SPECKLE INTERFEROMETRY AT SOAR IN 2014^*

The observations reported here were obtained with the high-resolution camera (HRCam)—a fast imager designed to work at the 4.1 m SOAR telescope (Tokovinin & Cantarutti 2008). For practical reasons, the camera was mounted on the SOAR Adaptive Module (SAM, Tokovinin et al. 2008). However, the laser guide star of SAM was not used; the deformable mirror of SAM was passively flattened and the images are seeing-limited. The SAM module corrects for atmospheric dispersion and helps to calibrate the pixel scale and orientation of HRCam, as explained in TMH14. The transmission curves of HRCam filters are given in the instrument manual.⁵ We used mostly the Strömgren y filter (543/22 nm) and the near-infrared I filter (788/132 nm). The response curve of the latter was re-defined as a product of the filter transmission and detector response furnished by the respective manufacturers.

The electron multiplication CCD (EM CCD) Luca-S used in HRCam failed in 2014 July. It was sent to its manufacturer (Andor) for repair and received back in 2014 December. For the two runs in the fall of 2014, we used the EM CCD Luca-R, kindly loaned by G. Cecil. It has a larger format of 1004 × 1002 pixels, with smaller pixels of 8 μm compared to Luca-S (658 × 496, 10 μm pixels). The second lens in HRCam was replaced by the achromatic doublet of 75 mm focal length to get the pixel scale of 14.33 mas. The detector software driver was also upgraded on this occasion.

Although the two Luca cameras look exactly the same and come from the same manufacturer, their CCDs are radically different. Luca-S uses a line-transfer CCD, while Luca-R has a frame-transfer CCD. We found that Luca-R had poor charge transfer efficiency (CTE) in the vertical direction. Weak signals such as cosmic-ray hits and dark current from hot pixels are smeared vertically over ∼5 pixels, while strong signals are transferred with less spread. This results in the signal-dependent loss of resolution in the vertical direction (along CCD columns), which was accounted for in the data processing by an additional adjustable parameter. This problem was revealed unexpectedly during the 2014 October run. In the next run, we placed stars closer to the line register, improving slightly the vertical CTE by reducing the number of charge transfers.

We studied the distribution of the bias signal of both cameras. It is very well modeled by the sum of a Gaussian component (readout noise) and a negative exponent that corresponds to the single-electron events amplified stochastically by the gain register. Most events are generated by the CCD clocking (clock-induced charge, CIC) typical for EM CCDs. The width (decrement) of the exponential term depends on the EM gain. The rate of CIC events is found as the ratio of the areas below these two terms. For Luca-R, the readout noise is 4.5 ADU and the CIC rate is 0.03. The CIC events are not smeared vertically because no charge transfer is involved. For Luca-S, the readout noise is 9 ADU, the exponential decrement is 30 ADU (for the gain setting of 200 used here), and the rate of CIC events is 0.13 per pixel. About half of the CIC events are removed by the threshold of 17 ADU applied in the power spectrum calculation.

The SOAR telescope suffers from 50 Hz vibration (see TMH14). The vibrations are non-stationary, causing variable loss of resolution if the 20 ms exposure time is used. Most data were taken with an exposure of 5 ms, sampling 1/4 of the vibration period. The resolution is then recovered and the effect of vibrations becomes less dramatic. It disappears completely at an exposure time of 2 ms, used on the brightest targets. With such short exposures, the power spectra extend to the cut-off frequency and are very symmetric. The Luca-R detector suffering from the CTE problem was used mostly with a 10 ms exposure as a compromise between vibrations and poor CTE at low signal levels.

The observations consist of taking a cube of 400 images of 200 × 200 pixels each. For binaries wider than 15, the 400 × 400 format was used. Each object and filter combination is normally recorded twice, these data are processed independently, and the result is averaged. We used 2 × 2 detector binning on the faintest targets observed in the I filter, with a minor under-sampling and loss of resolution. Measurements of binaries made with and without binning agree well mutually. Faint red dwarfs with I ∼ 12 still produced useful data.

The observing time for this program was allocated through NOAO (proposals 2013B-0172, 2014A-0038, 2014B-0019, eight nights total) and by the Chilean National TAC (proposal CN2014B-27, four nights).

Table 1 lists the observing runs, the calibration parameters (position angle offset θ₀ and pixel scale in mas), and the number of objects covered in each run. The calibration of angle and scale was done by referencing to the SAM imager, as described in TMH14.

Run 1 was affected by transparent clouds. Two hours on 2014 January 22 were added as a backup program. During this run, the Nasmyth rotator of SOAR had a large offset of +28. The seeing was mediocre to poor. The instrument was not dismounted between runs 1 and 2, but the Nasmyth rotator was re-initialized, explaining the difference in θ₀. The sky was clear, and two hours were added on March 7 as a backup from other program. The sky was also clear on the two nights of run 3, with seeing good to excellent.

Runs 4 and 5 used the substitute detector Luca-R and were affected by the CTE problem. For this reason the observations were performed mostly in the I filter with 10 ms exposure. All three nights of run 4 were clear, and mostly with good seeing (the median full width at half maximum, FWHM, of re-centered average images was 060, the best FWHM was around 045). The detector was removed between runs 4 and 5 in an attempt to improve the CTE by tuning the electronic parameters; however, this turned out to be impossible. During run 5, the seeing varied from average to poor. In the worst case, the star did not fit in the 3'' field, illuminating all pixels. In such cases, the image is truncated, the power spectrum contains bright vertical and horizontal lines, and the detection limits are poor. Under poor seeing, data obtained with the 400 pixel field and 2 × 2 binning are of better quality, without image truncation.

The two nights of run 6 were clear, with seeing average to poor. The HRCAM was returned to its original configuration with the Luca-S detector. Measurements with the SAM internal light source confirmed that the pixel scale did not change compared to runs 1–3.

The data processing is described in TMH10. As a first step, power spectra and average re-centered images are calculated from the data cubes. The auto-correlation functions (ACFs) are computed from the power spectra. They are used to detect companions and to evaluate the detection limits. The parameters of binary and triple stars are determined by fitting the power spectrum to its model, which is a product of the binary (or triple) star spectrum and the reference spectrum. We used as a reference the azimuthally averaged spectrum of the target itself in the case of binaries wider than 01. For closer pairs, the "synthetic" reference was used (see TMH10).

The signal-dependent vertical smearing of the image caused by the CTE problem required a modification of the reference spectrum. We model the smearing by an additional Gaussian term in the reference spectrum $P({f}_{x},{f}_{y})$:

$$\begin{eqnarray}P({f}_{x},{f}_{y}) & = & {T}_{\mathrm{DL}}(\kappa )\ {10}^{{p}_{0}+{p}_{1}\kappa }\ {e}^{-2{\pi }^{2}{({\rm{\Delta }}{{yf}}_{y}/2.35)}^{2}},\\ \kappa & = & \sqrt{{f}_{x}^{2}+{f}_{y}^{2}}/{f}_{c},\end{eqnarray} \tag{ 1 }$$

where f_x and f_y are spatial frequency coordinates along CCD lines and columns respectively, T_DL is the diffraction-limited transfer function, f_c is the cut-off frequency, $\kappa =| f| /{f}_{c}$, p₀ and p₁ describe the speckle signal attenuation (both are negative, p₀ is related to seeing), and the new parameter ${\rm{\Delta }}y$ is the FWHM of the CTE smear in the column direction in pixels. This "elliptical" synthetic reference was used for all binaries, both wide and close, observed in runs 4 and 5. The parameters of the reference $({p}_{0},{p}_{1},{\rm{\Delta }}y)$ were fitted to the power spectrum jointly with the binary parameters (θ, ρ, ${\rm{\Delta }}m$).

In the case of close binaries, the parameters of the elliptical reference and of the binary itself are mutually correlated, so the resulting measures could be biased. The close binary star B 430 (19155–2515) was observed with position angles of the instrument differing by 90°. The "fringes" in the power spectrum had different orientations relative to the CCD. However, the binary parameters (θ, ρ, ${\rm{\Delta }}$m) fitted to these power spectra are mutually consistent, (2837, 00673, 0.785) and (2850, 00690, 0.722). Therefore, the CTE effect is, to first order, accounted for by the modified reference model. Still, measurements of close binaries from runs 4 and 5 (2014.77 and 2014.86) should be taken with some caution.

Wide binaries resolvable in the re-centered long-exposure images are processed by another code that fits only the magnitude difference ${\rm{\Delta }}$m, using the binary position from speckle processing. This procedure corrects the bias on ${\rm{\Delta }}$m caused by speckle anisoplanatism and establishes the correct quadrant (flag * in the data table).

For run 5, we also calculated shift-and-add (SAA or "lucky") images, centered on the brightest pixel in each frame and weighted proportionally to the intensity of that pixel. All frames without rejection were co-added. Binary companions, except the closest and the faintest ones, are detectable in these SAA images, helping to identify the correct quadrant. Such cases are marked by the flag q in the data table. Quadrants of the remaining binary stars are guessed based on prior data or orbits, not measured directly.

Calibration of scale and angular offset in speckle interferometry is challenging because the accuracy of modern measurements with 4 m telescopes exceeds the accuracy of even the best orbits. Comparison with ephemerides is useful only as a sanity check. Recognizing this problem, we observed several wide binaries during each run. Their motions are slow and can be modeled by linear functions of time. These models can then be used to check the calibration of the archival data. This approach tests only the mutual consistency of speckle runs, rather than absolute calibration of the whole data set.

We selected 41 binaries wider than 05 that do not contain resolved subsystems and were observed during at least 4 runs each. The angle and separation of each binary is approximated by linear functions of time. If the fitted slope is less than twice its rms error, a zero slope is assumed. Deviations from these models are interpreted as calibration corrections. For each run, they are median-averaged. We then iterate by fitting new linear models corrected for the run's systematics, determining new calibration parameters of the runs, etc.

Figure 1 shows the offsets in angle and scale determined by this procedure for those of 26 runs where at least N = 2 calibrators were observed. Run 0 refers to observations at the Blanco telescope in 2008.5 (TMH10). Runs 1–6 in this paper correspond to numbers 21–26 in the plot. Vertical bars show the rms scatter between calibrators (not errors of the mean, which are smaller by $\sqrt{N}$). A typical rms scatter is 01 in angle and 0.3% in scale.

Zoom In Zoom Out Reset image size

This analysis reveals good mutual consistency of ∼5000 speckle measurements produced by HRCam to date. The largest systematics are found in runs 10 and 11 (2011.9 and 2012.1). The published data can be corrected by subtracting the angular offsets found here and dividing the separations by the new scale factors. We plan to observe more calibrators and publish improved systematic corrections in the future. The data of this paper rely only on the original instrument calibrations and are not corrected for the offsets found so far.

This study gives an estimate of the accuracy of speckle measurements. After correction for run systematics, the median rms deviations from the models for the 41 calibrator binaries are 01 in angle and 1.9 mas in separation. The actual errors should be a factor of ∼1.5 larger (∼3 mas), considering that the number of independent measurements is about twice the number of fitted parameters. This study will be repeated in the future with additional speckle runs and calibrators.

The data tables have almost the same format as in the previous papers of this series. They are available in full only electronically. Table 2 lists 1636 measures of 1218 resolved binary stars and subsystems, including 56 newly resolved pairs. The columns of Table 2 contain (1) the WDS (Mason et al. 2001) designation, (2) the "discoverer designation" as adopted in the WDS, (3) an alternative name, mostly from the Hipparcos catalog, (4) Besselian epoch of observation, (5) filter, (6) number of averaged individual data cubes, (7, 8) position angle θ in degrees and internal measurement error in tangential direction ρσ_θ in mas, (9, 10) separation ρ in arcseconds and its internal error σ_ρ in mas, and (11) magnitude difference ${\rm{\Delta }}$m. An asterisk follows if ${\rm{\Delta }}$m and the true quadrant are determined from the resolved long-exposure image; a colon indicates that the data are noisy and ${\rm{\Delta }}$m is likely over-estimated (see TMH10 for details); the flag "q" means the quadrant is determined from the SAA image. Note that in the cases of multiple stars, the positions and photometry refer to the pairings between individual stars, not the photo-centers of subsystems.

For stars with known orbital elements, columns (12)–(14) of Table 2 list the residuals to the ephemeris position and code of reference to the orbit adopted in the Sixth Catalog (Hartkopf et al. 2001, hereafter VB6).⁶ New and revised orbits computed in this work (Section 4) are referenced as "This work."

We did not use image reconstruction, and measured the position angles modulo 180° (except the SAA images in run 5). Plausible quadrants are assigned on the basis of orbits or prior observations, but they can be changed if required by orbit calculation. For triple stars, however, both quadrants of inner and outer binaries have to be changed simultaneously; usually the slowly moving outer pair defines the quadrant of the inner subsystem without ambiguity.

Table 3 contains the data of 441 unresolved stars, some of which are listed as binaries in the WDS or resolved here in other filters. Columns (1) through (6) are the same as in Table 2, although Column (2) also includes other names for objects without discoverer designations. For stars that do not have entries in the WDS, fictitious WDS-style codes based on the position are listed in Column (1). Column (8) is the estimated resolution limit, the largest of the diffraction radius λ/D and the vertical CTE smear ${\rm{\Delta }}$y (applicable to runs 4 and 5 only). Columns (8, 9) give the 5 σ detection limits ${\rm{\Delta }}$m₅ at 015 and 1'' separations determined by the procedure described in TMH10 (please note that this is not the resolution of the observations). When two or more data cubes are processed, the largest ${\rm{\Delta }}$m value is listed. The last column marks by colons noisy data mostly associated with faint stars. In such cases, the quoted ${\rm{\Delta }}$m might be too large (optimistic); however, the information that these stars were observed and no companions were found is still useful for statistics (Tokovinin 2014b).

Table 3.  Unresolved Stars (Fragment)

WDS (2000)	Discoverer	Hipparcos	Epoch	Filter	N	ρmin	5σ Detection Limit		Δm
α, δ (J2000)	Designation	or Other	+2000				Δm (015)	Δm (1'')	Flag
	or Other Name	Name				(arcsec)	(mag)	(mag)
00006–6641	GLI 289	HIP 55	14.7661	I	2	0.039	1.95	3.07	⋯
00052–6251	HIP 425	HIP 425	14.7661	I	2	0.039	2.78	4.10	⋯
00174–5131	HDS 40	HIP 1393	14.7662	I	2	0.039	2.56	3.97	⋯
00250–3042	HIP 1976	HIP 1976	14.8534	I	2	0.039	3.42	4.82	⋯
00301+0452	HIP 2358	HIP 2358	14.8561	I	2	0.039	2.28	3.69	⋯
00310–3138	HDS 69	HIP 2433	14.7662	I	2	0.039	1.69	4.01	⋯
00324+0657	MCA 1 Aa,Ab	HIP 2548	14.7632	I	2	0.039	3.01	4.25	⋯

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

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

Table 4 lists 56 newly resolved pairs. The last two columns of Table 4 contain the spectral type (as given in SIMBAD or estimated from absolute magnitude) and the Hipparcos parallax (van Leeuwen 2007, hereafter HIP2). Fragments of ACFs of newly resolved triple systems are shown in Figure 2. We comment on the newly resolved binaries below. The following abbreviations are used: PM—proper motion, CPM—common proper motion, RV—radial velocity, SB1 and SB2—single- and double-lined spectroscopic binaries, INT4—4th Interferometric Catalog (Hartkopf et al. 2001).

Zoom In Zoom Out Reset image size

01379–8259. HIP 7601 is a nearby (27 pc) dwarf also known as GJ 67.1 or HR 512. According to Wichman et al. (2003), it is a young spectroscopic triple detected in X-rays (1RXS J013755.4–825838). GCS also recognized the star as SB2. Spectroscopic monitoring (A. Tokovinin 2015, in preparation) shows that all three components are similar stars of approximately one solar mass. Here we resolved the outer subsystem AB and observed its fast motion. In fact it was already resolved at SOAR on 2011.036 at 2210 and 0044, but this low-quality observation has not been published. The available data indicate that the period of AB is 1.7 years; the pair completed nearly two revolutions since its first resolution in 2011.

02098–4052. HIP 10096 is SB2 according to GCS. The separation corresponds to a period of ∼5 years.

03046–5119. HIP 14307 and 14313 form the 38'' pair DUN 10 AB belonging to the FG-67 sample. The component B is resolved here at 019. The estimated period of Ba,Bb is ∼25 years. No motion is seen in 2 months. The subsystem Ba,Bb is also manifested by asymmetric line profiles (A. Tokovinin 2015, in preparation). The A component has never been observed at high angular resolution.

04386–0921. HIP 21265 is an X-ray source and an SB2 according to GCS (one observation only). It is resolved here at 56 mas, which corresponds to a period of ∼5 years.

04469–6036. First resolution of HIP 22229 at 0043, ${\rm{\Delta }}$I = 1.2. The measure is uncertain, but the elongation is confirmed with 2 ms exposure and is not seen in other observed stars, so it is not caused by vibration. This is a triple system containing an eclipsing binary AL Dor of Algol type with an eccentric orbit (Bulut & Demirkan 2007). The separation implies an orbital period of the outer system on the order of 5 years.

05354–0450. This is HIP 26237, HD 37018, HR 1892, 42 Ori, a young star in Orion which has not been observed at high angular resolution so far, according to INT4. We resolved the known binary AB = DA 4 and discovered the spectacular subsystem Aa,Ab at 016 (Figure 2).

06303–5252. HIP 30995 is an SB2 according to GCS. The new companion at 02 with ${\rm{\Delta }}$I = 4.1 mag is unlikely to correspond to the SB2, so the system is probably triple.

06497–7433. HIP 32735 has an acceleration detected in HIP2, but no RV data. The resolved 03 pair implies a period of ∼80 years. The spectral type K0IV given in SIMBAD may be inaccurate because the luminosity and color of the components correspond to main sequence stars with masses of 1.0 and 0.6 ${{\mathcal{M}}}_{\odot }$.

06499–2806. The 02 pair HDS 947 had not been observed since its discovery by Hipparcos. Here it is revealed as a triple system (Figure 2) with comparable separations between components. HDS 947 probably corresponds to AB, while the fainter C component is new.

07038–4334. HIP 34052 = HD 53680 = HIP 34065C = GJ 264 is a spectroscopic and astrometric binary for which Sahlmann et al. (2011) give an orbit with P = 1688 days = 4.62 years and an estimated semi-major axis of 015. There is also an astrometric orbit (Makarov et al. 2008) with P = 4.11 years that predicts θ = 327° for the moment of our first observation. The binary is resolved at 023, ${\rm{\Delta }}$I = 4.4 mag and shows some orbital motion. The binary makes a quadruple system together with components A and B (GJ 264.1 and 264) that form a 205 pair at 185'' from C. The B component was also observed here and found unresolved.

07294–1500 is another nearby multiple system. The main component HIP 36395 is a visual binary with a known 728-year orbit, also measured here. The C component (NLTT 17952) at 204 is physical, and yet another CPM component F is found at 1072'' (Tokovinin & Lépine 2012), while the WDS components D and E are optical. We observed C and resolved it at 009. The orbital period of Ca,Cb is on the order of 6 years, estimated masses are 0.6 and 0.4 ${{\mathcal{M}}}_{\odot }$. We also targeted F and did not resolve it. The magnitudes and colors of C and F are quite similar.

07304+1352 is a quadruple system. The 77 pair AB (STF 1102) is HIP 36485, the CPM component D = HIP 36497 = HD 59450 is located at 112'' from it, while the WDS components C and E are optical. The physical nature of AD is established by common PM, distance, and RV. D is a known SB1 with P = 2708 days = 7.4 years (Halbwachs et al. 2012) and an estimated semi-major axis of 93 mas, also an acceleration binary. We resolved the Da,Db pair at 011, ${\rm{\Delta }}$I = 2.6. The minimum mass of Db derived from its SB orbit is 0.27 ${{\mathcal{M}}}_{\odot }$, while we estimate the masses of Da and Db as 1.05 and 0.6 ${{\mathcal{M}}}_{\odot }$ from their luminosity. A previous non-resolution of D is reported in INT4; it was also unresolved with Robo-AO (Riddle et al. 2015).

07312+0210, HIP 36557 = HD 59688. According to observations by D. Latham (2012, private communication), this is a spectroscopic triple with an inner period of 70 days (double-lined, also detected by GCS, mass ratio 0.7) and an outer period of 2007 days or 5.5 years. The outer system is also detected by astrometric acceleration (Makarov & Kaplan 2005). We resolve it here at 0057, ${\rm{\Delta }}$I = 2.0, ${\rm{\Delta }}$y = 2.5 mag, and see the orbital motion. The estimated mass of Ab is 0.88 ${{\mathcal{M}}}_{\odot }$. The semi-major axis of the 70 days inner binary Aa,Ab is 7 mas, so accurate measurements of AB can detect the sub-motion to determine the orientation of the inner orbit.

08021–1710. This is the high-PM M-dwarf LP 784–12 (HIP 39293) at 30 pc from the Sun. A new distant component C was found at 18 in addition to the known pair HDS 1140 which closed from 04 in 1991.25 to 033 now (Figure 2). It is confirmed as physical by its fixed position during one year, the quadrant of the triple was determined in run 5.

08447–2126. Like the previous object, the late-type nearby binary HDS 1260 was discovered by Hipparcos and is expected to move rapidly (HIP 42910, $\mathrm{BD}-20^\circ$ 2665). It was targeted at SOAR for the first time. To our surprise, the object turned out to be a resolved triple, with the secondary being a 016 pair of equal stars (Figure 2). Hipparcos failed to recognize the triple nature of this star. The estimated period of BC is 15 years. The outer pair AB has closed from 08 to 05 and moved in position angle since its discovery. The separations between components are comparable, so this triple system may be interesting dynamically.

09299–3629. HIP 46572 is called a "high proper motion star" in SIMBAD, although its PM and RV are actually quite moderate. The binary AB has moved little in the 24 years since its resolution by Hipparcos. We discover the subsystem BC (Figure 2) with an estimated period of ∼100 years.

09586–2420. HIP 48906 is a double-lined binary according to the GCS, first resolved here at 64 mas. The period should be ∼20 years.

10056–8405. HIP 49442 = HD 88948 is a nearby dwarf in the 39 visual binary HJ 4310 AB. According to GCS, the RV of the main component A varies by 3.5 km s⁻¹. Here it is resolved into a 018 Aa,Ab pair with an estimated orbital period of 25 years. No astrometric acceleration was detected, however. The visual secondary B was targeted separately and found unresolved.

10070–7129. HIP 49546 is an astrometric binary of 1.5 year period (Goldin & Makarov 2006) with variable RV. The period corresponds to a semi-major axis of 25 mas. The star is resolved here tentatively at 26 mas (with 2 ms exposure). This resolution is below the diffraction limit and needs confirmation. The measured position angle of 346° is close to 342° predicted by the astrometric orbit.

10223–1032. HIP 50796 is a single-lined and astrometric binary according to Torres (2006), with a period of 570.98 days (1.56 years), K₁ = 20.76 km s⁻¹, e = 0.611. The Hipparcos parallax corrected for the binary motion is 20.6 ± 1.9 mas. The spectroscopic secondary companion is over-massive, most likely a close pair of M-dwarfs. If so, the new speckle companion at 166 with a period on the order of 500 years makes the system quadruple. The speckle companion might contribute to the IR excess found by Torres. The system is an X-ray source, and is possibly young.

10530+0458. D. Latham (2012, private communication) identified HIP 53212 = HD 94292 as SB2 with a period of 8.3 years and a highly eccentric orbit. It is resolved here securely at 50 mas.

12176+1427. HIP 59933 has a variable RV according to GCS. The 32 mas separation corresponds to an orbital period of ∼2 years. However, the separations in the y and I filters are somewhat discordant; further confirmation is needed.

12250–0414. HIP 60574 is a spectroscopic triple with periods 14 days and 22 years (D. Latham 2012, private communication), also an acceleration binary. We resolved the outer pair at 022 separation, matching the spectroscopic period. The lines of the visual secondary Ab could potentially be detected in the spectrum by cross-correlation, leading eventually to a full 3D orbit.

12528+1225. HIP 62933 (41 Vir) was observed on request by F. Fekel who studies its spectroscopic orbit. Apparently it is resolved for the first time.

13132–0501. HIP 64499 has a variable RV, with a preliminary spectroscopic orbit of 17 years period (D. Latham 2012, private communication). It is resolved at 01 and shows no motion in one month.

13321–1115. No previous indication of binarity was available for HIP 66018, apart from the CPM companion B at 84''. We discovered another faint component Ab at 089, ${\rm{\Delta }}$I = 4.6, likely to be physical (low background density). The B component (V = 14.8) was targeted, but its speckle signal was weak and no obvious close companions to B were found.

13344–5931. HIP 66230 has a variable RV in the GCS and an astrometric acceleration. We resolved it at 01, ${\rm{\Delta }}$I = 2.5, estimated period ∼10 years. The pair moved by 4° in one month.

13382–2341. HIP 66530 has a variable RV according to the GCS. It is resolved at 016 in the I band only, estimated period ∼20 years. This is a triple system, considering the CPM companion B at 28'' (LDS 4385 AB).

13401–6033. HIP 66676 (A) and HD 118735 (B, G6V, V = 9.17) at 77'' share common PM (although it is small, 58 mas yr⁻¹) which, together with photometry, indicates with high probability that it is a physical pair AB (Tokovinin & Lépine 2012). We targeted the secondary component B and found it to be a resolved triple. The faint star C, at 092 from B, is itself a close 016 pair Ca,Cb (Figure 2). Note that this is a region of the sky with very high stellar density, raising suspicion that the Ca,Cb pair might be a random background object. Re-observation of the triple in 2015 (to be published) shows, however, that it is physical because the center of C = (Ca,Cb) moves relative to B with a speed of 11 mas yr⁻¹ (or 3 km s⁻¹), compatible with the expected orbital motion of BC and much less than the system's PM of 58 mas yr⁻¹. The estimated period of Ca,Cb is ∼30 years, the period of BC is ∼300 years.

13495–2621. HIP 67458 is a double-lined chromospherically active binary with orbital period of 7.2 days (D. Latham 2012, private communication). We found a faint tertiary companion at 073 with an estimated orbital period on the order of 100 years. The speckle survey of chromospherically active stars by Mason et al. (1998) did not detect this tertiary, lacking the dynamic range of HRCam.

14014–3137. HIP 68507 is an acceleration binary with a variable RV resolved here at 006. The period of Aa,Ab is on the order of 5 years. There is a faint physical companion B at 67 found in 2MASS (Tokovinin 2011). Another visual companion at 14'', SEE 195, is optical, as revealed by its fast relative motion.

14382+1402. HIP 71572 is an acceleration binary without RV data. It is found to be a tight 90 mas pair with a small ${\rm{\Delta }}$m and an estimated period under 10 years (some motion is seen in one month). Very likely it can be studied as a double-lined SB.

14464–3346. HIP 72235B is located at 9'' from the primary star and shares its proper motion (AB = HDS 2082). Pre-discovery measurements of this Hipparcos pair were published by Wycoff et al. (2006). The RV of A may be variable (D. Latham 2012, private communication). The star B turns out to be a 04 binary Ba,Bb with masses of 0.70 and 0.17 ${{\mathcal{M}}}_{\odot }$ estimated from the luminosity. Its period is on the order of 100 years. The whole system could thus be quadruple.

15362–0623. HIP 76400 is identified by the GCS as an SB2 with a mass ratio q = 0.93, but there is no spectroscopic orbit available. We resolved the 019 pair with ${\rm{\Delta }}$I = 3.9, indicating a mass ratio of ∼0.5 and an orbital period of ∼30 years. Very likely the resolved binary does not match the spectroscopic double-lined system. Considering the CPM component B at 80'' (Tokovinin & Lépine 2012), this system could be a quadruple with a 3-tier "3+1" hierarchy.

15367–4208. HIP 76435 is a G5V star from the FG-67 sample. Its companion C (AC = FAL 78) at 135 is physical, while the Hipparcos companion B at 43 is not seen in the 2MASS images and has not been confirmed otherwise. We targeted C and resolved it into a close binary. Estimated masses of Ca and Cb are 0.70 and 0.66 ${{\mathcal{M}}}_{\odot }$, period ∼4 years.

16142–5047. HIP 79576 has a variable RV (GCS). The 79 mas separation implies an orbital period of ∼5 years. This is the high-PM star LTT 6467 with a low metallicity [Fe/H] = −0.78. Ivanov et al. (2013) found no CPM companions.

16195–3054. HIP 79980 and HIP 79979 form a 23'' CPM pair AB where the F6III primary is slightly evolved, while the G1/G2 secondary is closer to the main sequence, but still above it. The Hipparcos parallax of B, −4.7 mas, is obviously wrong, so we assume the parallax of the primary, 20.7 mas. The RV(B) is variable (GCS), and we resolve it into a 40 mas pair with an estimated period on the order of 1 year. The pair moved by 10° in a month. Binary motion is the likely cause of the incorrect Hipparcos parallax.

16454–7150. HIP 82032 is located in a crowded field, so the new faint Ab companion found here at 13 could be optical. The star was observed because of its suspected variable RV (GCS), but the newly found companion, even if physical, is too distant to explain this variability. Another visual companion B at 115 (AB = B 2392) is optical, as evidenced by its fast relative motion. The star is on the HARPS exo-planet program.

16563–4040. HIP 82876 is a distant O7V star. The 026 pair AB (HDS 2394) was measured among other neglected binaries. We found another faint companion C at 146 (Figure 2). The star has an extensive literature, including multiplicity surveys with speckle interferometry and RV (Chini et al. 2012). Owing to the large distance, no detectable orbital motion is expected. Indeed, the AB pair was measured with HRCam in 2008.5 at the same position as it is now. Those observations in the y band did not detect the companion C owing to a lower signal-to-noise ratio.

17054–3346. The RV of HIP 83612 varies by 52 km s⁻¹ (GCS). It is a very close pair with an estimated period of ∼1.5 years, and the measure near the diffraction limit derived from the elongated power spectrum is tentative. The Hipparcos parallax is likely biased by the binary. The visual component B = HIP 83609 (AB = WNO 5) at 25'' is optical.

17098–1031. HIP 83962 = HR 6375 has a variable RV according to the GCS, while N. Gorynya (2013, private communication) detected double lines. It is resolved tentatively at 33 mas with ${\rm{\Delta }}$y = 1.8 mag (the 5 ms exposure makes it unlikely that the asymmetry is caused by vibrations). The separation corresponds to an orbital period on the order of 1 year, which could bias the Hipparcos parallax. However, Eggleton & Tokovinin (2008) consider the star as single. The new pair was not resolved in 2014.3; presumably it became closer.

17264–4837. HIP 85342 and HIP 85326 form a physical pair AB at 127'' separation (common PM, RV, and parallax). The B component = HIP 85326 has a variable RV and an astrometric acceleration which could hardly be produced by the 1'' speckle companion Bb found here, owing to its long estimated period of ∼300 years. It seems that B is triple and the whole system is quadruple. This companion Bb was not detected in the previous speckle observations because it is red: ${\rm{\Delta }}$I = 2.4, ${\rm{\Delta }}$y = 4 mag; its color matches a dwarf star at the same distance as the system. However, the field is crowded and the newly found companion could still be optical.

17266–3258. HIP 85360 is an acceleration and spectroscopic binary. Its preliminary spectroscopic period (D. Latham 2012, private communication) corresponds to a semi-major axis of 80 mas. The star is chromospherically active and possibly young. The faint companion found here at 116 is most likely optical, as the field is extremely crowded. Re-observation within a year will resolve its status.

17341–0303. HIP 85963 has a variable RV and is an acceleration binary. The 91 mas separation implies an orbital period of ∼10 years; the estimated masses are 1.37 and 0.81 ${{\mathcal{M}}}_{\odot }$. Despite extensive literature (51 references in SIMBAD), there is no published spectroscopic orbit, while several high-resolution spectroscopic studies addressed the abundance.

17342–1910. B 1863 is a known close binary which has been unexpectedly found to be a triple (Figure 2). The new distant component C is detectable also in the y filter, but we measured only the inner binary in y. The star was observed at the Blanco telescope in 2008.5397, and the pair actually measured then was AC, at 1338, 0217, ${\rm{\Delta }}$y = 3.7 (same as now, see Table 2). The inner pair AB with a smaller ${\rm{\Delta }}$m was unresolved in 2008.5, while it is clearly resolved now. Owing to the large distance from the Sun, we expect only a slow motion, so even the inner pair observed since 1929 may not yet be ready for computing its first orbit.

17342–5454. HIP 85969 has a variable RV according to the GCS and confirmed by Jones et al. (2002). The 055 separation implies a period on the order of 80 years. The star is on the exo-planet program at the Anglo-Australian Telescope.

18243–0405. The neglected pair YSC 67 turned out to be a new triple (Figure 2). The outer 035 binary AB was known previously, now we detect elongation that implies an inner subsystem Aa,Ab. We compared with stars in the same area of the sky observed before and after to assure that the elongation is not of instrumental origin; it is seen in two filters.

18267–3024. HIP 90397 is an acceleration binary with a variable RV (GCS), resolved here at 69 mas (estimated period ∼4 years). The star is targeted by exo-planet programs.

18346–2734. HIP 91075 was noted as a double-lined binary by the GCS (one observation only). The separation of 86 mas implies a period of ∼10 years, while the double-lined observation matches the moderate magnitude difference (estimated masses 1.04 and 0.77 ${{\mathcal{M}}}_{\odot }$). The star is an X-ray source, so a future combined orbit could determine masses for testing evolutionary models of young stars. The sky around the object is quite crowded; the 9'' companion I 1026 is optical (it moves too fast).

19206–0645. HIP 95068 is the neglected Hipparcos binary HDS 2735 AB, a distant K-type giant. We did not resolve this 01 binary, which remains unconfirmed, but detected instead another faint star at 1''. The stellar background is crowded, the PM is small, and the status of the new companion remains uncertain.

19209–3303. HIP 95106 and 95110 form the 137 pair HJ 5107 AB. The RV variability of A was suspected by the GCS, it is now resolved as a 027 binary with estimated period of ∼35 years. The component B was also observed and found unresolved. It contains a spectroscopic pair (A. Tokovinin 2015, in preparation), the whole system is quadruple.

19221–2931. HIP 95203 is another acceleration binary with variable RV resolved here. The relatively large separation of 077 corresponds to a period of ∼180 years. The actual period can be as short as 60 years if the pair is seen now near its maximum separation (it would then have been closer at the time of the Hipparcos mission). Most likely, however, the faint visual companion found here and the spectroscopic/acceleration pair make a triple system. There is another companion HIP 95164 at 435''. The status of this wide pair (is it a real binary or just two members of a moving group?) is not clear, but the association of those stars leaves no doubt (common RV, PM, and parallax).

19409–0152. HIP 96834 has a spectroscopic orbit with a period of 1 year and expected semi-major axis of 27 mas (D. Latham 2012, private communication). We resolved this pair, although the measurement near the diffraction limit is uncertain. However, there are some unsolved questions. Why, despite the small magnitude difference ${\rm{\Delta }}$y = 1.2 mag, were double lines not seen? Why, despite the 1 year period, was the Hipparcos parallax not strongly affected and the star appears to be on the main sequence?

19453–6823. We resolved the secondary component of HDS 2806 AB into a close pair Ba,Bb (Figure 2). This is a K3 dwarf within 50 pc from the Sun. The pair Ba,Bb should be fast and turn around in about 10 years.

22259–7501. HIP 110712 and 110719 form the 206 pair DUN 238 AB. We observed both components. The A component has a variable RV according to the GCS, but not confirmed by Jones et al. (2002); it was unresolved. The newly found pair Ba,Bb has a period on the order of 10 years. Considering the distant companion C found by Caballero (2012), the system contains at least 4 stars.