyquist frequency. Some of this behaviour could be cause by intrinsic variations of the pulsation signals.
Despite these difficulties, TESS observations of ZZ Ceti stars proved that measurements of bright DAVs allow us to detect the signs of short-term amplitude/phase variations, and also derive new pulsation modes with high precision thanks to the (almost) continuous, homogeneous and high-quality data. However, the value of 20 s cadence observations is also obvious. The 2 min cadence induces a significant decrease of amplitudes, which is largely responsible for the non-detection of pulsations in several targets. Offering 20 s cadence mode observations for short period compact pulsators in the extended mission could infer a breakthrough in the study of these kinds of variables.  Bognár & Sódor 2016), and the DAVs observed by TESS during the the survey measurements of the southern ecliptic hemisphere (sectors 1-13) in the T eff − logg diagram. Black and green dots mark the stars found to be variable and not observed to vary by the TESS measurements, respectively. Their atmospheric parameters were derived by Fuchs (2017), except in the case of KUV 03442+0719, where we used the values presented in Bognár & Sódor (2016). Blue and red lines denote the hot and cool boundaries of the instability strip, according to Tremblay et al. (2015).
Acknowledgements. The authors thank the anonymous referee for the constructive comments and recommendations on the manuscript. We also thank the comments of Kosmas Gazeas (National and Kapodistrian University of Athens). This paper includes data collected with the TESS mission, obtained from the MAST data archive at the Space Telescope Science Institute (STScI). Funding for the TESS mission is provided by the NASA Explorer Program. STScI is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-26555. Support for this work was provided by NASA through the TESS Guest Investigator program through grant 80NSSC19K0378. ZsB acknowledges the support provided from the National Research, Development and Innovation Fund of Hungary, financed under the PD 17 funding scheme, project no. PD-123910. This project has been supported by the Lendület Program of the Hungarian Academy of Sciences, project No. LP2018-7/2019. KJB is supported by the National Science Foundation under Award No. AST-1903828. ASB gratefully acknowledges financial support from the Polish National Science Center under project No. UMO-2017/26/E/ST9/00703. GH acknowledges finan- Table 12. Atmospheric parameters of the ZZ Ceti stars observed by TESS during the the survey measurements of the southern ecliptic hemisphere (sectors 1-13). The T eff and logg values are from Fuchs (2017), except for KUV 03442+0719, where we used the values listed in Bognár & Sódor (2016). The external uncertainties for the observations presented by Fuchs (2017) are 52 K in T eff and 0.013 dex in logg, respectively, while 173 K and 0.06 dex for KUV 03442+0719, as given by Gianninas et al. (2011