© ESO, 2012

1. Introduction

The deep potential wells in clusters of galaxies make them unique laboratories in which to study astrophysical processes linked to gas physics, galaxy formation, and feedback. Furthermore, since clusters trace the highest peaks of the matter density field, the properties of the cluster population and their evolution are a sensitive cosmological probe.

The recent advent of increased sensitivity and survey capability has transformed galaxy cluster searches via the Sunyaev-Zeldovich (SZ) effect. Such surveys identify objects using the spectral distortion of the cosmic microwave background (CMB) generated through inverse Compton scattering of CMB photons by the hot electrons in the intra-cluster medium (Sunyaev & Zeldovich 1972). Crucially, the total SZ signal is expected to be closely related to the cluster mass (e.g., da Silva et al. 2004), and its surface brightness insensitive to distance. As a result, SZ surveys can potentially provide unbiased cluster samples over a wide range of redshifts that are as close as possible to being mass-selected. Such samples are essential for understanding the statistical properties of the cluster population and for its exploitation in cosmological studies. Examples of on-going cluster surveys in the SZ include the Atacama Cosmology Telescope (ACT Marriage et al. 2011), Planck¹ (Planck Collaboration 2011c) and the South Pole Telescope (SPT Carlstrom et al. 2011).

The Planck satellite has been surveying the sky in the microwave band since August 2009 (Planck Collaboration 2011a). Compared to other SZ surveys, Planck has only modest (band-dependent) spatial resolution of 5′ to 30′ (Mennella et al. 2011; Planck HFI Core Team 2011) but it possesses unique nine-band coverage from 30 to 857 GHz and, most crucially, it covers an exceptionally large survey volume. Indeed Planck is the first all-sky survey capable of blind cluster detections since the ROSAT All-Sky Survey (RASS, in the X-ray domain). Early Planck results on galaxy clusters were recently published in Planck Collaboration (2011b,c,g,d,e,f). These results include the publication of the high signal-to-noise ratio (S/N > 6) Early SZ (ESZ) cluster sample (Planck Collaboration 2011b).

The raw data product of any cluster survey is a list of potential candidates. Such a list is expected to include a fraction of false detections, e.g., for SZ detections, due to fluctuations in the complex microwave astrophysical sky. In the case of Planck, the moderate spatial resolution at SZ frequencies with respect to typical cluster sizes presents a further complication. A Planck cluster SZ measurement essentially provides only a position, a total SZ flux, and a coarse size estimate. In addition, the quality of the SZ flux estimates is degraded by the flux-size degeneracy, as discussed in Planck Collaboration (2011b). A follow-up programme is therefore required to scientifically exploit Planck candidate data. Such a programme should provide candidate confirmation, which is the final part of the catalogue validation, and a redshift measurement, the prerequisite to any cluster physical parameter estimate.

In this context, X-ray observations are extremely useful, as has been shown by the results from the initial validation follow-up of Planck cluster candidates with XMM-Newton (Planck Collaboration 2011c). These observations were undertaken in Director’s Discretionary Time via an agreement between the XMM-Newton and Planck Project Scientists. A pilot programme observed ten targets to refine the selection criteria for the ESZ cluster sample. A second programme focused on the validation of fifteen high-significance SZ sources (S/N > 5); eleven of the newly-discovered clusters from this programme are contained in the ESZ sample. These first observations provided a preview of the X-ray properties of the newly-discovered clusters (Planck Collaboration 2011c). In particular it was confirmed that, based on the detection of extended emission, XMM-Newton snapshot exposures (10 ks) are sufficient for unambiguous discrimination between clusters and false candidates for redshifts at least up to z = 1.5. In addition, it was shown that the spurious association of candidates with faint extended sources lying within the Planck position uncertainty (which can be up to 5′) can be identified via a consistency check between the X-ray and SZ flux. This latter constraint stems from the tight correlation between X-ray and SZ properties, since X-rays probe the same medium as the SZE. In this respect, X-ray validation presents a clear advantage over optical validation for Planck candidates. While optical observations offer important complementary information on the stellar component of clusters and on mass estimates derived from gravitational lensing of background sources, optical validation is hampered by the relatively large Planck source position uncertainty and the large scatter between simple optical observables (such as galaxy numbers) and the mass (or SZ signal), both of which increase the chance of false associations.

Fig. 1 Illustration of the three SZ quality grades as defined in Sect. 2. From left to right the three quality cases are: QSZ = A,B,C. The top row shows a 100′ × 100′ SZ map with a spatial resolution of 10′, centred on the candidate position, derived using the MILCA reconstruction method (Hurier et al. 2010). The colour table is identical for all clusters, with the Compton y parameter spanning the range  [ − 3 × 10-6,1 × 10-5] . The bottom row shows the associated SZ spectrum from aperture photometry measurements within R500 (see text for details). The red line is the SZ spectrum normalised to the Y500 value obtained from MMF blind detection.

A manageable confirmation programme for the compilation of a larger, final, cluster catalogue from the Planck survey requires a candidate sample with a high ratio of true clusters to total candidates (i.e., purity). The construction of such a sample relies both on Planck internal candidate selection and assessment of the SZ signal quality and also on cross-correlation with ancillary data and catalogues, as described in Planck Collaboration (2011b). In the present paper, in which we report on a further eleven XMM-Newton observations of Planck cluster candidates detected at 4.5 < S/N < 5.3, we address in more detail the internal quality assessment of cluster candidates in SZ. XMM-Newton validation, allowing unambiguous discrimination between clusters and false candidates, is essential for such a study.

X-ray observations can also constrain the redshift of the source through Fe K line spectroscopy, as demonstrated in Planck Collaboration (2011c). Here we also present new optical redshift determinations for XMM-Newton confirmed candidates, which we compare to the X-ray-derived values. We also discuss whether, in the absence of optical follow-up data, a combined X-ray/SZ analysis can improve the z estimate when X-ray data alone are insufficient to unambiguously determine the redshift.

We adopt a ΛCDM cosmology with H₀ = 70  km s^-1 Mpc^-1, 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.

2. Sample selection

The present candidates were chosen from the catalogue of detections in the all-sky maps from the first ten months of the Planck survey. This same catalogue was used for the construction of the ESZ sample, for which the reference method for blind cluster searches was the matched multi-frequency filter “MMF3”, developed by Melin et al. (2006). Complementary searches were also performed with an independent implementation of the MMF method and with the PowellSnakes algorithm (PWS; Carvalho et al. 2009, 2011). As described in Planck Collaboration (2011b), candidates then underwent a validation process, including internal SZ quality checks. The first part of this process included an initial quantitative assessment of the blind SZ signal detection, based on the S/N and the number of methods blindly detecting the candidate, N_det.

The quality of the SZ signal cannot simply be reduced to a single global S/N value. It depends not only on the intrinsic cluster SZ signal, but also on the detailed local properties of the various noise components, i.e., the background (e.g., CIB, CMB) and foreground environments (e.g., galactic dust, synchrotron, free-free emissions). Therefore, beyond the quantitative criteria stated above, we also performed a qualitative assessment of the SZ signal based on visual inspection of SZ maps and spectra.

We first examined frequency maps, using both raw maps made directly from the Planck all sky data, and maps that had been cleaned of dust emission. We used IRIS-100 μm (Miville-Deschênes & Lagache 2005) and Planck HFI-857 GHz maps as dust templates, and the “dust-cleaned” HFI-217 GHz map as a CMB template. These frequency maps were investigated for strong foreground dust contamination and the presence of submillimetre sources on the high frequency side. Radio source contamination and CMB residuals were searched for at low frequencies. In addition to the frequency maps, reconstructed SZ maps were built using three different reconstruction methods based on Independent Linear Component (ILC) analysis (e.g., Hurier et al. 2010). Finally, SZ spectra were built from the SZ flux estimation at each Planck frequency. Spectra were estimated both from the best detection outputs and also directly from aperture photometry on CMB- and dust-cleaned maps.

On the basis of the frequency maps, the reconstructed SZ maps, and the spectra for each cluster, we then defined three SZ quality grades, Q_SZ:

• Q_SZ = A,if all the following criteria are fulfilled:

  • clear compact SZ source detected in the SZ map;

  • obvious measurements of the SZ decrement at least at 143 GHz or 100 GHz;

  • low dust contamination (i.e., no increase in the 353 GHz and 545 GHz fluxes in the SZ spectrum or residual dust emission or submillimetre point sources in the frequency map), and a reasonable detection at 353 GHz;

  • no radio source contamination (checked in LFI maps) or CMB confusion (checked in the HFI 217 GHz map).

• Q_SZ = B,if all the following criteria are fulfilled:

  • visible SZ detection in the SZ map or significant measured SZ signal at 143 GHz. The 100 GHz signal can be more noisy;

  • dust emission well subtracted but for the effect of point source contamination at the cluster location, (i.e., increase of the 353 GHz and possibly the 545 GHz fluxes in the SZ spectrum or residual dust emission or submillimetre point sources in the frequency map) resulting in large uncertainties for dust emission removal;

  • no radio source contamination or CMB confusion.

• Q_SZ = C,if any of the three following criteria are fulfilled:

  • weak SZ spectral signature (due to large error bars or to inconsistent spectral shape) or visible signal in noisy SZ maps;

  • strong dust contamination (i.e., high 353 GHz and 545 GHz fluxes in the SZ spectrum or residual dust emission or submillimetre point sources in the frequency map);

  • possible contamination by radio sources seen down to the LFI-70 GHz channel.

The three cases are illustrated in Fig. 1. These criteria were checked using the maps and spectra obtained with the different methods described above. Convergence between methods helped us to define the quality grade for each candidate.

We chose candidates to examine our internal SZ quality assessment by exploring lower quality detections than in our previous publications. On the basis of a candidate list detected by at least two algorithms², we selected eleven candidates detected at 4.5 < S/N < 5.3 with the MMF3 algorithm. Here we are sampling a lower S/N regime than in our previous validation run (for which 5.1 < S/N < 10.6) or in the ESZ sample (for which S/N > 6). To investigate the pertinence of our SZ quality grade definitions, we selected typical cluster candidates from the three categories, in the following proportions: two, five, and four, respectively, for Q_SZ =  A, B and C. The SZ properties of the candidates are given in Table 1. Note that the objects in no way constitute a complete or even statistically representative sample. Hence, we cannot use them to draw any quantitative conclusions regarding, for example, the purity of the parent catalogue.

Two of the three lowest S/N candidates, PLCK G193.3 − 46.1 and PLCK G210.6+17.1 fall in the Sloan Digital Sky Survey (SDSS) area. They have no counterpart in published SDSS cluster catalogues, but our dedicated algorithm search for galaxy over-densities (Fromenteau et al., in prep.) indicated that they were each possibly associated with a z > 0.5 cluster. Inclusion of these two targets allowed us to further test SDSS-based confirmation at high z.

Fig. 2 XMM-Newton  [0.3–2] keV energy band cluster candidate images. North is up and east to the left. The bottom right-hand panel shows the image of the false Planck candidate. For confirmed clusters, image sizes are 3θ500 on a side, where θ500 is estimated from the M500–YX relation (Eq. (1)). Images are corrected for surface brightness dimming with z, divided by the emissivity in the energy band, taking into account galactic absorption and instrument response, and scaled according to the self-similar model. The colour table is the same for all clusters, so that the images would be identical if clusters obeyed strict self-similarity. The majority of the objects show evidence for significant morphological disturbance. A yellow cross indicates the Planck position and a red/green plus sign the position of a RASS-BSC/FSC source, respectively.

3. XMM-Newton observations

The data analysis and validation procedure is described extensively by Planck Collaboration (2011c). We present only a brief summary in this section.

3.1. Observations and data reduction

The candidates were observed between December 22, 2010 and May 16, 2011. The observation identification number and observation setup are summarised in Table 1. The nominal setup used the THIN filters (unless optical loading needed to be avoided) and extended full frame (EFF) mode for the epn camera.

Calibrated event lists were produced with v11.0 of the XMM-Newton Science Analysis System. Data that were affected by periods of high background due to soft proton flares were omitted from the analysis; clean observing time after flare removal is given Table 1. Three observations are affected by high background levels: PLCK G268.5 − 28.1, PLCK G200.9 − 28.2 and PLCK G266.6 − 27.3. The data treatment for the latter cluster is fully described in Planck Collaboration (2011g). For PLCK G268.5 − 28.1 and PLCK G200.9 − 28.2, the particle background after flare cleaning is 2 and 1.7 times higher than nominal for the epn camera, respectively. The epn data were thus discarded for the spectroscopic analysis, as this is very sensitive to the background estimate.

The cleaned data were pattern-selected and corrected for vignetting as described in Pratt et al. (2007). Bright point sources were excised from the data. The background treatment is as described in Pratt et al. (2010). In the spectroscopic analysis, the cluster component was modeled with an absorbed thermal emission model (mekal) with a hydrogen column density fixed at the 21-cm value of Dickey & Lockman (1990).

3.2. Candidate confirmation

The confirmation status of each XMM-Newton observation is given in Table 1 and the XMM-Newton images are shown in Fig. 2. Of eleven targets, ten candidates are bona fide clusters. In each case, the extended nature of the X-ray source, clearly detected within the Planck position error box, was confirmed by comparing the surface brightness profile with the XMM-Newton point spread function (PSF). The consistency between the SZ and X-ray properties (Sect. 4.3) provided the final confirmation check. The total epic count rates in the [0.3 − 2] keV band of each cluster and the maximum radius of detection are given in Table 2.

The offset between the X-ray position and the Planck position (Fig. 2) is similar to that observed for known clusters in the ESZ sample (Planck Collaboration 2011b) or for candidates that have previously been confirmed with XMM-Newton (Planck Collaboration 2011c). The median offset is , characteristic of the Planck reconstruction uncertainty, which peaks around 2′ (Planck Collaboration 2011b,c) and is driven by the spatial resolution of the instruments. The largest offset is or 0.8   R₅₀₀. This offset is observed for PLCK G200.9 − 28.2, a highly disturbed cluster with a flat X-ray morphology (Fig. 2), for which a true physical offset between the X-ray and SZ signal may also contribute.

One candidate, PLCK G113.1 − 74.4, proved to be a false detection (Fig. 2, last panel) as no extended source is detected within the Planck position error. The surface brightness profile of the RASS Faint Source Catalogue source, detected about 5′ South of the Planck position, is consistent with that of a point source.

3.3. Redshift and physical parameter estimates

To estimate the redshift from the X-ray data, z_Fe, we extracted a spectrum within a circular region corresponding to the maximum significance of the X-ray detection. Since the centroid of the Fe–K line complex depends on the temperature, the redshift was determined from a thermal model fit to the full spectrum in the  [0.3–10] keV band, as described in detail in Planck Collaboration (2011c). The quality of the redshift estimate was characterised by the quality flag Q_z as defined in Planck Collaboration (2011c). The redshift of most clusters is well constrained (Q_z = 2). Three clusters, PLCK G193.3 − 46.1, PLCK-G262.2+34.5 and PLCK G268.5-28.1, have ambiguous z_Fe estimates (Q_z = 1). They exhibit several χ² minima in the kT–z_Fe plane that do not differ at the 90% confidence level (see Sect. 5.3 for further discussion). For these systems we used the redshift corresponding to the most significant χ² minimum, listed in Table 2. For PLCK G193.3 − 46.1 and PLCK G262.2+34.5, this redshift corresponds to the optical photometric redshift subsequently derived from SDSS data (Sect. 4.2) and our optical follow-up (Sect. 5.1.2), respectively. The uncertainty on the redshift is not propagated through the physical parameter estimation procedure discussed below. The statistical uncertainty on z_Fe is small for the Q_z = 2 systems. The physical parameters for Q_z = 1 systems, especially PLCK G268.5-28.1, are less robust and should be treated with caution.

We then derived the gas density profile of each cluster from the surface brightness profile, using the regularised deprojection and PSF-deconvolution technique developed by Croston et al. (2006). Global cluster parameters were estimated self-consistently within R₅₀₀ via iteration about the M₅₀₀–Y_X relation of Arnaud et al. (2010) assuming standard evolution: (1)The quantity Y_X, introduced by Kravtsov et al. (2006), is defined as the product of M_g,500, the gas mass within R₅₀₀, and T_X, where the latter is the spectroscopic temperature measured in the [0.15 − 0.75] R₅₀₀ aperture. In addition, L₅₀₀, the X-ray luminosity inside R₅₀₀, was calculated as described in Pratt et al. (2009). The errors on M₅₀₀ given in the table correspond to statistical uncertainties only. Additional errors due to scatter around the relation (around 7% from simulations) and uncertainties on the relation itself are not taken into account.

The SZ flux was then re-extracted, calculating Y₅₀₀ 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. For most cases, the blind values are consistent with the recomputed Y₅₀₀, within the errors. However, as found in our previous studies (Planck Collaboration 2011b,c), there is a trend of SZ flux overestimation with size overestimation. For the present sample, the blind values are overestimated by a median factor of 1.3 for the size and 1.4 for Y₅₀₀.

We have checked for possible AGN contamination using the NVSS (at 1.4 GHz Condon et al. 1998) and SUMSS (at 0.84 GHz Bock et al. 1999) catalogues. A relatively bright radio source (560mJy) is found in the vicinity of PLCK G193.3 − 46.1 (at offset). However, LFI data do not show any significant signal so the source must have a steep spectrum. No other radio sources are found in any other candidates. We conclude that no significant contamination of the SZ signal is expected in any of the clusters. However, we cannot exclude the presence of radio faint AGN within each cluster area. Although they could contaminate the X-ray signal if present, the brightest X-ray sources are resolved and excised from the X-ray analysis.

Fig. 3 The new SZ-discovered Planck objects (red and green symbols) compared to clusters from the ROSAT All-Sky Survey catalogues in the LX–z plane. The X-ray luminosity is calculated in the  [0.1–2.4] keV band. Catalogues shown are REFLEX (Böhringer et al. 2004), NORAS (Böhringer et al. 2000), BCS (Ebeling et al. 1998), eBCS (Ebeling et al. 2000) and MACS (Ebeling et al. 2007). The solid line is the REFLEX flux limit, the dotted line is the HIFLUCGS flux limit of 2 × 10-11   ergs-1cm-2 and the dashed line is from the MACS flux limits.

4. XMM-Newton validation outcome

4.1. Planck sensitivity

The present validation run clearly demonstrates the capability of Planck to detect clusters of a wide range of masses up to high z. All targets in this run fall below the RASS X-ray flux limit. This is illustrated in Fig. 3, where the new clusters are shown in the L_X–z plane. They are plotted together with the clusters from large catalogues based on RASS data outside the Galactic Plane and the clusters confirmed in previous XMM-Newton validation observations (hereafter XMM-Val12 ). The new sample covers a wide range of redshift, 0.2 < z < 1. It includes two clusters at z > 0.5 and the first cluster blindly detected by Planck at z ~ 1 (see Planck Collaboration 2011g, for a detailed discussion of this cluster). The new clusters are less X-ray bright, at a given z, than those previously confirmed with XMM-Newton. This is not surprising, since we are probing a lower S/N, thus less massive, cluster candidate regime. The new clusters all lie below the RASS survey flux limits, even that of the most sensitive survey (MACS). The mass estimates range from as low as M₅₀₀ = 2.7±0.2 × 10¹⁴   M_⊙ for the nearby z = 0.22 cluster, PLCK G200.9 − 28.2, to M₅₀₀ = 7.8±0.6 × 10¹⁴   M_⊙ for PLCK G266.6 − 27.3 at z = 0.97. Interestingly, the two clusters are detected at very similar S/N. This reflects the Planck selection function, which depends on the integrated SZ flux, i.e., on the size and redshift of the cluster. Planck can detect both (1) low z, low mass clusters with large angular extent, and (2) compact high z, high mass objects. Consequently, the mass detection threshold of the Planck survey increases with redshift (at least in the redshift range probed by the present sample).

4.2. Candidate quality assessment

Only one of the eleven candidates, PLCK G113.1 − 74.4, is false. It is noteworthy that its S/N is the fourth-highest of the sample (S/N = 5.1). While this is rather high, its actual SZ detection falls into the lowest quality category, Q_SZ = C, an indication of the importance of the quality grades defined in Sect. 2, in addition to the S/N. The other three Q_SZ = C candidates are confirmed, including PLCK G268.5 − 28.1, detected at the same S/N as the false candidate, and PLCK G234.2 − 20.5, detected at S/N = 4.7, the second lowest S/N of the sample. Both clusters are detected by all three SZ detection methods, whereas the false candidate is only detected with the two MMF methods and not with the PWS algorithm. As expected, the probability that a candidate is a true cluster increases with N_det.

Fig. 4 Regularised scaled density profiles of the new confirmed Planck SZ clusters with redshift estimates (0.2 < z < 0.97, red lines). They are compared to those of similar mass systems from the representative X-ray samples REXCESS (Böhringer et al. 2007, blue lines), EXCPRES (Arnaud et al., in prep., cyan lines), and the new clusters at lower redshift (0.09 < z < 0.54) and higher SZ flux confirmed in previous validation runs (Planck Collaboration 2011c, green lines). The thick lines denote the mean scaled profile for each sub-sample.

As previously noted by Planck Collaboration (2011c), association of a cluster candidate with a RASS source within the Planck position uncertainty is not, by itself, sufficient for confirmation. The false candidate, PLCK G113.1 − 74.4, was associated with a RASS/FSC source that eventually proved to be a point source.

The lowest S/N candidate of all, PLCK G210.6+17.1, is confirmed, whereas it was detected by only two SZ detection methods and lies in the lowest quality category, Q_SZ = C. However, it is one of the two clusters that was flagged by our SDSS detection algorithm as being possibly associated with an SDSS cluster. The other SDSS cluster candidate, PLCK G193.3 − 46.1, is also confirmed. The XMM-Newton redshift measurements and the photometric redshift³ of the Brightest Cluster Galaxy (BCG) are fully consistent in the cases where there are matches with SDSS clusters identified by our internal algorithm. For PLCK G210.6+17.1 z_phot = 0.48±0.02 compared to z_Fe = 0.48±0.02. For PLCK G193.3 − 46.1, z_phot = 0.65±0.06 while z_Fe = 0.59±0.02. This supports the robustness of our SDSS analysis method and indicates that the SDSS can confirm candidates up to z ~ 0.6, and estimate their photometric redshifts.

4.3. X-ray versus SZ properties of newly detected clusters

The present study samples higher redshifts and lower SZ fluxes than the previous XMM-Newton validation observations (0.2 < z < 0.97 and 4 × 10^-4arcmin² < Y₅₀₀ < 1.5 × 10^-3arcmin², as compared to 0.09 < z < 0.54 and 6 × 10^-4arcmin² < Y₅₀₀ < 3 × 10^-3arcmin² for the previous observations). Our previous findings, detailed in Planck Collaboration (2011c), are confirmed and extended to higher z and/or lower Y₅₀₀. The new SZ-detected clusters have, on average, lower luminosities, flatter density profiles, and a more disturbed morphology than their X-ray selected counterparts.

The average scaled density profile (Fig. 4) is similar to that of the XMM-Val12 sample, and remains flatter than that of REXCESS, a representative sample of X-ray selected clusters (Arnaud et al. 2010). The gallery of XMM-Newton images (Fig. 2) shows a variety of morphologies with three out of ten clusters exhibiting extremely flat and asymmetric/or X-ray emission. One of those is PLCK G193.3 − 46.1 at z = 0.6, as shown in Fig. 5. Its double peaked X-ray morphology suggests an on-going merger of two sub-clusters along the NE-SW direction, which is supported by the available SDSS data. The galaxy distribution is not centrally peaked and its centroid is South/West of the BCG position. Neither the centre of the galaxy distribution nor the BCG position coincides with any of the X-ray peaks (see Fig. 5).

The new clusters follow the trends in scaling properties established from our previous follow-up (Fig. 6). They are on average less luminous at a given Y₅₀₀, or more massive at a given luminosity, than X-ray selected clusters. Eight out of ten of the new clusters fall on the low luminosity side of the L₅₀₀–Y₅₀₀ relation for X-ray selected clusters (Fig. 6 right panel). As shown in the left hand panel of Fig. 6, the Y₅₀₀–Y_X relation for most clusters remains consistent with the REXCESS prediction: (2)with C_XSZ = 1.416 × 10^-19   Mpc²/M_⊙keV. However, the SZ flux levels off around Y₅₀₀ ~ 4 × 10^-4arcmin². This turnover at low flux is clearly apparent when considering the weighted average Y₅₀₀ values in Y_X bins. It deviates significantly from the prediction in the two lowest Y_X bins, a deviation increasing with decreasing Y_X (grey area in the left panel of Fig. 6). This is reminiscent of the Malmquist bias resulting from a flux cut selection. Due to scatter around the mean relation between the observed flux (Y₅₀₀) and the “true” flux (estimated from Y_X), objects below the flux cut are detectable but, in order to be detected, they must be increasingly deviant from the mean relation with decreasing intrinsic flux. The effect is more prominent than that already observed for the XMM-Val12 sample (Planck Collaboration 2011c), whereas it is negligible for the ESZ-XMM-archive sample (Fig. 6). This is likely due to the increasing magnitude of the Malmquist bias as a function of decreasing flux (see also Planck Collaboration 2011e). Note that the scatter in the Y₅₀₀–Y_X relation, and thus the Malmquist bias, is likely dominated by measurements errors. Y₅₀₀ or Y_X are related to the same physical quantity, the thermal energy of the gas. The intrinsic scatter in Y₅₀₀ for a given Y_X is thus expected to be smaller than the  < 10% intrinsic scatter of either Y₅₀₀ or Y_X values at fixed mass (Kravtsov et al. 2006; Arnaud et al. 2007), and thus smaller than the statistical scatter in the S/N ≲ 5 regime. Outliers are present, though, as discussed below.

The most prominent outliers are the two lowest clusters, PLCK G235.6+23.3 and PLCK G268.5 − 28.1, which lie at 2.4σ and 2.8σ, respectively, above the expected relation (Eq. (2)). They could thus be due to statistical fluctuations. However, they correspond to Y₅₀₀/Y_X ratios 2.1 and 2.5 times higher than expected, respectively. The redshift of PLCK G268.5 − 28.1 is not well determined and may be under-estimated. We cannot thus exclude that its Y_X value is actually higher (see also Sect. 5.3). On the other hand, PLCK G235.6+23.3 is an unprepossessing cluster at z = 0.37 with no remarkable X-ray properties, and for which we have very accurate SZ and X-ray measurements. Only one such outlier in terms of Y₅₀₀/Y_X ratio appears in the ESZ-XMM-archival sample of 62 clusters: RXCJ0043.4 − 2037, a relaxed cluster at z = 0.29 (Finoguenov et al. 2005) as can be seen in Fig. 6. A complete follow-up of Planck candidates is required to quantify the intrinsic scatter in the Y₅₀₀–Y_X relation and its associated Malmquist bias. Only then can one compare the true dispersion in the Y₅₀₀/Y_X relation with that established from the ESZ-XMM-archival sample.

Fig. 5 SDSS colour composite image of PLCK G193.3 − 46.1 at z = 0.6 overlaid with isocontours of the wavelet filterered XMM-Newton image. The image size is . Green circles: cluster galaxies identified by the search algorithm. Red circle: Brightest Cluster Galaxy. Diamond: centroid of galaxy distribution. Cross: Planck SZ position.

Fig. 6 Scaling relations for the ten new confirmed clusters (red symbols). Black points show clusters in the Planck-ESZ sample with XMM-Newton archival data as presented in Planck Collaboration (2011e); green points represent previously-confirmed Planck clusters presented in Planck Collaboration (2011c). The blue lines denote the Y500 scaling relations predicted from the REXCESS X-ray observations (Arnaud et al. 2010). Left: relation between apparent SZ signal (Y500) and the corresponding normalised YX parameter. The grey area corresponds to weighted average Y500 values in YX bins with ±1σ errors. Right: relation between X-ray luminosity and Y500. For most data points, uncertainties on the luminosity are smaller than the point size.

5. Redshift determination

5.1. New optical redshift determinations

In this section we present new optical redshift determinations for ten confirmed clusters of the XMM-Val12 sample and for two of the present sample.

5.1.1. ENO observations

PLCK G171.9 − 40.7 and PLCK G100.2 − 30.4 were observed with the 0.82 m IAC80 telescope at the Observatorio del Teide (Tenerife, Spain) as part of a larger campaign for optical follow-up of newly detected Planck candidates. Images were taken in four Sloan filters, griz, with the CAMELOT camera. This camera is equipped with a 2048 × 2048 pixel CCD (0.304 arcsec per pixel), resulting in a field of view of .

Fig. 7 Colour composite images (Sloan gri filter) of PLCK G171.9-40.7 observed with ENO/ IAC80 telescope. North is up, East is right and the image size is . The isocontours of the wavelet filtered XMM-Newton image are overlaid.

The data reduction included all standard calibrations, i.e., bias and flat field corrections and astrometric calibration. Source detection was undertaken by running SExtractor (Bertin & Arnouts 1996) on the i-band images, and photometry on all bands was obtained in double-image mode. For source detection we used a detection threshold of 3σ in the filtered maps, which corresponds to a S/N ~ 6. All sources classified as stellar objects, based on a stellarity index greater than 0.8 in all bands (given by SExtractor) were excluded from our sample. We applied galactic extinction correction based on the dust maps by Schlegel et al. (1998).

PLCK G171.9 − 40.7 was observed in each Sloan griz filter in 4000 s exposures. The limiting magnitudes reached are 23.2, 21.1, 20.6 and 20.6 mag for g,r,i,z, respectively. The colour composite image in Fig. 7 clearly shows a galaxy overdensity coincident with the X-ray image. The BCG is only slightly offset from the X-ray peak. The final catalogue contains 384 sources, for which we obtained photometric redshifts using the BPZ code (Benítez 2000). The redshift estimate for each individual galaxy is based on all four filters, and is obtained by fitting a set of SED (spectral energy distribution) templates (see details in Benítez 2000). The BPZ code provides the Bayesian posterior probability distribution function (pdf) for the redshift of each object. We have calibrated the code for our set of four filters using a subsample of 5000 galaxies from SDSS DR8 with spectroscopic redshift, z_spec,SDSS. The standard deviation of the difference between z_spec,SDSS and the photometric redshift, z_phot,BPZ, obtained applying the BPZ code to this subsample, is Δz = 0.03. The deviation is similar for the whole sample and for the two different redshift intervals, 0 < z < 0.2 and 0.2 < z < 0.4. In a conservative approach, we used this deviation as systematic uncertainty on cluster redshift. The statistical uncertainty is negligible in comparison. For PLCK G171.9-40.7, we use 29 cluster members to infer the photometric redshift, and we obtain z_phot = 0.31±0.03.

The data taken for PLCK G100.2 − 30.4 were already presented in Planck Collaboration (2011c). Images have accumulated integration times of 3000 s in each filter and limiting magnitudes of 22.9, 21.7, 20.1 and 20.2 mag for g,r,i,z, respectively. Reduction and catalogue compilation followed the same steps as detailed above for PLCK G171.0 − 40.7. With respect to the results presented in Planck Collaboration (2011c), the main improvement is that the final images were photometrically re-calibrated using galaxies from SDSS DR8. The initial catalogue contains 452 sources for which photometric redshifts were derived. The object has a photometric redshift of z_phot = 0.34±0.03, estimated from the 72 identified cluster members.

Fig. 8 VRI colour composite image of PLCK G262.2 + 34.5 observed with the ESO/MPG 2.2m telescope, with exposure times of 0.5 h, 1.4 h, and 0.5 h in the V, R, and I-band, respectively.The isocontours of the wavelet filterered XMM-Newton image are overlaid. North is up and east to the left, and the image size is . Although the X-ray morphology of PLCK G262.2 + 34.5 is flat rather than centrally peaked, the X-ray centre coincides well with the location of the BCG. The field also contains a large number of X-ray point sources with optical counterparts.

5.1.2. ESO observations

Optical imaging observations of the XMM-Newton confirmed clusters discussed in Planck Collaboration (2011c) were also carried out on the ESO/MPG 2.2 m telescope at La Silla Observatory using the Wide-Field Imager (WFI), which has a field of view of 33′ × 34′ and pixel size . Each cluster was observed in the V, R, and I-bands in typical seeing conditions of 1.0–1.2″, for total exposure times of at least 0.5 h (consisting of 5 × 360 s dithered sub-exposures) per filter. The raw data were calibrated using standard techniques and individual exposures were re-registered and combined using the USNO-B1 catalogue as an astrometric reference. As an illustration, the VRI colour composite image of PLCK G262.2 + 34.5 is shown in Fig. 8.

Galaxies that were simultaneously identified in the combined V, R, and I images were plotted in a V − R vs. R − I colour − colour diagram. For each cluster, an overdensity of red galaxies, corresponding to the early-type cluster galaxies, was identified in colour–colour space. Galaxies associated with this overdensity in colour–colour space and also spatially coincident (to within  ~5′) with the X-ray cluster position were assumed to be early-type cluster members. Predicted V − R, V − I and R − I colors of early-type cluster galaxies as a function of redshift were calculated by convolving the “E0” template galaxy spectrum of Coleman et al. (1980) with the combined (filter+CCD) response curves for the V, R and I filters at WFI.

A photometric redshift estimate was then derived by comparing the median V − R, V − I, and R − I colors of the early-type cluster galaxies to these predictions and averaging the three resulting redshift values. We estimated how typical fluctuations in the photometric zero-point throughout the night translate into uncertainties in the measured V − I, V − R , R − I colors of galaxies. Given the predicted relation between these colors and the redshift of early-type galaxies, the estimated 1σ redshift accuracy is Δz = 0.02. The new photometric redshift estimates for ten clusters observed with WFI are given in Table 3. They were derived from at least 70 photometic redshifts per cluster (mean number of 120).

5.2. Comparison between optical and X-ray z estimates

Optical redshifts for twenty XMM-Newton confirmed Planck clusters are now available. This includes the fourteen measurements presented here or in Planck Collaboration (2011c), values from the literature for the four clusters discovered independently by ACT or SPT (Marriage et al. 2011; Williamson et al. 2011), and two photometric redshifts that we retrieved from SDSS data. The values and references are given in Table 3, together with XMM-Newton derived value from the X-ray spectra. For clusters with ambiguous X-ray redshift estimates (Q_SZ < 2), the values⁴ refer to the most significant χ² minimum used above to calculate physical properties.The optical and X-ray estimates are compared in Fig. 9. The agreement is excellent, with a weighted mean ratio of 1.002 and a standard deviation around equality of 0.08. The X-ray and optical spectroscopic redshifts (three clusters) are consistent within Δ(z) < 0.02.

Fig. 9 Comparison between the redshift estimated from optical data and that from XMM-Newton spectroscopy.

Fig. 10 Redshift determination from XMM-Newton spectroscopy for the highest quality data (top row, PLCK G234.2 − 20.5) and lowest quality data (bottom row, PLCK G268.5-28) in the sample. Left panels: variation of χ2 with z when fitting the EPIC spectra, all other parameters being let free. The dashed and full lines correspond to the 68% and 90% error range, respectively. Right panels: EPIC spectra (data points with errors), together with the best-fitting model thermal model (solid lines) with the position of the redshifted Fe K line marked. Only the data points above 2 keV are shown for clarity, but data down to 0.3 keV are used in the spectral fitting. For PLCK G234.2-20.5 (top right panel), the Fe-K line complex is clearly detected in the EPIC MOS1&2 (red and black points) and pn (green points) spectra. For PLCK G268.5-28 (bottom right panel), only MOS data can be used (see Sect. 3.1) and the spectra are of poor statistical quality. The redshift estimate is ambiguous and the χ2 distribution (bottom left panel) shows several minima. The MOS1&2 spectra, summed for clarity, are compared to the best fitting model for z = 0.47 (red line) and z = 0.86 (green line), corresponding to the two lowest minima.

5.3. Redshift estimate from a combined X-ray and SZ study

For three clusters in the present sample, PLCK G193.3−46.1, PLCK-G262.2+34.5 and PLCK G268.5-28.1, the X-ray z estimates are ambiguous (Q_z = 1) and the spectral fit as a function of z exhibits several χ² minima that cannot be distinguished at the 90% confidence level, as illustrated in Fig. 10 (Bottom left panel). This arises when the Fe-K line complex is detected at low significance and statistical fluctuations in the spectra of the same magnitude can mimic the presence of a line (see the bottom right panel of Fig. 10). Low statistical quality data arises because the cluster is intrinsically faint or the X-ray observations are affected by high background conditions. For comparison, the top row of Fig. 10 shows the results for PLCK-G234.2-20.5, for which the data quality are the highest in the sample.

Fig. 11 Variation with redshift of the ratio between the X-ray and SZ flux (left panel) and between YX and (right panel). Line: locus established from scaling relations (Planck Collaboration 2011e; Arnaud et al. 2010). The dotted lines correspond to a factor of two above or below the mean relation. Black points: data for new Planck candidates confirmed with XMM-Newton. Colour points: data for clusters with ambiguous X-ray redshift estimates, one colour per cluster. Each point corresponds to one of the redshift solutions for an individual cluster, as derived from the χ2(z) minima (see Fig. 10).

Optical follow-up observations are obviously required to obtain a precise redshift. However, better X-ray redshift estimates are useful for optimising any potential follow-up, e.g., for the use of the most appropriate optical facility or for deciding the pertinence of deeper X-ray follow-up based on known physical properties. In principle, the redshift can be constrained by combining X-ray and SZ data, following a method similar to that used historically to constrain the Hubble constant. The method relies on the different distance dependence of the X-ray and SZ measurements. Here we examined the redshift constraining power of both the Y_X–Y₅₀₀ and the L_X–Y₅₀₀ relations, using the relations established by Planck Collaboration (2011e) from ESZ clusters with archival XMM-Newton data. We consider the three clusters with ambiguous X-ray redshift estimates, including the two clusters, PLCK G193.3 − 46.1 and PLCK-G262.2+34.5, for which a photometric redshift is available (Table 3). Use of the latter allows us to undertake an internal consistency check.

The X-ray luminosity in the  [0.1–2.4]  keV energy band scales quasi-linearly with . Using the normalisation of the L_X– relation and its z dependence, given in Table 2 of Planck Collaboration (2011e), and taking into account the z dependence of the luminosity-distance, one can write: (3)where F_X is the X-ray flux at Earth in the same band and K(z) is the K correction. The K correction increases with z, with a typical value of K = 1.24 at z = 0.5 for a kT = 6keV cluster. We can neglect the temperature dependence of the K correction, which is much smaller than the typical dispersion of the L_X–Y₅₀₀ relation for the energy band and mass range under consideration.

The theoretical relation is plotted in the left hand panel of Fig. 11. For each cluster, we then estimated the X-ray flux and Y₅₀₀, fixing z to each possible value in turn. The flux estimates depend on physical cluster parameters such as size θ₅₀₀ and temperature, whose estimate depends in turn on z and requires data of sufficient quality. As can be seen in the figure, in practice the measured flux ratio depends weakly on the assumed z. This simply reflects the fact that the fluxes are the quantities most directly related to the raw measurements. If data are of insufficient quality, the ratio can also simply be estimated at a fiducial z and kT. More importantly, since the true ratio depends on z, this redshift can be constrained from the measured ratio and Eq. (3). Unfortunately, as can be seen in Fig. 11, the large dispersion around the relation limits the constraints one can achieve with this method. The variation beyond z ~ 0.2 is no more than a factor of two, meaning that one cannot distinguish between a factor of two under-luminous outlier at z = 0.2 and a “normal” cluster at z = 1. The lack of constraining power is exacerbated by the very nature of the clusters in question; those with poor z estimates are generally objects with low intrinsic X-ray fluxes.

We thus also examined the Y_X–Y₅₀₀ relation, which exhibits a lower intrinsic scatter. Planck Collaboration (2011e) showed that the Y_X–Y₅₀₀ relation is consistent with that derived from REXCESS. The ratio is fixed from Eq. (2), while its estimate from X-ray and SZ data depends on z. This dependence is complex and does not follow a simple analytical law. For each assumed z in turn, the parameters must be derived from X-ray data and SZ data re-processing. The estimated ratio increases with the assumed z, as illustrated in the right-hand panel of Fig. 11. Although its dispersion is smaller, the Y_X–Y₅₀₀ relation does not provide better constraints. For newly detected clusters, this method is limited by 1) the large statistical uncertainty of SZ data and 2) the possibility that the cluster is an outlier (this latter being all the more important because of the Malmquist bias). This is perfectly illustrated in the case of PLCK G268.5 − 28.5. A redshift as low as z = 0.15 is very unlikely. However, the cluster could either be at z = 1.2 if it perfectly follows the mean relations, or it could be an under-luminous, low Y_X cluster at z = 0.47 (the best X-ray estimate). The cases of PLCK G193.3 − 46.1 and PLCK G262.2+34.5 are very similar: only the lowest z solution can be excluded. On the other hand, the redshifts indicated by optical data, z_phot ~ 0.60 and z_phot ~ 0.23, respectively, are indeed allowed by the present analysis. However, higher z solutions yield X/SZ values closest to the theoretical relations. This again illustrates the limitation of the method in the presence of scatter.

In summary, we find that the use of the Y_X vs. Y_SZ and X-ray flux F_X vs. Y_SZ relations allows us to put lower limits on cluster redshifts.

6. Conclusion

We have presented a further eleven XMM-Newton X-ray observations of Planck cluster candidates, undertaken in the framework of a DDT validation programme. The sample was chosen from blind detections in all-sky maps from the first ten months of the survey and probes lower signal-to-noise and SZ quality criteria than published previously (Planck Collaboration 2011e). Ten of the candidates are confirmed to be bona fide clusters, all of which fall below the RASS X-ray flux limit. The objects lie at redshifts 0.22 < z < 0.94 and have masses (estimated from the M₅₀₀–Y_X relation) in the range (2.7±0.2) × 10¹⁴M_⊙ < M₅₀₀ < (7.8±0.6) × 10¹⁴M_⊙. We detect a first indication for Malmquist bias in the Y_SZ–Y_X relation, with a turnover at Y_SZ ~ 4 × 10^-4 arcmin².

This validation run clearly demonstrates the capability of the Planck survey to detect clusters of a wide range of masses up to high z, although with a mass detection threshold that increases with redshift. We emphasise that the present sample is neither complete not representative, being constructed to sample various SZ quality flags. While it is not a priori biased towards any specific type of cluster, it cannot be used to infer any statistical information on the parent catalogue, such as its underlying purity, or for quantifying the Malmquist bias.

We studied the pertinence of our internal quality grades assigned to the SZ detection, based on visual inspection of the reconstructed 2D SZ maps and SZ spectrum. The single false candidate has a relatively high S/N ~ 5, but the lowest SZ quality grade. This confirms that the quality of the Planck SZ detection cannot be reduced to a single global S/N and is an indication of the pertinence of our internal quality grade definition. On the other hand, real clusters do have C grade detections. Such a grade is clearly not sufficient to exclude a given candidate. However, A and B grade detections are strong indications for a real cluster.

We presented new optical redshift determinations of candidates previously with XMM-Newton, obtained with ENO and ESO telescopes. The X-ray and optical redshifts for a total of 20 clusters are found to be in excellent agreement. We also show that useful lower limits can be put on cluster redshifts using X-ray data alone via the Y_X vs. Y_SZ and X-ray flux F_X vs. Y_SZ relations.

In terms of physical properties, the present clusters are similar to the first new Planck SZ detections presented in Planck Collaboration (2011e), except at lower Y₅₀₀ and higher mean redshift. The majority show signs of significant morphological disturbance, which is reflected in their flatter density profiles compared to those of X-ray selected systems.They are, on average, under-luminous for their mass as compared to X-ray selected clusters.

In future work, we will explore even lower S/N detections and discuss information from ancillary data, such as that available from SDSS or RASS, as an indicator of candidate validity.

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), on observations made with the IAC80 telescope operated on the island of Tenerife by the Instituto de Astrofsica de Canarias in the Spanish Observatorio del Teide and on observations collected using the ESO/MPG 2.2 m telescope on La Silla under MPG programs 086.A-9001 and 087.A-9003. 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 the Laboratoire d’Astrophysique de Tarbes-Toulouse (LATT), 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. 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

All Tables

All Figures

Fig. 1 Illustration of the three SZ quality grades as defined in Sect. 2. From left to right the three quality cases are: QSZ = A,B,C. The top row shows a 100′ × 100′ SZ map with a spatial resolution of 10′, centred on the candidate position, derived using the MILCA reconstruction method (Hurier et al. 2010). The colour table is identical for all clusters, with the Compton y parameter spanning the range  [ − 3 × 10-6,1 × 10-5] . The bottom row shows the associated SZ spectrum from aperture photometry measurements within R500 (see text for details). The red line is the SZ spectrum normalised to the Y500 value obtained from MMF blind detection.

In the text

Fig. 2 XMM-Newton  [0.3–2] keV energy band cluster candidate images. North is up and east to the left. The bottom right-hand panel shows the image of the false Planck candidate. For confirmed clusters, image sizes are 3θ500 on a side, where θ500 is estimated from the M500–YX relation (Eq. (1)). Images are corrected for surface brightness dimming with z, divided by the emissivity in the energy band, taking into account galactic absorption and instrument response, and scaled according to the self-similar model. The colour table is the same for all clusters, so that the images would be identical if clusters obeyed strict self-similarity. The majority of the objects show evidence for significant morphological disturbance. A yellow cross indicates the Planck position and a red/green plus sign the position of a RASS-BSC/FSC source, respectively.

In the text

Fig. 3 The new SZ-discovered Planck objects (red and green symbols) compared to clusters from the ROSAT All-Sky Survey catalogues in the LX–z plane. The X-ray luminosity is calculated in the  [0.1–2.4] keV band. Catalogues shown are REFLEX (Böhringer et al. 2004), NORAS (Böhringer et al. 2000), BCS (Ebeling et al. 1998), eBCS (Ebeling et al. 2000) and MACS (Ebeling et al. 2007). The solid line is the REFLEX flux limit, the dotted line is the HIFLUCGS flux limit of 2 × 10-11   ergs-1cm-2 and the dashed line is from the MACS flux limits.

In the text

Fig. 4 Regularised scaled density profiles of the new confirmed Planck SZ clusters with redshift estimates (0.2 < z < 0.97, red lines). They are compared to those of similar mass systems from the representative X-ray samples REXCESS (Böhringer et al. 2007, blue lines), EXCPRES (Arnaud et al., in prep., cyan lines), and the new clusters at lower redshift (0.09 < z < 0.54) and higher SZ flux confirmed in previous validation runs (Planck Collaboration 2011c, green lines). The thick lines denote the mean scaled profile for each sub-sample.

In the text

	Fig. 5 SDSS colour composite image of PLCK G193.3 − 46.1 at z = 0.6 overlaid with isocontours of the wavelet filterered XMM-Newton image. The image size is . Green circles: cluster galaxies identified by the search algorithm. Red circle: Brightest Cluster Galaxy. Diamond: centroid of galaxy distribution. Cross: Planck SZ position.
In the text

Fig. 6 Scaling relations for the ten new confirmed clusters (red symbols). Black points show clusters in the Planck-ESZ sample with XMM-Newton archival data as presented in Planck Collaboration (2011e); green points represent previously-confirmed Planck clusters presented in Planck Collaboration (2011c). The blue lines denote the Y500 scaling relations predicted from the REXCESS X-ray observations (Arnaud et al. 2010). Left: relation between apparent SZ signal (Y500) and the corresponding normalised YX parameter. The grey area corresponds to weighted average Y500 values in YX bins with ±1σ errors. Right: relation between X-ray luminosity and Y500. For most data points, uncertainties on the luminosity are smaller than the point size.

In the text

	Fig. 7 Colour composite images (Sloan gri filter) of PLCK G171.9-40.7 observed with ENO/ IAC80 telescope. North is up, East is right and the image size is . The isocontours of the wavelet filtered XMM-Newton image are overlaid.
In the text

Fig. 8 VRI colour composite image of PLCK G262.2 + 34.5 observed with the ESO/MPG 2.2m telescope, with exposure times of 0.5 h, 1.4 h, and 0.5 h in the V, R, and I-band, respectively.The isocontours of the wavelet filterered XMM-Newton image are overlaid. North is up and east to the left, and the image size is . Although the X-ray morphology of PLCK G262.2 + 34.5 is flat rather than centrally peaked, the X-ray centre coincides well with the location of the BCG. The field also contains a large number of X-ray point sources with optical counterparts.

In the text

	Fig. 9 Comparison between the redshift estimated from optical data and that from XMM-Newton spectroscopy.
In the text

Fig. 10 Redshift determination from XMM-Newton spectroscopy for the highest quality data (top row, PLCK G234.2 − 20.5) and lowest quality data (bottom row, PLCK G268.5-28) in the sample. Left panels: variation of χ2 with z when fitting the EPIC spectra, all other parameters being let free. The dashed and full lines correspond to the 68% and 90% error range, respectively. Right panels: EPIC spectra (data points with errors), together with the best-fitting model thermal model (solid lines) with the position of the redshifted Fe K line marked. Only the data points above 2 keV are shown for clarity, but data down to 0.3 keV are used in the spectral fitting. For PLCK G234.2-20.5 (top right panel), the Fe-K line complex is clearly detected in the EPIC MOS1&2 (red and black points) and pn (green points) spectra. For PLCK G268.5-28 (bottom right panel), only MOS data can be used (see Sect. 3.1) and the spectra are of poor statistical quality. The redshift estimate is ambiguous and the χ2 distribution (bottom left panel) shows several minima. The MOS1&2 spectra, summed for clarity, are compared to the best fitting model for z = 0.47 (red line) and z = 0.86 (green line), corresponding to the two lowest minima.

In the text

Fig. 11 Variation with redshift of the ratio between the X-ray and SZ flux (left panel) and between YX and (right panel). Line: locus established from scaling relations (Planck Collaboration 2011e; Arnaud et al. 2010). The dotted lines correspond to a factor of two above or below the mean relation. Black points: data for new Planck candidates confirmed with XMM-Newton. Colour points: data for clusters with ambiguous X-ray redshift estimates, one colour per cluster. Each point corresponds to one of the redshift solutions for an individual cluster, as derived from the χ2(z) minima (see Fig. 10).

In the text