Dark matter searches with liquified xenon

Laura Baudis


1 Introduction

In the last decades, cosmological and astrophysical observations have established a standard model of cosmology, the predictions of which are in remarkable agreement with a vast range of observations. A key ingredient of this model is dark matter, making about 85% of the cosmic matter density. While the role of dark matter – a component which does not emit nor absorb or scatter any light – in our Universe and its distribution on galactic and extra-galactic scales is well understood, its composition remains a mystery. At the most fundamental level, we are yet to answer the questions: what is dark matter and how does it interact with visible matter? The question on the nature of dark matter is strongly connected to the physics of the early Universe when new elementary particles − a dark species – could have been produced in addition to neutrinos, electrons, quarks, photons and other known particles. Among these, weakly interacting massive particles (WIMPs), with masses from a few GeV/c2 (where the mass of the proton, the nucleus of the hydrogen atom, is about 1 GeV/c2) to ~ 100 TeV/c2, are particularly appealing, since they can address open questions in particle physics not related to the dark matter problem. Other compelling dark matter candidates are QCD axions, ultra-light particles with masses constrained by laboratory searches and astrophysical observations to the range ~1 μeV/c2 − 3 meV/c2. These arise as a solution to the strong CP problem in QCD, related to the fact that the strong interaction is time-reversal invariant within current experimental precision.

According to astrophysical measurements, the visible structure of our galaxy resides in an extended, roughly spherical dark matter halo, where the density at the Solar System location is around 0.3 GeV/cm3 while the average speed is ~220 km/s. As an example, for dark matter particle masses of 100 GeV/c2, this translates into a flux of roughly 105 particles per second and square centimetre on Earth. Thus, as we move, together with the Sun, around the Galactic Centre and through the dark halo, we encounter a wind of dark matter particles. Experimentalists hope to detect those very rare occasions when such a particle collides with an atomic nucleus or an electron in a detector and deposits a small amount of energy – this method to look for dark matter is called direct detection.

The density and velocity distribution of dark matter particles, together with their mass and interaction strengths, determine the expected scattering rates in a detector, as well as the deposited energies. In the case of WIMPs, the deposited energies are at the keV scale (the typical energy of X-rays) and below, requiring very low energy threshold detectors. From the fact that we have not observed any convincing dark matter signal so far, the expected scattering rates are smaller than one per tonne of target material and year for particle masses above a few GeV/c2. These are almost unthinkable low rates, millions of times below those expected from cosmic ray interactions at the Earth’s surface. Thus, in order to potentially observe the dark species, detectors must be constructed and operated in deep underground laboratories, such as the Laboratori Nazionali del Gran Sasso (LNGS) in the Gran Sasso National Park. However, an underground location is far from sufficient. The experiments must also be placed in extremely quiet environments: shielded from the natural radioactivity of their immediate surroundings, and purified from potential radionuclides which can emit $\alpha$-, $\beta$- and $\gamma$-radiation when they decay inside the target material and possibly mimic the expected signal.

After more than two decades of technological development, several technologies stand out as excellent dark matter detectors. Depending on the employed material and readout scheme, these observe scintillation light, ionisation or heat (phonons) when a particle scatters in a solid, liquid or gaseous target containing, e.g., Ge, Si, CaWO3, NaI, Ar, Xe, C3F8, etc. (see fig. 1). Among these, detectors based on the ultra-pure, liquified noble elements argon and xenon, as well as bolometers operated only a few tens of mK above the absolute zero are reaching unprecedented sensitivities in the search for feeble and rare dark matter interactions.

2 Liquified xenon as dark matter target

Xenon has outstanding properties as a dark matter target. In its liquid phase at a temperature around 165 K it is an ideal medium for building large, homogeneous, compact and self-shielding detectors, also due to its density, three times as high as the one of water (~3 g/cm3), and high atomic number (Z=54, A=131.3). Xenon is an excellent scintillator and good ioniser in response to the passage of radiation because of its low ionisation potential of 12.13 eV. Scintillation light (in the vacuum ultraviolet region at 175 nm) is produced by the formation and radiative decay of so called excimers, which are bound ion-atoms states, Xe2*, making the medium transparent to its own light. Due to the de-excitation of the singlet and triplet states of the excited dimers, the scintillation decay times have a fast (~4.2 ns) and slow (~22 ns) component. Although the intensity ratio of the singlet to triple states depends on the deposited energy density, the effect is difficult to exploit in practice because of the similar timescales. If an electric field is applied, typically a few 100 V/cm, ionisation electrons can also be detected, either directly, or through the process of proportional scintillation, or electroluminescence. Figure 2 shows a schematic view of the signal production in xenon.

The simultaneous detection of scintillation light and charge in a time projection chamber (TPC) leads to a good energy resolution and the identification of the primary particle interacting in the liquid, an essential feature for picking out a signal-like interaction from background events. Moreover, the three-dimensional mapping of the spatial position of scatters allows to select single events – as expected from dark matter particles and other rate interactions – from multiple interactions in the detector, which can be caused for instance by gamma-rays and neutrons. The 3D position resolution also allows for fiducialisation, without requiring the presence of physical surfaces (which are the main sources of materials background).

Another advantage of xenon as dark matter target is the presence of two isotopes with spin (129Xe, spin-1/2 and 131Xe, spin-3/2) in natural xenon, at a combined abundance of almost 50%, allowing to probe different types of dark matter particle interactions, in particular also spin-dependent couplings. Finally, natural xenon does not contain radioactive isotopes, apart from the very long lived 124Xe and 136Xe, with half-lives of 1.8×1022 years and 2.1×1021 years, respectively. These disintegrate via the exceedingly rare nuclear processes of double electron capture and double beta decay, and are interesting signatures on their own, especially the decay modes without the emission of neutrinos.

One of the primary challenges in current and future liquid-xenon detectors is the purification from radioactive isotopes which are either present in the atmosphere, such as 85Kr, or emanated by detector materials, such as 222Rn. These noble elements mix with the liquid xenon and can lead to background events due to $\beta$- and $\alpha$-radiation. Depletion by isotopic separation in large cryogenic distillation columns is required, and these explore the different volatilities of Kr:Xe:Rn. As we will see, cryogenic distillation has led to concentrations of <10–12 mol/mol for natKr/Xe (where 85Kr is present at a level of ~2×10–11 mol/mol in natural Kr) and <5 μBq/kg of 222Rn in Xe. In the case of radon, emanation by materials must also be minimised, via material selection, surface treatment and detector design. Yet another challenge in Xe TPCs is to reach ultra-low levels of electronegative impurities such as O2 and N2, which can absorb electrons as they drift through the liquid and thus reduce the charge-induced signal. To achieve the challenging value required for the O2 concentrations of <10–10 mol/mol, gas-phase and cryogenic liquid-phase purification with hot ion getters and low-radon-emanation O2 filters, respectively, are used. We will discuss these measures in the context of the XENON programme.

Worldwide, several programmes using liquid xenon have successfully incorporated large, two-phase TPCs and have recently reached unparalleled low backgrounds and sensitivities to dark matter interactions: the LUX-ZEPLIN, the PandaX and the XENON collaborations.

3 The XENON programme: overview

XENON is a phased approach to particle dark matter detection with two-phase xenon TPCs operated underground at LNGS of INFN. An interaction within the active volume of the detector will create ionisation electrons and prompt scintillation photons, and both signals are observed. The electrons drift in the pure liquid under an electric field, are accelerated by a stronger field and extracted into the vapour phase above the liquid, where they generate the proportional scintillation signal. Two arrays of photomultiplier tubes, one in the liquid and one in the gas phase, detect the prompt scintillation (S1) and the delayed, proportional scintillation signal (S2), see fig. 3. The ratio of the two signals is different for nuclear recoils created by WIMP dark matter or neutron interactions, and electronic recoils produced by β and γ-rays, providing the basis for background discrimination. Since electron diffusion in the ultrapure liquid xenon is relatively small, the proportional scintillation photons carry the x-y information of the interaction site. With the z-information from the drift time measurement, and the $t_0$ from the prompt scintillation signal, the TPC thus yields a three-dimensional event localisation, enabling to reject the majority of the background via fiducial volume cuts.

With the XENON10, XENON100, XENON1T and XENONnT experiments, the XENON collaboration has operated TPCs at the 10 kg, 100 kg, 1000 kg and 6000 kg active mass scale. An increase in mass is however only useful if a simultaneous decrease in backgrounds can be achieved. Thanks to an extremely low background level, XENON has led the field of direct detection for more than a decade. DARWIN, building upon the legacy of XENON, aims to construct and operate an even larger detector, with 40 t of liquid xenon in the TPC. Figure 4 shows an overview of the programme, including the total xenon mass, the electron drift length in the TPCs and the design cross section for scalar WIMP-nucleon interactions.

The XENON10 experiment was the

first two-phase, 3D position-sensitive xenon TPC at LNGS, and was operated underground for more than 6 months with a high degree of stability. A pathfinder experiment, it had been developed to test the concept, verify the achievable energy threshold, resolution and background rejection capability of this technology. With 89 2.5-cm square photosensors (41 in the liquid and 48 in the gas phase), an electron drift distance of 15 cm and a fiducial mass of 5.4 kg of LXe, XENON10 improved the then best constraints on dark matter nucleon interactions by almost one order of magnitude, reaching 4.5×10–44 cm2 for a WIMP mass of 30 GeV. The initial collaboration, comprised of 39 members from 10 institutions and led by Prof. Elena Aprile of Columbia University, was later enlarged to design and construct the next generations of detectors. The current XENON collaboration has 181 members from 27 institutions in 12 countries.

The XENON100 experiment was designed to increase the fiducial mass by about a factor of 10 and at the same time decrease the background by a factor of 100. This ambitious goal became reality thanks to the screening and selection of detector materials with low radioactivity levels, a 4 cm thick active liquid-xenon shield around the TPC (observed by 64 PMTs) and an improved passive shield at LNGS, with 5 cm of low-radioactivity copper lining the inner shield walls, and a 20 cm thick outer layer of water and polyethylene to moderate neutrons. The inner shield was constantly purged with high-purity, boil-off nitrogen to avoid radon penetrating the passive shield. The XENON100 TPC had an electron drift distance of 30.5 cm, a diameter of 30.6 cm and two arrays of 2.5 cm square photosensors, 98 above the xenon target in the gas phase, and 80 at the bottom in the liquid. In an inner fiducial mass of 34 kg and after 225 live days of data, it reached a background of 5.3×10–3 events/(keV kg day) and improved the sensitivity to dark matter by more than a factor of 10 compared to XENON10, reaching a minimum of 1.1×10–45 cm2 at a 55 GeV WIMP mass.

The next phase, XENON1T, aimed for a TPC with a 1 tonne fiducial mass (3.2 t LXe in total, with 2 t in the TPC) and more than one order of magnitude decrease in backgrounds. The latter was achieved by stronger material selection, the development – together with Hamamatsu – of low-radioactivity, 3-inch diameter photosensors with increased quantum efficiency, and a 740 m3 water Cherenkov shield surrounding the cryostat which houses the TPC. This new experiment was thus placed in Hall B of LNGS, unlike XENON10/100 which were located in the interferometer tunnel. The Cherenkov muon veto, 10.2 m in height and 9.6 m in diameter, was instrumented with 84 20.3-cm diameter photomultiplier tubes and served as an active shield against cosmic muons, as well as a passive shield for gamma and neutron radiation coming from the laboratory walls. The 96 cm diameter and 1 m high TPC was instrumented with two arrays (127 on the top and 121 at the bottom) of 3-inch photomultiplier tubes and lined with diamond polished PTFE to ensure a high reflectivity for the xenon scintillation light. XENON1T also had a new cryogenic and purification system, as well as a storage and recovery system (ReStoX). A vacuum-insulated stainless steel sphere with 2.1 m in diameter, ReStoX can hold up to 7.6 t of xenon in liquid or gas phase, being able to withstand pressures up to 73 bar. Finally a 5.5 m high distillation system reduced the natKr/Xe concentration to about 0.36 ppt, with a capability to reach a concentration of <0.03 ppt. The column exploits the fact that Kr has a 10.8 times higher vapour pressure than xenon and will thus be collected at the top, while the xenon depleted in krypton will be collected at the bottom. This step was essential to reduce the internal background coming from beta-decays of 85Kr. XENON1T took data until the end of 2018 and probed dark matter nucleon interactions over a large range of masses, and down to the very low cross section of 4.7×10-47 cm2 at a 30 GeV WIMP mass. Figure 5 shows the combined WIMP results from masses of 100 MeV to 1 TeV: the regions above the solid curves are excluded.

The current phase, XENONnT, with a total of 8.6 t of xenon started a first science run in July 2021 and is expected to take data until 2025. It contains a new, larger TPC with 494 3-inch diameter PMTs. The photosensors are improved versions of the XENON1T PMTs, and were tested in both liquid- and gaseous-xenon phase before being selected and assembled into the two arrays. Their quantum efficiency for the 175 nm Xe scintillation light has a mean of 34% and typical gain at a high voltage of 1.5 kV is 8×106 These sensors thus have an excellent response to single-photon signals and they are expected to operate in a stable mode in the cryogenic environment for long periods of time. Another advantage, which is essential for rare and low-energy searches, is their dark count rate, about 20 Hz/PMT at –100 °C.

To decrease the background even further and thus maximise the chance of a discovery, a new radon distillation column filters out the radon permanently emanated by the detector materials. In addition, a neutron veto, installed inside the Cherenkov water shield, will reduce the rate of neutrons scattering in the TPC and thus possibly mimicking a WIMP signal. Finally, a new liquid purification system allows for a faster cleaning from electronegative impurities and thus a much longer electron lifetime compared to XENON1T.

DARWIN, a 50 t liquid-xenon detector and the successor of XENONnT, will be the ultimate xenon-based direct detection experiment, designed to probe the theoretically allowed parameter space for WIMPs until the irreducible background from solar and cosmic neutrino interactions will limit a further increase in sensitivity.

4 Recent results from XENON1T

The XENON1T experiment reached an unprecedented low background level and a low energy threshold of 1 keV when observing both primary and secondary scintillation signals. While it did not discover WIMP dark matter, it set the world’s best constraints over a wide range of masses, from ~100 MeV to 1 TeV and above, thus excluding a large set of theoretical models for dark matter candidates. Apart from this achievement, XENON1T performed a series of exciting measurements in astroparticle and nuclear physics, some of which will be mentioned here. In 2019, the collaboration announced the first observation of the double electron capture process in 124Xe (see fig. 6). With a half-life of 1.8×1022 years, this is the slowest process ever measured directly. This study showed that the XENON1T detector was also capable to measure other rare physical phenomena with a high sensitivity. 124Xe has 54 protons and 70 neutrons, and in the double electron capture process two protons transform simultaneously into two neutrons by capturing two electrons from the innermost atomic shells, emitting two neutrinos. While the other atomic electrons reorganise themselves to fill the two holes in the innermost shells, energy in form of X-rays and Auger electrons is released, for example 64.33 keV which is twice the K-shell binding energy. XENON1T observed 126 ± 29 signal events, with 9±7 expected background events from 125I decays at 67.5 keV, where 125I comes from 125Xe decays via electron capture, itself produced by neutron capture on 124Xe during the neutron calibration campaign. Even more interesting yet much more difficult to observe is the neutrinoless double electron capture, which would occur if neutrinos were their own antiparticles, hence Majorana fermions. At the other end of the energy spectrum, namely at high energies relevant for the neutrinoless double beta decay of 136Xe, the XENON1T TPC reached an energy resolution of $\sigma / E = 0.8$% at 2.46 MeV, the Q-value of the decay, see fig. 7. While XENON1T is not competitive to dedicated neutrinoless double beta decay experiments, given the 8.9% abundance of 136Xe in natural xenon, this demonstration was crucial for XENONnT and in particular also for DARWIN, which will have an increased sensitivity due to lower backgrounds and higher isotope mass.

In 2020 the XENON collaboration reported a surprising excess of events at energies in the energy region 1-7 keV, namely 285 events instead of the expected 232±5 events. Extensive checks which took almost two years excluded known backgrounds and systematic effects as the source. One explanation however is a new source of background, not considered before, which could be caused by minute amounts of tritium in the detector. Tritium disintegrates via $\beta$-decay with an endpoint of 18.6 keV and a half-life of 12.3 years, and only a few atoms for every 1025 xenon atoms would be required to explain the excess. A more exciting explanation is the existence of axions which could be produced in the Sun in the electromagnetic field of charged particles. The energy of solar axions is in the keV range, related to the temperature of the Sun’s interior and the energy spectrum provides a better match to the observed spectrum (see fig. 8). Axions can escape the Sun’s core and travel to Earth and underground where they can interact with xenon trough the so-called axioelectric effect. This process is similar to the photoelectric effect, whereby the axion is absorbed and an electron is emitted (fig. 9). The XENON1T excess can be explained by a solar axion signal with a statistical significance of $3.4\sigma$ – the chance that the result is due to a statistical fluctuation is about 2 in 104. While the observation of solar axions would be a significant breakthrough, the data of XENON1T is also explained by the tritium hypothesis, with a statistical significance of $3.2\sigma$. With current knowledge, it is impossible to confirm or reject the tritium hypothesis independently. A larger detector with even lower background rates such as XENONnT is required.

5 Status of XENONnT

The XENONnT experiment has been constructed and assembled at LNGS in 2020 and 2021, and is about to start its first science run. It consists of three nested detectors: a large, Gd-loaded water Cherenkov detector acting as a muon veto with an optically separated smaller volume instrumented as a neutron detector and a TPC enclosed in a double-walled, vacuum-insulated stainless steel cryostat (a schematic view is shown in fig. 10). The total mass of xenon is 8.6 t, with 5.9 t inside the cylindrical TPC, enclosed by 24 polytetrafluoroethylene (PTFE) reflector panels on the sides and two PMT arrays with 253 and 241 3-inch PMTs on the top and bottom, respectively (fig. 11). The PTFE panels reflect the VUV Xe scintillation light which is then collected by the PMTs. Three top (screening, anode and gate) and two bottom electrodes (cathode, screening) are used to generate the drift and electron extraction fields, while the field uniformity is ensured by two concentric sets of copper field shaping rings. Connected to the cryostat are two double-walled vacuum-insulated pipes which lead to the cryogenics and purification systems in the service building and also house the signal and high-voltage cables of the PMTs and of various sensors. To achieve a high liquid-xenon purity regarding electronegative molecules and thus a high electron lifetime (which quantifies the charge loss, and is the mean time for the charge to drop by a factor of $e$) a new, cryogenic liquid-xenon purification system was built and installed. Based on efficient oxygen filters with low radon emanation and cryogenic liquid-xenon pumps, an electron lifetime larger than 7 ms (where the maximum drift time for a nominal drift velocity of 1.6 mm/μs is ~1 ms) was achieved after 10 days of operation with 8.6 t of xenon, a factor of ~10 higher than in XENON1T.

In XENONnT, the 85Kr activity must be even lower than in XENON1T and the krypton is removed by cryogenic distillation to a level of ~10-13 mol/mol natKr/Xe. 85Kr, a beta emitter with an endpoint energy of 687 keV and a half-life of 10.8 years is created in nuclear fission reactions and is present in the atmosphere. It thus originates from the xenon extraction from air, and must be removed once from the xenon budget of the detector. Radon, on the other hand, is continuously emanated by detector materials and thus an online distillation system must be employed. A dedicated radon removal system was installed at LNGS for XENONnT, with the goal of achieving a 222Rn level of 1 μBq/kg, which is about a factor of 10 reduction compared to XENON1T. As previously mentioned, this reduction factor cannot be achieved by the radon distillation column alone; it is also due to lower radon emanation of materials and a more favourable surface-to-volume ratio in the new detector. To mitigate the neutron background from detector materials, a new component was built and recently commissioned for XENONnT: the neutron veto. Surrounding the outer cryostat, it is based on gadolinium-loaded water, a technology first developed and employed in the Super-Kamiokande experiment in Japan. The Cherenkov light emitted due to particle interactions in the veto, in particular from high-energy gamma rays of 8 MeV after neutron capture on Gd, is observed by 120 8-inch PMTs placed in the water, and optically separated from the surrounding Cherenkov muon veto. The neutron tagging efficiency of this new detector is 87% with 0.2% of gadolinium in mass, allowing to reduce the radiogenic neutron background in XENONnT to sufficiently low levels.

After an extended commissioning campaign of all systems, including the TPC, the cryogenic and purification system and neutron veto, as well the new radon distillation column, calibration measurements for electronic and nuclear recoils were ongoing underground. These are followed by a first science run, to last for several months. XENONnT is expected so solve the puzzle of the excess of events at low energies observed in XENON1T data with a few months of statistics, and in particular distinguish, based on the spectral shape alone, a very small tritium component from a genuine solar axion signal, or something else. To achieve its design sensitivity to WIMP dark matter, the experiment will acquire data until 2025. With its large xenon mass in the TPC and an unprecedented low background, it has a fair chance to discover first, feeble signs of dark matter interactions.

6 The future and the DARWIN observatory

The aim of XENONnT is to increase the sensitivity to WIMP dark matter by one order of magnitude and thus to reach interaction cross sections on nucleons as low as 10–48 cm2 at 30 GeV/c2. To place this number into perspective, the cross section of neutrinos generated in beta decays, thus with energies around 1 MeV, is ~10-44 cm2, four orders of magnitudes larger. Nonetheless, to probe cross sections down to the region where cosmic neutrinos will start to limit the reach of dark matter detectors, an even larger target mass and backgrounds that are below those from solar neutrinos are called for. DARk matter WImp search with liquid xenoN (DARWIN) is a next-generation xenon observatory with the primary goal of exploring the experimentally accessible parameter space for WIMP dark matter. With a total liquid-xenon mass of 50 t (with 40 t inside the TPC), it will probe cross sections down to 10-49 cm2 and a wide range of dark matter masses. Should dark matter be discovered by the current generation of experiments (XENONnT, LUX-ZEPLIN, PandaX-4T) DARWIN will measure the induced dark matter interactions with higher statistics and thus better constrain the particle mass and its interaction strength with matter. A detector of this size and expected energy threshold as low as ~ 1 keV will have a plethora of other physics goals: it will detect solar pp-neutrinos in real time with high statistics, via electron-neutrino scattering, 8B solar neutrinos via coherent neutrino nucleus scattering, and search for the neutrinoless double beta decay of 136Xe, present with a natural abundance of 8.9% in xenon. DARWIN will also search for light dark matter candidates and will look for solar axions and for neutrinos from supernova explosions. Observing neutrinos of all flavours, such a measurement would provide complementary information to largescale water Cherenkov and liquid-Ar detectors.

In its baseline design, DARWIN will operate a 2.6 m tall and 2.6 m diameter two-phase TPC, with the primary and proportional light signal observed by two arrays of photosensors installed above and below the target. These sensors can be improved versions of the 3-inch diameter PMTs employed in XENONnT, if an even lower intrinsic radioactivity can be achieved. Albeit a proven and reliable technology, PMTs are bulky, can develop small vacuum leaks and still generate a significant fraction of the radioactivity budget in a large detector. Hence several alternatives are under study for DARWIN, such as silicon photomultiplier arrays, or hybrid sensors. Similarly to the previous generation of experiments, the TPC will be housed in a double-walled vacuum insulated cryostat, surrounded by neutron and muon water Cherenkov detectors. While the underground laboratory has not been yet fixed, LNGS is one of the prime locations. To reduce the muon-induced neutron background to negligible levels, a water tank will at least 12 m in diameter will be required.

The realisation of a detector the size of DARWIN is not feasible based on extrapolations from existing technologies and for this reason a dedicated R&D programme is in place to develop and test the required components. In particular, two full-scale demonstrators, one in vertical (see fig. 12) and one in horizontal direction were constructed and are under commissioning, with the principal aim of demonstrating electron drift over 2.6 m and testing electrodes of 2.6 m diameter, respectively. Other goals are measurements of the electrode cloud diffusion and of optical properties of liquid xenon over such large distances, development of high-voltage feedthroughs and characterisation of new photosensors under detector conditions as expected in DARWIN.

The new experiment will require another reduction in the overall background levels, in particular the radon background must be reduced to 0.1 μBq/kg, a factor 10 compared to the XENONnT goal. Apart from strong material selection and active removal by cryogenic distillation, various surface treatments that prevent radon emanation are being studied by the collaboration. The krypton goal is 0.03 ppt natKr/Xe, a value that has been demonstrated with gas samples from XENON1T’s distillation column.

The DARWIN collaboration includes 33 institutions in Europe, USA and Asia, who are working together towards a conceptual and technical design report, expected for 2022 and 2024, respectively. The construction phase is expected to start in 2024-25, the commissioning in 2026 and a first science data taking in 2027.

7 Summary and outlook

The dark matter problem is almost a century old.

We know that our cosmos is dominated by a new form of matter which forms galaxies, including our own Milky Way, and larger structures, but so far is invisible to any detector or telescope. While we can chart its spatial distribution and measure its evolution in time, the fundamental composition of this invisible, new form of matter remains a mystery. After decades of development, detectors using liquid xenon at their cores have matured into a robust technology for detecting the minuscule energy deposited when a hypothetical dark matter particle scatters off an atomic nucleus or electron in the target. Among these, those part of the XENON programme at LNGS have demonstrated the lowest backgrounds reached in any direct detection experiment, and obtained world leading sensitivities for a broad range of dark matter masses. The new XENONnT detector, starting a first science run in 2021, will improve the backgrounds by another order of magnitude and will have a realistic chance to detect first, faint signals of dark matter particles. DARWIN, a 50 t liquid-xenon detector to follow after XENONnT, is designed to reach the region where the irreducible, cosmic neutrino background will start to manifest itself, limiting the reach of even larger detectors. The pragmatic goal is to scrutinise the region in dark matter mass and interaction cross section that is accessible to experiment, while also breaking new grounds in other areas of astroparticle physics, in particular in neutrino physics. If direct evidence for a dark matter signal will indeed be established, the focus will shift towards measuring the properties of the dark species, such as their mass, interaction cross section and possibly spin. This may well launch the beginning of a new field, dark matter astronomy.