As particle physicists became increasingly interested in the problem of the missing matter of the Universe, some began to turn their attention toward ways that individual particles of dark matter might be detected, either directly or indirectly. Although many of the leading techniques were first conceived of in the 1980s, dark matter searches have continued with vigor ever since, occupying the attentions of generations of experimental particle-astrophysicists.
A. Scattering with Nuclei
In 1984, an article by Andrzej Drukier and Leo Stodolsky at the Max Planck Institute in Munich appeared in Physical Review D, discussing techniques that might be used to detect neutrinos scattering elastically off nuclei . Among other possibilities, the article proposed the use of a superconducting colloid detector, consisting of micron-scale superconducting grains maintained at a temperature just below their superconducting transition. Even a very small quantity of energy deposited by the recoil of an incident neutrino could cause a superconducting grain to flip into the normal state, collapsing the magnetic field and producing a potentially measurable electromagnetic signal. In January 1985, Mark Goodman and Ed Witten submitted a paper to the same journal, arguing that this technology could also be used to detect some types of dark matter particles  14. Although Drukier and Stodolsky's original detector concept was never employed at a scale sensitive to dark matter, the broader notion of experiments capable of detecting ∼1-100 keV nuclear recoils provided a path through which it appeared possible to test the WIMP hypothesis.
In their original paper, Goodman and Witten considered three classes of dark matter candidates: 1) those that undergo coherent scattering with nuclei (also known as spin-independent scattering), 2) those that scatter with nuclei through spin-dependent couplings, and 3) those with strong interactions. The first two of these three categories provide the basis for how most direct dark matter detection results have since been presented. If mediated by unsuppressed couplings to the Z boson (an important early benchmark), coherent scattering was predicted to lead to large scattering rates, typically hundreds or thousands of events per day per kilogram of target material. With such high rates, the prospects for detecting dark matter in the form of a heavy neutrino or sneutrino appeared very encouraging. Dark matter candidates that scatter with nuclei only through spin-dependent couplings, in contrast, were generally predicted to yield significantly lower rates, and would require larger and more sensitive detectors to test. Even as early as in this first paper, Goodman and Witten pointed out that such experiments would have difficultly detecting dark matter particles lighter than ∼ 1-2 GeV, due to the modest quantity of momentum that would be transferred in the collisions.
The first experiment to place constraints on the scattering cross section of dark matter with nuclei was carried out in 1986 at the Homestake Mine in South Dakota by a collaboration of scientists at the Pacific Northwest National Laboratory, the University of South Carolina, Boston University, and Harvard . Using a low-background germanium ionization detector (originally designed to search for neutrinoless double beta decay), they accumulated an exposure of 33 kg-days, yielding a limit that significantly constrained dark matter candidates with unsuppressed spin-independent scattering cross sections with nuclei (such as heavy neutrinos or sneutrinos) . Shortly thereafter, similar results were obtained by an independent collaboration of scientists from the Universities of California at Santa Barbara and Berkeley .
Despite the importance of these first dark matter scattering limits, the reach of such detectors quickly became limited by their backgrounds, making it difficult to achieve significant improvements in sensitivity. One possible solution to this problem, first suggested by Andrzej Drukier, Katherine Freese, and David Spergel , was to search for an annual variation in the rate of dark matter induced events in such an experiment, as was predicted to result from the combination of the Earth's motion around the Sun and the Sun's motion through the dark matter halo. Such a technique could, in principle, be used to identify a signal of dark matter scattering over a large rate of otherwise indistinguishable background events. The most well known group to employ this technique was the DAMA/NaI Collaboration (and later DAMA/LIBRA). The original DAMA/NaI experiment consisted of nine 9.70 kg scintillating thallium-doped sodium iodide crystals, located in Italy's deep underground Gran Sasso Laboratory. In 1998, they published their first results, reporting the observation of an annually modulating rate consistent with dark matter scattering . Over the past nearly two decades, DAMA's signal has persisted and become increasingly statistically significant as more data was collected , including with the more recent DAMA/LIBRA detector [44, 45]. At this point in time, it seems hard to reconcile dark matter interpretations of the DAMA/LIBRA signal with the null results of other direct detection experiments. On the other hand, no convincing alternative explanation for this signal has been so far identified.
During the period of time that DAMA/NaI was being developed and collecting its first data, experimental techniques were being pursued that could discriminate dark matter-like nuclear recoil events from various backgrounds. These efforts ultimately lead to the technologies employed by the CDMS (Cryogenic Dark Matter Search), EDELWEISS (Experience pour DEtecter Les Wimps En Site Souterrain), and CRESST (Cryogenic Rare Event Search with Superconducting Thermometers) Collaborations. These experiments each made use of two-channel detectors, capable of measuring both ionization and heat (CDMS, EDELWEISS) or scintillation and heat (CRESST), the ratio of which could be used to discriminate nuclear recoil events from electron recoils generated by gamma and beta backgrounds. All three of these experiments employed crystalline target materials, maintained at cryogenic temperatures, consisting of germanium and silicon, germanium, and calcium tungstate, respectively. Throughout most the first decade of the 21st century, the CDMS and EDELWEISS experiments lead the field of direct detection, providing the most stringent constraints and improving in sensitivity by more than two orders of magnitude over that period of time (see Fig. 8).
Figure 8. The past and projected evolution of the spin-independent WIMP-nucleon cross section limits for a 50 GeV dark matter particle. The shapes correspond to limits obtained using different detectors technologies: cryogenic solid state detectors (blue circles), crystal detectors (purple squares), liquid argon detectors (brown diamonds), liquid xenon detectors (green triangles), and threshold detectors (orange inverted triangle). Taken from Ref. .
In order to continue to increase the sensitivity of direct dark matter experiments, it was necessary for experiments to employ ever larger targets, gradually transitioning from the kilograms of detector material used by EDELWEISS and CDMS (9.3 kg in the case of SuperCDMS) to the ton-scale and beyond. Cryogenic solid state detectors, however, have proven to be costly to scale up into ton-scale experiments. In the late 1990s, Pio Picchi, Hanguo Wang and David Cline pioneered an alternative technique that exploited liquid noble targets (most notably liquid xenon). Like solid state detectors, such experiments discriminate nuclear recoils from electron recoils by measuring two quantities of deposited energy; in this case scintillation and ionization. Between 2010 and 2015, the XENON100 and LUX experiments (each of which utilize a liquid xenon target) have improved upon the limits placed by CDMS by approximately two orders of magnitude. It is generally anticipated that future experiments employing liquid xenon targets (XENON1T, LZ, XENON-NT) will continue along this trajectory for years to come.
As CDMS, EDELWEISS, XENON100, LUX and other direct detection experiments have increased in sensitivity over the past decades, they have tested and ruled out an impressive range of particle dark matter models. And although results from the CoGeNT [2, 1], CRESST , and CDMS  experiments were briefly interpreted as possible dark matter signals, they now appear to be the consequences of poorly understood backgrounds [25, 173] and/or statistical fluctuations. While many viable WIMP models remain beyond the current reach of this experimental program, a sizable fraction of the otherwise most attractive candidates have been excluded. Of particular note is the fact that these experiments now strongly constrain dark matter particles that scatter coherently with nuclei through Higgs exchange, representing an important theoretical benchmark.
B. Annihilation and Decay
In the 1978 Valentine's Day issue of Physical Review Letters, there appeared two articles that discussed – for the first time – the possibility that the annihilations of pairs of dark matter particles might produce an observable flux of gamma rays. And although each of these papers (by Jim Gunn, Ben Lee, 15 Ian Lerche, David Schramm and Gary Steigman , and by Floyd Stecker ) focused on dark matter in the form of a heavy stable lepton (i.e. a heavy neutrino), similar calculations would later be applied to a wide range of dark matter candidates. On that day, many hopeless romantics became destined to a lifetime of searching for signals of dark matter in the gamma-ray sky.
At the time, the most detailed measurement of the astrophysical gamma-ray background was that made using data from the Small Astronomy Satellite (SAS) 2 . Although the intensity of 35-100 MeV gamma rays measured by this telescope (∼ 6 × 10−5 cm−2 s−1 sr−1) was several orders of magnitude higher than that predicted from annihilating dark matter particles smoothly distributed throughout the Universe, it was recognized that inhomogeneities in the dark matter distribution could increase this prediction considerably. In particular, annihilations taking place within high-density dark matter halos, such as that of the Milky Way, could plausibly produce a flux of gamma rays that was not much fainter than that observed at high galactic latitudes, and with a distinctive gradient on the sky [144, 303]. Focusing on GeV-scale dark matter particles, Gunn et al. went as far as to state that such a signal “may be discoverable in future γ-ray observations”.
Several years later, in 1984, Joe Silk and Mark Srednicki built upon this strategy, considering not only gamma rays as signals of annihilating dark matter particles, but also cosmic-ray antiprotons and positrons  (see also, Refs. [304, 112, 172]). They argued that the observed flux of ∼ 0.6-1.2 GeV antiprotons  provided the greatest sensitivity to annihilating dark matter, and noted that ∼10 GeV WIMPs would be predicted to produce a quantity of cosmic-ray antiprotons that was comparable to the observed flux.
In 1985, Lawrence Krauss, Katherine Freese, David Spergel and William Press published a paper suggesting that neutrinos might be detected from dark matter annihilating in the core of the Sun  (see also, Ref. ). Shortly thereafter, Silk, Olive, and Srednicki pointed out that not only could elastic scattering cause dark matter particles to become gravitationally bound to and captured within the Sun, but that the number of WIMPs captured over the age of the Solar System could be sufficiently high to attain equilibrium between the processes of capture and annihilation . Observations over the subsequent few years by the proton decay experiments IMB, FREJUS, and Kamioka capitalized on this strategy, strongly constraining some classes of dark matter candidates, most notably including light electron or muon sneutrinos. Similar approaches using dark matter capture by the Earth were also proposed around the same time [190, 127].
In the decades that followed, measurements of astrophysical gamma ray, antimatter, and neutrino fluxes improved dramatically. In parallel, the scientific community's understanding of the astrophysical sources and propagation of such particles also matured considerably. Information from successive gamma-ray satellite missions, including COS-B , EGRET (Energetic Gamma Ray Experiment Telescope) , and the Fermi Gamma-Ray Space Telescope, gradually lead to the conclusion that most of the observed gamma-ray emission could be attributed to known gamma-ray source classes (such as active galactic nuclei), although it remains possible that a non-negligible component of the high-latitude background could originate from dark matter .
Motivated by their high densities of dark matter and low levels of baryonic activity, dwarf spheroidal galaxies – satellites of the Milky Way – have in recent years become a prime target of gamma-ray telescopes searching for evidence of dark matter annihilations. Fermi's study of dwarf galaxies has provided the strongest limits on the dark matter annihilation cross section to date, strongly constraining WIMPs lighter than ∼100 GeV or so in mass . Ground based gamma-ray telescopes have also used observations of dwarf galaxies to constrain the annihilations of heavier dark matter candidates. Although complicated by imperfectly understood backgrounds, gamma-ray observations of the Milky Way's Galactic Center are also highly sensitive to annihilating WIMPs. A significant excess of GeV-scale gamma-rays has been identified from this region, consistent with arising from the annihilations of ∼50 GeV particles [139, 87]. An active debate is currently taking place regarding the interpretation of these observations. Alternative targets for indirect searches have also been proposed, including Galactic dark matter subhalos not associated with dwarf galaxies [92, 244], and density “spikes” of dark matter around black holes [138, 345, 49].
Over approximately the same period of time, great progress has also been made in the measurement of the cosmic-ray antiproton spectrum, including successive advances by the CAPRICE [56, 57], BESS [27, 8], AMS , and PAMELA  experiments. When these measurements are combined with our current understanding of cosmic-ray production and propagation, they appear to indicate that the observed cosmic ray antiproton spectrum originates largely from conventional secondary production (cosmic-ray interactions with gas), although a significant contribution from dark matter remains a possibility. These measurements generally yield constraints on annihilating dark matter that are not much less stringent than those derived from gamma-ray observations.
Compared to antiprotons, measurements of the cosmic-ray positron spectrum have been more difficult to interpret. Building upon earlier measurements [221, 135, 55], the balloon-bourne HEAT experiment observed in 1994, 1995, and 2000 indications of an excess of cosmic-ray positrons at energies above ∼10 GeV, relative to the rate predicted from standard secondary production . This was later confirmed, and measured in much greater detail, by a series of space-based experiments: AMS , PAMELA , and AMS-02 . Although this positron excess received much attention as a possible signal of annihilating dark matter, this possibility is now strongly constrained by a variety of arguments (e.g. Ref. [47, 130, 296]), and plausible astrophysical explanations have also been proposed (e.g. Ref. ).
As large volume neutrino telescopes began to be deployed, such experiments became increasingly sensitive to dark matter annihilating in the interiors of the Sun and Earth. The AMANDA detector at the South Pole , along with Super-Kamiokande in Japan , each significantly improved upon previous limits, to be followed most notably by IceCube  and ANTARES . Constraints from neutrino telescopes are currently competitive with those derived from direct detection experiments for the case of WIMPs with spin-dependent interactions with nuclei.
Many of the strategies employed to search for annihilating dark matter have also been used to constrain the rate at which dark matter particles might decay. In addition to constraints on gravitinos and other potentially unstable particles, such searches are particularly interesting within the context of sterile neutrino dark matter. Sterile neutrinos with masses in the range of ∼1-100 keV are predicted to decay (into an active neutrino and a photon) at a rate that could generate a potentially observable X-ray line . In fact, considering the standard case of Dodelson-Widrow production (as discussed in Chapter V), the combination of constraints from X-ray observations and measurements of the Lyman-α forest [319, 284] disfavor sterile neutrino dark matter over this entire mass range. Models with enhanced production in the early Universe  can evade such constraints, however, and continue to receive considerable interest. In particular, reports of a 3.55 keV line observed from a collection of galaxy clusters [65, 61] have recently received a great deal of attention within the context of a decaying sterile neutrino.
C. Axion Experiments
For some time, there has been an active experimental program searching for dark matter axions, most notably in the form of the Axion Dark Matter eXperiment (ADMX). The idea behind this effort is to make use of the photon-photon-axion coupling, generically present in axion models, to convert dark matter axions in a strong and static magnetic field into a signal of nearly monochromatic microwave photons. This possibility was first suggested by Pierre Sikivie in 1983 , and was later expanded upon by Sikivie , along with Lawrence Krauss, John Moody, Frank Wilczek and Donald Morris . As the signal in such an experiment is maximized for a specific cavity frequency (corresponding to a specific axion mass), it is necessary that the resonant frequency of the cavity be tunable, making it possible to scan over a range of axion masses.
The first laboratory constraints on dark matter axions were presented in the late 1980s, by a number of groups [88, 342, 149]. While the frequency range covered by these experiments was well suited to axion masses favored by dark matter abundance considerations (covering approximately ma ≃ 4.5−16.3 µeV), their sensitivity was orders of magnitude below that required to test realistic axion models. In 2003, however, the ADMX Collaboration reported results that constrained realistic axion dark matter models, although only for a relatively narrow range of masses, 1.9−3.3 µeV . With anticipated upgrades [28, 28], ADMX is expected to be sensitive to a much larger range of axion masses and couplings, significantly constraining the axion dark matter parameter space in the coming years.
Acknowledgements. We are grateful to the many pioneering physicists and astronomers who helped us to reconstruct the methods, ideas and circumstances that led to the establishment of the dark matter paradigm. In particular, we would to thank Andrea Biviano, Lars Bergström, Albert Bosma, Shantanu Desai, Jaco de Swart, Jaan Einasto, John Ellis, Pierre Fayet, Ken Freeman, Katherine Freese, Steve Kent, Rocky Kolb, Keith Olive, Jerry Ostriker, Sergio Palomares-Ruiz, Jim Peebles, Joel Primack, Morton Roberts, Bernard Sadoulet, Bob Sanders, Gary Steigman, Alar Toomre, Juri Toomre, Scott Tremaine, Virginia Trimble, Michael Turner and Simon White. We also acknowledge many useful discussions with Jeroen van Dongen and Jaco de Swart. This work would not have been possible without the vast collection of articles and books, dating back to the 19th century, made freely available by NASA ADS and the Internet Archive Project. GB acknowledges support from the European Research Council through the ERC starting grant WIMPs Kairos. DH is supported by the US Department of Energy.
14 A similar paper by Ira Wasserman  was submitted shortly after Goodman and Witten's. Back.
15 In regards to Ben Lee, who died in a traffic accident in 1977, this article was published posthumously. Back.