The massive binary system Eta Carinae and the surrounding H ii complex, the Carina nebula, are potential particle acceleration sites from which very high energy (VHE; E≥ 100 GeV) γ-ray emission could be expected. This paper presents data collected during VHE γ-ray observations with the HESS telescope array from 2004 to 2010, which cover a full orbit of Eta Carinae. In the 33.1-h data set no hint of significant γ-ray emission from Eta Carinae has been found and an upper limit on the γ-ray flux of (99 per cent confidence level) is derived above the energy threshold of 470 GeV. Together with the detection of high energy (HE; 0.1 ≤E≤ 100 GeV) γ-ray emission by the Fermi Large Area Telescope up to 100 GeV, and assuming a continuation of the average HE spectral index into the VHE domain, these results imply a cut-off in the γ-ray spectrum between the HE and VHE γ-ray range. This could be caused either by a cut-off in the accelerated particle distribution or by severe γ–γ absorption losses in the wind collision region. Furthermore, the search for extended γ-ray emission from the Carina nebula resulted in an upper limit on the γ-ray flux of (99 per cent confidence level). The derived upper limit of ∼23 on the cosmic ray enhancement factor is compared with results found for the old-age mixed-morphology supernova remnant W28.
The Carina nebula is one of the largest and most active H ii regions in our Galaxy and a place of ongoing star formation. It is located at a distance of ∼2.3 kpc and harbours eight open stellar clusters with more than 66 O-type stars, three Wolf–Rayet stars and the luminous blue variable (LBV) Eta Carinae (Feinstein 1995; Smith 2006; Smith & Brooks 2008). The existence of a ∼106-yr-old neutron star indicates past supernova (SN) activity in the Carina complex (Hamaguchi et al. 2009; Pires et al. 2009). The age estimates of the member clusters Trumpler 14, 15 and 16 vary significantly, with an age spread of ∼2 Myr to ∼8 Myr, indicating several past episodes of star formation in the northern region; more recent star formation is going on in the southern part of the nebula (see Preibisch et al. 2011a, and references therein). Extended X-ray emission has been reported by e.g. Hamaguchi et al. (2007) based on observations with Suzaku, supplemented by XMM–Newton (Ezoe et al. 2008) and Chandra (Townsley et al. 2011) observations. These authors found a very low nitrogen-to-oxygen ratio, which, in addition to the presence of a neutron star, suggests that the diffuse plasma originates in one or several unrecognized supernova remnants (SNRs), in particular in the area surrounding Eta Carinae. The emission may also be attributed to stellar winds from massive stars. In their ∼1.4 deg−2 survey of the diffuse X-ray emission, Townsley et al. (2011) also found evidence for a significant contribution due to charge exchange. This mechanism would originate in a contact layer between the hot plasma and the cold molecular clouds (MCs).
Eta Carinae, a member star of Trumpler 16 (Tr 16), is one of the most peculiar objects in our Galaxy, whose environment shows traces of massive eruptions that occurred in past epochs. A giant outburst in the 1840s (known as the Great Eruption) and a smaller outburst in the 1890s produced the Homunculus and little Homunculus Nebulae, respectively (see e.g. Ishibashi et al. 2003). The material expelled from the central star in the Great Eruption has a combined mass of ∼12 M⊙ and moves outwards at an average speed of ∼650 km s−1, implying a kinetic energy of roughly (4–10) × 1049 erg (Smith et al. 2003). Smith (2008) found material that is moving ahead of the expanding Homunculus Nebula at speeds of 3500–6000 km s−1, which doubles the estimate of the kinetic energy of the giant outburst. For a long time it was believed that the central object, Eta Carinae, is a single, hypergiant LBV star – one of only very few found in the Galaxy (see e.g. Clark, Larionov & Arkharov 2005). However, observations now suggest Eta Carinae to be composed of a massive LBV star and an O- or B-type companion star (Hillier et al. 2001; Pittard & Corcoran 2002). The present-day period of the binary has been estimated to Porb= 2022.7 ± 1.2 d (Damineli et al. 2008), its eccentricity to be e∼ 0.9 (Nielsen et al. 2007) and the semimajor axis to be a= 16.64 au (Hillier et al. 2001). The LBV star has a very high mass loss rate of (Hillier et al. 2001; Parkin et al. 2009) and a terminal wind velocity of ; the companion star has a thin, fast wind ( and v2∼ 3000 km s−1; Pittard & Corcoran 2002). The total kinetic energy in stellar winds is of the order of a few ×1037 erg s−1 for the LBV and the OB star together.
When stellar winds of such stars collide, they form a stellar wind shock, where particles can be accelerated to non-thermal energies (e.g. Eichler & Usov 1993; Reimer, Pohl & Reimer 2006). There is strong evidence for the existence of non-thermal particles in Eta Carinae based on X-ray measurements performed with the instruments aboard the INTEGRAL (Leyder, Walter & Rauw 2008, 2010) and Suzaku satellites (Sekiguchi et al. 2009). In the high energy (HE; 100 MeV ≤E≤ 100 GeV) domain, the AGILE (Tavani et al. 2009) and Fermi Large Area Telescope (LAT; Abdo et al. 2009, 2010a,b; Nolan et al. 2012) collaborations have reported on the detection of a source coincident with Eta Carinae (henceforth 2FGL J1045.0−5941). Recently Farnier, Walter & Leyder (2011) confirmed with the Fermi-LAT data the position of the HE γ-ray source and extracted an energy spectrum which features a low-energy and a high-energy component. The HE component extends up to ∼100 GeV, close to the energy threshold of the HESS telescope array. The AGILE collaboration reported on a two-day γ-ray flare from the direction of Eta Carinae which occurred in 2008 October. Although this increased γ-ray flux could not be confirmed by Farnier et al. (2011), Walter & Farnier (2011) found that the HE component flux shows a drop in the yearly light curve. Both these findings point to a possible origin of the HE γ-ray emission in the colliding wind region of Eta Carinae.
TeV J2032+4130 (Aharonian et al. 2002), HESS J1023−575 (Aharonian et al. 2007a) and the extended very high energy (VHE) γ-ray emission seen from the vicinity of Westerlund 1 (Abramowski et al. 2012) seem to indicate that VHE γ-ray emission can be linked to massive stars in our Galaxy and motivates an investigation of Eta Carinae and the Carina region as a whole as potential VHE γ-ray emitters. A further motivation comes from the detection of γ-ray emission from binary star systems such as LS 5039 (Aharonian et al. 2006c), PSR B1259–63 (Aharonian et al. 2005), LS I +61 303 (Albert et al. 2006) and the probable TeV binary HESS J0632+057 (Aharonian et al. 2007b; Bongiorno et al. 2011). Note that, unlike Eta Carinae these objects have a compact object (a neutron star or black hole) as stellar companion. Furthermore, the recent detection of HE γ-ray emission up to 100 GeV from the direction of Eta Carinae might hint at particle acceleration up to the VHE γ-ray regime in which HESS is operating.
2 HESS OBSERVATIONS
2.1 HESS experiment
HESS is an array of four VHE γ-ray imaging atmospheric Cherenkov telescopes (IACTs) located in the Khomas Highland of Namibia. Each of these telescopes is equipped with a tessellated spherical mirror of 107 m2 area and a camera comprising 960 photomultiplier (PMT) tubes, covering a large field of view (FoV) of 5° diameter. The system works in a coincidence mode, requiring at least two of the four telescopes to detect the same extended air shower. This stereoscopic approach results in an angular resolution of ∼6 arcmin per event, a good energy resolution (15 per cent on average) and an efficient rejection of the hadronic background (selection cuts retain less than 0.01 per cent of the cosmic rays (CRs); Benbow 2005). HESS has a point-source sensitivity of within 25 h of observations (Aharonian et al. 2006a). This flux level corresponds to a 1 per cent integral flux of the Crab nebula for energies E > 0.2 TeV, and detection threshold of 5σ (Li & Ma 1983). The more advanced data analysis method that is used in this work is discussed later, and achieves a significantly better point-source sensitivity (Ohm, van Eldik & Egberts 2009).
2.2 Data set
Observations of the (Sagittarius-) Carina arm tangent have been carried out as part of the HESS Galactic plane survey (Aharonian et al. 2006b, 2008a). Additionally, observations pointing in the direction of Eta Carinae have been performed in the so-called wobble-mode, where the telescopes were alternately pointed offset in RA and Dec. from Eta Carinae (Aharonian et al. 2006a). The Carina region and its surroundings were observed with the HESS array for a total of 62.4 h between 2004 and 2010. After standard data quality selection, where data taken under unstable weather conditions or with malfunctioning hardware have been excluded, the total exposure time after dead time correction of 3 to 4 telescope data is 33.1 h (Aharonian et al. 2006a). Due to Eta Carinae’s very southern position on the sky, observations have been carried out at moderate zenith angles of 36°–54°, with a mean value of 39°. The average pointing offset from Eta Carinae was 0°.8.
2.3 Data analysis
The available data have been analysed with the HESS Standard Analysis for shower reconstruction (Aharonian et al. 2006a) and the Hillas-based Boosted Decision Trees (BDT) method for an efficient suppression of the hadronic background component.1 This machine-learning algorithm returns a continuous variable (called ζ) that was used to select γ-ray-like events. Compared to the HESS Standard Analysis, a cut on this parameter results in an improvement in terms of sensitivity of ∼20 per cent for spectral and morphological analysis. For the generation of sky images, the spectral analysis and the production of light curves, the ζstd-cuts with a 60 photoelectron (p.e.) cut on the image intensity has been applied (see Ohm et al. 2009). The usage of this set of cuts leads to an energy threshold of 470 GeV for these observations. The 68 per cent containment radius of the HESS point spread function (PSF) for the analysis presented here is 6.7 arcmin.
In order to search for a γ-ray signal from Eta Carinae and the Carina nebula, two different background estimation techniques have been employed, i.e. the ring background and the reflected background model (Berge, Funk & Hinton 2007). The former has been applied to produce two-dimensional sky images, whereas the latter method has been used to derive spectral information and light curves. Table 1 summarizes the properties of the different data sets used in this work and the orbital phases of Eta Carinae which are covered by HESS observations. Note that throughout the paper the orbital phase is defined as phase angle with reference zero-time MJD 52822.492 corresponding to the periastron passage, and a period of 2022.7 d (Damineli et al. 2008).
|Data set||Date||MJD||Phase||Live time (h)|
|Data set||Date||MJD||Phase||Live time (h)|
Observations have been carried out over a time span of six years, during which the reflectivity of the HESS mirrors varied and the gains of the PMTs changed. The energy scale of the instrument is calibrated by looking at the response to single muons (Aharonian et al. 2006a).
Two different circular regions have been selected a priori and have been searched for a signal in the HESS data. Both of them are shown in Fig. 1 and are centred on the Eta Carinae position at RA and Dec. −59°41′04″.3 (J2000). Given the size of the Eta Carinae system of (1 arcmin), any VHE γ-ray signal would appear point-like to HESS (Region 1, 0°.112 radius). The Carina nebula, on the other hand, is a large and complex reflection nebula which shows extended emission seen in mid-infrared (mid-IR), optical and X-ray wavelengths on scales of ∼1°× 2°.5. The second circular region (Region 2, 0°.4 radius) has a physical scale of 16 pc at 2.3 kpc distance and has been chosen such that the bulk of the diffuse X-ray emission (Townsley et al. 2011) and potential particle acceleration sites such as the massive young stellar clusters Tr 14, and Tr 16 are encompassed. Region 2 also encloses most of the Hα (Smith, Bally & Walborn 2010) and m emission which traces gaseous and dusty material.
All results presented in the following have been successfully checked for consistency with an analysis chain that is based on a different shower reconstruction method and γ-ray selection criteria (de Naurois & Rolland 2009), and on a different calibration. During data taking, increased and variable single-telescope rates and, after quality selection, an increased but stable system trigger rate have been observed. This can be ascribed to the very high night-sky-background (NSB) level caused by the strong ultraviolet emission from the Carina nebula. This NSB level is higher than in any other HESS FoV from which results have been reported so far. Systematic tests have been performed and show that predominantly events which result in shower images with intensities below 60 p.e. are affected. However, the high NSB level does not affect the results presented here, since only events with image sizes greater than 60 p.e. are used. Moreover, the main analysis and the cross-check analysis – which models the NSB for shower reconstruction (de Naurois & Rolland 2009) – give consistent results.
2.4 VHE γ-ray results
Fig. 1 shows the VHE γ-ray significance map of the 2°× 2° region centred on the optical position of Eta Carinae, and calculated according to Li & Ma (1983). The map has been obtained with the ring background method and for an integration angle of 6.7 arcmin. No evidence for significant VHE γ-ray emission is found from Region 1 or from Region 2. Assuming a point-like source at the position of Eta Carinae (Region 1), a total of 40 ± 26 excess events with a significance of 1.6σ are found. Within Region 2, 197 ± 101 excess events with a significance of 2.0σ are detected.
Upper limits (ULs) for the VHE γ-ray emission from Eta Carinae and the extended region of 0°.4 radius which covers the inner parts of the Carina nebula have been produced. Fig. 2 shows the 99 per cent ULs (following Feldman & Cousins 1998) on the VHE γ-ray flux from Eta Carinae and the Carina region, assuming an underlying power-law distribution dN/dE=Φ0(E/1 TeV)−Γ with photon index Γ= 2.0. Adjusting the assumed spectral index to Γ= 2.5 changes the presented ULs by less than 2 per cent. Also shown is the HE γ-ray flux from the point-like source 2FGL J1045.0−5941, coincident with Eta Carinae, as detected by the LAT instrument on-board the Fermi satellite (Abdo et al. 2009, 2010b; Farnier et al. 2011). Above the energy threshold of 470 GeV, the derived 99 per cent integral flux ULs are for a point-like source at the position of Eta Carinae and for the extended Region 2.
The light curve of the binary system Eta Carinae shows variability in the optical (e.g. Damineli et al. 2000), IR (e.g. Whitelock et al. 2004), X-ray (Corcoran et al. 2010) and HE γ-ray band (Walter & Farnier 2011) on time-scales of months to years. In order to search for a possible variability in VHE γ rays on similar time-scales, the data collected during the HESS observations between 2004 and 2010 have been split into six different data sets accordingly (see Table 1). Since no VHE γ-ray signal could be found in any of these data sets, flux ULs have been derived for the covered time periods using the same assumptions as before. The statistics, energy thresholds and ULs are summarized in Table 2. Fig. 3 shows the HESS flux ULs (99 per cent confidence level) above 1 TeV at the different orbital phases of Eta Carinae. Also shown are the RXTE/ASM light curve and the INTEGRAL/IBIS data points in the X-ray domain as well as the AGILE and monthly Fermi-LAT light curve in HE γ rays.2
|Data set||On||Off||α||Excess||Significance (σ)||Eth (TeV)||F99(>Eth) (×10−12 photon cm−2 s−1)||F99(>1 TeV) (×10−12 photon cm−2 s−1)||Phase|
|Data set||On||Off||α||Excess||Significance (σ)||Eth (TeV)||F99(>Eth) (×10−12 photon cm−2 s−1)||F99(>1 TeV) (×10−12 photon cm−2 s−1)||Phase|
On denotes the number of γ-ray-like events from Region 1, Off denotes the number of γ-ray-like events from the background control regions, α is the normalization factor between the On and Off exposures, Eth is the energy threshold in TeV, and F99(>Eth) and F99(>1 TeV) are the 99 per cent flux ULs above Eth and 1 TeV, respectively, following Feldman & Cousins (1998).
3.1 Eta Carinae
The detection of point-like HE γ-ray emission from 2FGL J1045.0−5941 was originally reported in the three-month bright source list (Abdo et al. 2009) and was confirmed by Farnier et al. (2011) based on 21 months of data. The spectrum presented by Farnier et al. (2011) shows two distinct features: a low-energy component which is best fitted by a power law with index Γ= 1.69 ± 0.12 and exponential cut-off at 1.8 ± 0.5 GeV and a HE component which extends to ∼100 GeV and is well described by a simple power law with index 1.85 ± 0.25. If the HE γ-ray flux shown in Fig. 2 extended to the TeV regime, it would have been detectable in the HESS data presented in this work. The non-detection of a significant VHE γ-ray signal from Eta Carinae at any orbital phase and in the complete HESS data set has some interesting implications for the origin of the HE γ-ray emission which are discussed below.
Walter & Farnier (2011) showed that the flux of the HE component (E > 10 GeV) decreases by a factor of 2–3 in the yearly light curve, which could point to a scenario in which the parent particle population is accelerated in the colliding wind region of the binary system (Tavani et al. 2009; Bednarek & Pabich 2011; Farnier et al. 2011). However, the low-energy component does not seem to vary on yearly or monthly time-scales. For the colliding wind model, the lower energy component (0.2 GeV ≤E≤ 10 GeV) detected by the LAT is interpreted as inverse Compton (IC) γ-ray emission produced in interactions of the accelerated electrons with the dense stellar radiation fields of the binary stars. The hard HE γ-ray component can be interpreted in the colliding wind region model as either π0-decay γ rays, which are produced in proton–proton interactions in the dense stellar wind material (Bednarek & Pabich 2011; Farnier et al. 2011, their Model B) or as a second leptonic IC contribution (Model A in Bednarek & Pabich 2011). Interestingly, the HESS flux ULs for the individual subsets above the threshold energies of ∼0.5 TeV are all well below the extrapolated hard HE γ-ray component measured by Fermi-LAT (which is at a level of ∼1 × 10−11 erg cm−2 s−1).3 This implies that the γ radiation spectrum has a cut-off below ∼1 TeV, caused either by a cut-off in the accelerated particle spectrum or resulting from significant γ–γ absorption in the radiation field close to the two stars in the colliding wind region model. Bednarek & Pabich (2011) concluded that in the case of accelerated protons, the resulting π0-decay γ-ray emission should extend to TeV energies at phases far from periastron. The HESS data do not show γ-ray emission in the multi-TeV range at any orbital phase. Note, however, that the maximum detectable photon energy critically depends on the γ–γ absorption at the location where the photon is emitted, and on the alignment between the γ-ray production region, the star and the observer. For Eta Carinae, the optical depth for TeV particles becomes smaller than unity only at phases far from periastron, where the radiation field densities of both stars are low enough to allow γ rays to escape the system (see e.g. fig. 3 in Bednarek & Pabich 2011).
In an alternative scenario, particles are assumed to be accelerated in the outer blast wave which originates in the Great Eruption (Ohm, Hinton & Domainko 2010). For a potential non-variable hadronic HE γ-ray component, as discussed in Skilton et al. (2012), the maximum particle energy of the parent proton population is limited by three different parameters: the time since the giant outburst, i.e. 167 yr, the blast wave speed, which is measured as 3500–6000 km s−1 (Smith 2008) and the magnetic field, which is only poorly constrained. Contrary to the effects close to the wind–wind collision region, γ–γ absorption at the location of the blast wave has no significant effect on the γ-ray spectrum, given that the optical depth τγγ at this location is orders of magnitude smaller than in the colliding wind region. For the parameters used in Ohm et al. (2010),4 the maximum energy of protons producing γ rays of 0.5–1.0 TeV energy would be of (5 –10) TeV for magnetic field strengths in the blast wave of and a blast wave speed of 3500 km s−1. For this set of parameters, the HESS measurement excludes larger magnetic fields and/or higher blast wave speeds for this model.
3.2 Carina nebula
The Carina nebula harbours many potential particle acceleration sites such as massive binary systems (e.g. WR 25, Eta Carinae or the recently discovered HD 93250; Sana et al. 2011), young massive stellar clusters (e.g. Tr 14 and Tr 16) and possibly one or more SNR shells. Electrons and hadrons accelerated at these places would diffuse out of the acceleration region and interact with interstellar radiation fields and/or gaseous material, producing γ-ray emission via π0-decay or IC processes. Potential HE or VHE γ-ray emission could therefore trace the regions where a SNR shell interacts with high-density gas in MCs (as observed e.g. for W28; Aharonian et al. 2008b; Abdo et al. 2010). Additionally, low-energy CRs could be traced by ionization of MCs (see e.g. Ceccarelli et al. 2011).
Townsley et al. (2011) investigated the complex structure and composition of the diffuse X-ray emission in the Carina nebula with the Chandra satellite. The spectrum of this emission is phenomenologically best described by a multi-component model of different thermal plasmas in collisional ionization equilibrium and in a non-equilibrium ionization state. The X-ray emission does not seem to show any hint of a non-thermal component which would be indicative of particle acceleration in this region. Possible explanations for these observations are e.g. that currently no particle acceleration is taking place and hence non-thermal emission is not expected, or that the potential synchrotron emission has a much lower flux level than the efficient plasma emission, or the SNR shock has been diluted in the ambient plasma. If, however, particle acceleration occurred in the past at e.g. the shocks from one or more potential SNR shells, electrons might have cooled via synchrotron or IC radiation to a level not detectable by Chandra or below the HESS UL, respectively. Note that for a far-IR luminosity of LCar∼ 7 × 106 L⊙ (Salatino et al. 2012) and a circular region of 16 pc radius, the IC cooling time for 1 TeV electrons would be τIC∼ 6 × 103 yr.
CR hadrons on the other hand diffuse out of the acceleration region and interact with the gaseous or dusty material, producing π0-decay γ rays. The HESS ULs can be used to constrain the CR density enhancement factor κCR in units of the local CR density using equation (10) from Aharonian (1991), assuming that all the gas located in Region 2 is irradiated by CRs at the same time. Following Preibisch et al. (2011b) and Yonekura et al. (2005), the total gas and dust mass in Region 2 can be estimated to ∼1.5 × 105 M⊙. At a distance of 2.3 kpc this gives κCR= 23/f, where f is the fraction of the MC mass effectively irradiated by HE CRs. Assuming f= 1, this value can be compared to the CR enhancement factors obtained from the HESS detection of VHE γ-ray emission from W28 (Aharonian et al. 2008b). W28 is an old [(3.5–15) × 104 yr; Kaspi et al. 1993], mixed-morphology SNR, which is seen to interact with MCs belonging to the same massive star-forming region (e.g. Brogan et al. 2006). Aharonian et al. (2008b) derive κCR(W28) = 13–32 for clouds with masses (0.2–1.5) × 105 M⊙ and distances between 2 and 4 kpc. However, there is at present no evidence for a SNR in the Carina nebula, although SN explosions must have already occurred in the past (say ∼106 yr ago), in view of the presence of a neutron star. In that case, the lack of GeV–TeV emission from the nebula may have two explanations, separate or combined: (i) the factor f being ≪1 due to diffusive or advective transport of CRs in the region (too slow to fill the region or so fast that they escape), in which case the UL to kCR≫kCR(W28) and/or (ii) the p–p collision time-scale (for an average gas density of 100–400 cm−3 in the 50-pc region) is about 10 times less than the age of putative SNRs.
The search for VHE γ-ray emission from the colliding wind binary Eta Carinae and the most active Hii region in the Galaxy, the Carina nebula, has been presented. No sign of VHE γ-ray emission could be detected by HESS for Eta Carinae and a 99 per cent UL on the integral γ-ray flux of 7.7 × 10−13 photon cm−2 s−1 above 470 GeV has been derived using a 33-h data set collected over 6 years and covering the full phase range of the binary. Given the detection of a HE γ-ray component by Fermi-LAT, which extents up to ∼100 GeV, and assuming a spectral index of the HE Fermi-LAT component as found for the average spectrum by Farnier et al. (2011), the derived HESS ULs imply a cut-off in the γ-ray spectrum below a few hundred GeV. HESS observations did not reveal significant VHE γ-ray emission from the Carina nebula either. The derived ULs allow us to estimate the CR enhancement factor in this region (<23) which is at a comparable level to the values obtained for the W28 complex, assuming that CRs illuminate the whole cloud complex. HESS II, which adds a 600 m2 telescope to the existing system, will be operational during the next periastron passage in mid-2014 and will be sensitive to lower energies. Together with the future Cherenkov Telescope Array (CTA; Actis et al. 2011), with its greatly improved sensitivity and broader energy coverage, both instruments will close the gap between the HE and VHE γ-ray range and will allow us to probe the cut-off region in the γ-ray spectrum of Eta Carinae and to search for any variability in this system at VHEs.
We thank the referee R. Walter for his helpful comments and suggestions. The support of the Namibian authorities and of the University of Namibia in facilitating the construction and operation of HESS is gratefully acknowledged, as is the support by the German Ministry for Education and Research (BMBF), the Max Planck Society, the German Research Foundation (DFG), the French Ministry for Research, the CNRS-IN2P3 and the Astroparticle Interdisciplinary Programme of the CNRS, the UK Science and Technology Facilities Council (STFC), the IPNP of the Charles University, the Czech Science Foundation, the Polish Ministry of Science and Higher Education, the South African Department of Science and Technology and National Research Foundation, and by the University of Namibia. We appreciate the excellent work of the technical support staff in Berlin, Durham, Hamburg, Heidelberg, Palaiseau, Paris, Saclay and Namibia in the construction and operation of the equipment. SO acknowledges the support of the Humboldt Foundation by a Feodor-Lynen research fellowship.