© ESO, 2013

1. Introduction

The Planck¹ satellite has been surveying the millimetre sky since 2009. Its two instruments together cover nine frequency bands: the Low Frequency Instrument (LFI; Mandolesi et al. 2010; Bersanelli et al. 2010; Mennella et al. 2011) at 30, 44, and 70 GHz, and the High Frequency Instrument (HFI; Lamarre et al. 2010; Planck HFI Core Team 2011) at 100, 143, 217, 353, 545, and 857 GHz. Before the HFI coolant ran out in January 2012, Planck had successfully performed nearly 5 surveys of the entire sky.

Planck allows the detection of galaxy clusters by their imprint on the cosmic microwave background (CMB) via the Sunyaev-Zeldovich (SZ) effect, a characteristic spectral distortion of the CMB due to inverse Compton scattering of photons by hot electrons in the intra-cluster medium (Sunyaev & Zeldovich 1972). The SZ signal of galaxy clusters is expected to correlate tightly with cluster mass (e.g., da Silva et al. 2004) and its surface brightness is independent of redshift. SZ selected cluster samples are thus particularly well-suited for statistical studies of the galaxy cluster population, either as a probe of the physics of structure formation, or for cosmological studies based on cluster abundance as a function of mass and redshift. Compared to other SZ surveys, such as those with the Atacama Cosmology Telescope (ACT, Marriage et al. 2011) or the South Pole Telescope (SPT, Carlstrom et al. 2011), the Planck survey covers an exceptionally large volume; indeed, it is the first all-sky survey since the ROSAT All-Sky Survey (RASS) in the X-ray domain. Planck allows the detection of clusters below the flux limit of RASS based catalogues at redshifts typically greater than 0.3 (Planck Collaboration 2012, Fig. 9). The first Planck SZ catalogue, the Early SZ (ESZ) sample, was published in Planck Collaboration (2011a). It contains 189 clusters and candidates detected at high signal-to-noise ratio (S/N > 6) in the all-sky maps from the first ten months of observations, 20 of which were previously unknown. At the release of the ESZ sample, 12 of those 20 had been confirmed as new clusters, 11 using XMM-Newton validation observations undertaken in director’s discretionary time (DDT) via an agreement between the XMM-Newton and Planck project scientists.

All cluster surveys include false detections. For Planck, these are mainly due to inhomogeneous, non-isotropic, and highly non-Gaussian fluctuations (galactic dust emission, confusion noise as result of the unsubtracted point sources, etc.) in the complex microwave astrophysical sky. After identification of known clusters, a follow-up programme is required for cluster confirmation and redshift estimation. It is essential to build as pure as possible an initial candidate sample in order for such a programme to be efficient and manageable. For this we rely both on internal Planck candidate selection and assessment of the SZ signal quality, and on cross-correlation with ancillary data, as described in Planck Collaboration (2011a). Beyond simple confirmation of new clusters, the XMM-Newton validation programme aims to refine this validation process and to yield a better understanding of the new objects that Planck is detecting. It consists of snapshot exposures (~10 ks), sufficient for unambiguous discrimination between clusters and false candidates (Planck Collaboration 2011b), for a total allocated time of 500 ks for 50 candidates.

In the first two follow-up programmes, described by Planck Collaboration (2011b), we observed 25 candidates in total and helped to define the selection criteria for the ESZ sample. They yielded the confirmation of 17 single clusters, two double systems, and two triple systems². The observations showed that the new clusters are on average less X-ray-luminous and more morphologically disturbed than their X-ray-selected counterparts of similar mass, suggesting that Planck may be revealing a non-negligible population of massive, dynamically perturbed objects that are under-represented in X-ray surveys. However, despite their particular properties, the new clusters appear to follow the Y₅₀₀–Y_X relation established for X-ray selected objects, where Y_X, introduced by Kravtsov et al. (2006), is the product of the gas mass and temperature.

In the third follow-up programme, described in Planck Collaboration (2012), we observed 11 candidates with lower SZ detection levels (4.5 < S/N < 5.3) than the previous programmes (5.1 < S/N < 10.6) in order to investigate the internal SZ quality flags. Probing lower SZ flux than previous campaigns, the third programme also demonstrated the capability of Planck to find new clusters below the RASS limit and up to high z, including the blind detection at S/N ~ 5 of PLCK G266.6−27.3, confirmed by XMM-Newton to be an M₅₀₀ ~ 8 × 10¹⁴ M_⊙ cluster at z ~ 1 (Planck Collaboration 2011d). We also detected tentative evidence for Malmquist bias in the Y_SZ–Y_X relation, with a turnover at Y_SZ ~ 4 × 10^-4 arcmin².

In the fourth and last XMM-Newton validation programme, presented here, we further probe the low SZ flux, high redshift regime. The sample includes 15 candidates, detected at signal-to-noise ratios between 4 and 6.1 in the 15.5-month nominal survey data. We use the results from all XMM-Newton validation observations to address the use of ancillary RASS information as an indicator of candidate reliability (Sect. 5). The evolution of cluster SZ/X-ray properties is discussed in Sect. 6. This paper, together with Planck Collaboration (2011b,d, 2012), presents our complete analysis of the DDT XMM-Newton validation programme.

We adopt a ΛCDM cosmology with , Ω_M = 0.3, and Ω_Λ = 0.7. The factor is the ratio of the Hubble constant at redshift z to its present-day value. The quantities M₅₀₀ and R₅₀₀ are the total mass and radius corresponding to a total density contrast δ = 500, as compared to ρ_c(z), the critical density of the Universe at the cluster redshift; . The SZ flux is characterised by Y₅₀₀, where  is the spherically integrated Compton parameter within R₅₀₀, and D_A is the angular-diameter distance to the cluster. Thus, as defined here, Y₅₀₀ has units of solid angle and is given in arcmin² in Table 2.

2. Sample selection

2.1. Planck catalogue

In this paper, candidates were chosen from the catalogue derived from the first 15.5months of data (the “nominal” mission). The processing status, calibration, and map versions were those of March 2011. The detection and quality assessment of the cluster candidates followed the general procedure described in Planck Collaboration (2011a). Briefly, a blind cluster search was performed with three methods: the matched multi-frequency filter “MMF3” developed by Melin et al. (2006); an independent matched multi-frequency filter “MMF1”; and the PowellSnakes algorithm (PWS; Carvalho et al. 2009, 2012). Candidates then underwent internal SZ quality checks, removing spurious detections (e.g., association with artefacts or galactic sources), and assessment of the SZ signal detection. The signal assessment included quantitative criteria such as the signal-to-noise ratio and the number of methods blindly detecting the candidate, N_det, as well as a qualitative assessment based on visual inspection of the frequency maps, reconstructed SZ images, and the frequency spectra for each cluster. The latter procedure is summarised in an SZ quality grade, Q_SZ, as described in Planck Collaboration (2012).

Previously known clusters were identified via cross-correlation with catalogues and NED/Simbad queries. Possible counterparts were searched for within a 5′ radius of the Planck position, allowing us to assign two further external reliability flags:

• association of a FSC (Faint Source Catalogue) ora BSC (Bright Source Catalogue)RASS source (Vogeset al. 1999, 2000) oran excess of counts (with corresponding signal-to-noise ratio) inthe RASS [0.5−2] keV image;

• galaxy over-density in the Digitized Sky Survey (DSS) red plates³, from a visual check. In the Sloan Digital Sky Survey (SDSS) area⁴, two independent galaxy detection algorithms were applied to the DR7 galaxy catalogues (Fromenteau et al., in prep.; Li & White, in prep.). Both algorithms use photometric redshift information. Quality match criteria were assigned based on cluster richness or the over-density signal-to-noise ratio.

2.2. XMM-Newton target selection

The resulting targets are listed together with their SZ quality flags in Table 1. The range of signal-to-noise ratios, 4 < S/N < 6.1, is wide, with nearly uniform coverage, so that the validation results can be useful for defining the final signal-to-noise ratio for the Planck Cluster Catalogue. We considered lower signal-to-noise ratios than the previous validation programme, with 9 targets at S/N < 5 and a median S/N of 4.9, as compared to 5.1 previously (for 10.5 months of survey data). A priori, this allows us to reach lower flux or higher redshift. To further push the sample towards high redshift, we discarded candidates with estimated R₅₀₀ size greater than 5′. Although the large positional uncertainty of Planck candidates makes the search for a DSS counterpart non-trivial, the brightest galaxies of clusters at z < 0.5 are generally visible in DSS (e.g., Fassbender et al. 2011). We thus also used DSS images to select high-z clusters. Half the targets, labelled PHZ (potentially at high z) in Table 1, have no visible counterpart in DSS red plates. These are obviously riskier candidates, particularly those with low N_det or Q_SZ.

As previous validation observations have shown, the association of a SZ candidate with a RASS FSC or BSC source is not in itself sufficient to confirm the candidate, as chance association with a point source is always a possibility. Conversely, a candidate with no counterpart in the RASS catalogue may well be a bona fide cluster. With this campaign, in combination with the previous observations, we also aim to address the use of RASS data as an indicator of candidate reliability. In the sample of 36 candidates observed previously, thirteen candidates were associated with a BSC source and seventeen candidates with an FSC source. Only six SZ candidates had no FSC/BSC counterpart, of which the three confirmed candidates were detected in RASS at a signal to noise ratio of 1.7 < S/N < 2.8. To better span the range of external RASS flags, we chose ten candidates with no FSC or BSC counterpart, six of which correspond to a RASS S/N < 1.5. Of the remaining five candidates, only one is associated with a BSC source and four are associated with an FSC source. The RASS association for all XMM-Newton validation targets is summarised in Table 3.

Finally one candidate, PLCKG210.6 + 20.4, was specifically chosen to further test our SDSS-based confirmation of very poor SZ candidates. PLCKG210.6 + 20.4 is the lowest SZ signal-to-noise candidate, detected at S/N = 4 by one method only, with a Q_SZ = C grade and no significant signal in RASS data. However, the galaxy-detection algorithms (Sect. 2.1) that we used indicated that the candidate is associated with an SDSS galaxy over-density at z = 0.5.

3. XMM-Newton observations and data analysis

Candidates were observed between 31 July 2011 and 13 October 2011. The observation identification number and observation setup are given in Table 1. Due to a slew failure in the satellite revolution 2132, the PLCKG208.6−74.4 observation was incomplete, with an EPN exposure time of 3.4 ks. The target was observed initially at the end of its summer visibility window, and could only be reobserved five months later. It was replaced with an additional visible candidate, PLCKG329.5−22.7.

Calibrated event lists were produced with v11.0 of XMM-SAS. Data that were affected by periods of high background due to soft proton flares were omitted from the analysis (Pratt et al. 2007); clean observing time after flare removal is given in Table 1. The status of each SZ candidate is also given in Table 1: 12 of the 15 candidates are confirmed to be real clusters, among which two are double systems. XMM-Newton images of unconfirmed candidates are shown in Fig. 1; confirmed candidates are shown in Fig. 2.

We derived redshifts and physical parameters of the confirmed candidates as described in Planck Collaboration (2011a); Planck Collaboration (2012). Cleaned XMM-Newton data were pattern-selected. Each photon was then assigned a weight equivalent to the ratio of the effective area at the photon energy and position to the central effective area, computed with SAS task evigweight. Images and spectra were extracted using this weight, assuring full vignetting correction (see Arnaud et al. 2001). Bright point sources were excised from the data and the background was handled as described in Pratt et al. (2010). The particle-induced background (PB) was estimated using a stacked event list built from observations obtained with the filter wheel in closed position. The cosmic X-ray background was modeled using a PB-subtracted spectrum of an annular region external to the cluster emission.

In the spectroscopic analysis, the hydrogen column density was fixed at the 21-cm value of Kalberla et al. (2005). The redshift was estimated by fitting an absorbed redshifted thermal model to the spectrum extracted within a circular region corresponding to the maximum X-ray detection significance. The quality of the z estimate was characterised by the quality flag Q_z as introduced in Planck Collaboration (2011b). Q_z was set to Q_z = 0 when the redshift could not be constrained due to the lack of line detection. Q_z = 1 corresponds to ambiguous z_Fe estimate, when the spectral fit as a function of z exhibited several χ² minima that could not be distinguished at the 90% confidence level. Q_z = 2 corresponds to a well constrained redshift (i.e., a single χ² minimum).

Surface brightness profiles centred on the X-ray peak were extracted from  bins in the [0.3−2] keV band for each instrument independently, background subtracted, co-added and rebinned to 3σ per bin. 3D gas density profile were obtained using the regularised non-parametric method of direct deprojection and PSF deconvolution of the surface brightness profile developed by Croston et al. (2006). Global cluster parameters are estimated self-consistently within R₅₀₀ via iteration about the M₅₀₀–Y_X relation of Arnaud et al. (2010), assuming standard evolution, The quantity Y_X, is defined as the product of M_g,500, the gas mass within R₅₀₀, and T_X, the spectroscopic temperature measured in the [0.15−0.75]R₅₀₀ aperture. In addition, L₅₀₀, the X-ray luminosity inside R₅₀₀, is calculated as described in Pratt et al. (2009). The SZ flux was then re-extracted, Y₅₀₀ being calculated with the X-ray position and size R₅₀₀ fixed to the refined values derived from the high-quality XMM-Newton observation. The X-ray properties of the clusters and resulting refined Y₅₀₀ values are listed in Table 2.

4. XMM-Newton validation outcome

4.1. False cluster candidates

For the three candidates shown in Fig. 1, no obvious extended X-ray sources were found within 5′ of the Planck position. We followed the maximum likelihood procedure described by Planck Collaboration (2011a) to find all extended sources in the field detected at the  ≳3σ level. We then assessed whether they could be the counterpart of the Planck candidate from their position and X-ray flux, using the relation between the X-ray flux in the [0.1−2.4] keV band, F_X, and the SZ flux Y₅₀₀ established by Planck Collaboration (2012): (1)Here K = K(z) is the K correction, neglecting its temperature dependance.

4.1.1. PLCKG196.4–68.3 and PLCKG310.5+27.1

PLCKG196.4−68.3 was classified as PHZ (potentially at high z). Analysis of the XMM-Newton data on PLCKG196.4–68.3 revealed two extended sources at 9.′9 and 11.′8 from the SZ position. The former corresponds to a RASS-FSC source. Both sources are too far away to be the X-ray counterpart of the Planck candidate. A RASS-FSC source is located at 5.′2 from the SZ position and likely contributes to the S/N = 1.7 signal derived from RASS data at the Planck source location. However, the comparison of its surface brightness profile with the XMM-Newton PSF shows that it is consistent with a point source. We thus conclude that PLCKG196.4−68.3 is a false detection.

PLCKG310.5 + 27.1 was also classified as PHZ. Two extended X-ray sources were detected at 10.′5 and 2.′5 from the SZ position, respectively. The former is too far away to be the X-ray counterpart, while the latter is very weak. Analysis of the surface brightness profile confirmed that it is extended. The detection radius is small, and the spectrum extracted from this region is too poor to put robust constraints on the redshift or the temperature. However, using the F_X–Y₅₀₀ relation (Eq. (1)) and the measured X-ray flux, we can put an upper limit on Y₅₀₀ assuming a redshift as high as z = 2 and taking into account a factor of two dispersion around the relation. For a temperature of kT = 4 keV and z = 2, we derive a flux within the detection radius of F_X = 2.8 × 10^-14 ergs^-1cm^-2. Assuming that this flux is close to the total, this gives an upper limit on the SZ flux of Y₅₀₀ ~ 9 × 10^-5 arcmin², nearly an order of magnitude smaller than the Planck value Y₅₀₀ ~ 6.7 ± 1.5 × 10^-4 arcmin². Moreover, the SZ significance drops under 2σ when the flux is re-extracted at the X-ray position. We conclude that this candidate is also a false detection.

Both of these false candidates were detected by two methods, with a medium quality grade of Q_SZ = B and at S/N = 4.7 and S/N = 4.8, respectively. A Q_SZ = B quality grade is thus not sufficient to ensure candidate validity at these signal-to-noise ratios. On the other hand, all Q_SZ = A candidates down to S/N = 4.6 that have been followed up by XMM-Newton have been confirmed.

4.1.2. PLCKG210.6+20.4

PLCKG210.6 + 20.4 is associated with an SDSS cluster. The SDSS search algorithm identified a galaxy over-density of 77 members at a photometric redshift of z ~ 0.57, consistent with the spectroscopic redshift of the brightest cluster galaxy (BCG) at z = 0.52. The barycentre of the concentration and the BCG are located 1.′5 and 5′ from the Planck position (see Fig. 1), respectively. The X-ray analysis revealed the presence of an extended source, centred on the BCG, detected at 3.3σ in the [0.3−2] keV image. However, the source is very faint and more reminiscent of a group of galaxies than of a rich cluster. This is confirmed by the X-ray spectroscopic analysis. Extracting and fitting the spectrum with an absorbed thermal model at z = 0.52, we measured a temperature within the detection radius of T_Rdet = 1.5 ± 0.5 keV and a flux of F_X = 2.31 × 10^-14 ergs^-1cm^-2. Using Eq. (1) as above, the upper limit on the corresponding SZ flux is Y₅₀₀ ~ 2.6 × 10^-5 arcmin², more than 10 times lower than the Planck value of 4.9 ± 1.2 × 10^-4 arcmin². The X-ray source is too weak to be the Planck counterpart and we conclude that the candidate is not a cluster.

In the previous XMM-Newton validation run, the two candidates potentially associated with z > 0.5 SDSS clusters were confirmed, including PLCKG193.3−46.1 at z ~ 0.6. This showed that SDSS can robustly confirm candidates up to such high z. It is instructive to compare PLCKG210.6 + 20.4 with PLCKG193.3−46.1. In both cases the search algorithm found a rich concentration of galaxies, as expected for Planck clusters. The masses, reconstructed from the luminosity function, are ~3 × 10¹⁴ M_⊙ and 9 × 10¹⁴ M_⊙, respectively, i.e., the false candidate has a larger mass. In both cases, the galaxy distribution appears rather loose (compare Fig. 1 right panel and Planck Collaboration 2012, Fig. 5). The XMM-Newton observation revealed that PLCKG193.3−46.1 is a double peaked cluster, i.e., a dynamically perturbed cluster with an ICM distribution consistent with the galaxy morphology. In view of the XMM-Newton image, the galaxy concentration at the location of PLCKG210.6 + 20.4 is likely a filamentary structure where only the part around the BCG is virialised and contains gas that is hot enough to emit in X-rays. This would also explain the large offset between the BCG position and the galaxy concentration barycentre, which is much larger than in the case of PLCKG193.3−46.1. These two cases illustrate the difficulty of distinguishing between massive clusters and pre-virialised structures with rather shallow SDSS data at high z. Beyond luminosity and mass estimates, important diagnostics include the offset between the SZ, BCG, and barycentre, as well as the galaxy distribution morphology, if available, and other ancillary data, such as significant RASS emission. These factors must all be considered for firm confirmation of low signal-to-noise-ratio SZ detections. On the other hand, we cannot be sure that the apparent SZ signal is purely due to noise, and cannot exclude a contribution from the pre-virialised structure itself, especially if it corresponds to a warm filament along the line of sight.

4.2. Confirmed candidates

Twelve of the 15 candidates are confirmed as real clusters, of which two are double systems as shown in Fig. 2. Physical parameters are given in Table 2. For the two double systems, the cluster closest to the Planck position is labelled A and the other is labelled B in Table 2.

4.2.1. Single clusters

The redshifts of eight clusters are well constrained by the XMM-Newton spectrum (quality flag of Q_z = 2). Three of these clusters, PLCKG219.9−34.4, PLCKG011.2−40.4 and PLCKG348.4−25.5 were classified as PHZ. The first two are indeed at z = 0.46 and z = 0.66, respectively, but PLCKG348.4−25.5 is at z = 0.25. Knowing the precise cluster location with XMM-Newton, we re-examined the DSS image. A bright galaxy is indeed located exactly at the position of the X-ray peak; however, the field is crowded and there is no obvious galaxy concentration around that BCG. This explains our initial mis-classification.

The redshift determination for three single clusters is more uncertain. There are several χ² minima that cannot be distinguished at the 68% confidence level (Q_z = 1). As proposed by Planck Collaboration (2012), we used the X-ray versus SZ properties to eliminate unphysical solutions, as well as DSS data. This is detailed in Appendix A.1. The XMM-Newton analysis gives three possible redshifts for PLCKG147.3−16.6: 0.4, 0.62, and 1.1, the last being the best-fitting value. The cluster has an interesting double-peaked morphology. It is likely an on-going merger of two nearly equal mass systems (Fig. 3). The analysis of imaging data obtained with the Telescopio Nazionale Galileo La Palma (TNG) telescope and the Nordic Optical Telescope, as well as spectroscopic data obtained at Gemini, are detailed in Appendix A.2. We confirm a redshift of z = 0.66 ± 0.05.

The spectral analysis of PLCKG208.6−74.4 gives a single χ² minimum at z = 0.9 ± 0.04, in very good agreement with SZ versus X-ray properties. However we assign a quality flag of Q_z = 1 since the statistical quality of the spectrum is poor due to the short exposure time. Furthermore the DSS image is ambiguous: although there is no visible galaxy at the X-ray maximum, the centroid of the large scale X-ray emission is close to a bright DSS galaxy.

In summary, of the seven candidates we classified as PHZ, two are false, four are indeed at z ≳ 0.5, and one is at a low redshift of z = 0.25. In addition to those clusters which were classified as PHZ, two further Q_z = 1 clusters, PLCKG239.9−40.0 and PLCKG208.6−74.4, are most likely at high z.

4.2.2. Multiple systems

In PLCKG196.7−45.5, two clusters, separated by  ≈5.5 arcmin, lie within the Planck position error box: PLCKG196.7−45.5A at 2.′34 and PLCKG196.7−45.5B at 3.′9 from the SZ position. In view of the Planck resolution, 5′ to 30′ depending on frequency (Mennella et al. 2011; Planck HFI Core Team 2011), both clusters certainly contribute to the SZ signal. It is likely a chance association, although given the uncertainty in the redshifts, a binary system cannot be ruled out (see Appendix A.1).

In PLCKG329.5−22.7, the cluster PLCKG329.5−22.7A lies about 1′ from the Planck position, while the second object is about 8′ away. From the Y_X values and redshift estimates, cluster B is expected to have a Y₅₀₀ flux 1.8 times smaller than that of cluster A, thus contributing 36% to the total flux. Its contribution to the blind signal may differ, as the blind signal is extracted using a single component model found roughly peaked at cluster A. Indeed comparison of such a single component extraction with that using a double component model (with flux ratio fixed to the X-ray constraint) suggests a contamination from cluster B of about  ~20%. In summary, PLCKG329.5−22.7A is the main contributor to the SZ detection, although PLCKG329.5−22.7B certainly contributes. The redshifts of the two clusters are well determined, z = 0.24 and z = 0.46, respectively, showing that they are not physically related. This double system is thus a chance association on the sky.

Overall, we have found four double systems and two triple systems among the 43 Planck candidates confirmed by XMM-Newton, i.e., 14% multiple systems. Since the XMM-Newton validation follow-up observations are neither representative nor complete, this fraction of multiple systems cannot be extrapolated to the population at large; however, it is more than five times larger than the fraction of cluster pairs separated by less than 10′ (63/1882 objects) in the whole MCXC X-ray catalogue compilation (Piffaretti et al. 2011). This is clearly a selection effect due to confusion in the large Planck beam, which it might be necessary to take into account for a precise estimate of the selection function.

4.3. Planck position reconstruction uncertainty

The Planck position reconstruction uncertainty is driven by the spatial resolution of the instruments. The positions determined by the Planck detection algorithm are compared to the precise XMM-Newton positions in Figs. 4 and 5, where we put together all validation observations of single systems. The mean offset between the Planck and the XMM-Newton position is , with a median value of , as expected from Planck sky simulations (Planck Collaboration 2011a, Fig. 7 left). For 70 and 86% of the clusters, this offset is less than 2′ and , respectively. The assumed positional uncertainty of up to 5′ is certainly conservative and an offset of 5′ is actually very unlikely. This needs to be taken into account when searching for possible counterparts in ancillary data or follow-up observations.

The offsets of five sources are greater than . Three of those objects are very diffuse, likely dynamically unrelaxed systems, at relatively low z, including the prominent outlier PLCKG18.7+23.6 at z = 0.09 (Fig. 4, purple point). As noted by Planck Collaboration (2011b) a real, physical offset between the X-ray and SZ peak may contribute to the overall offset for this type of cluster. In all cases but one, the offset remains smaller than the cluster size R₅₀₀. The notable exception is PLCKG11.2−40.4 (Fig. 4). The XMM-Newton position of this cluster is or ~1.8R₅₀₀ from the Planck position. The peak in the SZ reconstructed map is also  ~3′ away from from the Planck position. This cluster is detected by only one method and has a low quality grade Q_SZ = C, being located in a particularly noisy region of the Planck map. This is likely to complicate the estimate of the cluster position.

Finally, we note that the position reconstruction uncertainty is on average smaller than for the ESZ sample that peaks at  ~2′ (Planck Collaboration 2011a, Fig. 7 right). This is likely the result of the higher redshift range considered here. Indeed, at this redshift the sources are more compact and their position is easier to reconstruct. Furthermore, possible physical offsets are expected to become negligible as they become unresolved.

4.4. New clusters in the z–L_X and z–M₅₀₀ plane and Planck sensitivity

The present validation sample covers a wide range of redshift, 0.2 < z < 0.9, and SZ flux, 2.9 × 10^-4 arcmin² < Y₅₀₀ < 8.8 × 10^-4 arcmin². As expected from the lower signal-to-noise ratio considered and the deeper sky coverage (Sect. 2), the Y₅₀₀ range is lower than that of the previous validation sample, 4 × 10^-4 arcmin² < Y₅₀₀ < 1.4 × 10^-3 arcmin². Although not perfect, the strategy to preferentially select high-z clusters was successful, with five clusters found at z > 0.5, including three PHZ candidates. The full XMM-Newton validation sample (single objects only) is shown in the L_X − z plane in Fig. 6. We continue to populate the higher z part of the L_X − z plane and confirm Planck can detect clusters well below the X-ray flux limit of RASS-based catalogues, ten times lower than REFLEX at high z, and below the limit of the most sensitive RASS survey (MACS). The figure makes obvious the gain in redshift coverage as compared to the RASS-based catalogues.

We confirm our previous results on the Y₅₀₀–Y_X relation. Most clusters are consistent with the REXCESS prediction: (2)with C_XSZ = 1.416 × 10^-19 Mpc² M_⊙^-1keV. However, all clusters below a normalised Y_X ~ 5 × 10^-4 arcmin² lie above the predicted Y₅₀₀–Y_X relation and the bin average deviation increases with decreasing Y_X (Fig. 7). As noted by Planck Collaboration (2012), this is an indication of Malmquist bias.

Figure 8 shows the new Planck clusters confirmed with XMM-Newton in the z–M₅₀₀ plane (single objects only). The minimum mass increases with redshift, an indication of an increase of the mass detection threshold with z. Such an increase is expected from the fact that clusters are not resolved by Planck at high z; however, we clearly confirm that Planck can detect M₅₀₀ > 5 × 10¹⁴ M_⊙ clusters above z > 0.5. Two clear outliers in the z − M plane are evident in Fig. 8. They correspond to the lowest flux clusters PLCKG11.2−40.4 and PLCKG268.5−28.1 at z = 0.46 and z = 0.47, respectively (Fig. 7), lying in the region most affected by the Malmquist bias. PLCKG11.2−40.4 is the cluster mentioned in Sect. 4.3, which is detected with a large offset between the Planck position and the X-ray peak, due to its lying in a region with a noisy background. The blind signal is two times higher than the signal extracted at the X-ray position. This is a clear case of a detection boosted by specific local noise conditions.

5. Using RASS data in the construction of the Planck cluster catalogue

5.1. Position refinement

The positions of the associated FSC and BSC source are indicated in the individual XMM-Newton image of each candidate in Fig. 2, and for previous observations, in Figs. 3 and 2 published in Planck Collaboration (2011b and 2012). Comparing the positions of the SZ candidates and their FSC/BSC counterparts with the X-ray peaks determined from the XMM-Newton validation observations, we notice that the FSC/BSC position is a better estimate of the position of the cluster than the position returned by Planck alone. Most of the FSC/BSC sources are located within 1′ of the XMM-Newton position versus 2′ for the Planck-SZ position (see Fig. 5). Thus, the association with a faint or bright RASS source can be used to refine the SZ position estimate.

5.2. X-ray flux estimate

Figure 9 summarises the comparison between RASS and XMM-Newton unabsorbed fluxes computed in the [0.1−2.4] keV band. The XMM-Newton flux is given in Table 3. Fluxes measured in an aperture of 5′ centred on the Planck candidate position from RASS images are referred to as “blind”. Here the RASS count rate is converted to flux assuming a typical redshift of z = 0.5, temperature of kT = 6 keV, and the 21–cm N_H value. All other fluxes are recomputed in an aperture corresponding to R₅₀₀, centred on the X-ray peak as determined from the XMM-Newton validation observations, and using the measured temperature and redshift to convert XMM-Newton or RASS count rates to flux.

These figures indicate that the RASS blind fluxes and the RASS fluxes measured within R₅₀₀ are in relatively good agreement, with a slight underestimate at high fluxes (left panel). RASS and XMM-Newton fluxes measured within R₅₀₀ are also in relatively good agreement, although with a slight underestimate together with increased dispersion at low fluxes (middle panel). As a result, RASS blind fluxes slightly underestimate the “true” XMM-Newton flux measured within R₅₀₀, by  ~30% at 10^-12 ergs^-1cm^-2. The underestimate increases with decreasing S/N (right panel).

In view of this agreement, we conclude that the RASS blind flux can be used to estimate the exposure time required for X-ray follow-up of a Planck candidate, once confirmed at other wavelengths. The main limitation is the statistical precision on the RASS estimate.

5.3. Candidate reliability

The association of an SZ candidate with a RASS-B/FSC source is neither a necessary nor a sufficient condition for an SZ candidate to be a bona fide cluster. Putting together the results from all XMM-Newton validation observations for a total of 51 Planck cluster candidates, we find that three of the eight false candidates are associated with an FSC source, while eleven candidates are confirmed without association with a RASS-FSC/BSC source. On the other hand, it is striking that PLCKG266.6−27.3, the most distant cluster of the sample, with a z = 0.97, is detected at a S/N > 5 in RASS, and is in fact found in the RASS Faint Source Catalogue.

5.3.1. RASS source density

It is important to underline that the RASS is not homogeneous, and that neither the BSC nor the FSC are flux-limited or complete in any way. Using the RASS-BSC and FSC, we computed the source density map of each catalogue and the associated probability that a Planck candidate will be associated with a B/FSC source within a radius of 5′. The method is described in Appendix B, and the resulting probabilities are given in Table B.1.

Figures 10 and B.1 show the RASS-FSC and BSC source density maps with all XMM-Newton validation observations overplotted. The faint source distribution directly reflects the RASS scanning strategy, as evident in Fig. 10. In this context, the probability of chance association is also an indication of how well covered the region is and thus on the depth of the X-ray observation at this position. We found a mean probability of association with an FSC source of % over the whole sky, where is the area corresponding to a circle of 5′ and  is the mean density at the position of the candidate, respectively. The corresponding mean probability of association with a BSC source is  ~1%. However, in the best-covered regions of the RASS the probability can reach 95% for the FSC and 9% for the BSC, while in the least-covered regions these probabilities drop to 0.4% and 0.2%, respectively.

5.3.2. BSC source association

All 12 candidates associated with a BSC source are confirmed. This is not surprising. For the BSC, the probability of chance association is relatively low, varying from less than 1% to 9%, depending on the sky region. For one cluster, PLCKG305.9−44.6, the XMM-Newton validation observation reveals that a point source is located at the position of the BSC source. However, the source is labelled as extended in the BSC, and in fact the X-ray emission likely corresponds to a blend of the point source and extended cluster emission that was not resolved with the large ROSAT PSF. This is supported by a comparison of the XMM-Newton and RASS images.

Thus we conclude that the correspondence of a Planck SZ candidate with a RASS-BSC source is a very good indication of there being a real cluster at this position.

5.3.3. FSC source association

For the FSC catalogue, on the contrary, the conclusion is more uncertain because of the larger probability of chance association. Most (18 of 21, i.e., more than 85%) of the candidates associated with a faint source are indeed confirmed. For the triple system PLCKG214.6+36.9, the FSC source is classified as extended. Its position as given in the RASS catalogue lies between the three clusters and is close to that of a bright XMM-Newton point source. The FSC source is thus in fact a blend of the cluster(s) and of the point source. In only one case, PLCKG193.3−46.1, does the FSC not correspond to the cluster emission. The XMM-Newton and RASS data shows that it is a point source located  away from the cluster centre.

Taking into account PLCKG193.3−46.1 and the three false candidates associated with an FSC source, we found four cases of mis-associations out of 51 candidates, i.e., 8%. This is consistent with the mean probability of chance association of 6% computed above; however, the association with an FSC source is still an indicator of reliability even in the regions of high probability of chance association. For instance, the two highest-redshift clusters (z ≈ 0.9) are correctly associated with a faint source, despite both being in the ecliptic pole region where the probability of false association is high. The scanning strategies of Planck and RASS are very similar in that both surveys are deeper in the same regions. In well-covered regions, the association with the faint source catalogue allows us to probe less massive or higher redshift potential clusters. A possible indicator of false association might be the distance between the FSC source and the SZ position, although no strict criterion can be applied. Seventy-five per cent of the false associations correspond to a distance greater than 3′, compared to 2 out of 16 (13%) for true associations.

5.3.4. No association

Sixteen candidates are not associated with a B/FSC source. Five of these candidates are false and eleven candidates are true sources with no B/FSC source association. As mentioned above, the association with a B/FSC is not necessary for an SZ candidate to be a bona fide cluster. However, we note that the median probability of FSC chance association, a measure of survey depth as discussed Sect. 5.3.1, is 2.1% for clusters without association, to be compared to 6.7% for associated clusters (see also Fig. 10). These true clusters with no B/FSC counterpart are located in the shallower part of the RASS survey, which likely explains why they are not associated.

5.3.5. RASS flux and signal-to-noise limit for candidate validation

Unassociated and associated candidates follow the same general correlation between the RASS blind flux, F_X, and the SZ flux, Y₅₀₀ (Fig. 11). This correlation presents some dispersion, with deviations from the mean as large as a factor of three. This is expected from the large statistical errors, as well as from the intrinsic dispersion and z dependence of the F_X/Y₅₀₀ ratio (Planck Collaboration 2012) and the difference between the blind and true X-ray fluxes (Sect. 5.2).

Because of this large dispersion, it is not possible to determine a strict RASS flux (or signal-to-noise ratio) limit below which a candidate should be discarded. However, we note that all new clusters have an X-ray flux greater than ~2 × 10^-13 ergs^-1cm^-2 (grey area in Fig. 11). This flux is consistent with the Y₅₀₀ threshold Y_500,thresh ≈ 2−5 × 10^-4 arcmin², as defined from the region affected by the Malmquist bias (see Fig. 7). This RASS flux limit is more than 10 times lower than the REFLEX flux limit of ~2 × 10^-12 ergs^-1cm^-2, but still detectable with RASS⁵. For the confirmed candidates, the minimum signal-to-noise ratio computed from RASS data is  ~0.70. Below that limit, all the candidates were false. All candidates with RASS S/N > 3 are confirmed, and only one false candidate is found for RASS S/N > 2. The latter is an SZ candidate detected at low PlanckS/N = 4.

5.3.6. RASS reliability flag

In view of the above results, we conclude the following regarding the most relevant RASS reliability flags:

• positional association of a Planck SZ candidatewith a RASS-BSC source is a very strong indication that thecandidate is a cluster;

• positional association of a Planck SZ candidate with a RASS-FSC source at S/N > 2 is a good indication of a real cluster;

• an SZ candidate with no signal at all in RASS is false at very high confidence. Obviously, candidates with low signal-to-noise ratio in a well-covered region are particularly likely to be false.

6. A preview of cluster evolution

With this new XMM-Newton validation campaign, we have now assembled a sample of 37 new single Planck clusters covering a redshift range 0.09 < z < 0.97. With only snapshot XMM-Newton observations, the global properties and density profile of each object are measured accurately enough to allow a first assessment of evolution with redshift. The structural and scaling properties of the sample are illustrated in Fig. 12. We considered three redshift bins, z < 0.3 (10 clusters), 0.3 < z < 0.5 (19 clusters) and z > 0.5 (8 clusters). We confirm our previous finding regarding the scaling properties of these new Planck selected clusters, and do not find any evidence of departure from standard self-similar evolution.

The average scaled density profile (top left panel of Fig. 12) is similar for each z bin and is flatter than that of REXCESS, a representative sample of X-ray selected clusters (Arnaud et al. 2010). Once scaled as expected from standard evolution, the new clusters in each redshift bin follow the same trends in scaling properties (Fig. 12): they are on average less luminous at a given Y₅₀₀, or more massive at a given luminosity, than X-ray selected clusters. On the other hand, they follow the Y₅₀₀–Y_X relation predicted from REXCESS data (Eq. (2)).

To study possible evolution with z, we plot in Fig. 13 the ratio as function of z, including the 62 clusters of the Planck-ESZ sample with XMM-Newton archival data (Planck Collaboration 2011c). We exclude clusters at low flux, , to minimise possible Malmquist bias (see Sect. 4.4). The best fitting power law gives a slope α = 0.043 ± 0.036, with a normalisation of 0.97 ± 0.03 at z = 0.2. The relation is thus consistent with a constant ratio at the REXCESS value of 0.924 ± 0.004. A histogram of the ratio shows a peak exactly at the REXCESS position. The distribution is skewed towards high ratios, the skewness decreasing if low flux clusters are excluded. This skewness might be intrinsic to the cluster population. It might also reflect a residual effect of the Malmquist bias, clusters with intrinsic high Y₅₀₀/Y_X ratio being preferentially detected in SZ surveys.

7. Conclusions

We have presented results on the final 15 Planck galaxy cluster candidates observed as part of a 500 ks validation programme undertaken in XMM-Newton Director’s Discretionary Time. The sample was derived from blind detections in the full 15.5-month nominal Planck survey, and includes candidates detected at 4.0 < S/N < 6.1. External flags including RASS and DSS detection were used to push the sampling strategy into the low-flux, high-redshift regime and to better assess the use of RASS data for candidate validation. This last phase of the follow-up programme yielded 14 clusters from 12 Planck candidate detections (two candidates are double systems) with redshifts between 0.2 and 0.9, with six clusters at z > 0.5. Their masses, estimated using the M₅₀₀–Y_X relation, range from 2.5 × 10¹⁴ to 8 × 10¹⁴ M_⊙. We found an interesting double peaked cluster, PLCKG147.3−16.6, that is likely an ongoing major merger of two systems of equal mass. Optical observations with NOT, TNG, and Gemini confirmed a redshift of 0.65.

The full XMM-Newton validation follow-up programme detailed in this paper and in Planck Collaboration (2011b); Planck Collaboration (2012) comprises 51 observations of Planck cluster candidates. The efficiency of validation with XMM-Newton stems both from its high sensitivity, allowing easy detection of clusters in the Planck mass and redshift range, and from the tight relation between X-ray and SZ properties, which probe the same medium. The search for extended XMM-Newton emission and a consistency check between the X-ray and SZ flux is then sufficient for unambiguous discrimination between clusters and false candidates. We have confirmed the relation between the X-ray flux and the SZ flux, as a function of redshift, and estimated its typical scatter. This relation is used in the validation procedure. By contrast, optical validation is hampered by the relatively large Planck source position uncertainty and the large scatter between the optical observables (such as galaxy number) and the mass (or SZ signal), both of which increase the chance of false associations.

The programme yielded 51 bona fide newly-discovered clusters, including four double systems and two triple systems. There are eight false candidates. Thirty-two of the 51 individual clusters have high quality redshift measurements from the Fe K line. For other cases, the spectral fitting yields several χ² minima as a function of z, that cannot be distinguished at the 68% confidence level. We showed that the relation between the X-ray and SZ properties can be used to further constrain the redshift. The new clusters span the redshift range 0.09 to 0.97 and cover more than one decade in Y₅₀₀, from 2.9 × 10^-4 to 3.0 × 10^-3 arcmin². M₅₀₀ of single systems is in the range (2.5 × 10¹⁴−1.6 × 10¹⁵) M_⊙. These observations provided a first characterisation of the new objects that Planck is detecting:

• The newly-detected clusters follow the Y_X–Y₅₀₀ relation derived from X-ray selected samples. This is consistent with the prediction that both quantities are tightly related to the cluster mass.

• New SZ selected clusters are X-ray underluminous on average compared to X-ray selected clusters, and more morphologically disturbed. The dispersion around the M–L_X relation may be larger than previously thought and dynamically perturbed (merging) clusters might be under-represented in X-ray surveys. This has implications for statistical studies of X-ray selected samples, either to constrain cosmological models from cluster number counts or to probe the physics of structure formation from the cluster scaling properties. As discussed in detail by Angulo et al. (2012), precise knowledge of the actual scatter between the mass and the observable used in the detection is critical in both applications.

• We found no indication of departure from standard self-similar evolution in the X-ray versus SZ scaling properties. In particular, there is no significant evolution of the Y_X/Y₅₀₀ ratio.

Beyond new cluster confirmation and characterisation, we checked the pertinence of the validation process based on Planck internal quality assessments and cross-correlation with ancilliary data. There are eight false candidates in total, all of which were found at S/N < 5. These failures underline the importance of the number of methods detecting the clusters and were used to refine our internal quality flag definitions. All candidates with Q_SZ = A are confirmed. Galaxy overdensity in SDSS data can confirm candidates up to z ~ 0.6, although it remains difficult to distinguish between massive clusters and pre-virialised structures at high z. The quality of the SZ detection, ancillary data such as significant RASS emission, and the offsets between SZ, BCG, and other positions, must all be considered for firm confirmation. Using the full sample of 51 observations, we investigated the use of RASS-based catalogues and maps for Planck catalogue construction, finding that:

• Planck clusters appear almost always to be detectable in RASS maps, although there is not a one-to-one correspondence between a RASS-BSC or FSC source and the presence of a cluster.

• Association of a cluster candidate with a RASS-BSC source is a very strong indication that it is a real cluster.

• Whether or not there is a RASS-BSC or FSC source, S/N > 2 in the RASS maps is a good indication of a true candidate, while S/N < 0 is a good indication of a false candidate.

• The association with a faint or bright RASS source can be used to refine the SZ position estimate. The RASS blind flux can be used to estimate the exposure time required for X-ray follow-up of a Planck candidate, once confirmed at other wavelengths. The main limitation is the statistical precision on the RASS estimate.

The XMM-Newton validation observations could also be used for the verification of Planck performances, showing that:

• The mean offset between the Planck position and the cluster position is , as expected from Planck sky simulations, and this offset is less than for 86% of the clusters.

• Planck can detect clusters well below the X-ray flux limit of RASS based catalogues, ten times lower than REFLEX at high z, and below the limit of the most sensitive RASS survey (MACS).

• The Planck sensitivity threshold for the nominal survey is Y₅₀₀ ~ 4 × 10^-4 arcmin², with an indication of Malmquist bias in the Y_X–Y₅₀₀ relation below this threshold. The corresponding mass threshold depends on redshift, but Planck can detect systems with M₅₀₀ > 5 × 10¹⁴ M_⊙ at z > 0.5.

• Overall, there is a high fraction of double/triple systems in the XMM-Newton validation follow-up sample, illustrating the problems of confusion in the Planck beam.

These results illustrate the potential of the all-sky Planck survey to detect the most massive clusters in the Universe. Their characterisation, and the determination of their detailed physical properties, depends on a vigorous follow-up programme, which we are currently undertaking.

---

Acknowledgments

The Planck Collaboration thanks Norbert Schartel for his support of the validation process and for granting discretionary time for the observation of Planck cluster candidates. The present work is based on observations obtained with XMM-Newton, an ESA science mission with instruments and contributions directly funded by ESA Member States and the USA (NASA). This research has made use of the following databases: SIMBAD, operated at the CDS, Strasbourg, France; the NED database, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration; BAX, which is operated by IRAP, under contract with the Centre National d’Études Spatiales (CNES); and the SZ repository operated by IAS Data and Operation centre (IDOC) under contract with CNES. Based on photographic data obtained using The UK Schmidt Telescope. We further used observations made with the Italian Telescopio Nazionale Galileo (TNG) operated on the island of La Palma by the Fundación Galileo Galilei of the INAF (Istituto Nazionale di Astrofisica) at the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofisica de Canarias (Science Program ID AOT24/11-A24DDT3), on observations made with the Nordic Optical Telescope, operated on the island of La Palma jointly by Denmark, Finland, Iceland, Norway, and Sweden, in the Spanish Observatorio del Roque de los Muchachos of the Instituto de Astrofisica de Canarias (Science Program ID 43-016), observations obtained at the Gemini Observatory, which is operated by the Association of Universities for Research in Astronomy, Inc., under a cooperative agreement with the NSF on behalf of the Gemini partnership: the National Science Foundation (United States), the Science and Technology Facilities Council (United Kingdom), the National Research Council (Canada), CONICYT (Chile), the Australian Research Council (Australia), Ministério da Ciência e Tecnologia (Brazil) and Ministerio de Ciencia, Tecnologìa e Innovaciòn Productiva (Argentina), Gemini Science Program ID: GN-2011B-Q-41. The authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of Mauna Kea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain. A description of the Planck Collaboration and a list of its members, indicating which technical or scientific activities they have been involved in, can be found at http://www.rssd.esa.int/Planck_Collaboration. The Planck Collaboration acknowledges the support of: ESA; CNES and CNRS/INSU-IN2P3-INP (France); ASI, CNR, and INAF (Italy); NASA and DoE (USA); STFC and UKSA (UK); CSIC, MICINN and JA (Spain); Tekes, AoF and CSC (Finland); DLR and MPG (Germany); CSA (Canada); DTU Space (Denmark); SER/SSO (Switzerland); RCN (Norway); SFI (Ireland); FCT/MCTES (Portugal); and DEISA (EU).

References

Appendix A: Redshift estimates of confirmed candidates

A.1. Refinement of the XMM-Newton redshift estimate for Q_z =  1 cases

The redshift determination from XMM-Newton spectral analysis is uncertain for five clusters. There are several χ² minima that cannot be distinguished at the 90% confidence level (Q_z = 1). As proposed by Planck Collaboration (2012), we estimated the Y_X/Y₅₀₀ and F_X/Y₅₀₀ ratios as a function of z and compared them to expected values, to eliminate unphysical solutions.

Three possible redshifts were found for PLCKG352.1−24.0, 0.12, 0.4, and 0.77. The Y_X/Y₅₀₀ ratio method enables us to exclude the low redshift z = 0.12 solution. The z = 0.4 solution yields a Y_X/Y₅₀₀ ratio twice higher than expected, at the limit of the observed dispersion. Furthermore, we confirmed that there is no evidence of galaxy concentrations in the DSS red image at the precise XMM-Newton cluster location. We thus adopt the highest z value, z = 0.77, confirming the cluster to be at high z.

The best fitting redshift for PLCKG239.9−40.0, z = 0.74, yields the Y_X/Y₅₀₀ ratio closest to expectation and is adopted in the further analysis. The lowest z = 0.26 solution is very unlikely, yielding a Y_X/Y₅₀₀ ratio twice as high as expected. The other possible solution is z = 0.46: there are some very faint objects in the DSS images at the XMM-Newton position, although whether those are galaxies is unclear.

In the case of PLCKG147.3−16.6, all three redshift solutions, 0.4, 0.62, and 1.03, yield a Y_X/Y₅₀₀ ratio within the observed dispersion. The best fitting value, z = 1.03, and the second best solution, z = 0.62, are consistent at the 90% confidence level, with χ² values of 125.9 and 128.7 for 132 degree of freedom, respectively. The two models are shown in Fig. A.1. The optical measurement is described below (Sect. A.2).

The redshifts of the two components in PLCKG196.7−45.5 are uncertain. The Y_X/Y₅₀₀ and F_X/Y₅₀₀ ratio methods cannot be used for such double systems, since the individual SZ components are unresolved by Planck. Of the two solutions, z = 0.57 and z = 0.87 for PLCKG196.7−45.5A, the latter can be excluded: a clear concentration galaxies at the XMM-Newton location is visible in the DSS images, which thus cannot be at such high z (see Sect. 2.2). For PLCKG196.7−45.5B we adopted the best fitting value, z = 0.42.

A.2. Optical redshift estimate of PLCKG147.3–16.6

The optical data for PLCKG147.3−16.6 were taken using Director’s Discretionary Time with DOLORES (Device Optimized for the LOw RESolution), a low resolution spectrograph and imager permanently installed at the TNG telescope (Telescopio Nazionale Galileo La Palma). The camera is equipped with a 2048 × 2048 pixel CCD covering a field of view of (pixel scale of 0.252 per pixel). Exposure times of 3000 s in the r and i bands were split into 10 single exposures of 300 s each. Exposure times of 4000 s in the z band were split into eight separate exposures. Taking advantage of the dither-offsets between single exposures, no separate sky images were required. The images were bias and flat field corrected using IRAF⁶. For astrometric calibration we used astrometry.net. The average seeing derived from the final images is 0.84, 0.85, and 0.84 in the r, i, and z-bands, respectively. In the final images, we reach signal-to-noise ratios (over the PSF area) of 11, 23, and 8 for unresolved sources of 24th magnitude. The colour composite image allows us to pre-identify the cluster members.

The cluster was also observed using the 2.56-m Nordic Optical Telescope with the MOSCA camera, a 2 × 2 mosaic of 2048 × 2048 pixel CCDs. This camera covers a total field of 7.′7  ×  7.′7, and was used in 2 × 2 binned mode. This gives a pixel scale of 0217 per binned pixel. Total exposure times of 900 s were split into 3 dithered exposures of 300 s in each of the SDSS g- and i-bands in photometric conditions. The telescope was pointed such that the two peaks of the X-ray emission from the cluster would fall in the centreof the mosaic CCD chip that has the best cosmetic quality (named “CCD7”). After standard basic reduction and image registration, the combined images had FWHM of and in the g and i bands, respectively. Photometric calibration was based on an ensemble of stars in a field located inside the SDSS footprint, observed at similar airmass immediately following the observations of PLCKG147.3−16.6. Stellar objects were removed from the object catalogues based on their location in a size-magnitude diagram. A strong clustering of galaxies with red g − i colours was immediately detected around the position of the X-ray peaks. The colour-magnitude diagram in Fig. A.2 illustrates the red sequence formed by early-type galaxies at g − i ≃ 3.15 in this cluster. Predicted g − i colours of early-type galaxies as a function of redshift were calculated by convolving the EO template galaxy spectrum of Coleman et al. (1980) with the response curves of the SDSS g and i bandpasses. From this, a photometric redshift estimate of z_phot = 0.64 ± 0.03 was derived.

The calibrated g- and i-band photometry from NOT was used to select suitable spectroscopic targets for Gemini North Telescope by choosing galaxies at g − i ≃ 3.15. The observations (Program GN-2011B-Q-41) were made with GMOS-N, with two exposures of 1800 s each. The program was in Band 2 service mode, with relaxed observing conditions: the seeing was 17 the first night and 08 the second night, with cirrus both nights. The observations were reduced with the standard Gemini IRAF package. We obtained redshift measurements for 13 objects. Among those, 10 have redshifts between 0.64 and 0.68, for a cluster redshift measurement of 0.66 ± 0.05. If we exclude two objects at z = 0.68, we obtain z = 0.645 ± 0.005.

Appendix B: Density maps of RASS bright and faint sources

In this appendix we describe the procedure used to calculate the density maps of RASS-BSC and FSC sources, and the associated probability of false association with a Planck cluster candidate. We use the catalogues downloaded from Vizier⁷.

B.1. Source density maps

To compute the source density maps, we use HEALPix⁸ with a resolution of N_side = 64 (each pixel is 0.8 deg²). The HEALPix function ANG2PIX_RING was used to compute the pixel number corresponding to the coordinates of the FSC/BSC sources.

At each pixel, we compute the source density by summing the number of sources in the pixels inside a disc of increasing radius until a threshold number of 10 sources is reached. The source density is then the number of sources found, N_src, divided by the number of pixels, N_pix, normalised by the area covered by one pixel: (B.1)where 49152 is the total number of sky pixels for this resolution and 4π(180/π)² ≈ 41000 deg² is the total area of the sky. This gives the mean number of sources per square degree in each pixel.

The resulting source density maps are plotted in Figs. 10 and B.1. For the FSC density map, the mean source density per square degree ranges from 0.16 to 42.89. There is a clear correspondence between the source density and the depth of the RASS exposure, with regions of maximum source density lying in the regions of maximum RASS exposure at the ecliptic poles (Fig. 10). For the BSC density map, the mean source density per square degree ranges from 0.08 to 4.05, with a much less marked correspondence with the RASS exposure map (Fig. 10).

Figure B.2 shows the histogram of the number of pixels as a function of mean source density per square degree. We overplot on these histograms the mean  and the median (ρ_1/2) value of the number of sources per square degree. We find for the FSC and for the BSC.

B.2. Probability of association within search radius ℛ

We can convert the local FSC/BSC source densities into probabilities of chance association of an SZ candidate with a FSC/BSC source. The probability of finding a cataloged FSC/BSC source within a search radius ℛ of a Planck cluster candidate is the product of the FSC/BSC source density at the candidate location by the search area, . This yields a mean probability of association of an SZ candidate with a B/FSC over the full sky of for the FSC and 1% for the BSC. However, there is considerable variation depending on how well a given sky region is covered. In the most covered regions, the probability reaches nearly 95% of having an association within 5′ for the FSC and 9% for the BSC, while it decreases to 0.4% and 0.2% for the less covered regions for the FSC and BSC catalogues, respectively. We summarise these numbers in Table B.1.

All Tables

All Figures