By Mark Trapnell
The detection of exoplanets – that is, planets orbiting stars other than our Sun – has rarely been out of the news since the first successful detection back in 1992. A few months ago, headlines were full of items about the discovery of an exoplanet orbiting Proxima Centauri, possibly within that star’s “habitable zone” and so with consequences for life on a planet somewhere “near” to us.
I was fortunate to have studied a part-time course in Astronomy at University College London (UCL) over the last two years. One of the course’s modules was on exoplanets and included detailed descriptions of the different methods by which exoplanets are actually detected by astronomers. With one exception, those methods all require sophisticated systems which are way beyond the reach of even the most enthusiastic and well-heeled amateur. The exception is detection through the transit method.
The transit method measures the apparent brightness (or flux) of a star over a period of time, looking out for changes in that brightness. Regular small changes of a particular type and short duration may indicate the presence of a planet passing across the face of the star. Periodic measurement of the star’s brightness should result in a light-curve which shows a pronounced dip during the time of the planet’s transit.
More details on all this can be found easily with a quick Google search. NASA has a good website which includes a section on methods of exoplanet detection at https://exoplanets.nasa.gov.
My UCL course suggested to me that amateurs should be capable of detecting exoplanets using the transit method and I decided to have a go to see how easy or otherwise it might prove to do so.
Finding a previously undetected exoplanet is pretty much beyond an amateur. Even if we assume (as now seems likely) that most stars have orbiting planets, the transit method relies on the planet’s orbit passing at some point between us and the star. Because of all the possible orbits an exoplanet could have, the likelihood of that happening is about 1 in 100 for a large planet and about 1 in 200 for an Earth sized planet. So I am quite happy to try to detect planets that are already known to orbit stars and to be detectable by the transit method!
My online research came up with a short book written by Dennis Conti which is aimed at amateurs and explains the mechanics of exoplanet detection. It is free for download at http://astrodennis.com. I would recommend it to anyone interested. I think that anyone who is broadly competent at astrophotography should also have the ability to detect exoplanets.
The main question for me was what equipment was likely to be necessary for a successful detection. Some research suggested that a few amateurs have successfully detected exoplanets, but had almost invariably used large Schmidt-Cassegrain telescopes of upwards of 12 inches diameter, with 16 inches being normal. I prefer to use refractors. The largest refractor I own is a 6 inch f7 apochromat. I found little to suggest that exoplanets could be detected with such a small telescope. However, I decided to give it a go.
Here is what I think is needed:
- A telescope. Of course. My experiences suggest that a 6 inch refractor will do just fine, though I can see that a larger aperture scope might have some advantages, particularly in choice of target. However, for the amateur, larger aperture normally means greater weight and longer focal length, both of which challenge other parts of the system. Actually, I think a 5 inch refractor would probably have been fine too.
- A mount. In my case it was a German equatorial mount – a Paramount ME. I think it has to be (within reason) a good quality mount which is robustly set up and well calibrated with excellent polar alignment, tracking and pointing accuracy. Shortcomings with any of these will make the process much more difficult.
- A camera. I think pretty much any CCD (not video) astro camera will do. I admit I have no idea whether a DSLR would work. I have my doubts. I also don’t know whether a one shot colour camera would work. Again, I have my doubts. I used an elderly SBIG ST10XME which has a KAF3200 chip in it. Certainly, a large chip isn’t needed and in many ways is a disadvantage as it increases file sizes unnecessarily. The longer download times of older cameras is also a disadvantage as it restricts the amount of data that can be gathered over a finite exoplanet transit time. I stuck with the ST10 because I think it has the most linear CCD chip of my cameras. The key to the camera is that it has to be operated within its linear range (i.e. doubling the amount of photons hitting the detector doubles its output signal rather increasing it by some other random amount). More on this below.
I am very lucky to have a permanent small observatory near Benidoleig. Based on my experiences, I don’t think this is necessary, but it certainly helps.
Finding the target star, controlling the mount and acquiring images all needs software. For this I used Software Bisque’s The SkyX Professional. Other programmes will work equally well, I am sure.
What I found so helpful as to be close to essential (for me at least) was the additional use of a plate-solving programme. For example, if an image I have taken of a star field is loaded into The SkyX, the programme will compare the picture to its star database and when it finds a good match, will place and show that image at the right location in its planetarium programme. Other programmes do the same thing using online data resources.
Plate-solving of this kind is useful for several reasons: (i) finding the right star in the first place. The target stars are not the stars we all know and love and can point to in the sky. They are often unnamed or feature in an obscure catalogue and so can be quite hard to find in practice; (ii) pointing the telescope, once identified and shown in a planetarium programme, at the right star; and (iii) subsequently checking that data has indeed been acquired from the right star.
There are several databases of exoplanet hosting stars and the one I used (http://var2.astro.cz/ETD/) provides a 15 arc-minute square Deep Sky Survey (DSS) image of the target star and its immediate surrounds. I used The SkyX to plate-solve the DSS image so as to show me where in the sky the star is located. In fact The SkyX also then allows accurate movement of the telescope to the exact location of the DSS image. I was then also able to take a test exposure, plate-solve the resulting image and check that it coincided with the DSS image. These kinds of advances in relatively mainstream software now enable amateur astronomers to achieve far more than would have been possible only a few years ago.
Once the data has been acquired, the resulting images need to be calibrated and aligned. There are a number of programmes that will do this. I used CCDStack. See further later.
Finally, the target star’s brightness needs to be measured and the results plotted in a graph. Again, I am sure there are any number of available and suitable programmes. I had already used AstroImageJ during my studies at UCL. It is a popular and powerful photometry programme and is available as freeware at http://www.astro.louisville.edu/software/astroimagej/. It isn’t a particularly easy programme to use. I think it is designed primarily for professional scientists. However, following step by step instructions, after trial and plenty of errors over quite a few hours as I learned how the programme works, resulted in the output of a recognisable light curve.
My first attempt wasn’t too successful. I was in a hurry having arrived at my observatory only a couple of hours before the predicted exoplanet transit. As a result I made a mistake and set an exposure time for the camera that was too long and resulted in some of the camera’s pixels containing the target star image being saturated, something I didn’t discover until I examined the image data the next morning. The moment they are saturated, pixels are no longer linear and so do not react proportionately to an increase or decrease in signal. Indeed, I suspect they stop being linear some time before they saturate.
The following day I tried again on a different (and dimmer) target star. This time I was much more careful with exposure times and worried instead whether I had gathered sufficient signal to overcome background noise. As it turned out I had, though I suspect if I had increased exposure time slightly I might have had a cleaner light curve.
Here, broadly, are the steps I took:
- Identify a suitable target star. The exoplanet database I mentioned above is great. It allows the input of the user’s longitude and latitude and will then show a list of stars with predicted exoplanet transits for any given night together with data about sky coordinates, the predicted beginning and end of the transit and the likely fluctuation in measured brightness of the star during the transit. The key is to select a target star that shows a significant change in brightness. The star I chose (Wasp-52, named after the Wasp programme that first detected the exoplanet orbiting around it) would apparently show in excess of a 2% change when its exoplanet (Wasp-52b) transits. For some reason, the naming convention is that the first detected exoplanet associated with a particular star is given a “b” identifier, not an “a”!
- Check the position of the target star in the sky before and after the transit. Ideally it should be as close to the zenith as possible so as to mitigate atmospheric disturbance to the images. One disadvantage of a GEM is that the telescope and imaging equipment rotate around the telescope’s axis which is aimed at the pole star. It is inclined at the angle of the astronomer’s latitude. They rotate continuously while tracking a star and so soon after passing the meridian, the telescope may hit the ground or the pier that the mount sits on. To avoid this, GEMs have to reverse themselves around the meridian and perform a so-called “meridian flip”. For a GEM user, I assumed that ideally the target star should not cross the meridian during the transit measurement period so as to avoid the need for a meridian flip, which would involve reacquiring and centering the target star after the flip as well as causing other practical complications which at best would result in an interruption of at least several minutes in data gathering run. Wasp-52 crossed the meridian about 45 minutes before the exoplanet transit was due to start. The data run should ideally start an hour before transit and end an hour after transit so as to give a proper light curve. Meridian crossing meant that I would not be able to start until about 45 minutes before the predicted transit, which turned out not to be a problem.
- Focus the camera. Focus isn’t highly critical. Indeed, some techniques utilise deliberate defocusing to assist in smoothing out scintillation effects in the star image. This seems to be particularly useful if the target star only covers a few pixels on the detector. In my case, the star covered at least 10 pixels, so I didn’t bother defocusing and focused normally. It is worth noting that the data gathering run lasts several hours – nearly four hours in my case. I can see that a substantial temperature swing might cause focus problems, but I was lucky and the temperature dropped by only one degree during my data run.
- Plate-solve the DSS Wasp-52 image and move the telescope to the indicated position on the planetarium programme (in my case, The SkyX).
- Start auto-guiding. Auto-guiding takes images of a guide star every few seconds. The software detects any movement of the guide star on the CCD detector and communicates correction instructions to the mount. It is important to keep the star in pretty much the same position on the camera’s CCD chip during the entire transit run. This is because otherwise variations in the brightness of the image field could produce false results. This can be addressed to an extent by dark and flat-field subtraction, but research sources were all clear that it is better to start by eliminating movement on the chip as much as possible. I guided at 5 second intervals and achieved accuracy of about +/- 0.2 pixels through the data run.
- Start taking images. I set my exposure time at 25 seconds for each frame. The data run would last about 3 hours 15 minutes. So allowing for download times, that resulted in about 330 images! Here is one of my images, indicating the position of Wasp-52. Note how the star in the top right has saturated pixels, which give rise to “blooming” in this camera. This is the problem I had the night before with my target star, but in this case it had no effect on Wasp-52 which was not saturated.
- The next morning (the data run finished at about 3 am), calibrate all the images with darks and flats taken the previous evening. A dark is an image taken at the same temperature (CCDs are typically cooled to reduce noise, mine was running at -10 degrees C) and same exposure length as the main image, but with no light falling on the detector. It results in an image of the detector’s inherent noise which can then be subtracted from the actual image. Flats are images taken of an even (flat) light source (for example an illuminated white screen or early evening sky) which model the light variations inside the telescope and imaging equipment. The flat is then divided into the actual images to compensate for those variations. The sources are adamant that dark and flat calibrations need to be done very carefully, so I combined 30 separate darks and 30 separate flats to make respective masters and then used those masters for calibration of the 330 images in CCDStack. My laptop couldn’t cope with calibrating all 330 images at once – I had to batch process 50 images at a time. Even so, its not a huge task with modern programmes and the whole process probably took me no more than about three or four hours. If the camera moved between exposures, it would also be necessary to register (align) all the images so that the next stage, photometry, would work properly. Given my guiding results, I decided not to bother with registration as the 330 images were close to perfectly aligned already.
- Load the calibrated images into AstroImageJ and press “Go”! Well it isn’t quite that simple, but once the programme is properly configured with the coordinates of the target star, it is pretty much like that. The main other input is first to load one of the images into the programme and identify the target star as well as several other comparator stars in the same image by placing a pre-defined measurement “annulus” over each star. The annulus measures the star’s brightness as well as that of the sky background to determine the true brightness of the star itself.
The programme automatically assumes that the first star chosen (T1) is the target star and that the others (C2-6) are comparator stars. The idea is that the brightness of the target star (Wasp-52) is measured for each frame, as is that of each comparator star, which hopefully doesn’t have an orbiting planet affecting its brightness and isn’t a variable star, an orbiting binary or anything else that might cause signal fluctuations. That way, if the target and the comparator both vary at the same time in brightness by proportionate amounts, that variation can be ignored as resulting from external variations such as clouds passing across the target. If in any given image the target varies but the comparator does not, the programme registers the variation as it must have a source inherent to the target star. The literature does discuss the choice of comparator stars and checking beforehand that they aren’t variables or binaries. I just used a pin and assumed that if one of my five comparators wasn’t good, that would show in its data compared to the others and I could eliminate it. Not very scientific, but it seemed to work.
Once “Go” has been pressed, the programme analyses each image identifying variations and starts creating the light curve, dot by dot. It is a great experience to see the dots start appearing, and even better when they begin to head downwards in a curve indicating a successful exoplanet detection and then later head back upwards as the transit ends. Here is the curve I generated from my 330 images:
The top (horizontal) row of blue dots clearly shows the Wasp-52b exoplanet light curve. The two horizontal rows of magenta and orange dots underneath are the plots of two of the comparator stars. I actually used five comparators, but removed the other three from the measurement plot so it isn’t too cluttered. It can be seen from the vertical axis that there was an almost 3% dip in brightness of the target star (from 1.01 down to 0.98 on the relative flux scale) during the transit and that the transit lasted nearly two hours. All exactly as predicted! The radius of the host star is known from its spectral type and so the amount of the dip in the light curve allows calculation of the radius of the planet, which turns out to be about 1.3 x the radius of Jupiter. The period of the orbit is only a few days, so the orbit of Wasp-52b is clearly close to its star, making it a “hot Jupiter”, the easiest type of exoplanet to detect.
The quality of the curve isn’t bad, considering this was a first effort. The right hand side deteriorates in quality compared to the left hand side, but its trend is still very clear. The change in quality affects the target star and the comparators and I guess is due to the stars dropping in altitude during the imaging run and so becoming increasingly affected by atmosphere instability.
I proved to myself that a keen amateur with reasonably normal equipment and some CCD imaging experience can successfully detect an exoplanet from a back yard with a small telescope. It also turned out not to be that difficult, though admittedly I chose as easy a target as I could.
The sense of achievement as the planet’s light curve gradually appeared on my computer screen was immense. This felt like proper science!
The success is of course primarily attributable to the quality of equipment and software now available to amateurs at a reasonable cost and the immediate availability of help and information on the internet.
I am not sure what if anything I will do with that new found ability. In a few months time, I suppose I may have another go and pick a more difficult target. It seems that amateurs do have a role in exoplanet detection by providing further data in respect of planets already discovered where the professional community has moved on to other targets. But for the moment, I am happy to leave the fun to astronomers and their telescopes in Mauna Kea, La Silla or up in space.