Any planet is an extremely faint light source compared to its parent star. In addition to the intrinsic difficulty of detecting such a faint light source, the light from the parent star causes a glare that washes it out. For those reasons, only a very few extrasolar planets have been observed directly.
Instead, astronomers have generally had to resort to indirect methods to detect extrasolar planets. At the present time, several different indirect methods have yielded success.
This method consists of precisely measuring a star's position in the sky and observing how that position changes over time. Originally this was done visually with hand-written records. By the end of the 19th century this method used photographic plates, greatly improving the accuracy of the measurements as well as creating a data archive. If the star has a planet, then the gravitational influence of the planet will cause the star itself to move in a tiny circular or elliptical orbit. Effectively, star and planet each orbit around their mutual center of mass (barycenter), as explained by solutions to the two-body problem. Since the star is much more massive, its orbit will be much smaller. Frequently, the mutual center of mass will lie within the radius of the larger body.
Astrometry is the oldest search method for extrasolar planets and originally popular because of its success in characterizing astrometric binary star systems. It dates back at least to statements made by William Herschel in the late 18th century. He claimed that an unseen companion was affecting the position of the star he cataloged as 70 Ophiuchi. The first known formal astrometric calculation for an extrasolar planet was made by W. S. Jacob in 1855 for this star. Similar calculations were repeated by others for another half-century until finally refuted in the early 20th century. For two centuries claims circulated of the discovery of unseen companions in orbit around nearby star systems that all were reportedly found using this method, culminating in the prominent 1996 announcement of multiple planets orbiting the nearby star Lalande 21185 by George Gatewood. None of these claims survived scrutiny by other astronomers, and the technique fell into disrepute. Unfortunately, the changes in stellar position are so small and atmospheric and systematic distortions so large that even the best ground-based telescopes cannot produce precise enough measurements. All claims of a planetary companion of less than 0.1 solar mass (a planet) made before 1996 using this method are likely spurious. In 2002, however, the Hubble Space Telescope did succeed in using astrometry to characterize a previously discovered planet around the star Gliese 876.
Future space-based observatories such as NASA's Space Interferometry Mission may succeed in uncovering large numbers of new planets via astrometry, but for the time being it remains a minor method of planetary detection.
One potential advantage of the astrometric method is that it is most sensitive to planets with large orbits. This makes it complementary to other methods that are most sensitive to planets with small orbits. However, very long observation times will be required — years, and possibly decades, as planets far enough from their star to allow detection via astrometry also take a long time to complete an orbit.
In 2009 the discovery of VB 10b by astrometry was announced. This planetary object was reported to have a mass 7 times that of Jupiter and orbiting the nearby low mass red dwarf star VB 10. If confirmed, this will be the first exoplanet discovered by astrometry of the many that have been claimed through the years, however recent radial velocity results have failed to detect any evidence of a large-mass companion orbiting VB10.
Like the astrometric method, the radial-velocity method uses the fact that a star with a planet will move in its own small orbit in response to the planet's gravity. The goal now is to measure variations in the speed with which the star moves toward or away from Earth. In other words, the variations are in the radial velocity of the star with respect to Earth. The radial velocity can be deduced from the displacement in the parent star's spectral lines due to the Doppler effect.
The velocity of the star around the center of mass is much smaller than that of the planet because the radius of its orbit around the center of mass is so small. Velocity variations down to 1 m/s can be detected with modern spectrometers, such as the HARPS (High Accuracy Radial Velocity Planet Searcher) spectrometer at the ESO 3.6 meter telescope in La Silla Observatory, Chile, or the HIRES spectrometer at the Keck telescopes. An especially simple and inexpensive method for measuring radial velocity is "externally dispersed interferometry".
This has been by far the most productive technique used by planet hunters. It is also known as Doppler spectroscopy. The method is distance independent, but requires high signal-to-noise ratios to achieve high precision, and so is generally only used for relatively nearby stars out to about 160 light-years from Earth. It easily finds massive planets that are close to stars, but detection of those orbiting at great distances requires many years of observation. Planets with orbits highly inclined to the line of sight from Earth produce smaller wobbles, and are thus more difficult to detect. One of the main disadvantages of the radial-velocity method is that it can only estimate a planet's minimum mass. Usually the true mass will be within 20% of this minimum value, but if the planet's orbit is almost perpendicular to the line of sight, then the true mass will be much higher.
The radial-velocity method can be used to confirm findings made by using the transit method. When both methods are used in combination, then the planet's true mass can be estimated.
A pulsar is a neutron star: the small, ultradense remnant of a star that has exploded as a supernova. Pulsars emit radio waves extremely regularly as they rotate. Because the intrinsic rotation of a pulsar is so regular, slight anomalies in the timing of its observed radio pulses can be used to track the pulsar's motion. Like an ordinary star, a pulsar will move in its own small orbit if it has a planet. Calculations based on pulse-timing observations can then reveal the parameters of that orbit.
This method was not originally designed for the detection of planets, but is so sensitive that it is capable of detecting planets far smaller than any other method can, down to less than a tenth the mass of Earth. It is also capable of detecting mutual gravitational perturbations between the various members of a planetary system, thereby revealing further information about those planets and their orbital parameters.
The main drawback of the pulsar-timing method is that pulsars are relatively rare, so it is unlikely that a large number of planets will be found this way. Also, life as we know it could not survive on planets orbiting pulsars since high-energy radiation there is extremely intense.
In 1992 Aleksander Wolszczan and Dale Frail used this method to discover planets around the pulsar PSR 1257+12. Their discovery was quickly confirmed, making it the first confirmation of planets outside our Solar System.
While the above methods provide information about a planet's mass, this photometric method can determine the radius of a planet. If a planet crosses (transits) in front of its parent star's disk, then the observed visual brightness of the star drops a small amount. The amount the star dims depends on the relative sizes of the star and the planet. For example, in the case of HD 209458, the star dims 1.7%.
This method has two major disadvantages. First of all, planetary transits are only observable for planets whose orbits happen to be perfectly aligned from astronomers' vantage point. The probability of a planetary orbital plane being directly on the line-of-sight to a star is the ratio of the diameter of the star to the diameter of the orbit. About 10% of planets with small orbits have such alignment, and the fraction decreases for planets with larger orbits. For a planet orbiting a sun-sized star at 1 AU, the probability of a random alignment producing a transit is 0.47% However, by scanning large areas of the sky containing thousands or even hundreds of thousands of stars at once, transit surveys can in principle find extrasolar planets at a rate that could potentially exceed that of the radial-velocity method , although it would not answer the question of whether any particular star is host to planets.
Secondly, the method suffers from a high rate of false detections. A transit detection requires additional confirmation, typically from the radial-velocity method.
The main advantage of the transit method is that the size of the planet can be determined from the lightcurve. When combined with the radial velocity method (which determines the planet's mass) one can determine the density of the planet, and hence learn something about the planet's physical structure. The nine planets that have been studied by both methods are by far the best-characterized of all known exoplanets.[17 ]
The transit method also makes it possible to study the atmosphere of the transiting planet. When the planet transits the star, light from the star passes through the upper atmosphere of the planet. By studying the high-resolution stellar spectrum carefully, one can detect elements present in the planet's atmosphere. A planetary atmosphere (and planet for that matter) could also be detected by measuring the polarisation of the starlight as it passed through or is reflected off the planet's atmosphere.
Additionally, the secondary eclipse (when the planet is blocked by its star) allows direct measurement of the planet's radiation. If the star's photometric intensity during the secondary eclipse is subtracted from its intensity before or after, only the signal caused by the planet remains. It is then possible to measure the planet's temperature and even to detect possible signs of cloud formations on it. In March 2005, two groups of scientists carried out measurements using this technique with the Spitzer Space Telescope. The two teams, from the Harvard-Smithsonian Center for Astrophysics, led by David Charbonneau, and the Goddard Space Flight Center, led by L. D. Deming, studied the planets TrES-1 and HD 209458b respectively. The measurements revealed the planets' temperatures: 1,060 K (790°C) for TrES-1 and about 1,130 K (860°C) for HD 209458b.  In addition the hot Neptune Gliese 436 b enters secondary eclipse. However some transiting planets orbit such that they do not enter secondary eclipse relative to Earth; HD 17156 b is over 90% likely to be one of the latter.
A French Space Agency mission, COROT, began in 2006 to search for planetary transits from orbit, where the absence of atmospheric scintillation allows improved accuracy. This mission was designed to be able to detect planets "a few times to several times larger than Earth" and is currently performing "better than expected", with two exoplanet discoveries (both "hot jupiter" type) as of early 2008.
In March 2009, NASA mission Kepler was launched to scan a large number of stars in the constellation Cygnus with a measurement precision expected to detect and characterize Earth-sized planets. The NASA Kepler Mission uses the transit method to scan a hundred thousand stars in the constellation Cygnus for planets. Kepler will be sensitive enough to detect planets even smaller than Earth. By scanning a hundred thousand stars simultaneously, it will not only be able to detect Earth-sized planets, it will be able to collect statistics on the numbers of such planets around sunlike stars.
Kepler has already been able to detect the light from a known transiting extrasolar gas giant, HAT-P-7b. It is expected that Kepler will even be able to detect light from non-transiting gas giants on close orbits, though it will not be able to resolve that light into an image. Instead, the brightness of the host star seems to change gradually over time in a periodic manner, because like the Moon, the planet goes through phases from full to new and back again. The variation, although small, will be the signature of a planet. In addition to the reflected light from the star, some of the light from the planet will be thermally emitted by the planet itself. Thus the shape of the phase curve constrains the composition of the atmosphere based on the reflectivity of the planet, and also gives indications of the transport of heat on the planet from the day side to the night. This planet phase variation method may actually provide the greatest number of planets to be discovered by the Kepler satellite, since it does not require the planet to pass in front of the disk of the star.
Gravitational microlensing occurs when the gravitational field of a star acts like a lens, magnifying the light of a distant background star. This effect occurs only when the two stars are almost exactly aligned. Lensing events are brief, lasting for weeks or days, as the two stars and Earth are all moving relative to each other. More than a thousand such events have been observed over the past ten years.
If the foreground lensing star has a planet, then that planet's own gravitational field can make a detectable contribution to the lensing effect. Since that requires a highly improbable alignment, a very large number of distant stars must be continuously monitored in order to detect planetary microlensing contributions at a reasonable rate. This method is most fruitful for planets between Earth and the center of the galaxy, as the galactic center provides a large number of background stars.
In 1991, astronomers Shude Mao and Bohdan Paczyński of Princeton University first proposed using gravitational microlensing to look for exoplanets. Successes with the method date back to 2002, when a group of Polish astronomers (Andrzej Udalski, Marcin Kubiak and Michał Szymański from Warsaw, and Bohdan Paczyński) during project OGLE (the Optical Gravitational Lensing Experiment) developed a workable technique. During one month they found several possible planets, though limitations in the observations prevented clear confirmation. Since then, four confirmed extrasolar planets have been detected using microlensing. As of 2006 this was the only method capable of detecting planets of Earthlike mass around ordinary main-sequence stars.
A notable disadvantage of the method is that the lensing cannot be repeated because the chance alignment never occurs again. Also, the detected planets will tend to be several kiloparsecs away, so follow-up observations with other methods are usually impossible. However, if enough background stars can be observed with enough accuracy then the method should eventually reveal how common earth-like planets are in the galaxy.
Observations are usually performed using networks of robotic telescopes. In addition to the NASA/National Science Foundation-funded OGLE, the Microlensing Observations in Astrophysics (MOA) group is working to perfect this approach.
The PLANET (Probing Lensing Anomalies NETwork)/RoboNet project is even more ambitious. It allows nearly continuous round-the-clock coverage by a world-spanning telescope network, providing the opportunity to pick up microlensing contributions from planets with masses as low as Earth. This strategy was successful in detecting the first low-mass planet on a wide orbit, designated OGLE-2005-BLG-390Lb.
Disks of space dust (debris disks) surround many stars. The dust can be detected because it absorbs ordinary starlight and re-emits it as infrared radiation. Even if the dust particles have a total mass well less than that of Earth, they can still have a large enough total surface area that they outshine their parent star in infrared wavelengths.
The Hubble Space Telescope is capable of observing dust disks with its NICMOS (Near Infrared Camera and Multi-Object Spectrometer) instrument. Even better images have now been taken by its sister instrument, the Spitzer Space Telescope, which can see far deeper into infrared wavelengths than the Hubble can. Dust disks have now been found around more than 15% of nearby sunlike stars.
The dust is believed to be generated by collisions among comets and asteroids. Radiation pressure from the star will push the dust particles away into interstellar space over a relatively short timescale. Therefore, the detection of dust indicates continual replenishment by new collisions, and provides strong indirect evidence of the presence of small bodies like comets and asteroids that orbit the parent star. For example, the dust disk around the star tau Ceti indicates that that star has a population of objects analogous to our own Solar System's Kuiper Belt, but at least ten times thicker.
More speculatively, features in dust disks sometimes suggest the presence of full-sized planets. Some disks have a central cavity, meaning that they are really ring-shaped. The central cavity may be caused by a planet "clearing out" the dust inside its orbit. Other disks contain clumps that may be caused by the gravitational influence of a planet. Both these kinds of features are present in the dust disk around epsilon Eridani, hinting at the presence of a planet with an orbital radius of around 40 AU (in addition to the inner planet detected through the radial-velocity method).. These kinds of planet-disk interactions can be modeled numerically using collisional grooming techniques.
Recent spectral analysis of white dwarfs' atmospheres by Spitzer Space Telescope found contamination of heavier elements like magnesium and calcium. These elements cannot originate from the stars' core and it is probable that the contamination comes from asteroids that got too close (within the Roche limit) to these stars by gravitational interaction with larger planets and were torn apart by star's tidal forces. Spitzer data suggests that 1-3% of the white dwarfs has similar contamination.
As mentioned previously, planets are extremely faint light sources compared to stars and what little light comes from them tends to be lost in the glare from their parent star. So in general, it is very difficult to detect them directly. In certain cases, however, current telescopes may be capable of directly imaging planets. Projects to equip the current generation of telescopes with new, planet-imaging-capable instruments are underway at the Gemini telescope (GPI), the VLT (SPHERE), and the Subaru telescope (HiCiao). Specifically, this may be possible when the planet is especially large (considerably larger than Jupiter), widely separated from its parent star, and young (so that it is hot and emits intense infrared radiation).
In July 2004, a group of astronomers used the European Southern Observatory's Very Large Telescope array in Chile to produce an image of 2M1207b, a companion to the brown dwarf 2M1207. In December 2005, the planetary status of the companion was confirmed. The planet is believed to be several times more massive than Jupiter and to have an orbital radius greater than 40 AU.
The first multiplanet system, announced on 13 November 2008, was imaged in 2007 using telescopes at both Keck Observatory and Gemini Observatory. Three planets were directly observed orbiting HR 8799, whose masses are approximately 10, 10 and 7 times that of Jupiter. On the same day, 13 November 2008, it was announced that the Hubble Space Telescope directly observed an exoplanet orbiting Fomalhaut with mass no more than 3MJ. Both systems are surrounded by disks not unlike the Kuiper belt. An additional system, GJ 758, was imaged in November of 2009, by a team using the HiCIAO instrument of the Subaru Telescope.
Three other possible exoplanets have now been directly imaged: GQ Lupi b, AB Pictoris b, and SCR 1845 b. As of March 2006 none have been confirmed as planets; instead, they might themselves be small brown dwarfs.
When a double star system is aligned such that the stars pass in front of each other in their orbits, the system is called an "eclipsing binary" star system. The time of minimum light, when the star with the brighter surface area is at least partially obscured by the disc of the other star, is called the primary eclipse, and approximately half an orbit later, the secondary eclipse occurs when the brighter surface area star obscures some portion of the other star. These times of minimum light, or central eclipse, constitute a time stamp on the system, much like the pulses from a pulsar (except that rather than a flash, they are a dip in the brightness). If there is a planet in circum-binary orbit around the binary stars, the stars will be offset around a binary-planet center of mass. As the stars in the binary are displaced by the planet back and forth, the times of the eclipse minima will vary; they will be too late, on time, too early, on time, too late, etc.. The periodicity of this offset may be the most reliable way to detect extrasolar planets around close binary systems..
Short period giant planets in close orbits around their stars will undergo reflected light variations changes because, like the Moon, they will go through phases from full to new and back again. Although the effect is small — the photometric precision required is about the same as to detect an Earth-sized planet in transit across a solar-type star — such Jupiter-sized planets should be detectable by space telescopes such as the Kepler Space Observatory. This method may actually constitute the most planets that will be discovered by that mission because the reflected light variation with orbital phase is largely independent of orbital inclination of the planet's orbit. In addition, the phase function of the giant planet may be constrained which will, in turn, lead to constraints on the actual particle size distribution of the atmospheric particles.
Light given off by a star is un-polarized, i.e. the direction of oscillation of the light wave is random. However, when the light is reflected off the atmosphere of a planet, the light waves interact with the molecules in the atmosphere and they are polarized.
By analyzing the polarization in the combined light of the planet and star (about one part in a million), these measurements can in principle be made with very high sensitivity, as polarimetry is not limited by the stability of the Earth's atmosphere.
Astronomical devices used for polarimetry, called polarimeters, are capable of detecting the polarized light and rejecting the unpolarized beams (starlight). Groups such as ZIMPOL/CHEOPS and PLANETPOL are currently using polarimeters to search for extra-solar planets, though no planets have yet been detected using this method.
Several space missions are planned that will employ already proven planet-detection methods. Astronomical measurements done from space can be more sensitive than measurements done from the ground, since the distorting effect of the Earth's atmosphere is removed, and the instruments can view in infrared wavelengths that do not penetrate the atmosphere. Some of these space probes should be capable of detecting planets similar to our own Earth.
(On February 2, 2006 NASA announced an indefinite suspension of work on the Terrestrial Planet Finder due to budget problems. Then in June 2006, the Appropriations Committee of the U.S. House of Representatives partially restored funding, permitting development work on the project to continue at least through 2007. COROT was launched on December 27, 2006 and Kepler's launch was performed on March 7, 2009.)
NASA's Space Interferometry Mission, currently scheduled for launch in 2014, will use astrometry. It may be able to detect Earth-like planets around several nearby stars. The European Space Agency's Darwin probe and NASA's Terrestrial Planet Finder  probes will attempt to image planets directly. A recently proposed idea is the New Worlds Mission, which will use an occulter to block a star's light, allowing astronomers to directly observe the dimmer orbiting planets.
Huge proposed ground telescopes may also be able to directly image extrasolar planets. ESO is considering building the European Extremely Large Telescope, with a mirror diameter between 30 and 60 meters.
If it goes ahead sometime between 2025-2035, ATLAST would be able to study surface features of Earth-sized planets.