Spectroastrometry of close visual binaries at the Anton Pannekoek Observatory

(1)

Spectroastrometry of close visual binaries at the Anton

Pannekoek Observatory

By

Peter Bacily

Bachelor Thesis under supervision of Prof. Dr. Huib Henrichs

August 2013

Abstract

The rather recent technique of spectroastrometry, or using spectroscopy to derive spatial infor-mation, was applied to four stars with the LHIRES III spectrograph attached to the 30 cm Schmidt-Cassegrain telescope at the Anton Pannekoek Observatory. The goal was to detect their binarity, despite them having angular separations smaller than the average seeing. The binary stars we ob-served were: Boo, 44 Boo 52 Aql and HR 6803, ranging in magnitude 2.4 to 6.7, and in spectral type from A0V to K0II, with angular separations of 2.9, 1.3, 1.5 and 1.2 arcseconds respectively. With the average seeing at APO being roughly 2 arcseconds, three out of these four stars cannot be observed as binaries using the conventional method of looking trough an eyepiece or using imaging. On the first three, a spectroastrometry signal showed that they are indeed binaries, with a S/N of roughly 9, 32 and 10 respectively. This shows that it is indeed possible to detect binaries with a sub-seeing separation using the instruments available at APO. For the last star, the method failed due to the spectra not being exposed strongly enough. In addition, the effect of autoguiding function was investigated, which appeared to significantly reduces the noise, and hence significantly improves the detection limit.

(2)

1 Introduction

Traditionally the smallest possible spatial scale that could be studied was set by the size of the Airy disk formed by a stellar point source that was being studied. This limits for instance the minimum separation of binary stars that can still be resolved. This limit is determined by either the seeing or the diffraction limit (Bailey 1989). The emergence of spectroastrometry about 15 years ago, however, now allows for retrieval of spatial information on scales smaller than the seeing or diffraction limit. In this work we investigate whether the technique of spectroastrometry could be used to detect binary systems with sub-seeing separations with the equipment at APO.

1.1 Diffraction limit

The theoretical limit for the smallest separation that is observable with a telescope is given by the diffraction limit. Because of the diffraction of light on a circular lens or mirror a point source is not observed as a point like image but as the well-known Airy disk, consisting of concentric rings decreasing outwards in intensity. The size of this disk is related to the wavelength of the light and the diameter of the lens or mirror. When two point sources are close to each other their Airy disks start to overlap. When they get too close together it is impossible to distinguish them. The point when they are just barely distinguishable is called the Rayleigh criterion, which defines the smallest angular resolution of the telescope (Kaper 2009; Whelan & Garcia 2008).

sin(θmin) =

1.22λ

D (1)

in which θmin is the angular resolution, λ is the wavelength of the incoming light and D is the diameter

of the telescope mirror (Kaper 2009). For small angles this simplifies to:

θmin ≈1.22λ_D (2)

For the 30 cm Schmidt-Cassegrain telescope used at APO at 656 nm this gives θmin≈ 0.6 arcsec.

Figure 1: The effects of 2 point sources approaching each other in 1 dimension and in 2 dimensions. (http://www.xenophilia.com/zb/zb0012a.html)

1.2 Seeing

Even though the theoretical angular resolution of a telescope is given by the diffraction limit, in reality most earth-based telescopes are limited by the seeing. When light passes through the earth’s atmosphere it gets slightly diverted. The amount and direction of the diversion are dependent on the atmosphere itself. Because the atmosphere is not stable but constantly changes, the spot where the light ends up changes as well. This typically occurs more than 100 times per second. Since exposure times in astronomy

(4)

are typically much larger than 1/100th of a second (Chromey 2010), the light of a star ”moves” on the CCD. This causes a point source to turn into a blur. The Full Width Half Maximum of this blur is a measure for the seeing. At APO the seeing is typically 2-2.5 arcseconds, with the best ever recorded seeing being 0.9 arcseconds (Summer 2013). This is still much more than the diffraction limit and is therefore the dominating factor for the angular resolution. Some larger telescopes like the VLT use adaptive optics to counter the effects of seeing. However for smaller telescopes this is too expensive and complicated.

1.3 Spectroastrometry

Christy et al. (1983) suggested to use the shift of the photo center between the blue and red band of close binary systems to study their separation and called this technique ”Chromatic Position Difference”. Spectroastrometry as it is used today was first described by Bailey (Bailey 1989). Spectroastrometry is a technique that allows for the retrieval of spatial information on scales smaller than the seeing or the diffraction limit. A vertical cut through a long-slit spectrum of a star would ideally get a cross section of the Airy disk, with its photocenter exactly in the middle. A plot of these photocenters as a function of wavelength is called the spectroastrometry signal. For a single star this will be obviously a straight line since the photo-center is wavelength independent. However when a close binary is observed the combined photocenter will be determined by the relative intensity of the two spectra as a function of wavelength. This is particularly useful when the stars do not have the same spectral type, in which case the spectral line strengths will differ, end the photocenter will move closer to the stronger signal. For example, one star may have a very strong absorption line at Hα while the other one does not, and the spectroastrometric signal will show up as a peak only at the wavelength of the Hα line (Whelan & Garcia 2008). Separations even smaller than a pixel can be detected this way, for instance see Schnerr et al. (2006).

Figure 2: Schematic representation of spectroastrometry on a single star

Figure 3: Schematic representation of the process of performing spectroastrometry on two close stars

1.4 Visual binaries

Most double stars, or binaries, on the sky orbit each other. They have a set of characteristics that can be studied. First of all there are the characteristics of the individual stars, like their luminosity, spectral type, age, temperature, magnetic field, and so on. Binaries though, have some additional characteristics that are linked to their interaction with each other: their rotation period, angular separation and position angle. In this work we are interested in the last two. The angular separation simply means the distance between them on the sky, in angular units, usually measured in arcseconds. The position angle is defined

(5)

as the angle that the brightest star (star A) makes with the other star (star B), as measured relative to the North celestial pole, see Fig. 4.

Figure 4: Definition of position angle; the primary star (the brightest) is at the center. (http://en.wikipedia.org/wiki/Position angle)

2 Method

For the spectroastrometry of close visual binaries, long slit spectra of the binaries were obtained.

2.1 Instrumentation

Table 1: List of instruments used for spectroscopy

Instrument Type Specifications

Telescope Meade LX200 Schmidt-Cassegrain

Diameter D = 30.5 cm

Operating at focal length f = 353 cm Focal ratio f/D = 11.6

Spectrograph Shelyak LHIRES III Long slit spectrograph grating 1200 lines/mm Resolving Power R = 8000

Imaging Camera Atik 460EX 2749×2199 pixels of 0.454 µm square Spatial resolution: 0.265 arcsec/pixel Dispersion: 0.18 ˚A/pixel

Guiding Camera Watec 120N Camera Control MaximDL

(6)

Figure 5: Instrumental setup used for spectroastrometry. For specifications see Table 1

2.2 Target selection

Suitable targets were selected from the Struve Double Star Catalog (Talbot 1998). First all the stars that were not visible during the night were removed, then all the binaries of which the least bright star had a visual magnitude m2 > 8 were removed. After that all the binaries with an angular separation

larger than 2 arcseconds were removed. Lastly a manual search through the remaining stars was done to find the binaries with sufficiently different spectral types. The star Boo was added, which has an angular separation that is slightly larger than the seeing. It was used to get an impression of how angular separation affects the results. Due to time constrains only the brightest stars were used. This gives the following targets:

Name HR m1 m2 ∆m1,2 Separation Spectral types ∆ in sp. type Position angle

Boo 5506 2.4 5.1 2.7 2.900 _{K0II + A2V} _1.8 ₃₄₃◦

44 Boo 5618 5.2 6.1 0.9 1.300 _{F7V + K4V} _1.7 ₆₃◦

52 Aql 7544 6.3 6.8 0.5 1.500 _{A3V + F9III} _1.6 ₁₀₃◦

Her 6803 6.7 7.2 0.5 1.200 _{A0V + G0III} _2.0 ₂₂₆◦

Table 2: List of Targets

2.3 Instrument adjustments and finetuning

2.3.1 Telescope

There were two significant adjustments made to the Schmidt-Cassegrain telescope. The first one was the collimation of the telescope. Collimation is the process of getting the optical axis of the secondary mirror

(7)

coinciding with the optical axis of the system. Without proper collimation the image quality is greatly reduced, in particular because of coma distortion, see Fig. 6. The collimation is done by pointing to a bright star and adjust the three screws that hold the secondary mirror in place such that the rings in the Airy disks are entirely circular symmetric. These are very tiny adjustments, and had to be done at several occasions. This is typical for this type of optical system, as opposed to the RCOS telescope which does not need to be collimated. Without proper collimation spectroastrometry is seriously hampered.

Figure 6: The process of collimation to achieve all optical elements to be aligned along the same optical axis

The second adjustment was changing the focus range of the telescope. The motorized focuser (Van Slyke Instruments) of the telescope has a limited mechanical range. The telescope main mirror can also be used to achieve focus, but this is done by hand and is without any controllable position. Therefore the optimal focus setting of the main mirror was adjusted such that it is approximately in the middle of the mechanical range of the focuser when the spectrograph is in focus. Obviously the position of the main mirror should never be changed anymore, which therefore has been locked and labeled for this purpose. 2.3.2 Tuning the Spectrograph

The entire tuning process described in the LHIRES III user guide was followed. The most noteworthy setting was to make sure that the spectrograph was operating in first order, instead of the second order as it was used before. Furthermore, both the imaging and guiding camera on the spectrograph were replaced. The imaging camera was changed from a SBIG ST-2000 to the more sensitive ATIK 460EX. This camera has the bonus of not producing hardly any dark current at low temperature, resulting in a better S/N. The guiding camera was changed from the color Philips ToYoucam PCVC740K webcam, to the black and white Watec 120N, which is more sensitive and allows for integration of frames. This allows guiding on fainter stars.

2.3.3 Effect on measurements

In the initial setup β Cephei, a 3.2 magnitude star, separated by spectroastrometry by Schnerr et al. (2006), produced roughly 200 counts/pixel at an exposure time of 30 minutes with binning of 8, using a camera that had a pixelwidth of 7.4µm. After all the changes Bo¨otis, a 2.4 magnitude star, gave 65000 counts/pixel after an exposure time of 10 minutes with no binning using a camera that had a pixelwidth of 4.5µm. This means that the final setup was roughly 10000 times as sensitive as the first one.

7.42 4.52 30min 10min 8 1 65000 200 2.512 −0.8 ≈ 10000 (3)

This seems an excessively large difference, one of the causes might be that the intensity difference between the second order diffraction, which was being viewed in the first setup, and the first order diffraction, which was being viewed in the final setup could be very big. Another cause might be that in the first setup the stars were not positioned correctly in the slit. Although this seems unlikely since all the spectra taken with the setup showed roughly the same intensities, meaning the star would have had to have been outside the slit every time.

(8)

2.3.4 Rotator

An obviously essential part of the observations was the possibility to orient the slit of the spectrograph at any desired position angle. To calibrate the reported angle setting of the rotator we observed Albireo (β Cyg), which is a double star with both components clearly visible. We then rotated the spectrograph until the slit was along the position angle. Because the position angle is known we found that with the default settings of the rotator control software, the relation between the angle as set on the rotator and the actual angle it makes on the sky is:

SA = 180◦

− RA (4)

Where SA is the angle which the spectrograph slit makes with the sky and RA is the reported angle as given by the rotator. We therefore adjusted the rotator settings in the software, inverting the direction of rotation and setting the 0 point 180◦ _{further, resulting in}

SA = RA. (5)

This setting obviously depends on in which optical port the spectrograph is used: there is one extra mirror when used in a side port, as how it was used.

2.4 Obtaining the spectra

For each star two spectra were obtained. For the first one, the spectrograph is rotated to the position angle of the stars. That way both stars are positioned in the slit, and both spectra are expected to be present in the data. For the second spectrum the spectrograph is rotated 180 degrees more. This makes it possible to double check results, since the peak in the spectroastrometry signal should reverse when the spectrograph is rotated. The auto-guide function of MaximDL was used for the first set of spectra. Afterwards the same spectra were taken without auto-guide to see its effect. See appendix A for all the settings used for the auto-guide function. Due to time constraints the non-auto-guide images were not taken for every star, but only for Boo and 44 Boo. For every spectrum an exposure time of 600 seconds was used, and the temperature of the CCD was kept at 0◦_{C. All the spectra were taken in the night}

between 15 and 16 July 2013.

2.5 Image Data reduction

The Atik 460EX is a camera that does not produce a significant dark current, when cooled. Therefore the subtraction of a bias was sufficient. This was confirmed by taking a 10 minute dark and comparing it with the bias. All the spectra had a bias subtracted. When doing spectroastrometry, the intensity chance between individual pixels can greatly affect the results. Therefore it is very beneficial to have good flat fields. Unfortunately it appeared very difficult to obtain good flat fields. The flat fields for these spectra were taken a few days after the spectra were taken. They were made by shining a flashlight into the spectrograph. On first sight they appeared to be successful. Especially when we consider that large scale changes are not very important so if the light is not completely ”flat” or evenly distributed it will not damage the results, as long as pixels close to each other received roughly the same amount of light. However after applying the flat field correction it became clear that the flat field did more harm than good, as can be seen in Fig. 7. Therefore it was not used.

The wavelength calibration was done using a neon spectrum.

2.6 Spectroastrometry

The resulting spectra were imported into Mathematica. Every vertical line was fitted with a Gaussian of the form:

a · e−(x−c

b )2+ d (6)

Afterwards the photo-centers of the fits (c in the equation), were plotted with their corresponding wave-length. This is the spectroastrometry signal.

Observations by eye were also made by inspecting a cross section of the spectrum in the spatial direction to see if a double Gaussian was visible. This was done to compare spectroastrometry to the ”classic” method of observing visual binaries.

(9)

Figure 7: Image without (left) and with (right) flat field correction. This correction lowered the image quality, and was therefore not used.

3 Results

3.1 Observations by eye

For binaries with large separations the spectra themselves clearly show that there are two stars present, as is the case with Albireo. We can clearly see both spectra and we can, even by eye, see that they are different, with one having an absorption line at Hα while the other has an emission line. However when we try to do this for close binaries, like Boo, we can no longer distinguish the separate stars.

Figure 8: Long-slit spectra of Albireo with large separation

Figure 9: Long-slit spectra of Boo with small separation

(10)

which corresponds to roughly 11 pixels. The stars differ 2.7 magnitudes, which is a factor 2.52.7≈ 12 in intensity. So in the cross section we expect a second peak, roughly 11 pixels from the photo center, that is 1/12th of the size of the first peak.

Figure 10: Cross-sections of Boo at the expected position angle (left) and +180◦ _{(right). We expect}

any asymmetry at one side of a given spectrum to be present at the other side of the rotated spectrum. We can see that the spectra are slightly raised on opposite sides. This is most likely the second Gaussian we are looking for. It is evident though, that it is not a very clear case. It is very likely that without prior knowledge of this being a double star this observation is not sufficient to proof that it actually is.

3.2 Spectroastrometry results

3.2.1 Boo, 44 Boo and 52 Aql

Because the two stars of the binaries have different spectral types, their hydrogen lines will have different strengths. This means that in our spectroastrometry signal we expect to see a peak or dip at these hydrogen lines. We will look at Hα at 656.3 nm, since this is the strongest hydrogen line and should therefore give the strongest signal.

Figure 11: Spectroastrometry signal in Boo around Hα. Top row: with autoguiding, bottom row: without autoguiding. Figures at the left are taken at the assumed position angle. Figures at the right are taken at the assumed position angle + 180◦_.

(11)

Figure 12: Spectroastrometry signal in 44 Boo around Hα. Top row: with autoguiding, bottom row: without autoguiding. Figures at the left are taken at the assumed position angle. Figures at the right are taken at the assumed position angle + 180◦_.

Figure 13: Autoguided spectroastrometry signal in 52 Aql around Hα. Figure at the left was taken at the assumed position angle. Figure at the right was taken at the assumed position angle + 180◦_.

For Boo, 44 Boo and 52 Aql we can clearly see the peaks and dips we expect to see (see Figs. 11 – 13. We can therefore conclude that these are in fact visual binaries. Two out of the three stars have a sub-seeing separation, which shows that it is possible to get sub-seeing information with the 30 cm telescope at APO. It is also clearly visible that not using the auto-guiding function increases the noise levels.

3.2.2 HR 6803

Spectroastrometry of the HR6803 spectra was not successful (see Fig. reffig:HR6803Ha). If we compare the spectroastrometry signal we get for HR 6803 with the other 3 stars, we first notethe large amount of noise that is roughly 10000 times larger than in the other spectra. From this we can conclude that the fitting of the Gaussians failed. After inspecting the original spectra it was found that the signal was roughly 200 counts on top of the noise, which was roughly 150 counts. This means that the point with the most counts on the vertical lines is very often not the peak of the signal, but at a pixel that had a lot of noise. This completely prohibited the fitting process. It is possible to perform spectroastrometry on HR6803, but a longer exposure time is required.

(12)

Figure 14: Autoguided spectroastrometry signal in HR6803 around Hα. Figure at the left was taken at the assumed position angle. Figure at the right was taken at the assumed position angle + 180◦_.

3.3 Signal to noise

Because the fitting of a Gaussian significantly complicates the error calculations the easiest way to determine the signal to noise ratio is by measuring it directly from the spectroastrometry signal. For this purpose a 2nd degree polynomial was fitted to the spectroastrometry signal. The standard deviation from the fit by the data is taken as the noise level. Unfortunately the deviation caused by the signal is then also included. Therefore we expect the actual S/N to be better than the S/N calculated. A 2nd degree polynomial is chosen because it approximates the data better than a first degree polynomial. This is most likely due to slight imperfections in the optics. The signal is measured by eye, by reading off the number of counts from the graph. The results are summarized in the following table:

Table 3: Signal to noise ratio of spectroastrometry

Name Number Slit at PA Slit at PA+180◦ _Auto-guide _Noise _S/N

Boo 1 X Yes 0.024 6 2 X No 0.032 1.6 3 X Yes 0.018 5 4 X No 0.043 9 44 Boo 1 X Yes 0.048 31 2 X No 0.112 11 3 X Yes 0.055 33 4 X No 0.060 20 52 Aql 1 X Yes 0.107 10 2 X Yes 0.130 6

4 Conclusion

Using the 30cm telescope at APO it was possible to detect binaries of sub-seeing separations (1.3 arc-sec) using the spectroastrometry method. This is quite an achievement for a telescope this size. We can conclude that auto-guiding significantly reduces noise and is therefore highly recommended for spec-troastrometry. It should also be noted that having low counts (only a few times more than the average background) ruins the spectroastrometry signal, so for fainter stars, a longer exposure time is a strong requirement.

5 Discussion

An important question during this project was: ”What is the smallest separation that is still detectable at APO?” Unfortunately this is not an easy question to answer. This is because the S/N of the spec-troastrometry signal depends on a lot of variables: the intensity of the spectrum, the accuracy of the

(13)

guiding system, the difference between the stars’ spectral type, the difference between the stars’ appar-ent magnitude, and their angular separation. Therefore it is nit straightforward to determine what the smallest, still observable angular separation is. However I feel confident saying an angular separation of 0.5 arcseconds is easily obtainable. With an exposure time of 600 seconds a S/N of over 30 was achieved for 44 Boo. The star in question had an angular separation of 1.3 arcsec. If the exposure time is increased to an hour and the 50 cm telescope at APO is used, we obtain a spectrum that has 28 times as many counts. We assume that the S/N of the spectroastrometry signal is proportional to the square root of the amount of counts: S/N ∝ √counts. Doing this would give a S/N of roughly 150 for 44 Boo. It is hard to imagine a relationship between angular separation and the strength of the spectroastrometry signal, where going from 1.3 arcsec to 0.5 arcsec would bring the S/N from 150 all the way down to 2-3. Therefore it is fairly safe to assume that under the right condition binaries with a separation < 0.5 arcsec can be detected at APO.

5.1 Possible new projects

The successful usage of spectroastrometry opens up the possibility to some very interesting projects at APO. The most obvious project would be determining the limit of the smallest separation detectable at APO by using spectroastrometry on targets that are closer together. A possible problem with a project like this is that there might not be any suitable targets. A second possible future project would be to determine the position angle of visual binaries by taking spectra at different angles and finding where the signal is the strongest.

(14)

A

Maxim DL settings for autoguiding

Figure 15: Setting for maxim DL auto-guide

(15)

Figure 17: Setting for maxim DL auto-guide, continued

References

Bailey, J. 1989, Optical Astronomical Instrumentation, 3355, 932

Christy, J. W., Wellnitz, D. D., & Currie, D. G. 1983, Lowell Observatory Bulletin, 9, 28 Chromey, F. R. 2010, To Measure the Sky

Kaper, L. 2009, Astrofysica (Universiteit van Amsterdam)

Schnerr, R. S., Henrichs, H. F., Oudmaijer, R. D., & Telting, J. H. 2006, A&A, 459, L21 Talbot, J. 1998, http://www.skymap.com/struve.htm

Whelan, E. & Garcia, P. 2008, in Lecture Notes in Physics, Berlin Springer Verlag, Vol. 742, Jets from Young Stars II, ed. F. Bacciotti, L. Testi, & E. Whelan, 123

(16)

B

Samenvatting

In dit project is onderzocht of spectroastrometrie uitvoerbaar is op het APO. Spectroastrometrie is de techniek om aan te tonen dat een ster eigenlijk een dubbelster is. Als de twee sterren van een dubbelster te dicht bij elkaar staan zijn ze niet meer van elkaar te onderscheiden. Deze grens van het scheidend vermogen wordt bepaald door twee limieten: de diffractielimiet en de seeing. De diffractielimiet is de theoretische limiet van de telescoop en zal bij grotere telescopen beter zijn. In de praktijk wordt op aarde meestal de limiet bepaald door de seeing. Dit is de onscherpte die wordt veroorzaakt door de breking van licht in een turbulente atmosfeer. Met spectroastrometrie is het mogelijk om dubbelsterren te herkennen terwijl ze dichter bij elkaar staan dan het scheidend vermogen van de telescoop. Het werkt als volgt: sterren zijn puntbronnen, en worden vanwege de begrenzingen van een telescoop afgebeeld als de zogenaamde Airy disk, bestaande uit concentrische ringen die in intensiteit naar buiten afnemen. Dubbelsterren worden afgebeeld als een superpositie van de individuele Airy disks. Als de twee sterren dezelfde intensiteit hebben zal het fotocentrum precies in het midden tussen de twee sterren liggen. Als een van de sterren helderder is, zal het fotocentrum in die richting verschoven liggen. Nu passen we dit toe op dubbelsterren met verschillende spectra. Dat betekent dat de verhouding van hun intensiteit bij verschillende golflengtes anders zal zijn. Als we dus een spectrum opnemen met de dispersierichting loodrecht op de verbindingslijn van de dubbelster, zullen ze bij bepaalde golflengtes, in de door ons gekozen sterren zijn dat de waterstoflijnen, een andere intensteitsverhouding hebben dan bij andere golflengtes, en zal het fotocentrum daarom een fractie van de doorsnede verschoven zijn. Als we dus de positie van het fotocentrum uitzetten tegen de golflengte zullen we bij dubbelsterren met verschillende spectra piekjes in de lijn krijgen. Dit is de techniek van spectroastrometrie. In dit project is spectroastrometrie uitgevoerd op vier bekende dubbelsterren met magnitude 2.4 tot 6.8 en spectraaltype A0V tot K0II met een scheiding uiteenlopend van 2.9 tot 1.2 boogseconden. De seeing op het APO is gemiddeld rond de 2-2.5 boogseconden, dus een aantal van de onderzochte dubbelsterren zijn niet als dubbelster te zien. Met behulp van spectroastrometrie is het voor het eerst gelukt om voor drie van deze vier met de APO instrumentatie het dubbelsterkarakter eenduidig vast te stellen.

Figure 18: Boven: twee sterren met een overlappende Airy disk. Daarnaast zijn hun spectra te zien. Het spectrum van de ene ster heeft een absorptielijn waar de andere een emissielijn heeft. In de praktijk overlappen de spectra. Beneden: de positie van het fotocentrum uitgezet tegen de golflengte. Bij de golflengte waar de onderste ster een absorptielijn heeft is deze zwakker ten opzichte van de bovenste ster, en het fotocentrum verschuift dan naar boven. Bij de golflengte waar de bovenste ster een emissielijn heeft is deze sterker, en het fotocentrum verschuift daar dus ook naar boven. (http://www.hs.uni-hamburg.de/DE/Ins/Per/Lesage/spectro astro.html)

Spectroastrometry of close visual binaries at the Anton Pannekoek Observatory