blank page
Since we are the only intelligent life we know of, we generally assume, as mentioned in the Introduction, that whatever intelligent species may exist elsewhere also originated on a planet. If the quest for other intelligence is to succeed, the fraction of stars with planetary systems must be reasonably large.
What do we know about this fraction? Extraordinarily little, and the little we know is clouded by controversy. A near neighbor (Barnard's star) was supposed to show small perturbations induced by planetary companions. Refinements of astrometric technique have recently shown these perturbations to be in serious question. No other astrometric perturbation was as large; as a consequence we do not know whether any planets, other than our own neighbors, exist in the Galaxy. Is our solar system unique'?
Is this state of knowledge concerning the frequency of occurrence of planetary systems likely to change in the near future, or must we pursue a search for extraterrestrial intelligence in the absence of basic data? The answer depends on a willingness to invest time, thought, and money in an effort to overcome our ignorance. A first step, an investment of time and thought, is under way. A group of scientists (see Section III-15) under the leadership of Jesse Greenstein, has attempted to define how observations might shed some light on the frequency of low-mass companions to stars. A report to NASA will be based on the two Workshops on Extrasolar Planetary Detection held at Santa Cruz (March 24 25, 1976) and at NASA Ames (May 20-21, 1976), and on a meeting of the astrometric community (U.S. Naval Observatory, May 10-11, 1976). This report will outline technical problems and a coherent program that might produce answers. The time scale for detection is not short, since characteristic indirect methods involve years (planetary revolutions) and direct methods may require an orbiting space telescope. Classical means for detecting low-mass companions to stars involve accurate positional astrometry, useful for nearby stars, or accurate determination of radial velocity changes. Both techniques can be improved with modern technology. Direct detection, at optical or infrared wavelengths, is also subject to orders-of-magnitude technological improvement. A listing of science-related activities associated with detection of other planetary systems is presented in Section III-6.
If a star has a companion, the star will revolve about the center of mass of the system at the angular velocity of its dark companion. The radial component of velocity is observed spectrographically (independent of distance from the observer), the tangential component is observed with respect to an inertial frame, defined by other stars, as a sinusoidal term superposed on the tangential motion of the star (decreasing with distance from the observer).
What level of accuracy is required to detect the radial velocity effect arising from the orbital motion of a star around the barycenter of a star-planet system? Jupiter's motion around the barycenter of the solar system causes a reflex movement of the Sun of approximately 12 m/sec (with a period of 12 years). The effect of the Earth on the Sun amounts to about 0.09 m/sec. Thus, to detect Jupiter-like planets around other similar stars by radial velocity determinations we need accuracy on the order of 10 m/sec, while 1 m/sec would be desirable. For Earth-like masses, 0.01 m/see would be none too high an accuracy if it were possible to achieve.
How far are we from such accuracy? Typical radial velocity measurements are accurate to about 1000 m/see, well above the level required. However, an autocorrelation system can already attain a few hundred meters per second on faint stars. One such system is that employed by Griffin (Cambridge) and Gunn (Palomar). This system is photoelectric and was designed to work on faint stars rather than give high accuracy. It regularly gives standard deviations of 250 m/sec on stars as faint as tenth magnitude and in integration times of 5 min. Griffin has expressed the view that this technique can eventually produce accuracies of about 10m/sec. More complicated wavelength calibrations, impressed on the spectrograph, are planned at Arizona and are yet to be tested.
What are the foreseeable limits in terms of accuracy for radial velocity determinations? The general feeling of the Workshop on planetary detection was that a level of accuracy in the neighborhood of 10 m/see can be obtained if great care is taken. Progress to higher levels of accuracy might be achievable by means of conventional photographic spectroscopy, if certain precautions are met. Such elaborate spectrographs are probably reliable over periods of years, if maintained in a fixed position on a dedicated telescope with an aperture of 1-2 m. Construction time scales for such a radial velocity machine and telescope are short (2-3 years) compared to the detection time (several planetary revolutions, or 2 to 20 years). A fundamental problem is the noise in stellar radial velocities; solar granules have motions comparable to 1000 m/sec, but should average out in integrated light. The Workshop felt that this question should be studied over the next few years, with emphasis on the Sun. We know little about the stability of the solar radial velocity, but magnetographs can study the wave motions and non-radial pulsations, and the work of Hill and others (following Dicke) should determine short-period radial pulsations. The gap in knowledge concerns month-to-year variations in solar diameter, the effect of spottedness, etc. The Workshop felt that a program aimed at determining the stability of the solar velocity is valuable and feasible.
Another, and perhaps better known way to discover dark companions to a given nearby star is that of astrometric observation. In order to set the level of accuracy we require, it should be noted that the displacement of the Sun due to its motion around the Jupiter-Sun barycenter corresponds to about 10-3 arcsec as viewed from a distance of 5 parsecs. The displacement of the Sun due to the Earth is considerably smaller, about 10-6 arcsec, again as viewed from a distance of  5 parsecs. With good seeing the disk of a star is one tenth arcsec or larger. Current photographic techniques are static, but new measuring instruments have increased the speed and accuracy of bisection of such images. Present technology gives astrometric precision of about 0.003 arcsec in normal points for a year's observation. At the May conference of astrometers, this accuracy and some ideas for improving it were discussed. There are 15 stars (intrinsically less massive than the Sun) for which this precision suffices to detect the presence of a planet similar to Jupiter in mass and distance from its parent star (in time scales of a decade). No system allows detection of perturbations as small as those produced by planets of Earth-mass.
Future technology to improve astrometric accuracy may involve arrays, for example, of charge-coupled devices, with dimensional stability, high quantum efficiency, and linearity. Turbulence in the Earth's atmosphere is the major factor limiting ground-based astrometry, but interferometry can increase accuracy to 50 µarcsec (according to Currie), even with a 1-arcsec seeing disk. The Workshop agreed that spaceborne telescopes could produce accuracies of a few micro-arcsec. Difficult technological problems need to be overcome before this can be attained. The electronic sensors need to be tested on ground-based dimensionally stable, special purpose astrometric telescopes in good seeing.
It seems clear that an infusion of technology is required in both radial velocity measurements and astrometric measurements in order to be able to reach the levels of accuracy required to detect planetary companions around nearby stars. However, no fundamental reasons preclude these technological advances, in a relatively short time and at modest cost, before they are used in space.
One aspect of the indirect sensing methods is that one must observe the system under study for at least one orbital period of the planetary companion in order to have sufficient information to say with any confidence that an object of planetary mass exists. Because such a long time is required for positive identifications (at least several years), more direct means of detecting planets are desirable as a guide to SETI. The radial velocity technique can be applied to many more stars, but lacks information on the inclination of the orbital plane; it therefore tends to be of a statistical nature. Both techniques will have numerous important scientific offshoots, for example, data on the number of stars of low mass and very low luminosity.
Detecting planets directly is very difficult. One must rely on either thermal (infrared) emission from the planet, reflected starlight, or line emission characteristic of a planetary atmosphere. In spite of the obvious difficulties associated with each of these three kinds of radiation, direct detection may become feasible by one or more techniques.
What are the problems associated with trying to detect a planet by the light it reflects from its parent star? Consider Jupiter and the Sun. The absolute visual magnitude of Jupiter is about +26, while that of the Sun is about +5. An isolated object of +26 is near the limit of detectability  with any telescope, beyond a few parsecs. This magnitude difference of 21 corresponds to a factor of about 2.5 x 108 in brightness. At 5 parsecs, the separation between Jupiter and the Sun is I arcsec. The practical problem then is to reduce the brightness of the star s diffraction pattern by some nine orders of magnitude within I arcsec of the center of the image. This formidable task has been studied by Oliver. Diffraction theory indicates that if one uses Sonine functions to define an apodization mask, one could reduce brightness by a factor of 1O9 at 1 arcsec from the central image on a one to two meter space telescope of extraordinarily high surface accuracy. Proper image processing will probably allow considerable relaxation in the tolerances needed in the apodized optics.
Detecting thermal radiation from a planet should be relatively easier than optical light. In the Rayleigh-Jeans (long wave) part of the spectrum, surface brightness is linear with temperature and the signal is proportional to area. At a wavelength of 10 µ, Jupiter is only three to four orders of magnitude dimmer than the Sun, a much more favorable situation than in visible light. But detectors at 10 µ are noisier and far less sensitive than their optical counterparts. The Workshop estimated that an isolated object, as bright as Jupiter at a temperature of 300 K, could be detected at 1 parsec using a 1.5-m telescope. The diffraction pattern at 10 µ is larger than the pattern at O.5 µ in the ratio of wavelengths, that is, 20 times larger. Thus. high resolution is required, as well as low noise level detectors. A space system for infrared interterometry appears quite promising and is well worth study.
An indirect way of detecting planetary systems in the process of formation is by means of their far infrared radiation. Protoplanetary material in the form of dust grains has much larger area, for the same mass, than the resulting solid planets after the dust has condensed. Thus, polarization and infrared observations of young stars in the process of formation may reveal the existence of sufficient quantities of solids in orbital motion to suggest at least the future possibility of planetary-sized bodies. Angular resolution is less important than, for example, polarization measurements of thermal (10-30 µ) radiation showing periodic variability of position angle. Another possibility is the detection of the spectroscopic signatures of solid minerals, near young stars, which would be incompatible with formation in or from stars. Protoplanetary disks seen edge-on may be found by their obscuration of the parent star.
Townes announced at the first Workshop that the Berkeley group has discovered nonthermal CO2 features near 10 µm wavelength of unexpected intensity, with possibly a small amount of maser-type gain, in the atmosphere of Venus. If nonthermal emission mechanisms can lead to large maser-type gain, the signal-to-noise problem is greatly eased, because the thermal input to the planet is concentrated into a relatively narrow wavelength band. If this emission is not found in the star, and is not a common feature of interstellar gas clouds, then direct detection of the planet may be possible. Are high-gain molecular amplifiers common in planetary atmospheres? The Workshop felt that a list of potential atmospheric masers or lasers should be studied, and an attempt made to see from space whether the Earth (or Jupiter) is masing in some molecular lines. The airglow planetary emission is far too weak to be detected over interstellar distances. The advantage of a planetary maser or laser is that the planet has orbital velocities of several thousands of meters per second, and its periodic variation is therefore easily detected if the maser line has sufficient energy pumped into it.
Many stars near the Sun are binaries, visually resolved. Others are close enough pairs to be detected by velocity variation. Contraction of interstellar gas clouds down to the dimensions of a star leads to excessive rotation, if the gas has preserved angular momentum. Direct observations show some stars rotating at speeds near breakup. The star may fission into nearly equal stellar masses, the momentum going into orbital motion, or it may leave behind it (as it contracts) a protoplanetary disk, that is, a possible planetary system. Stability of planetary orbits was not considered in detail, but the Astrometric conference thought it clear that if a planet must revolve about a binary star at distances large compared to the separation, it may be too cold. An orbit around one star, small compared to the two stars' separation, may be too hot. A recent radial velocity study of bright F and G stars by Abt and Levy suggested that nearly all stars are members of binary systems. This statement required considerable extrapolation from the actual numbers of stars observed to have variable velocity. Many systems, with periods less than 100 years, clearly were of the fission type, while wide pairs were consistent with independent formation. A rediscussion of the data by Branch suggested that as many as five out of six stars were members of multiple systems. This frequency is also suggested by the scatter in color-luminosity diagrams of clusters. The question of how many single stars (with or without planetary companions) remain is important both for astrophysics and for SETI. Branch's corrections to the findings of Abt and Levy increased this number to 0.15, compared to <0.10 as found by Abt and Levy. In both estimates there is a large gap over which extrapolation is necessary, from the smallest detectable mass in a binary (0.10) to that of Jupiter (0.001).
The most common stars in space are the low mass M dwarfs, which are astrometrically most easily studied. They are subject to flares (bursts of optical, ultraviolet, and probably x-rays and cosmic-rays). Little is known about the incidence and effect of flares, especially among the old M stars. Because they have low mass they are most subject to gravitational perturbations, astrometric or in radial velocity. Much conventional study of M dwarfs can be encouraged with expected large rewards.
With respect to astrometric techniques for detecting planets, a thorough study of the effect of atmospheric seeing on positional determination should be undertaken, and an examination should be made of possible advantages to be gained by way of electronic detection, as compared to use of photographic plates. The 1976 Ames Summer Study on astrometric technique provided important input on these points. It was also felt by the Workshop that astrometric systems in space would have accuracies at least an order of magnitude better than ground-based systems, but that many technical problems had to be overcome before space-borne systems could become a reality. With regard to radial velocity techniques for detecting planets, it was concluded that a determination of the stability of the radial velocity of integrated sunlight would be very valuable, and that the ideal radial velocity instrument needs to be defined. An attempt should be made to  obtain an independent determination of the frequency of binary occurrence, and to examine consequences of a binary system on the stability of planetary orbits. Preliminary bench testing of simple apodizing systems could tell us whether the difficult problem of high mirror surface accuracy can be overcome, for direct means of detecting planets. An effort should be made to determine which planetary molecules might possibly give rise to planetary masers, or other forms of non-thermal emission.
The prospects of increasing our confidence concerning the frequency and distribution of other planetary systems are good, if we are willing to invest the effort. As a consequence of the Workshops, several novel approaches to the problem have come to light, as have potential improvements to classical means of detecting planets.