Rotation in NGC 2264: a study based on CoRoT★ photometric observations

Abstract

1 INTRODUCTION

Over the last few years, the determination of stellar rotation rates for large samples of stars with different masses and ages in young open clusters, has made substantial progress, providing a large observational sampling of the angular momentum evolution of pre-main-sequence stars (Bouvier 2008, 2009). The precise mechanisms governing angular momentum evolution of pre-main-sequence low-mass stars are still not well understood, but basically they can be schematized with two main competing processes: star contraction and star–disc interaction. Stars in the pre-main-sequence phase are still contracting; thus, increasing their angular velocity to conserve the angular momentum. On the other hand, the observed existence of slow rotators and of a wide dispersion in rotation rates of cluster stars on the zero-age main sequence (ZAMS) can be explained only with the presence of a competing mechanism of angular momentum loss (i.e. spin-down of the star), different from star to star, during the pre-main-sequence phase. It is generally believed that this mechanism can be explained by the magnetic interaction between stellar magnetospheres and circumstellar discs, in a scenario known as disc locking, first proposed by Camenzind (1990) and Koenigl (1991) and explained in detail by Shu et al. (1994), which assumes that the angular momentum deposited on an accreting star (due to mass accretion from disc to star; Edwards et al. 1993) is exactly removed by torques carried along magnetic field lines connecting the star to the disc. The wide dispersion of rotational velocities observed on the ZAMS is the result of different disc lifetimes (Bouvier et al. 1993). Several observational results indicate a relation between the presence of discs and rotational evolution, in particular population of stars with discs, on average, rotate more slowly than those without discs and does exist a statistically significant anticorrelation between angular velocity and disc indicators such as near-infrared (NIR) excess and Hα equivalent width (EW; Bouvier et al. 1993; Edwards et al. 1993; Herbst et al. 2000, 2002; Lamm et al. 2005; Rebull et al. 2006). Moreover, the disc-locking scenario predicts that the torques arising from the magnetic connection between the star and the disc remove substantial angular momentum enforcing an equilibrium angular spin rate (Choi & Herbst 1996) which is in agreement with the constant rotation period in the 2–8 d range, characteristic of the majority of young stars. However, there have been several conflicting theoretical and observational evidences concerning the role of disc-locking scenario in the evolution of low-mass pre-main-sequence stars. Dahm & Simon (2005) pointed out that the P_rot distribution histograms for weak T-Tauri stars (WTTSs, whose periodic variability is believed to be induced by large starspots) and classical T-Tauri stars (CTTSs, whose variability may be also due to accretion spots and shadowing of the photosphere from dusty disc structures) in NGC 2264 are very similar and do not indicate that CTTSs are rotating more slowly than their WTTS counterparts. Furthermore, Dahm & Simon (2005) did not find a correlation between P_rot and theoretical age, as might be expected if stars were spinning up after decoupling from their discs. Stassun et al. (1999) and Cieza & Baliber (2006) did not find a correlation between accretion and rotation in Orion nebula cluster (ONC) and IC 348 low-mass stars, respectively (though they do not conclude that their results are inconsistent with disc locking). We have to note that most of the samples for which there is no clear evidence of a connection between the existence of discs and slow rotation, suffer from several biases, such as small sample size, sample biased towards small-mass or high-mass stars, the use of NIR photometry as disc indicator or the use of vsini values instead of rotation periods, which make them unsuited to perform this kind of test, on star–disc interaction outcomes. In particular, the presence of a NIR excess does not guarantee that the star is actually accreting mass from a disc. The studies of Cieza & Baliber (2007) and Rebull et al. (2005), based on Spitzer mid-infrared (mIR) observations, however, as well as demonstrating that objects which currently show mIR excesses are more likely accreting than not, also found differences in the rotational properties of accreting and non-accreting stars for NGC 2264 and the ONC, respectively, and represent the best test case to date, providing the strongest evidence that star–disc interaction regulates the angular momentum evolution of pre-main-sequence stars.

The idea and the basic assumptions of disc locking, sketched above, are a simplification of a much more complex phenomenon, and indeed several discussions on the shortcomings of the theory and its confrontation with observations have been put forward. In particular, Matt et al. (2010, 2012) critically examined the theory of disc locking, noting that the differential rotation between the star and disc naturally leads to an opening (i.e. disconnecting) of the magnetic field between the two. They find that this significantly reduces the spin-down torque on the star by the disc, thus, disc locking cannot account (at least, alone) for the slow rotation observed in several systems and for which the model was originally developed. Matt et al. (2010, 2012) supported the idea that stellar winds may be important during the accretion phase, they may be powered by the accretion process itself and be the key driver of angular momentum loss (Hartmann & MacGregor 1982; Paatz & Camenzind 1996; Matt & Pudritz 2005). A strong magnetically driven wind, as proposed by Matt & Pudritz (2005), is an idea which deserves further study, as well as the development of a more realistic theoretical model able to explain the full range of observed rotation periods and magnetic phenomena and the achievement of a sufficient amount of accurate data to empirically constrain them.

NGC 2264 is one of the best known studied star forming regions in the solar vicinity (d ≈ 760 pc, age ≈  3 Myr) and is considered a benchmark for the study of star formation processes in our Galaxy. NGC 2264 luckily falls in the small portion of the sky accessible by COnvection ROtation and planetary Transits (CoRoT; Baglin et al. 2006), and thus represents a unique chance for the mission and the study of young stars still in a formation phase. Its distance and age make it an ideal CoRoT target, its size is well suited to the CoRoT field of view, with a large fraction of cluster members falling in the appropriate magnitude range for accurate photometric monitoring in the CoRoT observations.

NGC 2264 has been extensively observed at all wavelengths from radio to X-rays (see Dahm 2008 for a review on the region), for studies of the star formation process through the observation of its outcomes: the initial mass function (e.g. Sung et al. 2008), the star formation history and the spatial structure (e.g. Teixeira et al. 2006; Sung, Stauffer & Bessell 2009). NGC 2264 is also a primary target for the study of the evolution of the stellar angular momentum and its relation to circumstellar accretion (e.g. Lamm et al. 2005) of the evolution (and dispersal) of circumstellar discs (e.g. Alencar et al. 2010) and of the correlation between optical and X-ray variability of young stars (Flaccomio et al. 2010). Among several optical, infrared (IR) and X-ray surveys, both photometric and spectroscopic, on NGC 2264, the most recent are Sung et al. (2008), who have provided the widest area and deepest publically available optical photometry; Sung et al. (2009), who have published Spitzer (IRAC + MIPS) photometry; Rebull et al. (2002) and Dahm & Simon (2005), who have published spectral types, Hα and Li EWs, from low-dispersion spectra, for a total of ∼500 members; Fűrész et al. (2006), who have published radial velocities for 436 stars.

Lamm et al. (2004) performed a photometric monitoring of about 10 600 stars to search for periodic and irregular variable pre-main-sequence stars and found 543 periodic variables with periods between 0.2 and 15 d, and 484 irregular variables. Lamm et al. (2005) used this extensive study to conclude that the period distribution in NGC 2264 is similar to that of the ONC, though shifted towards shorter periods (confirming the conclusion of Soderblom et al. 1999, based on the analysis of the rotation rates of 35 candidate members, that the stars in NGC 2264 are spun-up with respect to members of the ONC).

The period distribution found by Lamm et al. (2005) is unimodal for masses lower than 0.25 M_⊙  while it is bimodal for more massive stars. Lamm et al. (2005) also found evidence for disc locking with a constant period, among 30 per cent of the higher mass stars (with a locking period of ∼8 d), while disc locking is less important among lower mass stars, whose peak in the period distribution at 2–3 d suggests that these stars have undergone a low rate of angular momentum loss from star–disc interaction, while not completely locked [an evolution scenario defined by Lamm et al. (2005) as ‘moderate’ angular momentum loss].

The bimodality is interpreted as an effect of disc–star interactions in pre-main-sequence stars, slow rotators being interpreted as stars that are magnetically locked to their discs, preventing them from spinning up with time and accounting for the broad period distribution of ZAMS stars. This assumption is supported by some observational results showing that WTTSs are rotating faster than CTTSs with inner discs (Bouvier et al. 1993; Edwards et al. 1993). Nevertheless, the hypothesis that accreting stars rotate more slowly than non-accreting ones is still a matter of debate, since conflicting evidence exists (see e.g. Stassun et al. 1999; Dahm & Simon 2005; Lamm et al. 2005; Rebull et al. 2005, 2006; Cieza & Baliber 2006, 2007), as explained above.

The CoRoT satellite has allowed us to conduct a large-scale survey of photometric variability of NGC 2264. Thanks to the accurate high-cadence photometry and large field of view of CoRoT, we could study rotation and activity of about 8000 stars in a 3.4 deg² region. The entire star-forming region fits into a single CoRoT field of view, and the campaign resulted in continuous 23 d LCs for 301 known cluster members brighter than V ≈ 16 (M = 0.3–0.4 M_⊙). The resulting optical broad-band LCs are the first accurate, highest cadence (32 or 512 s), longest duration, data set available for ≈3 Myr old stars.

So far, the CoRoT NGC 2264 data have been used to study the correlation between optical and X-ray variability in young stars (Flaccomio et al. 2010); asteroseismological properties of two high-mass cluster members (Zwintz et al. 2011); and to identify and study the behaviour of NGC 2264 members that are AA Tau like (Alencar et al. 2010).

Alencar et al. (2010) demonstrated that the peculiar photometric behaviour of AA Tau, which consist in a flat maximum in the LC interrupted by deep quasi-periodical minima (due to obscuring material with a variable structure which is located in the inner disc region, near the corotation radius), that vary in depth and width from one rotational cycle to the other, is quite common among CTTSs (Bouvier et al. 1999, 2003, 2007; Ménard et al. 2003; Grosso et al. 2007). The interpretation for AA Tau can now be considered quite solid, and its extension to a high fraction of CTTSs simply requires that the size of the obscuring clump of material (e.g. the height of the inner disc warp) is larger than previously thought.

In a rather similar scenario of circumstellar material orbiting the star and consequent time-variable shading, Flaccomio et al. (2010) found evidence of a correlation between soft X-ray and optical variability of CTTSs (no correlation is apparent in the hard band), while no correlation in either band (soft and hard) is present in WTTSs. Flaccomio et al. (2010) suggested that this observation is consistent with a scenario in which a significant fraction of the X-ray and optical emission from CTTSs is affected by temporally variable shading and obscuration.

The conclusions of both Alencar et al. (2010) and Flaccomio et al. (2010) point towards a different origin of the observed periods, suggesting a difference both in the physical and morphological properties of CTTSs and WTTSs.

In this paper, we derive accurate rotational periods of known NGC 2264 members, testing whether a relationship between accretion and rotation exists in pre-main-sequence stars.

In the following, we describe the observational and data reduction strategy (Sections 2 and 3). In Section 4, we discuss the results obtained and our major conclusions are summarized in Section 5.

2 CoRoT OBSERVATIONS

The first short run CoRoT observation (SRa01; P.I. F Favata) lasted from 2008 March 7 to 31 (23 d) and was devoted to the observation of the very young (∼3 Myr) stellar open cluster NGC 2264, which covers most of the mass sequence from ∼3 to ∼0.1 M_⊙.

A total of 8150 stars were observed, with right ascension (RA) between 99| ${.\!\!\!\!\!\!^{\circ}}$|4 and 100| ${.\!\!\!\!\!\!^{\circ}}$|9 and declination (Dec.) between 7| ${.\!\!\!\!\!\!^{\circ}}$|6 and 10| ${.\!\!\!\!\!\!^{\circ}}$|3 and R magnitudes from 9.2 to 16.0. The sample includes 301 cluster members (see Section 3 for details regarding membership criteria). We used the so called N2 data delivered by the CoRoT pipeline (Samadi et al. 2007) after correction of the electronic offset, gain, electromagnetic interference and outliers. The pipeline includes background subtraction and partial jitter correction. Low-quality data points, e.g. taken during the South Atlantic Anomaly crossing or due to hot pixels events, are flagged. Some of the stars have LCs in three separate but ill-defined bands (red, green and blue). In these cases, our analysis was conducted on the white-light data obtained by summing the three bands. The LCs are sampled at a rate of 512 s or oversampled at 32 s. All the LCs presented here were rebinned to 512 s. Using CoRoT photometry we are able to reveal luminosity variation, with a precision down to 0.1 mmag h⁻¹ (magnitude between 11 and 16), during continuous observations, allowing us to measure photometric periods also in relatively quiet stars (for comparison, the luminosity variations of the Sun range between ∼0.3 and 0.07 mmag at maximum and minimum activity, respectively; Aigrain & Irwin 2004).

3 SAMPLE EXTRACTION AND LIGHT-CURVE ANALYSIS

We selected 301 cluster stars (whose spatial distribution is shown in Fig. 1 as big dots) satisfying one or more of the following membership criteria (the number of objects which fulfill the various criteria is indicated near each criterium).

• Detection in X-rays by Chandra ACIS or XMM–Newton (Ramírez et al. 2004; Flaccomio, Micela & Sciortino 2006; Dahm et al. 2007; Flaccomio et al., in preparation) and location on the cluster sequence in the (I, R−I) diagram, when I and R magnitudes are available (191).

• High levels of Hα  emission, indicative of accretion, according to photometric indices (Lamm et al. 2004; Sung et al. 2008) (104).

• Hα  with spectroscopic EW greater than 10 Å  and/or indicated to be in strong emission by Fűrész et al. (2006) (86).

• Classified as Class I or Class II according to Sung et al. (2009), based on Spitzer mIR photometry (76).

• Strong optical variability + high Hα emission, indicative of high chromospheric activity, according to Lamm et al. (2004) (87).

• Radial velocity members according to Fűrész et al. (2006) (192).

Figure 1.

Open in new tabDownload slide

Spatial distribution of the CoRoT targets in the SRa01 (dots) with target stars satisfying one or more membership criteria (big dots), as described in the text.

After selecting the cluster members, we classified them as CTTSs if their Hα  EW was greater than 10 Å, and WTTSs if smaller than 5 Å. The threshold between these two classes is a function of spectral type, as suggested by Martín (1998) and deeply analysed by Barrado y Navascués & Martín (2003). The information regarding the Hα  EW is available from the work of Dahm & Simon (2005) for 164 members, 86 with EW_Hα ≤ 5 Å, 19 with EW_Hα between 5 and 10 Å, and 59 with EW_Hα ≥ 10 Å. We decided to exclude intermediate EWs (5 <  EW_Hα <  10 Å), though not a usual procedure when dealing with Hα, to keep the two samples of CTTSs and WTTSs well separate and thus avoid possible ambiguity in the classification. In Fig. 2, we show the V versus (V−I) colour–magnitude diagram for the 8150 stars in the SRa01 observations with isochrones from Siess et al. (1997) (transformed to the observational plane using the Kenyon & Hartmann 1995 compilation), together with the NGC 2264 members.

Figure 2.

Open in new tabDownload slide

The colour–magnitude diagram of the 8150 SRa01 stars observed by CoRoT (grey dots). The large black dots indicate the 301 cluster stars satisfying one or more of the membership criteria. The crosses are WTTSs and the squares are CTTSs, defined following our criterium (WTTSs for EW_Hα ≤ 5 Å  and CTTSs for EW_Hα ≥ 10 Å). The curves denote isochrones (yr) from Siess, Forestini & Dougados (1997) (transformed to the observational plane using the Kenyon & Hartmann 1995 compilation).

We have analysed the CoRoT LCs as in Affer et al. (2012), we refer the reader to this work for a full description of data reduction and analysis. In brief, we prepared the LCs by correcting the following systematic effects: spurious data points (mainly due to cosmic rays and/or to the satellite crossing of the South Atlantic anomaly), as flagged by the reduction pipeline were removed; we rebinned the data to 2 h to smooth out the orbital period of the satellite (1.7 h); spurious long period trends (due to pointing and instrumental drift) were removed by fitting a third degree polynomial to the data and then dividing the original data points by this function.

The presence of periodic signals was detected using the Lomb Normalized Periodogram (LNP) technique (Scargle 1982; Horne & Baliunas 1986). With this algorithm we calculated the normalized power P(ω) as function of angular frequency ω = 2π ν and identified the location of the highest peak in the periodogram. In order to decide the significance of the peak we have followed Eaton, Herbst & Hillenbrand (1995), randomizing the temporal bins from the original LC. By calculating the maximum power on a large number of randomized data sets, the conversion from power to False Alarm Probability (FAP) can be determined. In detail, we constructed 1000 LCs resampled from the original ones randomizing the position of blocks of adjacent temporal bins (block length, 12 h; e.g. Flaccomio et al. 2005). By shuffling the data we break any possible time correlation and periodicity of the LC on time-scales longer than the block duration. We calculated the Scargle periodogram for all the randomized LCs and we compared the maximum from the real periodogram to the distribution obtained from the randomized LCs, at the same frequency, in order to establish the probability that values as high as the observed one are due to random fluctuations. Given some threshold FAP_*  we state that the detected candidate periodicity is statistically not significant if FAP >  FAP_*. The calculation we performed on CoRoT LCs led often to small FAPs, indicating that LNPs of our LCs present peaks that in most cases cannot be explained by pure stochastic noise, or non-periodic variability on time-scales shorter than 12 h. In fact, if LCs present variations on time-scale smaller than the size of the temporal block we used in the simulations, these variations will be still present in the simulated curves and will be not recognized as significant (more details in Affer et al. 2012). We have chosen a bin size of 12 h as a reasonable compromise between the expected time-scale of stochastic variations and the shortest expected periodic signal. Once significant peaks (at FAP 1 per cent; since we simulated 1000 LCs, the 1 per cent FAP power is the power that was exceeded by the highest peak in 10 simulations) were determined from the LNPs, we also used an autocorrelation analysis (Box & Jenkins 1976) to validate the periodicity of the LCs and to eliminate or correct spurious periods due to aliasing effects (in 5 per cent of the cases, with a 95 per cent confidence interval) or residual effects due to the choice of temporal blocks used in the LCs’ simulations (block length, 12 h).

Following Alencar et al. (2010), morphologically we can divide the LCs in four groups: (1) spot-like periodic LCs, whose periodicity can be interpreted as rotational modulation of surface features such as cool and/or hotspots (Fig. 3, top-left panel); (2) AA Tau-like systems, whose quasi-periodicity is likely caused by the obscuration of the stellar photosphere by circumstellar disc material (Fig. 3, top-right panel), the stability of the spot-like LCs on the time-scale of the observations, makes them easily distinguishable from the AA Tau-like ones; (3) irregular LCs, whose non-periodic brightness modulations (some of them with peak-to-peak variations up to 1 mag) are probably due to a complex mixing of non-steady accretion phenomena, and obscuration by non-uniformly distributed circumstellar material, as suggested by Alencar et al. (2010) (Fig. 3, bottom-left panel); and (4) non-variable LCs which do not display an obvious periodicity, most of which look like noisy LCs with small variability amplitude ≤1 per cent (Fig. 3, bottom-right panel).

Figure 3.

Open in new tabDownload slide

Four examples of morphologically different LCs (rebinned to 512 s): spot-like periodic (top-left panel); AA Tau-like system (top-right panel); irregular (bottom-left panel) and non-variable ‘noisy-like’ (bottom-right panel). Time is in days from 2000 January 01 (JD-245 1544.5).

To evaluate a short term variability amplitude of the LC, we calculated the running median flux, obtained using a temporal bin set up by 15 time points (thus, a time-scale of 15 × 512 s), we calculated for each time value the difference between the instantaneous flux and the running median flux derived at the corresponding time, obtaining an array. We measured the amplitude variation of the LC as the difference between the maximum and the minimum values of this array. We found that the variability amplitude range is 2 to 171 per cent for CTTSs and 1.0 to 38 per cent for WTTSs in our sample. In Fig. 4, we show the results of the complete analysis for two LCs, one belonging to a CTTS and the other to a WTTS.

Figure 4.

Open in new tabDownload slide

Two examples of NGC 2264 members: the four left-hand plots are for a CTTS, CoRoT ID 0223964667, while the four right-hand plots are for a weak line T-Tauri star, CoRoT ID 0223956264. For each example, the top panels show the LC and the relative LNP, with the dotted curve superimposed indicating the 1 per cent significance threshold determined by simulations (the periods indicated in the top-right of the LNP plot are the five most significant periods yielded by the Lomb-Scargle periodogram), while the bottom panels show the LC folded with the most significant period for each example, and the autocorrelation plot with the 95 per cent confidence interval (dotted horizontal lines) with the vertical lines indicating the position of the two most significant periods found with the LNP. Time in the LCs is in days from 2000 January 01 (JD-245 1544.5).

4 RESULTS: ROTATIONAL PERIODS

Of the 301 monitored cluster members, we found that 189 are periodic variables, with regular LCs, possibly resulting from the rotational modulation of the light by stellar spots, 20 are AA Tau-like systems, 45 are irregular LCs and 47 display no significant variability (noise dominated LCs). Among the 86 WTTSs, we measured periods for 76 stars (74 regular and 2 AA Tau like), 6 were found to be irregular and 1 non-variable. For 3 stars we did measure a period, the variability is clearly due to spot, nonetheless we discarded these periods as they are not significant according to both the LNP and autocorrelation methods. Among the 59 CTTSs, we measured periods for 33 stars (21 regular and 12 AA Tau like), 23 were found to be irregular and 1 non-variable, for two stars the periods found were discarded as not significant. In Table 1, we list the samples used in this work, indicating the morphological division performed (spot like, AA Tau like, etc.) and the membership-IR classification (Class II and Class III) following Sung et al. (2009).

All the rotational periods derived for NGC 2264 members are reported in Table 2. For each star we list: the CoRoT ID; the derived period; the variability amplitude AmpVar; the R magnitude; the B − V magnitude; the RA and Dec.; the membership-IR classification (Class II and Class III) following Sung et al. (2009) and other membership criteria listed in Sung et al. (2008); the Hα EWs from Dahm & Simon (2005) and the mass from the Siess, Dufour & Forestini (2000) tracks.