A&A 428, 925-934 (2004)
DOI: 10.1051/0004-6361:20040457

A photometric pilot study on Sonneberg archival patrol plates

How many "constant'' stars are in fact long-term variables?[*]

N. Vogt1,2 - P. Kroll 1 - E. Splittgerber 1

1 - Sonneberg Observatory, Sternwartestr. 32, 96515 Sonneberg, Germany
2 - Instituto de Astronomia, Universidad Catolica del Norte, Avenida Angamos 0610, Antofagasta, Chile

Receiced 16 March 2004 / Accepted 7 June 2004

The light curves of 216 arbitrarly chosen field stars and of 23 known variables in the Aur/Tau/Ori region were derived ( $7\hbox{$.\!\!^{\rm m}$ }8 \leq B \leq 12\hbox{$.\!\!^{\rm m}$ }2$) from scanned, blue-sensitive archival patrol plates, covering a total of 34 years (1961-1995). We achieved a photometric accuracy of 0.07 ... 0.12 mag in spite of rather unfavourable locations of most stars near the plate borders. 17 field stars turned out to be variables, most of them with time scales of 1000-8000 days in the form of slow waves with amplitudes between 0.1 and 0.3 mag, i.e. below the threshold of traditional variable searches on photographic plates. About 50% of these new long-term variables exhibit drifts indicating periodic or erratic variability at much longer time scales than covered here. For the 23 known variables we achieved improvements in their periods and amplitudes and detected long-term variations (drifts, waves) in about 50% of them. The above fraction of low-amplitude long-term variables among field stars implies that a total of about 45 000 new variables should be detectable in the Sonneberg patrol plate archive. They will represent a new, hitherto not investigated population of variable stars with a possibly significant impact on our understanding of the stellar interior and evolution.

Key words: stars: variables: general - stars: evolution

1 Introduction

For more than a century, the photographic plate has been the most reliable light detector and data storage device available for astronomical observations. However, it is being nearly completely replaced by digital detectors like CCDs in practically all astronomical applications, even those requiring wide fields. This implies that most of the older photographic observations are becoming obsolete, unless they document variations on long time scales.

This is the case if one considers the patrol plate archive of the Sonneberg Observatory, which, following Harvard, is the second largest in size and plate number (Bräuer & Fuhrmann 1992). The Sonneberg Observatory was continuously active in sky patrol observations from the 1930ies until today in a very homogeneous manner, using the same optics and very similar plate scales, sizes and emulsions for many decades. Therefore, the Sonneberg archive which contains more than 275 000 plates and films, today represents a unique collection of sky patrol coverage of the entire northern and equatorial sky (down to declination $-33^\circ$), without any major gaps.

There was a considerable impact of this effort on variable star research. Hoffmeister, together with his collaborators and successors, detected, classified and investigated a total of more than 10 000 variable stars in the Milky Way, about 25% of all variables known at the time when the latest edition of the General Catalogue of Variable Stars (GCVS) was published (Kholopov et al. 1985). The technique used for this purpose: blink comparison and eye estimates of magnitudes. Obviously, only a very small fraction of all the information contained in the plate archive could be extracted this way. Now, with the advent of rapid and precise scanners, the entire plate archive can be digitized and subsequently analysed in a more general manner. As a first step in this direction we determine here the photometric variations of 216 arbitrarily chosen fields stars and, in addition, 23 known variables over a period of 34 years. Our study is restricted to rather small areas (about 74 square degrees in total) in the region of Aur/Tau/Ori. The main aim of this patrol project is to determine the photometric accuracy of scanned stellar images on patrol plates, as well as to find out what kind of information on the long-term variability of all classes of stars can be derived and expected from a comprehensive analysis of patrol plates.

2 Plate material, scanning and reduction procedure

The Sonneberg Sky Patrol (SSP), originally proposed by Paul Guthnik (1879-1947), is recording the entire accessible sky with 14 short-focus cameras simultaneously in two colours "pg'' and "pv'' (Bräuer & Fuhrmann 1992; Bräuer et al. 1999). The plate size is  $13 \times 13~{\rm cm}^2$, the scale 830''/mm resulting in a useful field size of about  $26^\circ \times 26^\circ$. The cameras are centred at declinations  $-20^\circ, 0^\circ,
+20^\circ, +40^\circ, +60^\circ$ and $+80^\circ$, recording the sky every $1^{\rm h}$ in right ascension at  $\delta = +40^\circ$ and south of it, and every $2^{\rm h}$ in the $60^\circ$ and $80^\circ$ zones. The limiting magnitudes are of the order of  $14\hbox{$.\!\!^{\rm m}$ }5$ in pg (blue) and  $13\hbox{$.\!\!^{\rm m}$ }5$ in pv (red), but the average limit achieved is about 1 mag brighter than these values.

Plate scanning was performed with the digitalization machine DIA ("Digital Image Analyser'') described by Kroll & Neugebauer (1993). DIA was able to scan an SSP plate with a resolution of  $15~\mu {\rm m}$ and 8 bit data depth within 45 min. More than 5000 plates have been recorded this way between 1994 and 1999, mainly in the Taurus/Orion region for the study of T Tauri stars (Heines 1999). Each plate was subdivided into 4 quarter subsections for the scanning procedure, called 00, 01, 10 and 11 for the NE, NW, SE and SW quarter resp. of each plate.

\end{figure} Figure 1: Approximate locations of the sub-fields on the SSP field 5h +20$^\circ $ (plate center).
Open with DEXTER

For our pilot project we selected a total of 11 smaller sub-fields (about  $2.8^\circ \times 2.8^\circ$) in the quarter sections 00 and 10 of the SSP field  $5^{\rm h} +20^\circ$. The 7 fields in quarter 00 are called: 00.A, 00.B, 00.C, ... 00.G, the remaining 4 in quarter 10: 10.A, 10.B, 10.C and 10.D. These sub-fields were chosen in order to include some known bright variable stars. Their positions on the sky and plate are shown in Fig. 1 in a schematic way. Most of them are rather far away from the plate centre and near the edges and/or corners of the plate. This choice was made to study the photometric accuracy under "worst case conditions''. Our study refers only to the blue pg plates.

In each of these sub-fields 20-30 field stars were selected according to the following criteria:

accurate B and B-V values are available from the HIPPARCOS and TYCHO photometric catalogue (van Leeuwen et al. 1997);
Single star image on the SSP plates, well separated from its neighbour, no blends;
There is no strong background variation present in the surroundings of the scanned star image;
The entire magnitude range of  $7\hbox{$.\!\!^{\rm m}$ }8 \leq B \leq 12\hbox{$.\!\!^{\rm m}$ }2$ and colours from B-V = 0.0-1.8or more is represented in the sample.
For the measuring procedure we used the software package described by Kroll & Neugebauer (1993). The basic algorithm of this method is to perform a seven-parametric, three-dimensional Gaussian fit over the pixel grey values of a single star imprint. It was shown that the logarithm of the volume, $I = \ln V$, of the Gaussian bell is a convenient measure of the magnitude of the star. This software, however, requires the manual selection and measurement of each star image on each of the nearly 500 SSP plates, a task which was performed by ES and NV, eliminating, this way, also contamination by emulsion faults, scratches, spots, as well as meteor, satellite and airplane tracks and other artefacts (see, e.g. Kroll 1999).

For each sub-field, the I-values were linked to the HIPPARCOS and TYCHO catalogue magnitudes, mB and mV, of the constant stars in order to define a transformation between I and the photographic magnitude  $m_{\rm phot}$:

\begin{displaymath}m_{\rm phot} = a_{0} + a_{1} I + a_{2} I^2 + a_{3} (m_{B} - m_{V}).\end{displaymath}

By setting  $m_{\rm phot} = m_{B}$, for each sub-field and each plate, the coefficients, a0 ... a3, were determined with the least-square method. This fit also gives the standard deviation $\sigma $ corresponding to the differences between catalogue magnitudes and reduced magnitudes of all stars in a sub-field. In order to apply the above fit to unknown program stars one has to know their B-V colour. In a few cases in which this colour value was not known we applied a mean value of B-V = 0.6. If the colour is different from this, some zero point shift in  $m_{\rm phot}$ will result. This, however, will not affect the variability discussion given here.

Originally, we had included stars between $6^{\rm m}$ and  $13^{\rm m}$ in our measurement program. In the course of the reduction procedure the coverage of very bright and very faint stars in most sub-fields was too poor to get reliable photometric values. In addition, systematic variations from the above parabolic fit arise as soon as the total magnitude range exceeds about 4.5 mag. Therefore, we limited the final reduction to stars in the range  $7\hbox{$.\!\!^{\rm m}$ }8 \leq m_{B} \leq
12\hbox{$.\!\!^{\rm m}$ }2$.

The colour coefficients a3 were determined, in a first step, for different sets of about 60 plates each taken within a three year interval. However, the coefficient did not vary significantly from epoch to epoch, so we could use a mean colour coefficient for each sub-field. Table 1 lists their values, together with the approximate mean distance of each sub-field from the plate center. Apparently, there is no correlation; however, the most distant field 00.A has the largest value of a3.

The mean distances from the plate center and the colour coefficients for each sub-field are listed in Table 1, together with the mean standard deviations $\sigma $ from the calibration fits.

Table 1: Center distances, colour coefficients and mean scatter $\sigma $ of the sub-fields.

They range from 0.07 to 0.12 mag, and there is no close correlation with the position on the plate. However, the field nearest to the plate center (00.G) also reveals the smallest scatter, as expected. In general, we derived our photometric data with a mean error of 0.096 mag, and we expect that this value could improve to about 0.06-0.08 mag if all stars of an entire plate are measured, since most stars in our sample are located near the plate edges.

3 Constant stars and new variables

The above reduction procedure reveals light curves of a total of 239 stars, with an average of about 450 measurements distributed more or less homogeneously over 34 years (1961-1995). 23 of them are known variables that will be discussed in Sect. 4.

All light curves were analysed in various ways. We calculated seasonal means and searched for drifts, waves and erratic variability in all accessible time scales. In addition, we applied a period search routine developed by Schwarzenberg-Czerny (1989), searching for periodicities between 1 and 10 000 days. A critical comparative analysis of all light curves revealed a total of 17 new variable stars while the remaining 199 stars of our sample have to be considered as constant within the time interval and accuracy considered here. The latter are listed in Tables 2 and 3 (see online material). The scatter around their mean magnitude value is always of the order of 0.08-0.12 mag, as expected from the photometric accuracy determined in Sect. 2. Only these constant stars were used for the magnitude calibration.

The most important properties of the 17 new variable stars are given in Table 4. Eight of them were drifting, showing a linear increase or decrease in brightness over the entire 34-year time interval with a total amplitude between 0.09 and 0.25 mag (- sign refers to increasing, + sign to decreasing brightness with time). This may indicate the presence of periodic or erratic variations at much larger time scales than covered here. Five stars show periodic long-term variations with periods between 3500 and 11 000 days, and amplitudes up to 0.25 mag. In addition, in seven cases erratic variability with similar or shorter time scales (down to about 20 days) is present while in one case (S10953 = GSC 708.0904) the enhanced scatter suggests unresolved short-term variability. The amplitude of the erratic variations ranges from 0.15 to 0.5 mag. A special case is S10955 = GSC 714.0246 for which the period search routine has revealed a strictly periodic variation with 1.58625 days, displaying the typical light curve of an eclipsing binary (Fig. 4), probably of $\beta$ Lyrae type with small amplitude (0.2 mag).

Table 4: New variable stars.

The most interesting light curves of these new variables are shown in Figs. 2 to 9, together with that of a nearby constant star from the same sub-field and similar brightness.

In general, most observed amplitudes of the new variables do not exceed 0.3 mag, the typical threshold for a detection with blink comparator or similar visual inspection methods, as applied traditionally at the Sonneberg Observatory. Therefore, it is not surprising that the variables listed in Table 4 have previously not been detected. However, four of them are listed in the New Catalogue of Suspected Variables (NSV: Kukarkin et al. 1982).

The spectral type o the 17 new variable stars are distributed as follows: B(4 stars), A(8), F(1), G(0), K(1) and M(3). This could be a hint of a bimodal frequency distribution in spectral types of long-term variables with low amplitude, with maxima around types A and M, since the fraction of A type among the constant stars in our sample is only 20%, that of M stars only 1%. The maximum at A stars is surprising and would require more data for confirmation. This preliminary result suffers from the low star numbers.

4 Known variable stars

The 23 previously named variable stars included in our study are listed in Table 5; a sample of the most interesting light curves is given in Figs. 10 to 20.

\end{figure} Figure 2: Light curve of the new variable star S10951 = Hip 26669 ( lower panel) and the constant star Hip 26239.
Open with DEXTER

\end{figure} Figure 3: Light curve of the new variable S10952 = Hip 27699 ( upper panel) and the constant star Hip 27686.
Open with DEXTER

\end{figure} Figure 4: Light curve of the new eclipsing binary S10955 = GSC 714.0246 ( lower panel) and the constant star Hip 25814 vs. phase of the period  $1\hbox{$.\!\!^{\rm d}$ }58625$.
Open with DEXTER

\end{figure} Figure 5: Light curve of the new variable S10956 = GSC 714.1144 ( central panel) and the two constant stars Hip 26729 ( upper panel) and GSC 701.0006 ( lower panel).
Open with DEXTER

Table 5: Known variable stars.

These variables can be subdivided into the following classes (GCVS designations in brackets):

Eclipsing binaries of Algol type (EA): 5 stars;
Mira-type variables (M): 2 stars;
Other red semiregular or irregular giant/supergiant late type variables (LB,LC,SR,SRA): 6 stars;
Orion-type variables (INS,INA): 5 stars;
Be and shell stars (BE,GCAS): 4 stars;
 $\delta$ Scuti variable (DSCT): 1 star.
In 4 of the 5 eclipsing binaries our data confirm or are compatible with the published ephemeris. Only for FP Aur is this not the case: The GCVS gives a period of  $0\hbox{$.\!\!^{\rm d}$ }947236$ and an amplitude of only 0.3 mag. Our data give a mean brightness of  $\overline{B} = 11\hbox{$.\!\!^{\rm m}$ }419$ in HJD 24 37500 to 41700 ( $\sigma = 0.15$ mag) and  $\overline{B} = 11\hbox{$.\!\!^{\rm m}$ }640$ in HJD 24 42400 to 49800 ( $\sigma = 0.28$ mag), but no indication of eclipses. The enhanced scatter may be due to another kind of short-term variability.

For one Mira star, U Aur, our data give exactly the published ephemeris values. In the other case, BK Ori, the GCVS gives a period of  $354\hbox{$.\!\!^{\rm d}$ }2$, valid after HJD 24 38800, and an epoch of maximum HJD 24 40925. At earlier epochs, the period varied between 326 and 346 days. Our data fit best the element maximum = HJD 24 40908 + 346.3 E (see Fig. 13).

As expected, the third group of late type giants and supergiants demonstrate a wide range of behaviour, from unresolved short-term variability (V440 Ori), to erratic variations at all time scales from 20 to 3000 days and amplitudes up to 0.4 mag, as well as possible long-term waves with quasi-periods up to 8000 days. A special case is AB Tau: the GCVS gives an epoch HJD 24 37340 for the light maximum and a period of  $142\hbox{$.\!\!^{\rm d}$ }0$. Our data fit well to the epoch, but the period search routine reveals  $143^{\rm d}$ as the best period. The enhanced scatter (see Fig. 19) implies additional short-term variability.

One of the Orion-type variables (HH Aur) turned out to be constant in our data set. The remaining variables show either erratic variations at time scales of 20 to 1000 days with amplitudes up to 0.8 mag, or waves with quasi-periods of 2400-3500 days and amplitudes up to 0.25 mag.

The group of Be stars is characterized by erratic variations in all time scales between 20 and 6000 days, with amplitudes up to 0.6 mag. The only $\delta$ Scuti star in our sample, V 356 Aur, seems to be constant with a slightly enhanced scatter. Its  $0\hbox{$.\!\!^{\rm d}$ }19$-period (amplitude 0.1 mag) reported in the literature apparently was not resolved by our data.

Due to our period search routine it was possible to discover new, hitherto unknown long-term periodicities in four cases (V399 Aur, HK Ori, V 1374 Ori and V 1409 Ori), with  $2400^{\rm d} \leq P \leq 8000^{\rm d}$ and amplitudes between 0.1 and 0.5 mag. In another four cases (U Aur, BK Ori, SV Tau and AB Tau) we were able to modify and/or improve the published periods. In addition, eight stars (35% of our sample of previously known variables) show significant drift variations in the entire 34 year interval covered, with amplitudes between 0.09 and 0.5 mag. Three of them belong to the group of Orion variables, the remaining ones are more or less evenly distributed among the other variable types. All these findings support the importance of this kind of supplementary information which can be derived by a study like ours even for known variables: they all are either well known, or recently discovered by HIPPARCOS due to striking short-term variations. Their long-term behaviour has never been investigated because there is essentially no way to do this other than via sky patrol plate archives.

\end{figure} Figure 6: Light curve of the new variable S10957 = GSC 1873.0775 (NSV 2444: central panel) and the two constant stars GSC 1873.0519 ( upper panel) and GSC 1873.0489 ( lower panel).
Open with DEXTER

\end{figure} Figure 7: Light curve of the new variable S10959 = GSC 2403.0702 ( upper panel) and the constant star Hip 25512.
Open with DEXTER

5 Discussion

Our sample consisted of a total of 316 field stars and 23 known variables. 17 of the field stars, i.e. 7%, are variable with amplitudes between 0.1 and 0.3 mag, below the threshold of traditional visual searches on photographic sky patrol plates. Richter (1968) estimated that 2% of the bright field stars of about $6^{\rm m}$ show variability exceeding the above threshold in amplitude while this fraction is a factor of 10 lower (0.2%) for  $16^{\rm m}$ stars. He explained this difference by the fact that bright stars are mainly giants which have a much stronger tendency to vary while in the faint star sample relatively stable main sequence stars are dominating. On the other hand, Jackisch (1963) found in a photoelectric study on micro-variability that about 40% of supergiants, 26% of giants and 16% of main sequence stars show variations with amplitudes of more than 0.02 mag. Our fraction of 7% lies between these earlier findings, as is the amplitude range covered by us, which links those of Richter (1968) and Jackisch (1963).
\end{figure} Figure 8: Light curve of the new periodic variable S10962 = GSC 2405.1545 ( lower panel) and the constant star Hip 27404 vs. phase of the period 3580 days.
Open with DEXTER

\end{figure} Figure 9: Light curve of the new variable S10964 = GSC 2408.0661 ( lower panel) and the constant star Hip 26326.
Open with DEXTER

This comparison, however, has to be considered with caution for two reasons. Firstly, our fraction of 7% variables is a lower limit because many "constant'' stars show drifts and/or possible waves with amplitudes between 0.05 and 0.1 mag which, however, failed the test of statistical significance with the presently available data. This problem will be solved as soon as the red plates are included in the analysis. They will provide a simultaneously observed, independent data set in a band pass whose variations should be similar to those on blue plates. Secondly, none of the above cited studies investigated the long-term behaviour. 15 of the 17 new variables show variations with time scales of the order of 1000 days or longer. This makes any comparison with published results difficult.

Similar arguments are valid if we compare our study with those made with other modern techniques. Recently, many new variable stars have been detected and investigated as a by-product of the search for gravitational microlensing effects such as MACHO and OGLE. These surveys record simultaneously millions of stars on large CCD arrays with higher photometric accuracy, lower limiting magnitude and better time resolution than provided by SSP. OGLE has detected a total of about 200 000 variables stars in the Galactic Bulge (Wozniak et al. 2002) and 68 000 variables in the Magellanic Clouds (Zebrun et al. 2001). Thus, the question arises to what extent we should study relatively bright stars on photographic sky patrol plates (with all the problems of calibration, image distortions, blends etc.) if we can get similar information from other ongoing research projects?

\end{figure} Figure 10: Light curve of the Mira type variable U Aur vs. phase of the period  $408\hbox{$.\!\!^{\rm d}$ }09$.
Open with DEXTER

\end{figure} Figure 11: Light curve of the LC type variable V362 Aur ( upper panel) and the constant star GSC 1860.1270.
Open with DEXTER

We would have to wait more than 50-100 years to be able to cover the time scales available in Harvard, Sonneberg and few similar existing plate archives. Archival studies cover mainly stars of the solar neighbourhood, i.e. a rather homogeneous population of stars whose fundamental data such as spectral type, luminosity class, radial velocity, parallax, proper motion, UV and IR spectrum etc. are known or at least will be known very soon due to scheduled survey projects and space missions. In contrast, OGLE and MACHO observe a mixture of stellar populations at far distances, consisting of very faint stars without any hope of easily obtaining the fundamental data mentioned above and required for a meaningful astrophysical discussion of their variation. Both methods, instead, could complement each other. In the near future plates will definitely be replaced by large CCDs, but the policy to observe the entire sky or at least a large fraction of it should be maintained.

The future impact of a complete analysis of all SSP plates is obvious: according to Allen (1973) there are about 29 stars per square degree brighter than  $12\hbox{$.\!\!^{\rm m}$ }0$ in average. SSP has covered the entire sky down to about $-30^\circ$ declination, containing about 28 000 square degrees in total. This means that a total of roughly 800 000 stars in the magnitude range of our pilot study have been monitored in Sonneberg during the past five or more decades. These data will soon be available in digital form: about 30% of the plates have already been scanned with the new HP flat-bed scanners, and we expect to digitize the entire Sonneberg archive within the next five years. If we modify our software so that full-automatic photometric measurements are possible and if we eliminate, in a first attempt, 20% of the targets due to blends, background variability etc. we could study the long-term behaviour of about 650 000 stars just with the methods presented here. 7% of them, i.e. about 45 000 are expected to show long-term variability according to our pilot results; this number is of the same order as that of all known galactic variable stars in the GCVS. This means, on the other hand, that more than 50% of stellar variability is unknown to us because never investigated with proper methods. We will be able to determine, for the first time, the entire frequency distribution of variable stars, including its hitherto unknown tail at low amplitudes and time scales of over 1000 days. Such a study will have important consequences for our understanding of the stellar interior, evolution and variability.

\end{figure} Figure 12: Light curve of the $\gamma $ Cas type variable V438 Aur ( lower panel) and the constant star Hip 27605.
Open with DEXTER

\end{figure} Figure 13: Light curve of the Mira type variable BK Ori vs. phase of the period  $346\hbox{$.\!\!^{\rm d}$ }3$.
Open with DEXTER

\end{figure} Figure 14: Light curve of the INSA type variable HK Ori ( lower panel) and the constant star GSC 709.1948.
Open with DEXTER

\end{figure} Figure 15: Light curve of the INSA type variable HK Ori ( lower panel) and the constant star GSC 709.1948 vs. phase of the period 2400 days.
Open with DEXTER

\end{figure} Figure 16: Light curve of the eclipsing binary OS Ori ( lower panel) and the constant star GSC 701.0006 vs. phase of the period  $2\hbox{$.\!\!^{\rm d}$ }383525$.
Open with DEXTER

\end{figure} Figure 17: Light curve of the Be type variable V1374 Ori ( upper panel) and the constant star GSC 701.0053 vs. phase of the period 5685 days.
Open with DEXTER

\end{figure} Figure 18: Light curve of the eclipsing Algol type binary SV Tau vs. phase of the period  $2\hbox{$.\!\!^{\rm d}$ }1669051$.
Open with DEXTER

\end{figure} Figure 19: Light curve of the SRA type variable AB Tau ( lower panel) and the constant star GSC 1869.1671 vs. phase of the period 143 days.
Open with DEXTER

\end{figure} Figure 20: Light curve of the Be type variable V1163 Tau ( upper panel) and the constant star Hip 26845.
Open with DEXTER

6 Conclusions

We can summarize the main results of our pilot study as follows:

We thank 4pi Systeme - Gesellschaft für Astronomie und Informationstechnologie mbH, Sonneberg, Germany, for financial and logistic support for this project.



Online Material

Table 2: Number of measurements, mean B magnitude and scatter $\sigma $ of the constant HIPPARCOS stars.

Table 3: Number of measurements, mean B magnitude and scatter $\sigma $ of the constant TYCHO stars.

Copyright ESO 2004