A program in observational astronomy to search the Virgo Cluster of Galaxies for supernovae (SNe) using inexpensive consumer digital camera "backyard astronomy" equipment, and covering some 160 galaxies over ~100 square degrees of sky with 3 pixel positional accuracy (20 arc-seconds, or ~1/60th of a millimeter on the sensor inside the camera).
Motivation | Explanation | Remedy | VISNS v.1 | Results | v.2 | References | Sky Chart | Donate!
[ This Month in the History of Astronomy ]
An investigation conducted in early 2015 of the relevant literature showed that only one-third (14 of 43)* of the SNe ever discovered in our nearest neighboring large cluster of galaxies, the Virgo Cluster, had been found since 1990 -- even though the rate of SNe discovery overall has since increased by something like a factor of 20-30x:
Note the logarithmic vertical scaling, which compresses the size of the
increase that started in the
late-1980s. The increase from 1935 well into the post-WWII era is due to Zwicky's pioneering work.
The graph raises the obvious question: were all of the Virgo Cluster's SNe being discovered several decades ago, and if not are they all being found today?
* - The numbers here strictly apply only to the sample of the ~125 brightest galaxies in the cluster on my observing list in the region around M87/Vir A, which is usually taken to be the center of the main core of the cluster. See sky map and footnote below.
The advent 25-30 years ago of computer controlled telescopes and specialized astronomical digital cameras has caused the number of SNe discovered annually around the world and across the sky to increase from some fifteen or two dozen in the four decades following WWII to many hundreds per year by the 2000's. This classic example of 'technoscience' in action (Sullivan, 2009) has led to a Nobel Prize and a re-modelling of the model of the universe, which now emphasizes 'dark energy' as dominating even 'dark matter' as the principle constituent of the cosmos.
One would naturally expect the number of supernovae discovered in the Virgo Cluster to have scaled up similarly, but in fact it hasn't. The number has stayed at zero to two per year throughout this time, and, for example, while there were 573 SNe found in 2007 none at all were in the Virgo Cluster.
This may actually be easy to understand: even a small (~½-meter) observatory class telescope can image to levels perhaps ~100x fainter than is needed to find SNe in the nearby (astronomically speaking) cluster. This corresponds to a distance ~10x greater, or 500+ million light years (Mly) rather than ~50-60 Mly. There is obviously a correspondingly much greater volume of space to survey at the fainter levels and greater distances, so many more galaxies (and SNe) can be found there. See this demonstration of exactly what I'm talking about.
There are other possible considerations: some of the Virgo galaxies may be so close as to be too big for a modest camera system. Cassegrain telescopes (which are perhaps most common) have long focal lengths and big image scales: a 16" f/16 or 20" f/12 has F > 6000mm. So even with a moderately large CCD sensor (~25mm diagonal) its field of view will only be ~10-12 arc-minutes across. This is comparable to the size of M100. The photo at right is from the National Geographic / Mt. Palomar Sky Survey plate scans, made with F=3450mm; the field is ~8'x12'. M100 turns out to be the Virgo Cluster record holder with 5 SNe over the last 100+ years, so the brighter galaxies in the cluster do get observational attention, but only about one in six from my fields has ever had a SNe discovered in it. With a wide field one can image many galaxies at once, whereas a longer focal length and larger scale means they have to be done one at a time, which is ultimately more expensive in terms of observation time.
Whatever the exact reason, the rate of SNe discovery in the Virgo Cluster has not kept up with recent trends and advances in technology and capabilities. Solid state sensors are intrinsically 30x or more more efficient than the silver based emulsions of the past, so apertures 5-6x times smaller can do approximately as well. The 70mm diameter objective of a telephoto lens, with a digital camera, can perform about equivalently to a 14"-16½ telescope of several decades back shooting (slow) glass plates.
All this leaves an excellent sample of galaxies that are otherwise close enough for fairly detailed investigation but presumably also are not being very thoroughly searched for SNe, though doing so is not an extravagant undertaking.
VISNS is intended to fill this gap and remedy the evident neglect of the Virgo Cluster by current searches.
Much more modest equipment is sufficient for the task of finding SNe in the Virgo Cluster, which turns out to be within the capabilities of inexpensive consumer digital cameras with telephoto lenses.
If a conservative (pessimistic or long) distance to the Virgo Cluster is taken to be m-M=31½ (65 million light years; wikipedia lists m-M=31.1 or 54 Mly), and if SNe range in maximum absolute (intrinsic) brightness from M = -16½ up to M = -19, then their observed apparent magnitudes will peak at m=12½ down to m=15. This is roughly in line with the historical record: only three SNe in the cluster have ever been discovered brighter than m=12½ and but two brighter than m=12.
Experiments with an APS-C sized sensor DSLR -- a circa 2006-7 'pro-sumer' model camera (a Canon EOS 30D, otherwise a backup body I wasn't using) with an F=135mm f/2 telephoto lens (which I didn't use much and was thinking of unloading) -- were able to record, in a single 30 second exposure and without any special treatment, about a dozen cluster galaxies with m=13½ or more, like NGC 4516 (m=13.6):
The DSS section is 20 arc-minutes square and scaled equivalently to the v.1 image.
The Large Magellanic Cloud would have m≅13 at my assumed distance.
Since galaxies are fuzzy they are almost always more difficult to detect than stars of comparable brightness, the latter's light being more concentrated. So these numbers suggest modest equipment like this -- using a photo tripod and a portable equatorial drive -- has a limiting magnitude in the m = 14 to 14¼ range. This is in good accord with expectations based on sensitivity estimations (how these are made), and it is enough to ensure that none of the brighter (Type I) SNe go undetected.
[In the Fall of 2016 the equipment was tested on Pluto, then listed as being fortuitously right at m=14.2, but was unsuccessful. Pluto was 30+° lower in the sky compared to the Virgo Cluster, and this extra atmospheric path length would have cost ~0.2 mags in 'extinction', putting it just below the limiting single frame detection level.]Extra sensitivity can be achieved with the software technique called 'frame stacking': a longer exposure time is synthesized simply by adding multiple 30-second exposures together. This can be better than a single long exposure for several reasons: 1) it avoids sky background saturation (the night sky level is low but not zero); 2) it can remove drive error/drift by re-registering each unguided frame against a common reference, making a less precise drive usable; and 3) as well it permits the removal of the occasional frame badly contaminated by an airplane or satellite track, or other forms of transient backyard light pollution (car headlights, neighbor's porch lights, etc.).
Anyway, right now I'm not using frame stacking, though as you'll see below the capability is there since I'm taking more than one frame of each field.
Another one takes aim at the Pleiades...
After setting up the equipment it's about as easy in practice to take several exposures as it is to take only one, so I settled on eight per field. In theory this can give as much as two more magnitudes of sensitivity. The other way to look at it is in terms of signal-to-noise ratio (S/N), which improves as the square root of the integration time (or almost 3x for eight): as net exposure time increases either a 'noise bump' (marginal detection) turns into an actual object or it melts back into the background noise.
The VISNS v.1 field of view (6 1/3° x 9½°) required three fields to cover the central core of the cluster, with a 50% vertical overlap. After setting up the equipment it's about as easy in practice to do a few fields as it is to do just one. The main, middle field (centered on M87) was then entirely covered twice, which guarantees each galaxy is imaged on two different places on the sensor, and not too close to the edge of a frame in one or the other of the fields.
With VISNS v.1 all the galaxies on Sky & Telescope's Pocket Sky Atlas Close-up Chart C of the Virgo Cluster (Sinnott, 2006; see below) were located in single 30-second exposure. These total 64 galaxies: 47 in the M87 field, plus 8 more in the N offset (M100) field and 9 in the S offset (M49) field.
The Revised Shapley-Ames Catalog of Bright Galaxies (RSACBG) then supplied another 36 galaxies in the fields: 27 in the main catalog and 9 more in its Appendix A (Additional Bright Galaxies).
With these 100 galaxies and their known positions it was then possible to calculate the 'plate constants' (-the term goes back to the silver emulsion on glass plates era), which are the numbers (coefficients) for the equations that transform from pixel (x,y) locations in the frames into celestial coordinates on the sky (longitude/latitude, or Right Ascension and Declination, RA/Dec). As an aside, this is one of the few instances where galaxies have been used as coordinate references this way; usually they are considered too fuzzy and their centers too poorly defined, compared to the simplicity of stars, for such purposes. But at a scale of ~10 arc-seconds per pixel it works, as we'll see. The pixel (x,y) values for each galaxy come from manually placing computer crosshairs over each one in one of the frames, centering it by "optical conjecture".
A computer program then finds the plate constants using standard linear least squares (sometimes called linear regression). The program also has to solve for the amount of rotation needed to square the pixel grid with the graph paper on the sky that the astronomical coordinate system represents. Recall, this is just a camera temporarily set up on a tripod and it's only possible to get things so level without going to too much trouble needlessly, since the program can find and remove whatever rotation there was.
Technical details: How It Works.
With the plate constants in hand one can transform any (x,y) location in the frames into its astronomical coordinates (RA/Dec). By simple scrutiny of the images for the three fields, an additional three dozen objects that looked possibly non-stellar were ID'd as galaxies this way, by checking against catalogs like NASA's Extragalactic Database (NED).
The inverse transformation takes astronomical coordinates (RA/Dec) and converts them into pixel (x,y) locations. Many of the fainter galaxies from the RSACBG were in fact first ID'd this way, as below about magnitude 12 2/3 to 13 the galaxies are dwarfs and can be less easy to distinguish casually from the many foreground stars (in the Milky Way) at that level. This is especially true for the S0 and E galaxies, which are intrinsically more condensed and thus can look star-like at this scale.
The inverse transformation also can be used to calculate each of the ID'd galaxy's expected locations in the frames, based on their cataloged celestial coordinates. When this is done for all the galaxies and compared against their actual, measured locations, the RMS fit error results (RMS = square Root of the Mean Square difference) -- a standard measure of the overall accuracy of the plate constants and the fit. This number is typically right at ~3.0 pixels after excluding the few galaxies that are off by ten or more pixels, which are almost always in the frame corners or near a far edge, where lens distortion starts to become apparent.
~3.0 pixels is an extraordinarily good result. For one thing, a Bayer array sensor like this, where there is a red, green, or blue filter over every pixel (as shown at right), so that the camera's onboard computer chip has to synthesize the other two-thirds of the picture's RGB data (of three color values per pixel) during readout via some sort of interpolation scheme, is only theoretically capable of ~1½ pixel resolution at best.
For another thing, the smallest star images in the best focused frames are some 3-4 pixels in diameter, about as expected for a mass produced lens. Also, recall I'm only estimating the galaxy locations in the frames to ±½ pixel, and as well the precision of the cataloged positions have a comparable uncertainty: the RSACBG gives an RA to the nearest whole second (of time), which is ±½ second, or ±7.5 arc-seconds -- which is ~3/4ths of a pixel.
Each little uncertainty contributes to the total, so a net RMS fit error of ~3.0 pixels is about as good as could be anticipated, especially using such a mathematically simple, linear fit.
[One implication is that the lens used has a very good image over most of
the sensor area, out to 80-85% of the distance from the optical axis to the
frame corners, which includes 90+% of the area. (For a 3:2 aspect ratio, by
comparison, a 'long' or far edge is 83.2% of the distance to a corner.)
Other ways of saying this: the lens has a 'flat', distortion-free, or very
rectilinear image; equivalently, its magnification (scale) doesn't change
over the bulk of the image, to within the lens's ability to image point
Also note that the lens's focal length falls out of the calculations. One of the reasons for doing this was curiosity about how accurate the FL's stamped on lenses are. Ansel Adams stated without any apparent reason that they were good to ~1%, but both my lenses are off by more than that: the F=135mm lens by 1.2% and the F=200mm lens by 2.8%; and both are on the short side. With a 3 pixel RMS and a frame dimension of 3½-4K , these are accurate to better than a part in a thousand, or ~0.10-0.15 mm.]
All this was shaken out of VISNS v1.0 (2015) and v1.1 (2016).
The former consisted of 19 nights of observations from about the first week of February to the 3rd week of April; between weather and bright moonlight this was about two nights per week.
The prime observing season for the cluster, when it is highest up (on the meridian) sometime during astronomical night, runs from the first week in January until a little past the middle of May. This is ~19 weeks. The overall observing season can run a few weeks more on each end, so for v1.1 I was able to go from 12/19/2015 to 6/25/2016: I got in 32 nights of observations over the 27 weeks, or a little more than one per week (5.9 days).
A typical observing run settled in as forty-four 30-second exposures, after a short (1/20th second) exposure on Denebola, to check focus; I usually stop the lens down a 'click' or two from wide open for better image quality (depth of focus):
When everything is operating smoothly and there are no snags it takes an average of about 40 seconds per exposure, or ~40-45 minutes total, including setup and tear-down time. The original goal was to keep it to an hour or less, given frigid mid-winter nights at 4 AM.
Getting a handle on fundamental infrastructure such as the plate constants was a cloudy night goal of VISNS v.1, and with 50+ nights of imaging 125+ galaxies, or 6000+ 'observations' in all, it was then time to proceed to the project's primarily goal of finding supernovae.
How many SNe should be expected?
With VISNS v.1 working out so well, VISNS v.2 is seeking funding primarily for 1) dedicated equipment, including a reasonable upgrade; and 2) general continuing support of the project -- about half for each.
With regard to #1, the observing gear consists of five parts; from the
ground up they are:
a) a sturdy (4x5) tripod with custom machined center post
b) an Ioptron Sky Tracker equatorial drive
c) a Manfrotto geared tripod head (for ~0.1° offset/pointing accuracy)
d) the camera body
e) the camera's lens
Using my own equipment I've noticed that during the Virgo observing season half of the year I tend to want to leave everything assembled and ready to take outside at the next good opportunity. Even though it all breaks down easily for other uses, the threshold for doing so gets raised, inhibiting some of my other uses for the gear.
In particular, getting good/best focus, starting from scratch, can be an annoying, slightly tedious hassle. At one point I experimented with making and using Bahtinov focusing masks (see photo at right).
So once I get decent focus I like to tape it in place and not even touch the lens or the camera until forced to, which can be several weeks. The location of the exact focal plane for modern, fast telephoto lenses is temperature dependent, especially when they are shot nearly wide open; one no longer simply racks the lens focus in one direction to a physical stop at infinity (instead, the lens can focus "beyond infinity" a small amount), auto-focus doesn't work because there's nothing in a dark night sky to focus on, and as well the viewfinder may not be precisely the same distance from the lens as the sensor so visual focusing can be off slightly. (Pro level cameras have an auto-focus micro-adjustment feature, which can be calibrated through tests for each lens used on the camera and then dialed in once somewhere in the menu setting, since the body recognizes which lens is mounted on it.)
All these little difficulties would be obviated by having equipment dedicated soley to the project.
The overall success of VISNS v.1 also justifies upgrading the camera to an APS-H sensor size (now defunct), which is 25% larger in each dimension, or +56% in area. The bigger field of view with a given lens it provides versus an APS-C sensor is not needed, so instead a longer focal length lens is called for, F=200mm (f/2.8) rather than 135mm (f/2). This will cover approximately the same area on the sky while giving a larger image scale: ~6 arc-seconds per pixel rather than ~10.
Initial tests show a noticeable improvement in both image quality and sensitivity. Even though the two lenses have about the same light gathering power -- the slightly larger aperture diameter of the 200mm gives 0.10 more magnitudes -- the larger sensor and its better noise characteristics and onboard computer chip yields objects at least half a magnitude fainter.
The faintest galaxies found with VISNS v.1 were at m=13.7 to 13.9, with one (IC 3267) listed at v=14.2, whereas VISNS v.2 shows several IC (Index Catalog) galaxies in the 14.5 to 14.7 range. This is just enough of an improvement to be able to turn up, in principle at least, even the fainter classes of SNe at the distance of the Virgo Cluster in a single 30-second exposure and without any frame stacking, so this upgrade seems easily justifiable.
With regard to #2, I'd feel like this project was getting good crowd source support if ~$100 per night of observing came in, which is ~$20-30 per day (~$1 per hour); hopefully there'll be a lot of $2 and $5 donations.
As a reward program for keeping VISNS going...
For a $10 donation I'll sent you one of the data frames along with its dark frame (and a few other support files, like the galaxy ID list) if you'd like to play around with the data yourself. You'll need software that can read Canon .CR2 format files.
For $25 I'll sent you one complete set of eight light frames from one of the fields along with its dark frames, which is about a quarter of a night's take.
And for $100 I'll give you a copy of an entire night's worth of data which I haven't had a chance to thoroughly search yet.
Please donate via BitCoin (preferred), plastic card, or PayPal
(look for "Send Tip" button). Yes, your $$$'s really can sent me
out in the cold in the middle of the night.
Your money is no good on other planets (or galaxies) so you may as well spend it now.
Buy Some Astronomy Today!
[ Back to Top ]
The blue lines show the VISNS v.1 frame boundaries, the green lines v.2.
The central v.1 frame shows 47 galaxies here.
The ~100 square degrees covered turns out to be about the same amount of sky as is being surveyed by the Kepler satellite.
Source: Sky & Telescope's Pocket Sky Atlas (Sinnott, 2006).
M87 is 10+° E and 2° S of the star Denebola, which marks the tail of Leo, in an otherwise empty (to the eye) part of sky. Third magnitude ε Virginis (Vindemiatrix) is ~8° further in the same direction.
The overall cluster core is about ~40 minutes wide in RA (right ascension), or ~10°, and is elongated somewhat, slanting diagonally at an ~25° angle from the SSE up to the NNW. My various statistics apply only to the galaxies I've been able to find in my single 30-second exposure frames. That is, I haven't used some master list of cluster members, which may contain galaxies or areas on the sky that I'm not observing.
Galaxies of roughly similar brightness to those in the Virgo Cluster (or brighter) stretch along this general direction for some 50° more all the way up to the cup of the Big Dipper -- and, in the other direction, almost the same distance down to Centaurus. The research of the last several decades has shown that the Milky Way, and the Local Group of Galaxies of which it's a part, is on the outskirts of this Virgo Supercluster of Galaxies, and due to gravity is falling towards it -- but not enough that the expansion of the universe isn't still increasing the distance between us and the cluster to the tune of ~1000 km/sec, or ~2¼ million miles per hour!
The region of sky around the Virgo Cluster is generally devoid of stars brighter than the fifth or sixth magnitude because we're looking up out of the thin plane of the Milky Way. The North Galactic Pole (NGP) is only ~15° away to the N-NNE. So those in the Virgo Cluster would see the Milky Way nearly face on, much as we see M100, which is sometimes said to be one of the galaxies the Milky Way probably most resembles. And, in case you're wondering, the Milky Way's sense of rotation as seen from the Virgo Cluster (or NGP) is clockwise.
About 6° W of the NGP (right, if you're facing S) is the visually faint Coma Berenices star cluster, which is both easy to find in a decently dark sky and is a good test of sky darkness and transparency. (At my site I measure 20½ magnitudes per square arc-second or a little better (20¾ max) on clear, moonless nights.) The Coma Berenices star cluster is almost exactly twice as far away as the much more famous Hyades cluster.
BTW - the close approaching Asteroid 2014 JO25 came right through this region of the sky on the night of April 19-20, 2017, passing right by the NGP and the galaxy M64 -- all just to the E of the Coma Berenices cluster -- on its NNE to SSW path along the E edge of my fields.
BTW #2 - the Big Eclipse in August, 2017, will take place with the sun and moon very near Regulus, at "the other end" of Leo, almost exactly a tenth of the way around the sky (36°) from M87.
Flash and the Pan's observations of the people at the Great Frisbee Eclipse of 1979.
Tilt your vision to the right if you don't see this incredible
Easter Island-like head looking up at the sky in this boulder:
So that the curious and intrepid can see for themselves I can provide
proprietary GPS coordinates to the select few. I don't want crowds
heading into this trailless, pristine, unique place so the price is $400.
(Four people max; scientists, artists, etc. can receive special consideration.)
There is nothing else like it anywhere in Colorado!
(No, it's not even in the otherwise excellent book Weird Colorado.)
Is it natural, as it appeared to me, or is it 3-D Ute rock art? See for yourself.
Buy just the 11"x14" photo (bordered & titled, $34.95 + S/H).
If you have ~3x more $$'s and want a stunning large picture of the Milky Way, click here.
If you happen to be in town you can see an example of my other super-telephoto work, framed and on the wall, upstairs at the Coffee Trader. (I recommend their signature "Mountain Splendor" blend btw.)
Or maybe you'd be interested in some of my Large Format 4x5 Film Photography. (currently on hold until after the eclipse)
Purchasing a print keeps me producing works like these Large Format Photography Beginner's Notes.
Back to Top of Page