The U.C. Berkeley
Infrared Spatial Interferometer Array
[A more technical and comprehensive description of the ISI System has been written by Hale et al., ApJ 537:998-1012, 2000]
I. Introduction: [top]
To the naked eye, and even to most telescopes, stars appear to be unresolved points of light. That means that no matter how much you try to magnify the image of a star, you will not see any detail, just a bigger blob. This is because the stars' enormous distances make their angular sizes smaller than the diffraction limited resolution of the telescopes. The resolution of any telescope is limited by the size of the aperture, as compared to the wavelength of radiation received. For example, the 100-Inch Telescope at Mount Wilson Observatory has an angular resolution in visible light of about 0.05 arcsec. This, by comparison, is about the angular size of Alpha Orionis, a star which is ten times the size of Earth's orbit and has one of the largest apparent diameters. In addition to the resolution problem, fluctuations in the atmosphere make the star "twinkle", and their images are not as small as they might be theoretically.
Therefore, your resolving power, or ability to see detail, depends not on magnification but instead only on the diameter of your telescope. So, in order to be able to resolve the disk of even one of the largest known stars you would need a telescope which is about 10 feet across. This may not seem terribly difficult these days, but that's also not the whole story. You see, looking through the earth's atmosphere causes everything to be smeared out such that no matter how big your telescope is, you still can't see any detail beneath a certain size limit, which is roughly 20 times greater than that needed to resolve even the largest stars. And that is when the atmosphere is behaving at about its best. Sometimes the air can be especially turbulent, smearing out stars even more. Just to give an example, below are two snapshots taken of the star Betelgeuse using just one of the ISI 1.65m, or 65 inch telescopes, on two different nights. (You can click on the image if you'd like to see a larger version.)
The reason why it looks so much larger on the right is due to excessive smearing caused by the turbulent atmosphere, which was particularly bad that night. Astronomers call this "seeing", and Mount Wilson has been reported to have the best seeing conditions of any site in the continental United States. Unfortunately, even when the atmosphere is at its best, such as when the picture at the above left was taken, the 100 inch telescope on its own still can't quite resolve even Betelgeuse.
What do we do about the fact that we can't actually resolve any stars? There is a technique called adaptive optics which looks to be very promising for removing the effects of the turbulent atmosphere. Or we could launch telescopes into space where there is no atmosphere to worry about at all. Yet, using either of these two techniques (adaptive optics and/or space telescopes) we still have the limitation of being able to make telescopes only so big. Well, here is where interferometry comes in. Pioneered for astronomy by Albert A. Michelson, first in 1890 and then more famously in the 1920s, the technique was then rather dormant for nearly 50 years, waiting for technology to catch up.
What Michelson did was the following. He went to the biggest and best telescope at the time, which happened to be the 100 inch telescope right here on Mount Wilson. He then covered up the entire aperture of this large telescope except for two small holes just six inches in diameter. Now if you think of light as a wave, much like waves on a pond, then as light waves from a star pass through these two little holes, when they reach the eyepiece the crests of a wave coming from one hole can add with crests from the other hole to form higher crests, troughs from one can add with troughs from the other to make deeper troughs, or the crest of a wave from one hole can add with the trough of a wave from the other hole to cancel each other out. Or you can have anything in between. At different places across the image plane, the light from the two apertures in the mask will travel different distances to each point, creating areas of alternating bright and dark spots known as "fringes".
Now, this is a simplified sketch which isn't quite exact for a couple of reasons. First, the 100 inch is a reflecting telescope, and I drew it as a refractor. That's just easier to draw. But more importantly, I mentioned earlier that you need a telescope about 10 feet in diameter in order to resolve even the largest stars, and the 100 inch isn't quite big enough across. So here is exactly how Michelson did it. In fact, below here is a copy of a drawing from his own 1921 paper (click on the drawing if you'd like to see a larger version). You see, by placing a steel beam across the end of the telescope (shown at the bottom left of the figure) he could put 6 inch flat mirrors out on the ends of the beam (indicated by M1 and M4), to effectively make the apertures farther apart. Mirrors M2 and M3 then re-directed these beams onto the 100 inch primary.
What he then saw in the eyepiece was a series of bright and dark patterns, known as an interference pattern because we are interfering light from two apertures. The exact kind of interference pattern you get at the end depends on the wavelength of light, the angular size of the star and the separation between the two small apertures. Then, with geometry, if we know two of these quantities, say the wavelength of light and the separation of the two apertures, then we can solve for the other quantity, the size of the star. (Refer back to the previous figure.) So you see, while a 100 inch telescope is limited in resolving power such that we can't image even the largest star, if you cover up about 97% of it, then suddenly you can! It was (and is) a very powerful technique!
But measuring the diameters of just a few of the very largest stars was where stellar interferometry would have to sit for about thirty years, until technology could catch up with the precise demands of the technique.
Francis Pease attempted to push the technique further, by making a 50 foot beam -- but was ultimately unsuccessful because the technology at his time could not meet the demands required to keep the optical path lengths within the interferometer constant to within a fraction of a wavelength of light.
II. The Prototype Instrument [top]
Jumping ahead to the 1970s... Fast electronics, narrow bandwidth detectors, and lasers were all available. Interferometry was somewhat difficult but these technological advancements allowed it to be carried out at visible and radio wavelengths of the spectrum.
However, there was an important gap in between these two wavelength regions that hadn't been studied yet. Thermal radiation in the mid-infrared is a very important wavelength for a couple of reasons. First, it allows us to see relatively cooler material surrounding a star, and being able to observe material lost by a star can tell us a lot about that star's evolution. Furthermore, this wavelength also allows us to peer through dust and see an unobstructed view of stellar photospheres. In visible light a process known as limb darkening can fool us into thinking that a star is much smaller than it really is. Or, at some wavelengths a molecular cloud around a star can fool us into thinking that a star is really bigger than it is. In either case, observing in the mid-infrared can overcome both of these obstacles.
Here we have a very powerful technique (interferometry) and an unexplored and promising wavelength (mid-infrared). And so it was decided to combine them.
A fellow named Charles Townes, along with a few graduate students, Mike Johnson, Al Betz and Ed Sutton, designed and built the world's first mid-IR interferometer using the two auxiliaries of the McMath Solar Telescope at Kitt Peak National Observatory. It was a challenging experiment; it used completely new technology and took nearly 10 years before publication of measurements on stars, but it was a success. They then decided that a larger, dedicated system needed to be built, and after more years required to design and build special telescopes, the ISI was completed in 1988.
III. Early 2-Telescope ISI (c.1988) [top]
Now, the ISI interferometer works along the same basic principle as Michelson's interferometer, except for one thing (and a handful of improvements, of course). That main difference stems from the mid-infrared wavelength which, being invisible to the eye, requires some special techniques to detect. Besides being undetectable by the eye, when the ISI prototype was built, it was still a brand new wavelength region and so even electronic detectors were originally scarce at best.
The other very obvious difference that you can immediately see in this photograph is that instead of using two small sub-apertures of one larger telescope, we have two completely dedicated telescopes, each one acting just as one of the little holes over the 100 inch in Michelson's first experiment. You see, while an interferometer with a baseline of 32 meters as shown here has a greater resolving power than a 32 meter telescope, it just doesn't have all the light collecting area in between, so the bigger we can make each individual element, the more light we can receive, giving us a greater sensitivity to detect the weaker stars. (Of course, never mind the fact that no one has yet built a 32 meter telescope.)
Note that in the photo above, the 100-inch telescope can be seen at the upper right, and the end of the building seen at the upper left is the Pease building, which housed his 50ft interferometer. The top of that building would roll off (towards the camera) to reveal the interferometer. Our on-site workshop is now in the lower part of that building, and certain parts of the ISI were even constructed using material from Pease's 50ft beam. So we are certainly carrying on a tradition here in this interferometry hot-spot.
IV. Telescope Optics [top]
The overall design of each ISI telescope is itself somewhat unique, known as a Pfund type telescope. This type of telescope had been built long before the ISI but was never very common. However, the design has an advantage of being very low to the ground and hence relatively stable against vibrations, and is also very compact and so a decent-sized aperture fits reasonably well into a standard semi-trailer. What you see here, in both the photo above and the schematic below is a 2 meter flat mirror on a steerable altitude-azimuth mount. This mirror, known as a siderostat, is used to redirect light from a star onto the stationary 1.65 meter (65 inches) parabolic mirror. You can just see the edge of the parabolic mirror cell in both the photo and the schematic. The filled parabola then focuses the starlight through a hole in the flat mirror, onto an optical table mounted behind it.
(click the above schematic for a larger version)
Another unique feature is the trailer design. This allows us to be able to move the telescopes to different discrete baselines, giving us a variety of different resolving powers. It is also worth mentioning that the telescope optics are mounted firmly on the ground through holes in the trailer frame. Therefore, during observations the optics are mechanically decoupled from the trailer and rest firmly on concrete pads buried in the ground. But, when it comes time to move the telescopes we can raise the trailer frame on its axles and lift the optics in order to drive them to a new location.
You might ask, why not just set them as far apart as possible and have the best resolving power in the world? Well, a lot can be learned at a variety of different spatial scales. If I wanted to study trees, for example, one thing I might do would be to put a cross section under a microscope. But what about the bark? For that I might look at a piece of it with a hand magnifying lens, and then just look at it with my naked eye. Of course, I wouldn't really have the full picture until I stood back several feet and looked at the whole tree. By the same principle we want to study stars at various resolutions, not always just the highest one.
V. Infrared Heterodyne Detection [top]
How do we manage to detect this infrared light, which is invisible to the eye? Inside each ISI telescope is an infrared laser. Light collected by the telescope comes inside to an optical table, where it is brought together with light from the laser using a beamsplitter. The two beams (laser and star) are then combined with a beamsplitter onto an infrared-sensitive photodiode, which produces an electric current proportional to the light intensity on it. Here is a schematic of the star and laser beam paths...
(click on the image for a larger version)
There is a crucial detail in the mixing of these two beams. You see, the two light waves, those from the laser and from the star, are very close in wavelength, or frequency. To help you understand what happens when the star and laser light beat together, think now of two musicians, standing beside each other, playing a horn. Now one of those horns is just slightly out of tune. In addition to hearing that high pitched concert A that they are trying to play together, you also hear a slowly variable warble. Those two notes they are playing are beating together to produce a more slowly variable tone, or a lower frequency. When we combine light from an infrared laser with infrared light from a star, they also beat together to produce a more slowly variable signal, or a lower frequency. In other words, this process of mixing converts a high frequency light wave down to a lower frequency such that it becomes a radio wave. Now our light signal can be carried around on wires. This method of down-converting the frequency of a signal you want to detect by mixing it with a fixed-frequency "local oscillator" is called heterodyne detection. (The term "local oscillator" refers to the laser in this case.)
The schematic above shows how combining radiation from a star with radiation from a laser produces a beat between the two signals in the radio portion of the spectrum. The bottom panel of this figure is an actual mechanical drawing of one of our optical tables. Click on the thumbnail below to see an actual photo of this same optical table.
VI. Producing Interference [top]
Remember how Michelson produced an interference pattern that he could see with his eye, by overlapping the two light beams from each little aperture? Now, we are going to do the same basic thing here, however from the way I just described how we detect starlight and convert it into a radio signal, we are going to overlap two radio signals instead of light beams. They are exactly equivalent. All of the information about the star that is carried in the infrared light wave is still in the new radio signal. The only thing that is different is the wavelength.
So take another look at Michelson's interferometer below, on the left. The yellow line represents a wavefront of light from a distant star. See how the entire baseline is always exactly perpendicular to the star? That was a big advantage to placing his beam across the top of the 100 inch telescope. Each wavefront from a star hits each of the two little apertures at exactly the same moment, which is very important for interferometry to work. That's also why Pease's 50 foot interferometer did not work very well, because his long beam flexed far too much, causing him to interfere two different wavefronts, which just doesn't work. (Note that when he kept his small mirrors close together, about the same distance as Michelson's Beam, or 20 feet, then it worked, but it didn't provide anything new.)
(click on this image to see a larger version)
Now look at the ISI arrangement in the same figure above, on the right. Only for one very brief instant, when the star is exactly on the meridian is it perpendicular to the baseline. Most of the time during the night the star is on one side or the other. Take the example that I've drawn here, where the star is off to this side. A light wave from a star hits telescope #1 first, and has to travel an extra distance indicated by the blue line before it arrives at telescope #2. In order for the interference method to work properly, we have to interfere the same wavefront at each telescope. So, to do this we must delay in time the signal from telescope #1 until the wavefront reaches the other telescope. Only then can we interfere them together.
VII. Delay Line [top]
How do we delay the signal? Here you will see where one of the advantages of heterodyne detection comes in. Remember we converted the light from the star into a radio signal? To delay the stellar signal in time, all we have to do is run it through a longer piece of wire, so that the signal from Tel1 takes a little bit longer to get to this common combining point than the signal from Tel2 does. And we do this continually throughout the night, as the star's position slowly changes due to earth's rotation, we switch in and out various lengths of cable to make up for it. The device that does this is called a delay line, and in our case is a small box full of various lengths of cables and a series of relays to switch them in and out.
Note that we also gain an advantage just from the Earth's rotation. The ISI employs a technique known as Earth rotation aperture synthesis. This means that for a fixed separation of the telescopes on the ground the rotation of the Earth changes the apparent separation between them, as seen from the vantage point of the object being observed. Therefore, a whole range of angular resolutions can be covered in one single night, from the maximum telescope separation down to about one half their separation.
You may have read about the new CHARA interferometer, currently under construction and also on Mount Wilson. They do the same sort of thing as the ISI except that they do it all optically. They directly combine the actual starlight beams from their telescopes without first converting them into a radio signal like we do. That means that their delay line has to be a series of adjustable mirrors to bounce the beams back and forth in order to delay them in time. So instead of having a small box they have an enormous building the length of a football field, dedicated just to this task.
VIII. Correlator [top]
So now what about the actual interference patterns? We certainly don't look at them with an eyepiece. Instead, once the signals from the telescopes are appropriately delayed, we multiply them electronically using something that is known as a correlator. This produces an electronic interference pattern, analogous to the visual interference pattern that Michelson saw with his eye. We then digitize this electronic interference pattern and analyze it with a computer. A stellar interferometer such as the ISI actually measures the visibility, or contrast between bright and dark interference fringes produced by combining the light from two (or more) apertures which are separated by some distance. The measured fringe visibility at a given baseline can be used to derive the angular sizes of features which could not otherwise be resolved. Since the interference fringe pattern from a star can be very weak, and weak signals can be difficult to detect, we modulate the fringe signal with a known, fixed-frequency modulation.
This spectrum shows a typical interference fringe signal produced with the Infrared Spatial Interferometer. The bright and dark fringes are made to pass across the object at a constant speed of 100 fringes per second, by means of radio frequency technology. This yields an alternating current (AC) signal at the output of the fringe correlator as a function of time.
When this signal is Fourier transformed, one obtains a power spectrum, i.e. the power in the fringe signal as a function of frequency, centered at 100 Hz. The normalized integrated power under the entire curve, between 98 and 102 Hz, is a measure of the visibility or contrast between bright and dark fringes of a particular source for a given baseline.
The wings to the left and right of the central peak are not noise, but part of the signal that is moved slightly off the center, caused by atmospheric turbulence, or in other words by rapid changes in the arrival time of the light waves from the stars to the two telescopes
IX. The Limitations of Two Telescopes [top]
While interferometry is an incredibly powerful technique since it allows us to measure the size of distant stellar objects which would otherwise be impossible, the "high-resolution axis" is only a one-dimensional line that connects the two telescopes. This axis makes a single cut across the object being observed and we measure the visibility along that axis only. Therefore, we are unable to make any determinations about the shape of an object. One would have to continually move the telescopes around so that the interferometer baseline presented itself at different angles to the source. While this is at least possible with the ISI's movable telescopes, it is not practical to move them so much or so quickly. If you take too long to move the telescopes then you can't be sure that the star hasn't changed between measurements.
X. The New 3-Telescope ISI Array (c.2003) [top]
The best answer is to have more telescopes. Simply adding one additional telescope, from two to three, triples the number of baselines from one to three. This is how the ISI looks today...
The ISI presently consists of three telescopes. The above photo shows them all in a line which we used for initial testing purposes, with 4m, 8m, and 12m baselines. However, soon we will move them apart in such a way that they form a triangle and we can measure three baselines at three different angles simultaneously. Three telescopes also provide us with a new observable, something known as the closure phase. This quantity tells us about the symmetry of an object and is only available from a closed triangle of elements, such as three telescopes. Normally, atmospheric fluctuations disturb the fringe phase between any two interferometer elements, but these fluctuations cancel out in a closed triangle.
The small square building at left in the above photo, with the pipes sticking out of the top, houses our Master Local Oscillator (MLO). As described in the heterodyne section above, each telescope uses a laser local oscillator for detecting the starlight. But, in order to ensure that we interfere the same wavefront from each telescope, the lasers inside the telescopes must each have a constant or known phase relationship with one another. We therefore have a master laser which is split three ways and sent to each telescope via the periscopes at the top of the MLO building. This way the telescope lasers can lock on to the master laser and we can maintain a fixed phase relationship between each. The updated schematic of the 3-Telescope ISI system then looks like this:
The other building, the end of which can be seen at the extreme left in the photograph, is the ISI central control room. This is where the entire array is operated and monitored from, including pointing the telescopes and monitoring status of the many critical subsystems. The delay line and correlator are also housed here, where the signals from each of the telescopes are brought together to form interference fringes.