StraKLIP: A novel pipeline for detection and characterization of close-in faint companions through Karhunen-Lo\^eve Image Processing algorithm

TL;DR

The paper introduces StraKLIP, a new pipeline utilizing Karhunen-Loève Image Processing to detect and characterize faint companions close to stars in space-based imaging data, achieving high sensitivity even at small separations.

Contribution

StraKLIP is a novel pipeline that applies Karhunen-Loève PSF subtraction to wide-field space images, enabling detection of faint companions at small angular separations, even in datasets not originally intended for high contrast imaging.

Findings

Successfully tested on HST datasets from Orion Nebula Cluster.

Detects brown dwarf-mass companions near the planetary mass limit.

Achieves reliable detection at separations as close as 0.1 arcseconds.

Abstract

We present a new pipeline developed to detect and characterize faint astronomical companions at small angular separation from the host star using sets of wide-field imaging observations not specifically designed for High Contrast Imaging analysis. The core of the pipeline relies on Karhunen-Lo\^eve truncated transformation of the reference PSF library to perform PSF subtraction and identify candidates. Tests of reliability of detections and characterization of companions are made through simulation of binaries and generation of Receiver Operating Characteristic curves for false positive/true positive analysis. The algorithm has been successfully tested on large HST/ACS and WFC3 datasets acquired for two HST Treasury Programs on the Orion Nebula Cluster. Based on these extensive numerical experiments we find that, despite being based on methods designed for observations of single star at a time, our pipeline performs very well on mosaic space based data. In fact, we are able to detect brown dwarf-mass companions almost down to the planetary mass limit. The pipeline is able to reliably detect signals at separations as close as $\gtrsim 0.1 "$ with a completeness of $\gtrsim 10\%$, or $\sim 0.2"$ with a completeness of $\sim 30\%$. This approach can potentially be applied to a wide variety of space based imaging surveys, starting with data in the existing HST archive, near-future JWST mosaics, and future wide-field Roman images.