A Vanishing Star Revisited
VLT Observations of an Unusual Stellar System
20 July 1999
ESO's new Very Large Telescope allows the discovery of a rather peculiar binary system, made up of two dwarf stars orbiting each other. The discovery provides new insights on the theory of stellar evolution.
VLT Observations of an Unusual Stellar System
Reinhold Häfner of the Munich University Observatory (Germany) is a happy astronomer.
In 1988, when he was working at a telescope at the ESO La Silla observatory, he came across a strange star that suddenly vanished off the computer screen. He had to wait for more than a decade to get the full explanation of this unusual event.
On June 10-11, 1999, he observed the same star with the first VLT 8.2-m Unit Telescope (ANTU) and the FORS1 astronomical instrument at Paranal . With the vast power of this new research facility, he was now able to determine the physical properties of a very strange stellar system in which two planet-size stars orbit each other.
One is an exceedingly hot white dwarf star , weighing half as much as the Sun, but only twice as big as the Earth. The other is a much cooler and less massive red dwarf star , one-and-a-half times the size of planet Jupiter.
Once every three hours, the hot star disappears behind the other, as seen from the Earth. For a few minutes, the brightness of the system drops by a factor of more than 250 and it "vanishes" from view in telescopes smaller than the VLT.
A variable star named NN Serpentis
The unsual star in question is designated NN Serpentis , or just NN Ser . As the name indicates, it is located in the constellation of Serpens (The Serpent), about 12° north of the celestial equator. A double letter, here "NN", is used to denote variable stars . It is a rather faint object of magnitude 17, about 25,000 times fainter than what can be perceived with the unaided eye. The distance is about 600 light-years (180 pc).
In July 1988, Reinhold Häfner performed observations of NN Ser (at that time still known by its earlier name PG 1550+131 ) with the Danish 1.54-m telescope at La Silla. He was surprised, but also very pleased to discover that it underwent a very deep eclipse every 187 minutes. Within less than 2 minutes, the brightness dropped by a factor of more than 100 (5 magnitudes). During the next 9 minutes, the star completely disappeared from view - it was too faint to be observed with this telescope. It then again reappeared and the entire event was over after just 11 minutes.
Why eclipses are so important for stellar studies
An eclipse occurs when one of the stars in a binary stellar system moves in front of the other, as seen by the observer. The effect is similar to what happens during a solar eclipse when the Moon moves in front of the Sun. In both cases, the eclipse may be partial or total , depending on whether or not the eclipsed star (or the Sun) is completely hidden from view. The occurence of eclipses in stellar systems, as seen from the Earth, depends on the spatial orientation of the orbital plane and the sizes of the two stars. Two eclipses take place during one orbital revolution, but they may not both be observable.
The physical properties of the two stars in a binary system (e.g., the sizes of the stars, the size and shape of the orbit, the distribution of the light on the surfaces of the stars, their temperatures etc.) can be determined from the measured "light-curve" of the system (a plot of brightness vrs. time). The stars are always too close to each other to be seen as anything but a point of light. The light-curve thus describes the way the total brightness of the two stars changes during one orbital revolution, including the variation of the combined light of the two components as they cover each other during the eclipses.
Already in 1988, it was concluded that the eclipse observed in NN Ser must be caused by a bright and hot star (a white dwarf ) being hidden by another body, most probably a red dwarf star . Because of the dramatic effect, this object soon became known as the "Vanishing Star" , cf. ESO Press Release eso8809 (8 December 1988).
Critical information missing for NN Ser
One particularly critical piece of information is needed for a light-curve study to succeed, that is whether the eclipse is "total" or "partial" . If during the eclipse one star is entirely hidden by the other, we only see the light of the star in front. In that case, the measured amount of light does not change during the phase of totality. The light-curve is "flat" at the bottom of the minimum and the measured brightness indicates the intrinsic luminosity of the eclipsing star. Moreover, for a given orbit, the duration of the totality is proportional to the size of that star.
This crucial information was not available for NN Ser . The brightness at minimum was simply too faint to allow any measurements of the system with available telescopes during this phase. For this reason, the properties of the eclipsing star could only be guessed.
Reaching for the bottom
The new VLT observations have overcome this. Thanks to the powerful combination of the 8.2-m ANTU telescope and the multi-mode FORS1 instrument, it was possible to measure the complete lightcurve of NN Ser , also during the darkest phase of the eclipse.
This extreme observation demanded most careful preparation. Since there is very little light available, the longest possible integration time must be used in order to collect a sufficient number of photons and to achieve a reasonable photometric accuracy. However, the eclipse only lasts a few minutes and it would only be possible to exposure and read-out a few, normal exposures from the CCD camera, not enough to fully characterize the light curve at minimum.
Reinhold Häfner decided to use another method. By having the telescope perform a controlled change of position on the sky ("drift") during the exposure, the light from NN Ser before, during and after the eclipse will not be registered on the same spot of the camera detector, but rather along a line. He carefully chose a direction in which this line would not cross those of other stars in the neighbourhood of NN Ser . This was ensured by rotating FORS1 to a predetermined position angle.
The drift rate was fixed as one pixel (0.20 arcsec) per 3 seconds of time, a compromise between the necessary integration time and desired time resolution that would give the best chance to document the exact shape of the light-curve . In theory, this would then allow the measurement of the intensity along the recorded trail of NN Ser and hence its brightness at any given time during the eclipse.
But how deep would the eclipse be? Would the resulting exposure on each pixel at minimum light be long enough to register a measurable signal?
Seeing the light from the cool star!
As the images show, ANTU and FORS1 did manage this difficult observation! Aided by an excellent seeing of 0.5 arcsec, i.e. a good concentration of the light on each pixel, the recorded signal from NN Ser - although very faint - is well measurable at all times during the eclipse . In the mean, about 70 counts/pixel were registered at the minimum, down from about 18,000 outside the eclipse. The ratio is then about 250, corresponding to just over 6 magnitudes. The measured magnitude during eclipse is 23.0 in the V-band (green-yellow; wavelength 550 nm).
Of even greater importance is the fact that the light-curve is found to be perfectly flat at the bottom, i.e. the eclipse is most certainly total . The white dwarf star is therefore being completely hidden as it moves behind the cooler and larger star, and we see only the latter during the eclipse. As explained above, this then allows to determine many of its properties. For instance, the fact that the light-curve has no obvious "soft shoulders" at the beginning and end of the total phase indicates that the white dwarf abruptly disappears from view. Thus the faint star cannot have a very extended atmosphere, otherwise the brightness change would have been more gradual.
The total phase was found to last 7 m 37 s and each of the partial phases only 1 m 26 s. This shows that the orbit must be nearly perpendicular to the plane of the sky. This angle is referred to as the orbital inclination ; for NN Ser , it must be in the interval between 84° - 90°.
A preliminary analysis indicates that the diameter of the cool star is between 200,000 and 245,000 km, i.e. about 1.5 times that of planet Jupiter. The white dwarf is even smaller; its diameter is between 25,000 and 31,000 km, or about twice the size of the Earth.
The distance between the two stars is 660,000 km, or half the size of the Sun. Thus NN Ser is really a very small system - it would easily fit into our central star!
The surface temperatures are widely different, about 55,000 and 2,800 degrees, respectively.
By adding to this analysis earlier measurements of the orbital velocity of the white dwarf star, it is possible to estimate the mass of the cool star as between 0.10 and 0.14 solar masses. The white dwarf is significantly heavier, about 0.57 solar masses.
Stellar objects with masses below approx. 0.08 solar mass are believed to be brown dwarfs , i.e. "still-born" stars in which nuclear fusion did not ignite. Since the mass of the cool star in NN Ser is near this limit, could it perhaps be such an object?
A spectrum of the cool star
The VLT has already delivered the answer: it turns out to be no . The cool component of NN Ser may be a very small and faint object, but it is a real star that harbours nuclear processes in its interior.
The temperature is on the high side for a brown dwarf, but the definite proof can only be obtained from the spectrum.
ANTU and FORS1 were able to obtain a spectrum of NN Ser during the total eclipse, i.e. at a time when the visual magnitude was 23.0. The exposure had to be limited to 5 min only, in order to ensure that there would be no contamination by extra light from the much brighter white dwarf companion star, as this is the case during the partial phases of the eclipse.
Despite the difficult circumstances, it was possible to record a faint spectrum in the 600 - 900 nm (red - near-IR) wavelength interval. Although it is quite noisy, several molecular bands of TiO (titanium oxide) are well visible; VO (vanadium oxide) bands may also be present.
They allow the classification of the spectrum as that of a very-late-type star, of spectral type M6 or later . This is in reasonable agreement with the mentioned temperature around 2800 degrees. In any case, this spectrum is quite unlike that of a brown dwarf, thus confirming that the cool companion star in NN Ser is a normal hydrogen-burning red dwarf star .
NN Ser: a "missing link" in stellar theory
The binary system NN Ser is now in an evolutionary stage that is referred to as the pre-cataclysmic phase. It will be followed by the cataclysmic phase , during which a gas stream will flow from the larger star to the smaller one. This phenomenon is characterized by frequent and abrupt increase in brightness.
While many stars are known that are now in that unstable phase, only a few stars have ever been found to be in the preceding, transitory phase. Of these, NN Ser is the only one that has such a deep eclipse and for which it has now become possible to determine quite well the properties of the two components.
NN Ser thus represents a most welcome example of a "missing link" in the theory of stellar evolution. It is therefore of great interest to perform further observations of such a rare object. They will include attempts to obtain more spectra to define the spectral type of the cool star very accurately. This will allow a critical check of current theories of atmospheres and evolutionary computations for the smallest and lightest stars.
But for now, Reinhold Häfner looks forward to further nights at Paranal with the ESO astronomers there. "We worked together in a wonderful way during these demanding observations", he said, "and without their great support all of this would have been next to impossible!"
 These observations were carried out during "guaranteed observing time", allocated to the three German institutes that built the FORS instrument. More details about this instrument and related issues are available in eso9814.
 Astronomers designate variable stars according to the constellation in which they are seen in the sky and the order in which they are recognized as having variable brightness. For historical reasons, the first variable star in a given constellation (that is not already known by a greek letter, e.g. "Delta Cephei") is designated as "R" (e.g. "R Coronae Borealis"), the second as "S", etc. until "Z". Then follow "RR", "RS",..."RZ", "SS"..."SZ" until "ZZ" and only then from the beginning of the alphabet, "AA"..."AZ", "BA".. etc. until "QZ".