Chapter 6
The LuckyCam Survey for VLM Binaries
I. M5.5-M8 dwarfs within 40pc

In this chapter I describe the first part of the LuckyCam VLM binary survey, a survey of M5.5-M8 M-dwarfs within 40pc.

6.1 Introduction – VLM Binaries

There are compelling reasons to search for companions to nearby stars. In particular, the properties of binary systems provide important clues to their formation processes. Any successful model of star formation must be able to account for both the frequency of multiple star systems and their properties (separation, eccentricity and so forth) – as well as variations in those properties as a function of system mass. In addition, the orbits of binary systems provide us with the means to directly measure the mass of each component in the system. This is fundamental to the calibration of the mass-luminosity relation (MLR: Henry & McCarthy 1993Henry et al. 1999Ségransan et al. 2000).

The stellar multiplicity fraction appears to decrease with decreasing primary mass (eg. Siegler et al. 2005). Around 57% of solar-type stars (F7–G9) have known stellar companions (Abt & Levy1976Duquennoy & Mayor1991), while imaging and radial velocity surveys of early M dwarfs suggest that between 25% & 42% have companions (Henry & McCarthy1990Fischer & Marcy1992Leinert et al.1997Reid & Gizis1997). Later spectral types have been studied primarily with high resolution adaptive optics imaging: Close et al. 2003 and Siegler et al. 2005 find binary fractions of around 1020% for primary spectral types in the range M6–L1. Bouy et al. 2003 and Gizis et al. 2003 find that 10–15% of L dwarfs have companions, and Burgasser et al. 2003 find that 10% of T dwarfs have binaries. These very low mass (VLM) M, L and T systems appear to have a tighter and closer distribution of orbital separations, peaking at around 4 AU compared to 30 AU for G dwarfs (Close et al.2003).

However, each of these surveys have inevitably different (and hard to quantify) sensitivities, the effect of which is especially evident in the large spread in the derived multiplicity of early M-dwarfs. In particular, high-resolution imaging surveys are sensitive only to companions wider than ~0.1” while radial velocity surveys are much more sensitive to closer (shorter period) companions. Maxted & Jeffries (2005), by examining a small sample of radial velocity measurements, estimate that accounting for systems with orbital radius <3 AU could increase the overall observed VLM star/BD binary frequency to 32–45%. Basri & Reiners (2006) also find that taking into account spectroscopic binaries doubles the binary fraction (in a more heterogenous sample of VLM stars and brown dwarfs).

The orbital radius distribution of VLM binary systems is also quite uncertain (figure 6.1). A well-constrained peak is found at around 4 AU, but several systems have been found at much wider radii. It is important to enlarge the sample of VLM binaries to ascertain how common these wide systems are, and whether they form a separate population of large-radius systems or are simply the tail of the distribution which peaks at 4 AU.

For these reasons, we decided to use LuckyCam to target several well-understood samples of low-mass and very low mass stars. The surveys have the aim of both increasing the known number of VLM binary systems (and finding more exotic systems such as triples and substellar companions), and constraining the binary statistics in a number of different samples.

In this thesis, I define VLM stars to have a V-K colour of >6 (M5 and later, Leggett (1992)) and therefore a mass of < 0.11M, following the stellar models described in Baraffe et al. (1998) for ages >2 Gyr. The distinction between VLM and low mass stars is somewhat arbitrary, and some authors (eg. Burgasser et al. (2006)) would place the cut-off at the slightly lower primary mass of 0.10M .

6.1.1 Why high-resolution imaging?

VLM binary stars are an excellent example of a science programme which cannot be performed with seeing-limited instruments. Figure 6.1 shows the distribution of the orbital radii of known VLM binaries as of 2005 (prior to the LuckyCam survey). Only 45 VLM binaries were known (the figure at the time of writing is only on the order of 80, spanning a wide range of system total mass, mass ratios and separations). Almost all are at smaller orbital radii than 10AU.


Figure 6.1: The histogram of known VLM binaries, from data collated in Siegler et al. (2005). Poisson error bars are shown on the histogram bins; note the uncertainty in the large-radius orbital distribution.

Generating a larger sample of VLM binaries for statistical studies of the nature of these systems, the aim of the LuckyCam survey and others like it, requires including target stars up to a sufficiently large distance from the Sun that a large number of M-dwarfs are within the survey volume. As figure 6.1 shows, at 10pc distance almost all VLM binaries have separations smaller than 1.0 arcsec, and 60% are at separations smaller than 0.5 arcsec. Essentially all the stars within 10pc have been surveyed for close companions; significantly extending the VLM binary sample thus requires observations with consistently better than 0.5 arcsec resolution if any large number of new systems are to be resolved. Detailed studies of the nature of the systems and the measurement of accurate astrometry requires still higher resolution. A high-resolution imaging system, capable of around 0.1 arcsec resolution, is thus required.

6.1.2 Why LuckyCam?

VLM binary surveys have been pursued using HST (eg. Bouy et al. (2003); Gizis et al. (2003)) and AO (eg. Close et al. (2003); Siegler et al. (2005)) systems. Obtaining time for an extensive survey using HST is difficult, given the popularity of the telescope. Adaptive optics surveys can be very time intensive, as each of the tens-to-hundreds of targets requires a separate re-lock of the AO systems. For example, the NAOS-CONICA AO system on the VLT requires 5-10 minutes for AO acquisition per target before data can be taken (VLT NAOS-CONICA User Manual* , 2006). Furthermore, the M-dwarfs in question are very red objects, and are thus faint in the visible wavelengths used by most AO systems for wavefront sensing (NAOS-CONICA has both visible and infrared WFSs).

LuckyCam offers capabilities that make it an excellent instrument for VLM binary surveys. It is a completely passive system, so data is taken as soon as the telescope is pointed. A complete observation of a candidate VLM binary target usually takes 6-7 minutes, including telescope pointing, target finding and 100 second integrations in two filters. The faint guide star capabilities of LuckyCam also allow observations of M-dwarfs that are fainter than most AO systems can use (apart from those with infrared wavefront sensors, as these objects are very red). For example, the binaries newly discovered with LuckyCam in chapter 7 are all fainter than the previously known ones in the same sample of targets.

In this chapter I present results from a 32-star VLM binary sample, completed in only 5 hours of on-sky time. I present five new VLM binaries and evaluate the utility of LuckyCam for programmes of this type, including the use of calibrated high-spatial resolution photometry.


Figure 6.2: The observed sample (crosses), plotted in a V/V-K colour-magnitude diagram. The background distribution shows all stars in the LSPM-North catalogue. Circles show 38 spectroscopically confirmed M6 and later dwarfs from Cruz et al. 2003, which also appear in the LSPM-North; all but two (both M6) are recovered by the selection criteria.

Table 6.1: The observed sample. The quoted V & K magnitudes are taken from the LSPM catalogue. K magnitudes are based on 2MASS photometry; the LSPM-North V-band photometry is estimated from photographic BJ and RF magnitudes and is thus approximate only, but is sufficient for spectral type estimation (figure 6.2). Spectral types and distances are estimated from the V & K photometry and the young-disk photometric parallax relations described in Leggett (1992). Spectral types are accurate to approximately 0.5 spectral classes and distances to ~ 30%.
LSPMID------2MASS-ID-------------V--V-K--PM--/ arcsec/yr-Estimated-spectral type--Photom.-dist/pc--Newly-detected-companion?--
LSPMJ1235+1318   12351726+131805   18.0  7.7  0.219          M6.5                             14              *
LSPMJ1235+1709   12351850+170937   19.3  7.5  0.570          M6.5                             29
LSPMJ1246+0706   12460939+070624   17.7  7.1  0.549          M6.0                             19

LSPMJ1303+2414   13034100+241402   19.6  7.9  0.370          M7.0                             24
LSPMJ1305+1934   13053667+193456   18.4  7.3  0.554          M6.5                             22
LSPMJ1314+1320   13142039+132001   15.9  7.2  0.307          M6.0                            7.7              *
LSPMJ1336+1022   13365393+102251   18.7  7.3  0.381          M6.5                             26
LSPMJ1341+0805   13413291+080504   18.5  7.4  0.269          M6.5                             22

LSPMJ1354+0846   13540876+084608   19.3  8.2  0.219          M7.0                             17
LSPMJ1423+1318   14231683+131809   17.9  7.3  0.174          M6.5                             18              *
LSPMJ1423+1426   14234378+142651   17.9  7.7  0.638          M6.5                             13
LSPMJ1428+1356   14280419+135613   18.3  8.3  0.605          M8.0                             10
LSPMJ1432+0811   14320849+081131   16.3  7.2  0.455          M6.0                            9.2

LSPMJ1440+1339   14402293+133923   19.0  7.7  0.337          M6.5                             22
LSPMJ1454+2852   14542356+285159   18.6  7.0  0.212          M6.0                             32
LSPMJ1516+3910   15164073+391048   17.1  7.3  0.224          M6.5                             12
LSPMJ1554+1639   15540031+163950   19.9  7.8  0.529          M7.0                             30

LSPMJ1605+6912   16050677+691232   19.3  7.5  0.224          M6.5                             29
LSPMJ1606+4054   16063390+405421   17.6  7.6  0.735          M6.5                             12
LSPMJ1622+4934   16225554+493457   19.4  7.2  0.316          M6.0                             39
LSPMJ1626+2512   16263531+251235   19.3  7.5  0.271          M6.5                             29
LSPMJ1646+3434   16463154+343455   16.6  7.0  0.550          M6.0                             13

LSPMJ1647+4117   16470576+411706   18.5  7.4  0.289          M6.5                             22
LSPMJ1653+0000   16531534+000014   18.6  7.8  0.287          M7.0                             17
LSPMJ1657+2448   16572919+244850   18.8  7.5  0.391          M7.5                             23
LSPMJ1703+5910   17031418+591048   18.8  7.0  0.572          M6.0                             35
LSPMJ1735+2634   17351296+263447   19.1  9.0  0.349          M9.0                            9.2              *

LSPMJ1741+0940   17415439+094053   18.7  7.8  0.435          M7.0                             17
LSPMJ1758+3157   17580020+315726   18.2  7.1  0.158          M6.0                             24
LSPMJ1809+2128   18095137+212806   18.3  7.1  0.193          M6.0                             25              *
LSPMJ1816+2118   18161901+211816   19.0  7.2  0.171          M6.0                             32
LSPMJ1845+3853   18451889+385324   19.4  8.4  0.408          M8.0                             16

6.2 The Sample

We selected a distance, flux and colour limited sample of stars from the LSPM-North Catalogue (Lépine & Shara2005), which is the result of a systematic search for stars with declination > 0 and proper motion > 0.15”/year in the Digitized Sky Surveys. Most stars in the catalogue have 2MASS IR photometry as well as V-band magnitudes estimated from the photographic BJ and RF bands.

6.2.1 Proper motion biases

The selection of high-proper-motion stars ensures that the stars are nearby. Historically, the vast majority of stars now known to be in the solar neighbourhood were first identified as high-proper-motion stars (Lépine2005). However, stars with motion vectors pointing towards or away from the sun are not detected, and some distant stars can have large proper motions if their relative velocity is also large, such as stars on Galactic halo orbits (Lépine2005). For this reason, to construct a well-defined sample, I use colour, magnitude and photometric distance cuts to narrow down the parameter space of the targets. Giant stars are excluded by the proper motion and magnitude cut, because even halo stars, with velocities typically 2-5× larger than disk objects, are only likely to be 2-5× more distant than the rest of the candidates (Lépine & Shara2005).

Lépine (2005) finds that the census of nuclear-burning stars within 33pc is ~68% complete in the LSPM (and 82% complete within 25pc), where the main source of incompleteness is the lower proper motion limit. High-velocity stars (halo, old disk) tend to be selected from a larger distance, and may thus bias the sample (Lépine & Shara2005). However, in both the survey presented in this chapter and that in chapter 7 we select stars which are significantly brighter than the V=+19.0m 90% completeness limit of the LSPM North survey (figure 6.2), excluding the more distant population of high-velocity stars.

6.2.2 Target List

The properties of the selected stars are detailed below:

  1. V - K > 7; thus selecting approximately M6 and later stars (Leggett1992).
  2. Distance < 40pc. Absolute magnitudes are estimated from the V-K colours quoted in the LSPM and the V-K vs. MK relations described in Leggett (1992). Distances are then estimated by comparing the estimated absolute magnitude to the observed K magnitude.
  3. mi < +15.5; Lucky Imaging requires a mi = 15.5 guide star for full performance. All targets serve as their own guide star.
  4. We removed all stars from the remaining sample that had been to our knowledge previously observed at high angular resolution.

The remaining sample consists of 91 stars in the R.A. and declination range that was accessible during the survey period (June 2005). 32 were selected for these observations (picking the brightest targets first, as well as those close to the zenith during the observations), and are detailed in table 6.1. The region of colour-magnitude space in which they are found is shown in figure 6.2; the distributions of magnitudes and colours are detailed in figure 6.3.

The V-band LSPM photometry has been estimated from observations in the photographic BJ and RF bands (as detailed in Lépine & Shara 2005), and its use therefore requires some caution. To test its utility for late M-dwarf target selection I have confirmed that a sample of spectroscopically confirmed late M-dwarfs (Cruz et al.2003) is fully recovered by the V-K selection (figure 6.2). In addition, LuckyCam resolved SDSS i’ and z’ photometry gives confirmation of estimated spectral type for the objects in the full survey. In all checked cases the spectral type and distance estimated from LSPM-North V-K photometry matches that derived from LuckyCam SDSS i’ and z’ photometry. The LSPM V & K photometry is used extensively in the survey described in chapter 7, where more detailed checks on its accuracy are made.

6.3 Observations

We performed observations with the Cambridge Lucky Imaging system, LuckyCam, on the 2.56m Nordic Optical Telescope in June 2005, during 5 hours of on-sky time spread over 4 night observing run. Each target was observed for 100 seconds in each of the SDSS i’ and z’ filters. SDSS standard stars (Smith et al. 2002) were observed for photometric calibration; globular clusters and similar fields were imaged for astrometric calibration. The seeing measured by the Isaac Newton Group RoboDIMM at the observatory site varied between 0.5” and 1.0” during the observations, with a median of ~ 0.8”.

Each target observation was completed in an average of 10 minutes including telescope pointing, 100 seconds of integration in each filter, the observation of one standard star for every three targets, and all other overheads.


Figure 6.3: The distribution of the target sample in K magnitude and V-K colour, both from LSPM-North photometry.

For this survey we operated LuckyCam at 30 frames per second, with each frame being 552x360 pixels. The image scale was 0.04~/pixel, giving a field of view of 22×14.4 arcseconds2. The observations totalled approximately 100GB. The dataset was reduced with the standard Lucky Imaging pipeline (chapter 3).

6.3.1 Binary detection and photometry

Obvious binaries with a low contrast ratio and/or > 0.5” separation were detected by eye in reduced images including <= 10% of frames. I limited the companion detection radius to 1.5”, allowing use of the remainder of the 20”x14” fields as control areas. Because the chance probability of an object falling within the small detection radius is very low, any detections are likely to be physically associated with the target star.

In order to detect fainter companions I fit and subtract a model Moffat profile point spread function (see chapters 4 and 5) to each target, using 50% of the recorded frames to increase the SNR of faint companions at the expense of some resolution. Candidate companions were detected using a sliding-box method, with custom software implementing the detection criteria.

We stipulated the detection of a faint companion to require a 10σ deviation above the background noise, which is due to both photon and speckle noise and varies with distance from the primary star. The background noise at each radius was specified to be the upper 1σ excursion from the average RMS noise at several azimuthal positions.

In addition, I implemented the following criteria to confirm the detection of faint companions:

  1. the candidate must maintain a constant flux per frame as more frames are included in the reduced images - most persistent speckles appear in only a fraction of frames and so fail this test.
  2. detection must be repeated in the same position in each filter.
  3. the candidate must appear point-like (extended objects are thus not detected).
  4. the candidate must not be visible in the PSFs of other stars observed within a few minutes of the target, thus any removing persistent speckles and ghost images.

I measured resolved flux measurements and errors for binaries wider than 0.4” with simple aperture photometry. However, at closer radii more sophisticated strategies were required, especially since four of the five detected binaries had primaries fainter than i’=15m. As detailed in chapter 5, if a Lucky Imaging guide star is faint, its PSF is altered by frame selection proceeding partially on the basis of high excursions of photon-shot-noise. Since the companion and primary now have different PSFs, point spread function subtraction is difficult (although possible with sufficiently similar calibrator binary observations). The closest binary detected in these observations was sufficiently bright to avoid these problems, however.

Extensive experimentation confirmed that simple aperture photometry also provides accurate flux measurements at close radii for Lucky Imaging PSFs, provided that care is taken in the choice of foreground and background aperture sizes. The contrast ratios of the two close binaries with relatively faint primaries were reduced in this manner. The accuracy and precision of the derived contrast ratios was measured by repetition over several different frame selection fractions (and thus several different PSFs). The accuracy of the algorithms was also checked against simulated binary images. Note that these reductions were performed before the development of the more sophisticated PSF fitting routines described in chapter 5 and used in chapter 7.


Figure 6.4: Calibration of the LuckyCam photometry against SDSS standards from Smith et al. (2002). The dashed line has a gradient of 1.0. As all the targets are very red the standards with the highest i’-z’ available were used; there is no i’-z’ colour term detectable within the photometric errors. As shown in chapter 2, the photometric accuracy of the LuckyCam system has also been verified with much fainter targets.


Figure 6.5: The 10σ contrast ratios achieved in a variety of conditions. The i’=+14.3m star is shown in both 0.5” and 0.8” seeing; the imaging performance is not dependent on this small difference in seeing for bright stars. However, poorer seeing reduces the average light per pixel sufficiently to affect the faint guide star imaging performance (similar to the rapid falloff in image quality with guide star brightness described in chapter 4). The detected binaries are also shown; the measured contrast ratio uncertainties are often smaller than the plotted points. At radii < 0.2” the cell-size of the faint companion detection algorithm does not adequately sample the shape of the PSF; detectable contrast ratios in this area are approximately 2 magnitudes, as demonstrated by the clear detection of the faint 0.15 arcsec companion to Ross 530 (section 6.4.5).


Figure 6.6: Simulated faint companion PSFs around an i’=14m star. At each radius in each of the 2800 frames in this observation I added to the background a version of the primary PSF, rescaled to the 10σ detection flux shown in figure 6.5 and with appropriate Poisson and L3CCD multiplication noise added. For clarity the primary’s PSF has been subtracted. The simulated companions become brighter at lower radii to allow detection above both photon and speckle noise from the primary’s PSF. The residual ring around the central star is largely due to the Airy ring expected in near diffraction limited images, which is not included in the model PSF.

For this survey, I measured photometry in the SDSS system from the total integrated flux in a 3” radius aperture, calibrated against SDSS standards (Smith et al. 2002, figure 6.4). I then used the measured contrast ratios to derive resolved photometry. In each observation the L3CCD gain is calibrated as described in chapter 2. The calibrated magnitudes are then calculated as:

mag = ZP - 2.5log10(gain× DN ∕sec)

where the raw photometry is given in data numbers (DN) per second, gain is measured in photons per DN, and ZP is the photometric zero point of the system. Airmass corrections are not calculated because all observations were performed within 30o of the zenith; the approximately 10% uncertainty in the L3CCD gain calibration dominated the remaining errors.

6.3.2 Sensitivity

Beyond 1.0” radius from the primary detection sensitivity is primarily limited by the sky background; at smaller radii both azimuthal variations in the target star’s PSF and its photon shot noise limit the detection sensitivity. The SDSS i’ detection contrast ratios for two typical stars are shown in figure 6.5 and example faint companion simulated PSFs are shown in figure 6.6.

The survey is sensitive to the detection of brown dwarf companions around all the surveyed stars. For example, around a star with mi 14, the survey is sensitive to mi = 5 at radii > 1.0” and mi = 4 at > 0.5”. At an age of 5.0Gyr an object at the hydrogen-burning limit has Mi 16 . Typical stars in this sample have Mi 13 - 15. Figure 6.5 thus implies the survey is sensitive to brown dwarf companions at > 1.0” around all surveyed stars – and much closer (or, equivalently, much fainter) brown dwarf companions around many of the stars.

6.4 Results & Analysis

I detected five previously unknown binaries in the sample, with separations ranging from 0.13” to 1.12”. Resolved i’ and z’ images of each binary are given in figure 6.7; their directly observed properties are summarised in table 6.2.


Figure 6.7: Lucky imaging of the five new binaries. All images are a 10% selection from 3000 frames taken at 30FPS, with the exception of the 0.13” binary LSPM J1314+1320 which required a more stringent selection. The very close binary’s i’ image is the result of a 2.5% selection from 3000 frames and its z’ image is a 1.0% selection from 10,000 frames taken at the higher frame rate of 50FPS. In each case the image greyscale covers the full dynamic range of the images. All images are orientated with North up and East left. In many cases the guide star has a very compact core, due to the Lucky Imaging system aligning images partially on the basis of high photon-shot-noise excursions.

6.4.1 Confirmation of the binaries

In the entire June 2005 VLM binary survey observation set (including those targets from the chapter 7 sample that were observed in June 2005), covering (22′′× 14.4′′) × 48 fields, there is only one field object. The object would be characterised as a companion if it had fallen close to a target star (the object is in the sample not presented in this chapter). Limiting our detection radius to 1.5”, and thus reducing the likelihood of chance associations, the probability of one or more false associations in our dataset (i.e. the probability of any number of the new binaries not being physically associated) is therefore only 1.5%.

I also find below that the estimated photometric parallaxes of each binary’s components are equal (within the stated errors), and they are therefore at equal distances, further decreasing the likelihood of contamination. Thus, I conclude that all of these candidate binaries are physically associated systems. As all of these systems have high proper motion, confirmation of common proper motion can be easily determined from repeat measurements on year timescales.

6.4.2 Distances, ages and masses

I show derived properties of the individual binary components in table 6.3. Masses of each component are estimated by comparing the component absolute magnitudes with the models presented in Baraffe et al. (1998), custom-integrated over the SDSS i’ passband (I. Baraffe, private communication). In the absence of published age estimates for these targets the full range of ages found in the solar neighbourhood (0.5-10.0 Gyr (Gizis et al.2002), with 5 Gyr as our adopted best-estimate value) is assumed (figure 6.8).

None of the 5 systems have trigonometric parallaxes, so in table 6.3 I estimate distances using the i’-z’ vs. Mi relations given in Hawley et al. (2002). In all cases the binary components were found to be at equal distances, within the stated errors, and I combine their values for a single more precise system distance.

6.4.3 Sensitivity to white dwarf companions

Kilic et al. (2006) describes typical cool white dwarfs in the SDSS survey, which are expected to have i-z colours in the range -0.5 to +0.5 magnitudes and i’ absolute magnitudes of ~13-16 mags. A typical white dwarf companion would thus be detectable around all stars in this survey, and would have i-z colour differing by at least .5 magnitudes from a late M-dwarf companion (although cooler white dwarfs can have similar colours to late M-dwarfs). However, it should be noted that a white dwarf companion with higher temperature than the M-dwarf primary would bias the V-K colour of the system bluewards, and so the survey V-K colour selection introduces a bias against such systems.

The four newly detected companions in this chapter which have measured i-z colours are all consistent with being late M-dwarfs. Farihi et al. (2005) find that ~3% of white dwarfs have companions that are M5 or later; futher LuckyCam observations of many late-M-dwarfs as part of the VLM binary survey will further constrain the frequency of such systems.

6.4.4 Notes on the new systems

None of these stars have been previously investigated in detail in published work; no previous high resolution imaging or spectroscopy (resolved or unresolved) is available.

LSPM J1235+1318 A close to equal-mass 0.21” binary with an estimated distance of 24.4 ± 2.5pc and an estimated orbital radius of 5.1 ± 0.9AU.

LSPM J1314+1320 Resolved at a separation of only 0.13”, and the system is one of the two nearest binaries presented here (9.8 ± 2.0pc), with an orbital radius of only 1.3 ± 0.3AU. SDSS i’ photometry could not be calculated because (in this observation only) the camera gain was set too low to accurately calibrate. This system’s distance has been estimated from the V & K magnitudes listed in the LSPM North and the young disk photometric distance relations from Leggett (1992), assuming that the two components have equal V-K colours.

LSPM J1423+1318 A 0.57” binary estimated to be M5.5/M5.5 with an estimated distance 33.1 ± 3.4pc and an orbital radius of 18.9 ± 2.0AU.


Figure 6.8: The allowed mass ranges for the components of LSPM J1735+2634. The solid horizontal lines delimit the allowed mass range for the system’s primary, given its calculated absolute magnitude; the dashed lines show the mass range of the secondary. The isochrones are from Baraffe et al. (1998); I here assume an age range covering the distribution in the solar neighbourhood, 0.5-10.0 Gyr (Gizis et al.2002).

Table 6.2: Observations of the new binary systems. Errors are 1-sigma and derived from the variation in fit values with different frame selection fractions (and thus PSFs). The position angle (P.A.) error includes a 1.0o uncertainty in the orientation calibration. The companion to LSPM J1735+2634 is very faint and so only upper and lower limits to the i’ and z’ contrast ratios are given.
LSPMID       2MASS  ID              i′            z′         Separation (arcsec)  P.A. (deg)
LSPMJ1235+1318   12351726+131805     0.10 ±0.15  0.07± 0.15  0.21± 0.03          257.0± 2.5
LSPMJ1314+1320   13142039+132001     0.93 ±0.25  0.97± 0.25  0.13± 0.02          46.0± 2.0
LSPMJ1423+1318   14231683+131809     0.47 ±0.03  0.48± 0.05  0.57± 0.01          275.6± 2.0

LSPMJ1735+2634   17351296+263447     0.61 - 1.21 0.62- 1.11  0.29± 0.01          171.2± 2.1
LSPMJ1809+2128   18095137+212806     0.44 ±0.09  0.40± 0.10  1.12± 0.01          262.7± 2.0

Table 6.3: The new binary systems’ component properties. Spectral types and distances have been derived from the i’-z’ relations in Hawley et al. (2002). The spectral type relations have a plateau in i’-z’ at L0-L3, limiting precision for the later spectral classes. I further constrained the earliest spectral types (M5-M6) by comparing the derived i’ and z’ absolute magnitudes to those expected for those spectral types (Hawley et al.2002). Orbital radius uncertainties are 1σ and are estimated in quadrature from the system distance and separation uncertainties. Masses are estimated from the allowed ranges given by the models of Baraffe et al. 1998 (figure 6.8), between the 1σ photometric errors and the range of ages in the solar neighbourhood. Values of q > 1.0 imply that the true primary of the system may have been identified as the secondary. LSPM J1735+2634B is very faint, leading to difficulties in estimating resolved photometry, and so I only note upper and lower limits to the system’s observed and derived properties. The L3CCD gain for the i’ observation of LSPM J1314+1320 was set too low to accurately calibrate, so I do not give i’ photometry for this system.
LSPMID---------SDSS-i’ mag--SDSS-z’ mag---i’-z’-------Spectral type--Mass-/ M-⊙---q-(Ms-∕Mp-)--Dist. /-pc-orbital-rad.-/-AU--
LSPMJ1235+1318A    15.1 ±0.20    13.8 ± 0.20     1.25 ±0.23  M5  - M7       0.097 - 0.107 0.89 - 1.08 24.4± 2.7  5.1± 0.9

LSPMJ1235+1318B    15.2 ±0.13    13.9 ± 0.13     1.30 ±0.18  M6  - M7       0.097 - 0.106

LSPMJ1314+1320A    ⋅⋅⋅          11.9 ± 0.32     ⋅⋅⋅         ⋅⋅⋅           ⋅⋅⋅           ⋅⋅⋅        9.8± 2.0   1.3± 0.3
LSPMJ1314+1320B    ⋅⋅⋅          12.8 ± 0.21     ⋅⋅⋅         ⋅⋅⋅           ⋅⋅⋅

LSPMJ1423+1318A    15.1 ±0.11    13.9 ± 0.12     1.21 ±0.16  M5  - M6.5     0.108 - 0.122 0.82 - 0.99 33.1± 3.4  18.9± 2.0
LSPMJ1423+1318B    15.6 ±0.10    14.4 ± 0.11     1.20 ±0.15  M5  - M6.5     0.099 - 0.109

LSPMJ1735+2634A    15.3 - 15.5   13.6 - 13.8     1.52 - 1.85 M7  - L3       0.077 - 0.086 0.85 - 0.96 10 - 12    2.8- 3.6

LSPMJ1735+2634B    16.1 - 16.5   14.4 - 14.7     1.69 - 2.11 M8  - L4       0.066 - 0.082

LSPMJ1809+2128A    15.6 ±0.14    14.4 ± 0.15     1.17 ±0.21  M5  - M6       0.124 - 0.109 0.82 - 1.03 41.8± 4.4  46.8± 5.0

LSPMJ1809+2128B----16.0-±0.11----14.8-±-0.12-----1.21-±0.16--M5----M6-------0.101-- 0.112-----------------------------------------

LSPM J1735+2634 This system contains a possible brown dwarf companion, and is located at only 10-12pc. The faintness of the companion leads to difficulties in estimating resolved photometry of this system, so only upper and lower limits are noted for many of this system’s properties.

The companion has a lower mass limit of 0.066M and thus may be a brown dwarf; the allowed mass ranges are detailed in figure 6.8. Changing seeing conditions and the very red colour of the system gave a more clearly resolved image in the z’ filter; in this image the secondary appears very elongated approximately (not exactly) East-West. This elongation suggests that it is possible that the secondary is an unresolved (~0.04 arcsec) brown dwarf binary. This very close pair would have a very short period and can thus be rapidly confirmed with followup observations of the elongation direction.

This system is an excellent target for resolved infrared photometry or spectroscopy to confirm the nature of the secondary. With an expected orbital period of approximately 15-30 years it is also well suited for astrometric followups to measure a dynamical mass.

LSPM J1809+2128 A 1.12” binary; proper motion 0.19”/year, estimated distance 41.8 ± 4.4pc. With an orbital radius of 46.8 ± 5.0AU this estimated M5/M5.5 system is one of the very few known VLM binaries wider than 30AU (Siegler et al.2005Phan-Bao et al.2005).


Figure 6.9: The companion to Ross 530. 5% selection from 900 frames in SDSS z’. The 0.15” binary is very clearly resolved; the diffraction limit of the telescope in z’ is 0.09”.

6.4.5 A companion to Ross 530

During the observations I also detected a companion to Ross 530, one of the SDSS standard stars (figure 6.9). The system is very clearly resolved at only 0.15” separation.

Ross 530 is known to be a metal-poor spectroscopic binary (Latham et al.2002), but to our knowledge this is the first resolved image of this system. From this single observation it is unclear if the single-lined spectroscopic binary has been resolved or if a new companion has been detected.

6.5 Discussion

6.5.1 The binary frequency of M5.5-M8.0 stars in this survey

5 new binaries were detected in a 32 star sample, giving a raw binary fraction of 16-4+8%. However, this survey is based on a magnitude-based distance limit that assumes all the targets are single stars. Unresolved binaries in the LSPM survey appear to be brighter and thus closer than single stars for a specific colour, as there are two luminous bodies. The observed binary fraction is thus biased as a result of leakage of binaries into the sample from further distances.

I compensate for the distance bias by comparing the volume containing the single stars in the sample to the larger space which would contain any binaries. Lacking a useful constraint on the contrast ratios of low mass binaries I again assume a range of possible distributions - from all binaries being equal magnitude systems to a flat distribution of contrast ratios (Burgasser et al.2003). Including the probability distribution of the raw binary fraction and the range of contrast ratio distributions yields a distance bias corrected binary fraction of 7-3 +7 %.

The derived binary statistics are very similar to the sample of 36 M6.0-M7.5 M-dwarfs described in Siegler et al. (2005), who obtain a distance-corrected binary fraction of 7-2 +4 .

6.5.2 Contrast Ratios

I did not detect any companions at SDSS i’ contrast ratios >1 magnitude, although the survey is sensitive to up to 5 magnitude differences (figure 6.5), well into the brown dwarf regime.

The new binaries are all close to equal mass (figure 6.10), in common with other VLM binary surveys which are sensitive to faint companions (eg. Close et al. 2003Siegler et al. 2005).

6.5.3 The distribution of orbital radii

The new binaries’ orbital radii are shown in figure 6.11, compared to the distribution of known VLM binary systems with primaries later than M6 (as of 2005, from Siegler et al. 2005). Three of the five new systems fall within 1-5AU, the most common radius for known VLM binaries (the surveys are incomplete at very small radii). However, one binary (LSPM J1809+2128) is one of only very few (Siegler et al.2005Phan-Bao et al.2005) known VLM binaries wider than 25AU. It is important to enlarge the sample of VLM binaries to ascertain how common these wide systems are, as well as to constrain the fraction of higher order multiple systems.


Figure 6.10: The binaries’ mass ratios; an example curve shows the minimum detectable binary mass ratio for a typical star in our survey, an M6.5 primary at approximately 20pc. To generate the curve, the contrast ratios shown in figure 6.5 were used to calculate a minimum detectable companion mass from the 5.0 Gyr models given in figure 6.8.

6.6 Summary


Figure 6.11: The 1σ range of orbital radii for each detected binary, compared with a histogram of the previously known sample collated in Siegler et al. (2005) (and the 33 AU VLM binary described in Phan-Bao et al. 2005). Poisson 1σ error bars are shown for the histogram - and illustrate the necessity for increased sample sizes. For reasons of clarity, the 200AU system described in Luhman (2004) is not displayed.