KINEMATIC AND SPATIAL SUBSTRUCTURE IN NGC 2264^*

Because of the proximity of NGC 2264 to the Galactic Plane, careful selection of candidate members was necessary to reduce contamination as much as possible. This continuing study uses the same target selection from Fűrész et al. (2006) which drew sources from X-ray (Ramírez et al. 2004), Hα surveys, and some additional sources showing UV or (infrared) IR excess from Park et al. (2000). We have added additional targets from Rebull et al. (2002) whose V-band magnitudes fall within the range of magnitudes expected for cluster members. Selection from these catalogs attempts to identify pre-main sequence cluster members in different stages of evolution; stars from the Hα and Park et al. (2000) optical study will preferentially find Classical T Tauri Stars or Class II accreting objects while the X-ray sources will find Class III or weak T Tauri Stars without substantial accretion activity. The additional targets from Rebull et al. (2002) are meant to fill unused fibers, enabling us to identify additional members not selected based on other criteria.

A subset of stars whose membership was questionable because their radial velocities differ significantly from the molecular gas in the region (the "blueshifted population"; Sections 3 and 4.2) were selected as the main targets for multi-slit observations with the Inamori-Magellan Areal Camera & Spectrograph (IMACS; Bigelow & Dressler 2003) (Section 2.4) to verify membership via detection of Li i absorption.

We have observed NGC 2264 with the multi-fiber echelle spectrograph Hectochelle (Szentgyorgyi et al. 1998) on the MMT during several epochs over the past six years. Hectochelle uses robots to position 240 fibers on a 1° field of view (FOV). The fibers subtend 14 on the sky yielding a spectral resolution of $R\;\sim$ 35000. There are limitations of fiber spacing in a given configuration; no targets may be closer than 30''. We used an order-separating filter (RV 31) to cover the wavelength range of ∼5150–5300 Å.

The spectra were reduced with an automated pipeline developed by G. Fűrész that utilizes standard spectral reduction procedures within the IRAF.⁵ A more detailed description of Hectochelle data reduction can be found in Sicilia-Aguilar et al. (2006). An additional manual step not taken care of by the pipeline was sky subtraction. A number of fibers (∼20) must be allocated for sky observations because Hectochelle observations are generally conducted in bright time and the scattered moonlight must be subtracted to measure velocities for faint sources. To subtract the sky observations, we normalized the fiber throughputs using the flat-field exposure, and then subtracted the average spectrum taken from all the sky fibers.

Fűrész et al. (2006) observed NGC 2264 in 2004 March and 2005 December; our new observations were conducted in the fall of 2007, 2008, spring 2009, fall 2009, and spring 2010; see Table 1 for details of the observations. When possible two epochs were taken of each field in a particular observing season to identify short period radial velocity variability. Because Hectochelle observations are conducted in queue mode, the second epochs could be days or even a month apart from the first epoch.

We also used MIKE Fibers (Bernstein et al. 2003; Walker et al. 2007) on the Magellan Clay telescope to observe NGC 2264 in the spectral range ∼5150–5210 Å. This instrument uses 256 manually plugged fibers that fit within a pre-drilled plate on a 25' FOV. Each fiber subtends 1 25 on the sky, resulting in a resolution $R\;\sim$ 18000 and fibers must be spaced at least 14'' apart. MIKE Fibers has two independent spectrograph channels; at 5200 Å we are able to use both channels with 128 fibers going to each channel. The two channels are essentially separate spectrographs with different gratings, optics, and CCDs in the same enclosure.

Because MIKE Fibers has a smaller FOV than Hectochelle, we selected two fields with the highest stellar density for these observations: one in the northern part of the cluster near the O7V star S Mon, and another in the southern part centered between the Spokes and Cone "clusters." As in the case of the Hectochelle observations, data were taken on two epochs several days apart.

The data were reduced using IRAF. We used the task ccdproc, to carry out initial reductions on the raw CCD imaging data: subtracting the overscan, trimming, and subtracting the bias. We combined our individual exposures with "imcombine" set for cosmic ray rejection. The spectra were extracted using "twodspec," by first tracing the apertures in a flat field frame taken of a continuum source to create a map of aperture traces to extract the science spectra. The wavelength solution was calibrated by fitting a 4th order Legendre polynomial to the Thorium–Argon (Th–Ar) lamp spectrum taken before and after the sets of science observations.

Low-resolution spectra for the purpose of determining membership from Li i 6707 Å absorption were obtained with IMACS on the Magellan Baade telescope, during December 20–22 2010 (UT). The setup was the same as used in our study of the L1641 region of the Orion A cloud (Hsu et al. 2012). We used the IMACS f/2 camera in multi-slit spectroscopy mode with the 300 line grism at a blaze angle of 175. With a 06 slit, this configuration yields a resolution of 4 Å and spectral coverage from approximately 4000–9000 Å. The standard observation time for each field is 5 × 10 minutes, but we increased the time to 6 × 10 minutes for a few observations at higher airmasses. A total of 160 stars were observed, selected primarily from the stars that appear slightly blueshifted from the gas (see Sections 3 and 4.2).

We used the rvsao package in IRAF (Mink & Kurtz 1998), to measure the radial velocities of our observed targets. The radial velocity of an object is determined by cross-correlating the observed spectrum with a template spectrum of known velocity. The cross-correlation signal-to-noise (S/N) or quality is given by R, defined as

$$\begin{eqnarray}&&R=\frac{h}{{{2}^{1/2}}{{\sigma }_{a}}}\end{eqnarray} \tag{ 1 }$$

where h is the peak height of the correlation function and ${{\sigma }_{a}}$ is the rms noise estimated from the antisymmetric portion of the correlation function. The velocity measurement error also depends on the spectral resolution; it is of the form

$$\begin{eqnarray}&&{{\sigma }_{v}}\sim \frac{C}{1+R}\end{eqnarray} \tag{ 2 }$$

where C is 20 km s⁻¹ for MIKE Fibers and 14 km s⁻¹ for Hectochelle. These values were determined empirically by adding random noise to a high S/N spectrum as in Hartmann et al. (1986).

We used libraries of synthetic stellar spectra as velocity templates rather than observed velocity standards, enabling us to explore a wider range of stellar parameters than a few observed templates. We used the library by Coelho et al. (2005) for velocity measurement of spectra from both Hectochelle and MIKE Fibers. The Coelho et al. (2005) templates are computed with a resolution of R ∼ 100,000, much higher resolution than the observed spectra. However, a cosine-bell filter function is applied to the Fourier transform of the template and observed spectrum within the xcsao task. This effectively reduces the template resolution to that of the Hectochelle or MIKE Fibers instrument. The filter applied to the data from both instruments (and templates) has an inner cutoff at a wave number of 10 and the filter function increases to 1 at a wave number of 40. The purpose of the inner cutoff is to limit the effect of large-scale features in the cross-correlation. The outer cutoff values depend on the instrument and limit on the smallest features considered in the cross-correlation. Tonry & Davis (1979) suggests an outer cutoff of ∼N_pixels/(2π × FWHM/2.355), yielding ∼400 for Hectochelle and ∼80 for MIKE Fibers. However, we found that outer wavenumber cutoff beginning at 120 and reaching zero at 450 worked the best for MIKE Fibers and a cutoff beginning at 600, reaching zero at 1000 worked best for Hectochelle.

We used a subset of the spectral templates with surface gravity log(g) = 3.5, effective temperatures (T_eff) ranging from 3500–7000 K in steps of 250 K, and solar metallicity. There was no need to explore a wider parameter space with the templates because the most important factor determining the quality of a correlation was the T_eff of the template.

The results from the correlation with the highest R value are selected, matched to the appropriate target coordinates, and stored in a Starbase database (Roll 1996). The right ascension and declination are used to correlate the new observations with the existing catalog and the coordinates are used to generate a truncated 2MASS ID number (2MASSID), throwing out fractional seconds in right ascension and declination. Once the targets have been matched in right ascension and declination, the 2MASSID is used to identify sources in further analysis.

To combine radial velocities taken in different epochs, it is necessary to determine zero-point velocity shifts for an entire observation. This compensates for possibly different calibration schemes, different wavelengths, different instruments, as well as temperature variations within the spectrograph (Sicilia-Aguilar et al. 2006; Fűrész et al. 2006; Fűrész et al. 2008). We also accounted for the well-characterized fiber-to-fiber velocity offset caused by the calibration lamps not illuminating the fibers in the same way as astronomical objects.

We adopted the results of Fűrész et al. (2006) as baseline velocities to shift our observations to match. This is because the Fűrész et al. (2006) velocities are referenced to the radial velocity standard star W23870 (Latham et al. 1991), having an absolute velocity accuracy of ±0.2 km s⁻¹ (Stefanik et al. 1999). We applied a constant shift for each field given by the mean of a Gaussian fit to the distribution of radial velocity differences for each target in a field. The zero-point shifts for the Hectochelle and MIKE Fibers fields are generally quite small, less than 0.5 km s⁻¹ and are given in Tables 1 and 2. In contrast, observations of the ONC in Tobin et al. (2009) had zero-point shifts ∼1 km s⁻¹. Given than many NGC 2264 fields were conducted in the same semester and even same nights as the Orion fields, the zero-point shifts may be overestimated in Tobin et al. (2009). We believe that the NGC 2264 fields track the zero-point shift better than the Orion fields due to the inclusion of many bright non-members in the Galactic Plane that give velocity precisions of 0.5–0.2 km s⁻¹(note that this is the formal error and does not include systematic effects).

We have converted the heliocentric radial velocities of each target to the kinematic LSR velocities (Kerr & Lynden-Bell 1986). In order to compare the velocity structure of the stars and gas, the LSR velocities are used exclusively in the plots, but the heliocentric radial velocities are given in Table 3.

Table 3.  Velocity Summary

2massID	R.A.	Decl.	$\overline{RV}$ a	N Epochsb	Lithium EW	Spectral Type	K − [3.6]	[3.6] − [4.5]
	(J2000)	(J2000)	(km s− 1)		(Å)
0640153 + 094242	06:40:15.35	09:42:42.4	12.08 ± 0.42	8	0.38	65.9	0.14	-0.06
0640160 + 095737	06:40:16.04	09:57:37.7	−5.04 ± 0.48	3	0.00	61.0	0.24	0.00
0640176 + 094758	06:40:17.62	09:47:58.0	0.54 ± 0.94	2	0.00	65.7	0.19	0.05
0640200 + 092828	06:40:20.09	09:28:28.5	5.56 ± 0.04	2	0.67	67.0	0.33	0.34
0640210 + 094628	06:40:21.06	09:46:28.0	6.96 ± 0.94	2	0.00	66.0	0.20	0.04

Notes. Abridged table of target IDs, positions, average radial velocities, Lithium 6707 Å Equivalent Widths, Spectral Types, and near infrared colors. Entries with −99 signify that these data are unavailable for a given target.

^aVelocities from Fűrész et al. (2006) are included in average. For a measurement to be included in the average R must be >6.0; R is defined in Equation (1). ^bNumber of observations with R > 6.0.

Only a portion of this table is shown here to demonstrate its form and content. Machine-readable and Virtual Observatory (VO) versions of the full table are available.

Download table as:  Machine-readable (MRT)Virtual Observatory (VOT)Typeset image

The multi-epoch observations have enabled us to also identify a number of candidate spectroscopic binaries from radial velocity variability and/or double-peaked correlation functions. We have removed these from the cluster kinematic analysis because they will broaden the cluster velocity distribution. These candidate spectroscopic binaries will be presented in a subsequent paper along with observations from later epochs.

Spectral types and Li i 6707 Å absorption equivalent widths were determined for a subset of target stars observed with IMACS. We used SPTCLASS, a semi-automatic spectral-typing program (Hernández et al. 2004), which uses empirical relations of spectral type and absorption line equivalent widths for classification (see Hsu et al. 2012 for details). SPTCLASS also automatically measures Li i equivalent widths, but we made manual determinations because of the uncertainty in automatically setting continuum levels at this resolution when strong TiO bands are present.

The overall kinematic structure of NGC 2264 bears a striking similarity to that found in the Orion Nebula Cluster (ONC). As shown in Figure 3 of Tobin et al. (2009), the position–velocity plot of both stars and gas shows a northern region redshifted with respect to the central region. The right panel of Figure 2 shows qualitatively similar behavior, although more clearly in the gas than in the stars. One also observes the same asymmetry between the stellar and gas velocities, with a population of stars blueshifted with respect to the molecular gas. The one-dimensional velocity dispersions of the two clusters are also similar, σ = 3.1 km s⁻¹ for Orion (Fűrész et al. 2008) and σ = 2.5 km s⁻¹ for NGC 2264. NGC 2264 does differ from the ONC with its two components in the molecular gas in the north near S Mon (three if the clump west of S Mon is considered), and the gas in the ONC has only one component. The stars in the northern part of the ONC also have a smaller velocity dispersion with respect to the center of the ONC, while in NGC 2264 the stars in the north have a larger spread in velocity than the northern ONC stars.

Proszkow et al. (2009) suggested that the transition from redshifted stars and gas in the northern regions of the ONC to more blueshifted velocities near the cluster center could be the signature of gravitationally-driven infall. The interpretation is qualitatively similar to using "caustics" in position–velocity space to infer gravitational motion in galaxy clusters (e.g., Geller et al. 2013). To see whether a similar interpretation could hold for NGC 2264, we compute a "binding velocity" ${{v}_{b}}\sim {{(2{\rm GM}/r)}^{1/2}}$. Crutcher et al. (1978) estimated a total gas mass of ∼8000 ${{M}_{}}$ spanning a total length of ∼10 pc. Using this mass and setting $r\sim 5$ pc yields v_b ∼3.6 km s⁻¹.

However, while this scenario was plausible for Orion, it seems less likely to explain the redshifted population in the north of NGC 2264. This group has V(lsr) ∼ 10 km s⁻¹ which is redshifted by about 5 km s⁻¹ with respect to the gas and stellar velocities at the position of the Spokes Cluster. Unless the gas mass (which is the dominant contributor) is underestimated by roughly a factor of four, it seems very likely that this redshifted population is not bound to the rest of the cloud. Rather, we suggest this group is more likely to be the result of blowout by the wind from S Mon; this would also explain the "hole" in the gas at $\delta \sim 9\buildrel{\circ}\over{.} 8$, $V({\rm lsr})\sim 7$ km s⁻¹.

With respect to the rest of the cluster, the binding velocity calculated from the molecular gas is larger than the one-dimensional velocity dispersion (${{\sigma }_{v}}$ = 2.5 km s⁻¹) within a restricted range of R.A. (bottom right panel of Figure 3). However, if the velocity dispersion is isotropic, the three-dimensional velocity dispersion would be ∼4.3 km s⁻¹ and thus suggests that the region is unbound. In addition, if our suggestion that a redshifted population is missing from our optical sample is correct, the 1 d velocity dispersion of the total stellar population is larger than we measure here. Resolution of these question will wait until we have proper motions from Gaia. For now, we conjecture that the true situation is a mix of possibilities, with some fraction of the stars that will eventually be left behind in a bound cluster and some dispersing.

As discussed in the previous section, it appears that a substantial fraction of the blueshifted stars are causally-related to NGC 2264 if not members, based on the indicators of youth—IR excesses and Li absorption. One possibility for explaining why this population is not at the same velocity as the molecular gas is that it might be older, resulting from a previous episode of star formation. To test this idea, we constructed color–magnitude diagrams for the various kinematic groups. As seen in Figure 5, the color-magnitude diagram (CMD) of the "field" stars with velocities < −5 km s⁻¹ and >18 km s⁻¹ from the cluster mean shows a much larger dispersion around the median color–magnitude line of the cluster. We observed an excess of faint stars, as would be expected for foreground main sequence stars, and brighter, redder background giants. Moreover, the blueshifted stars tend to be brighter and bluer in color, while the redshifted stars are fainter, redder, and have more dispersion in magnitude.

Zoom In Zoom Out Reset image size

The CMDs of the various probable cluster member velocity groups are essentially indistinguishable except for the redshifted group with 8 km s⁻¹ $\lt V({\rm lsr})\lt 15$ km s⁻¹, which exhibits a broader dispersion. This could be due to contamination because the northern region, where much of the 8 km s⁻¹$\lt V({\rm lsr})\lt 15$ km s⁻¹population is found, has less extinction to cut out background stars (and is also sparser). The blueshifted group (−2 km s⁻¹ $\lt V({\rm lsr})\lt 2$ km s⁻¹) appears to have a larger fraction of redder stars than the main cluster as well. However, because these photometric data have not been de-reddened, it is impossible to decouple evolutionary effects or population differences from extinction. Notice that the reddening vector shown in Figure 5 is nearly parallel to the median CMD of the cluster. On the whole, there is a significant spread in V at a given $V-{{I}_{C}}$ color, but much of this is undoubtedly due to an unresolved binary population; other sources of error are probably present but poorly understood (Jeffries 2012). In any case, there is little evidence for any systematic age differences larger than about ∼2–3 Myr, comparable to the median estimated age of the cluster (Sung & Bessell 2010).

Systematic ejection of stars toward earth seems unlikely if not implausible, especially because the ONC shows the same excess of blueshifted stellar velocities relative to the molecular gas velocities. We next consider other possibilities for the asymmetry between the stellar and gas velocities. The first possibility recognizes that we have an optically biased sample of radial velocities; stars moving away from us would move more deeply into the molecular cloud and thus have much higher line-of-sight extinctions. It seems very likely that some bias exists; as shown in Figure 3, we observe relatively few stars at the reddest velocities of the molecular gas, which would be surprising if not attributable to sample bias. Whether there exists a substantial population of redshifted stars with velocity shifts of comparable magnitude to the blueshifted population is not clear. If such a population did exist, the overall velocity dispersion would increase to ${{\sigma }_{v}}\sim 5$ km s⁻¹, making it even less likely that any part of the region is gravitationally-bound.

An alternative possibility to explain the asymmetry between stars and gas is preferential dispersal of blueshifted (nearside) gas by stellar energy input. We argued above that S Mon is blowing away gas, producing the "hole" in the ¹³CO emission at ∼7 km s⁻¹, $\delta \sim 9\buildrel{\circ}\over{.} 8$. It seems conceivable that this O7 star could have dispersed additional molecular gas. Schwartz et al. (1985) argued that S Mon is responsible for photoionizing the rim of the Cone Nebula at a projected distance of about 6 pc. Furthermore, they argued that S Mon and the Cone Nebula might actually lie a substantial distance in the foreground from other dense molecular gas in the region (basically, the Spokes Cluster region). Whether or not this picture is correct, it is possible that S Mon photoevaporated and photoionized the gas originally on the near side of the original molecular cloud.

The simulations of photodissociation and ionization of initially molecular gas with magnetic fields and turbulence by Arthur et al. (2011) show substantial evacuation of a region of size 4 pc and initial density $n={{10}^{3}}\;{\rm c}{{{\rm m}}^{-3}}$ in only $4\times {{10}^{5}}$ yr for an assumed O9 star, which emits a factor of 3–4 fewer ionizing photons than an O7 star is expected to provide (Osterbrock & Ferland 2006, Table 2.3). Thus, it is possible that there had been and episode of star formation in the near side of the molecular cloud, prior to the formation of S Mon. Once S Mon formed and began photoionizing the molecular gas associated with the previous generation of star formation, the molecular gas removal caused the previously formed stars to become unbound and expand (e.g., Lada & Lada 2003). This scenario would produce a population of stars systematically blueshifted with respect to the cloud velocity, because we are able to observe the blueshifted stars due to their low extinction, but the redshifted stars move into the molecular cloud and become too extincted to determine radial velocities in the optical. We therefore suggest that the observed kinematic differences between the stellar population and the molecular gas result from a combination of optical bias against stars moving into the molecular gas and blowout of molecular gas on the near side by S Mon. The low extinction observed toward some members (Walker 1956; Park et al. 2000) is clear evidence for some gas dispersal to have taken place.

While we cannot determine a systematic age difference between the blueshifted population and the rest of the cluster, Sung & Bessell (2010) argued that the "field" and "halo" populations of NGC 2264 are older than the rest of the cluster. These regions are designated as such because they they are spread across the NGC 2264 region on the sky and the blueshifted population of stars is found at all positions in NGC 2264 (see Figure 4). It is noteworthy that Orion also has several know foreground populations as well, projected on the same region of the sky as the ONC (e.g., Bally 2008; Alves & Bouy 2012), possibly contributing to the blueshifted population observed there.