The Hubble Ultracool Multiplicity (HUM) Survey. I. Characterizing Sensitivity to Companions at Sub-Diffraction Limit Separations with HST WFC3/IR

TL;DR

This work develops and validates a double-PSF fitting technique that uses empirical, position-dependent PSFs to detect close companions in HST WFC3/IR data, including a thorough false positive and completeness calibration across eight filters and multiple S/N values. By creating and analyzing thousands of synthetic single and binary sources, the authors quantify detection limits and demonstrate sub-diffraction-limit sensitivity, achieving separations as small as $\sim$0.8 pixels (roughly $0.75\lambda/D$) at high S/N and up to $\sim$3.3 pixels for wider, high-contrast cases. Application to four known ultracool binaries reveals a mix of robust detections and near-threshold candidates, with improvements over previous surveys—up to about $2.5\times$ better in some filters—and confirms the method's capability to push BD multiplicity studies toward smaller separations. The results justify applying the approach to a large archival BD sample (the HUM survey) to constrain companion frequency, separation distribution, and mass ratios, while discussing limitations from undersampling and comparing with KPI/AMI. Overall, the paper establishes sub-diffraction, high-angular-resolution capabilities for WFC3/IR imaging and sets the stage for a comprehensive BD multiplicity census using archival HST data and future space-based facilities.

Abstract

We characterize the sensitivity of a double point-spread function (PSF) fitting algorithm -- employing empirical, position-dependent PSF models -- for detecting companions using the infrared channel of the Wide Field Camera 3 (WFC3/IR) on the Hubble Space Telescope (HST). The observed separation distribution of known brown dwarf (BD) binaries is potentially biased towards separations larger than the angular resolution limits of current techniques. Previous imaging analyses suffer from incompleteness at separations $<2λ/D$; our aim is to probe within this limit to identify previously missed companions. We evaluate the performance of our technique on artificial data across 8 WFC3/IR filters and a broad range of signal-to-noise ratios (S/N), determining our ability to accurately recover injected companions and identifying the region of parameter space where false positive fits are likely. Here, we demonstrate the capability of this technique to recover companions at sub-pixel separations on the WFC3/IR detector -- below the diffraction limit in multiple filters. For F160W at a typical S/N of 75, we resolve companions separated by 0.8 pixels (104 mas, $0.759λ/D$) at 1.5 magnitudes contrast with $>90\%$ confidence. We achieve the closest angular resolution for any detection method with WFC3/IR imaging to date. Compared to previous BD multiplicity surveys with WFC3/IR, we achieve a 2.5$\times$ improvement in separation sensitivity at contrasts of 0-3 magnitudes in F127M. We have demonstrated that applying our improved technique to archival HST images of field BDs will thus probe down to separations of 1 au, in one of the largest high angular resolution surveys of such objects to date.

Figures (8)

• Figure 1: False positive probability map in separation and contrast space of synthetic data in F160W with a S/N of 75. Overplotted is the 0.1% false positive probability line and the black stars represent the best-fit binary parameters from the double-PSF fitting algorithm to 1089 artificial single PSFs. Since all 1089 fits fall below the 0.1% probability line, fits above the line are inconsistent with a single PSF. The line appears higher than the black points because fits to singles can have individual samples of lower contrast than the best fit solution that have a false positive probability $>0.1\%$.
• Figure 2: Completeness probability map in separation and contrast of synthetic data in F160W with a S/N of 75. Overplotted is the 90% completeness contour and the black stars represent the true synthetic companion parameters which were used to bin the data. The best-fit double-PSF recovered parameters were used to make the heatmap.
• Figure 3: 40 False-positive probability curves plotted by filter. As S/N increases, the area of the parameter space where detections are 99.9% likely to not be false increases as well.
• Figure 4: Completeness probability contour plot in separation and contrast space of F160W with a S/N of 15. Since the completeness probability $> 90\%$ parameter space is fragmented, we would recover a true companion of this confidence less frequently than if the region was continuous. However, this does not mean recovery is impossible.
• Figure 5: 40 completeness probability curves plotted by filter. As S/N increases, we are able to accurately recover companion parameters with higher contrasts and closer separations.