STARBASE 451 :: Research

PROJECT ALPHA: VLASS TRIPLE RADIO SOURCE CLASSIFICATION ACTIVE

Radio Astronomy / AGN
Target: ApJS

▶

PAPER: Achong, Hooper, Morris, Goetz, Einwalter, Bach, Benedetto & Gordon (2026)
"A Quick Look at the 3 GHz Radio Sky. III. Characterizing the Triple Radio Sources in VLASS"
In preparation — Target journal: ApJS

ABSTRACT: Human visual classification of 1,836 triple radio sources identified in the Very Large Array Sky Survey (VLASS) at 3 GHz. Each source is classified along three independent axes — morphology, source bending, and flags — using a detailed workflow and glossary developed by a 7-person team across Lycoming College and UW-Madison. The goal is to characterize bent-jet AGN for use in IGM density measurements and as signposts to galaxy clusters and groups at cosmological distances.

— — — — — — — — — — — — —

MORPHOLOGY CLASSIFICATION SCHEME

TRIPLE (t)

Radio source with three distinct components — two lobes plus a visible radio core. The most common class in the sample.

CORE IDENTIFIEDBIDIRECTIONAL

DOUBLE (d)

Two-lobe source with no identifiable core component. The third "component" detected by DRAGNHunter may be an artifact, an unrelated source, or substructure.

NO CORETWO LOBES

X-SHAPE (x) / S-SHAPE (s)

X-shaped sources show a secondary wing axis misaligned from the primary jets. S-shaped sources show an undulating jet axis. Both indicate jet reorientation events.

JET REORIENTATIONWING EMISSIONPRECESSION

HEAD-TAIL (h-t) / ONE-SIDED (1-side)

Head-tail sources have emission swept so far back the two strands become nearly parallel — classic ram pressure morphology. One-sided sources show emission on only one side of the AGN, often due to relativistic beaming.

RAM PRESSURERELATIVISTIC BEAMINGIGM INTERACTION

COMPOSITE (c) / INCOMPLETE (i) / BLOB (b) / WEIRD (w)

Composite: multiple unrelated sources in the same detection. Incomplete: sub-structure of a larger source. Blob: indistinct diffuse emission. Weird: beyond all classification vocabulary — must be seen.

EDGE CASESARTIFACTS

BENDING CLASSIFICATION

Bending is rated on a 0–3 confidence scale focusing on systematic bending consistent with ram pressure from IGM motion — not jet reorientation (x/s-shapes).

VALUE	MEANING
0	No bending apparent
1	Possible bending — modest confidence, low S/N features
2	Reasonably clear systematic bending
3	High confidence — clearly bent jet axis or hotspot sequence
(asym)	Asymmetric bending — different confidence on each side of core
(t-p)	Trailing plume — lobes swept back further than jet bending implies

DATA & TOOLS

SOURCE / TOOL	ROLE	STATUS
VLASS (3 GHz)	Primary radio survey — 1,836 triple sources	ACTIVE
WISE (infrared)	Host galaxy identification	ACTIVE
PanSTARRS (optical)	Optical host confirmation	ACTIVE
DRAGNHunter	Automated triple component detection (Gordon et al. 2023)	ACTIVE
SAOImageDS9	Visual classification — 4-panel montage per source	ACTIVE
opends9.py	Driver script: auto-loads DS9 windows for each source	ACTIVE
Python / Astropy	Bending angle measurement, radius of curvature, catalog	IN PROGRESS
TOPCAT	Catalog cross-matching and table analysis	ACTIVE

CLASSIFICATION PROCESS

7 classifiers split into two teams by institution. Lycoming College (faculty + 3 undergrads including me) started from the lowest R.A. values. UW-Madison (senior scientist + 2 postbac scholars) started from the highest and worked downward. Teams met weekly. Each classifier worked independently before comparing notes, preserving original classifications in separate spreadsheets. Once teams crossed each other in the catalog, overlapping sources were discussed and a consensus classification was recorded.

MISSION PROGRESS

Human classification of all 1,836 sources

Bending angle measurements (Kaylan)

Radius of curvature automation (Melissa & Thane)

Artifact analysis & multi-parameter space

Host galaxy & environment cross-matching

Paper write-up

KEY REFERENCES

[1] Lacy et al. (2020) — VLASS overview

[2] Gordon et al. (2021) — VLASS component catalog

[3] Gordon et al. (2023) — VLASS DRAGN (double/triple) source catalog

[4] Morris et al. (2022) — IGM density from bent radio AGN

[5] Freeland & Wilcots (2011) — Bent jets as IGM probes

CONTRIBUTION: DRAGNs IN THE FOREST — RANDOM FOREST ARTIFACT IDENTIFICATION ACKNOWLEDGED

Radio Astronomy / ML
Submitted to ApJ

▶

"DRAGNs in the Forest: Identifying Artifacts with Random Forest Models in the VLASS DRAGNs Catalog"
Einwalter, Hooper, Morris, Bach & Gordon (2026) — arXiv:2512.20999 — Submitted to ApJ

I contributed visual classification of triple radio sources from the VLASS DRAGNs catalog as part of a related project. The paper trains random forest models to identify imaging artifacts in VLASS Quick Look DRAGNs — multi-component radio sources that may include spurious detections from the survey's simplified imaging algorithm. Acknowledged in the paper for assisting with visual classification of the triples dataset.

// PAPER SUMMARY //

The VLASS DRAGNs catalog (Gordon et al. 2023) contains 17,724 double and triple radio sources, ~11% of which are spurious artifact detections. This paper trains random forest classifiers on the 1,836 triple sources to identify 0-, 2-, and 3-artifact classes, then applies the model to classify the 15,888 double sources. The best model (log-log LAS S/N and flux S/N selected training set) achieves a weighted F1 score of 97.01%, producing a catalog with 99.3% completeness and 97.7% artifact-free purity — outperforming the existing DRAGNhunter Q-flag method.

// MY ROLE //

TASK	DESCRIPTION	STATUS
Visual classification	Manual inspection of VLASS triple sources for artifact identification	COMPLETE
Acknowledgement	Credited in paper acknowledgments alongside Mia Benedetto	PUBLISHED (arXiv)

// KEY RESULTS //

97.01%

Weighted F1 Score

99.3%

Catalog Completeness

97.7%

Artifact-Free Purity

17,724

DRAGNs Classified

// REFERENCES //

[1] Einwalter et al. (2026) — DRAGNs in the Forest, arXiv:2512.20999

[2] Gordon et al. (2023) — VLASS DRAGNs catalog (ApJS 267, 37)

[3] Gordon et al. (2021) — VLASS component catalog (ApJS 255, 30)

PROJECT BETA: X-SHAPED RADIO GALAXY CATALOG ONGOING

Radio Astronomy / AGN
Senior Research / Fellowship

▶

ABSTRACT: X-shaped radio galaxies (XRGs) exhibit a distinctive double-axis morphology: a primary jet axis and a secondary "wing" axis misaligned from the main lobes. This project builds a quantitative catalog of XRG properties — jet misalignment angles, wing-to-lobe flux ratios, and morphological classification metrics — starting from the Cheung (2007) sample. The goal is to characterize XRG structure with reproducible, numerical methods rather than subjective visual labels.

— — — — — — — — — — — — —

DATA SOURCES

SURVEY	WAVELENGTH	USE
FIRST	Radio (1.4 GHz)	Primary radio morphology
NVSS	Radio (1.4 GHz)	Total flux, extended emission
VLASS	Radio (3 GHz)	High-resolution morphology
LoTSS	Radio (144 MHz)	Low-frequency extended structure
SDSS	Optical	Host galaxy identification
Pan-STARRS	Optical	Host galaxy backup / overlap

PIPELINE WORKFLOW

STEP 1 — Source Selection

~20 XRG candidates from Cheung (2007). Download radio FITS + optical host galaxy images. Output: initial source list CSV.

CHEUNG 2007~20 SOURCES

STEP 2 — Image Visualization

Load images into DS9. Overlay radio contours on optical. Identify core, lobes, and wings. Verify X-shape morphology visually.

DS9RADIO/OPTICAL OVERLAY

STEP 3 — Jet Geometry Measurement

Measure position angles of primary jet axis and secondary wing axis. Compute misalignment angle: ΔPA = |PA_jet − PA_wing|. Store in CSV tables.

POSITION ANGLESMISALIGNMENT ΔPA

STEP 4 — Flux Measurements

Define regions in DS9 for each lobe and wing. Integrate flux per region. Compute wing-to-lobe flux ratio and lobe asymmetry metrics.

REGION FLUXWING/LOBE RATIOLOBE ASYMMETRY

STEP 5 — Classification Metrics

Design numerical morphological metrics: wing prominence, jet bending angle, lobe symmetry, wing/lobe flux ratio. Goal: reproducible quantitative labels, not subjective judgements.

WING PROMINENCEBENDING ANGLESYMMETRY

STEP 6 — Automated Pipeline

Python scripts xrg_pipeline.py and flux_tools.py load FITS files, normalize catalogs, extract axes via PCA, and compute flux automatically.

xrg_pipeline.pyflux_tools.pyPCA AXIS EXTRACTION

STEP 7 — Catalog Production

Final output: standardized dataset with columns:

Source	Survey	Jet PA	Wing PA	ΔPA	Lobe Flux	Wing Flux	Ratio
— catalog in progress —

STEP 8 — Colloquium Presentation

Preparing a ~35 minute research talk (~30 slides). Structure: AGN background → XRG overview → data sources → methodology → pipeline → results → implications & future work.

~35 MIN TALK~30 SLIDESIN PREP

MISSION PROGRESS

Source selection (Cheung 2007 sample)

Radio/optical overlays

Jet geometry measurements

Flux measurements

Python pipeline (xrg_pipeline.py)

Catalog production

Colloquium presentation

PROJECT GAMMA: GLOBULAR CLUSTER PROPER MOTION EARLY STAGE

Stellar Dynamics
Observational Astronomy

▶

ABSTRACT: Analysis of proper motion and structural changes in globular clusters using multi-epoch archival FITS imagery. The project tracks changes in cluster angular size and stellar color profiles over time using archival data from MAST, with potential extension to Gaia proper motion vectors for individual stars.

— — — — — — — — — — — — —

STEP 1 — Data Collection

Download multi-epoch FITS images from MAST database. Multiple observation dates required to detect changes.

MASTMULTI-EPOCH

STEP 2 — Visualization

Load images into DS9. Examine cluster structure and compare epochs side-by-side.

DS9

STEP 3 — Angular Size Measurement

Track changes in cluster core and tidal radius across epochs. Detect structural evolution.

CORE RADIUSTIDAL RADIUS

STEP 4 — Color Analysis

Measure color differences across the cluster to analyze stellar population distributions and age gradients.

COLOR PROFILESSTELLAR POPULATIONS

POSSIBLE EXTENSIONS

If DS9 proves insufficient: switch to Gaia proper motion catalog data and use Python for full vector motion analysis of individual cluster members.

GAIA DR3PYTHON VECTOR ANALYSIS

MISSION PROGRESS

Literature review

Data collection from MAST

Visualization setup

Measurements & analysis

METHOD: RADIO-OPTICAL SOURCE CROSS-MATCHING ACTIVE

Multi-wavelength Astronomy
Supporting XRG Research

▶

ABSTRACT: Overlay radio emission maps onto optical host galaxy images to determine host galaxy location, jet orientation relative to the host, and confirm source identity. This is a core supporting technique for the XRG morphology catalog project.

— — — — — — — — — — — — —

STEP 1 — Load Optical Image

Retrieve SDSS optical frame for the target host galaxy field.

SDSS

STEP 2 — Load Radio Map

Load FIRST or NVSS radio FITS image of the same field.

FIRSTNVSS

STEP 3 — Overlay Contours

Render radio contours on top of the optical background image in DS9. Adjust contour levels to reveal morphological structure.

DS9 CONTOURS

STEP 4 — Verify Alignment

Confirm that the radio core position matches the optical host galaxy nucleus. Determine jet orientation axis relative to host morphology.

ASTROMETRIC CHECKJET ORIENTATION

Output: annotated radio-optical composite images used as figures in the XRG catalog and colloquium presentation.

TOOLS: FITS PROCESSING AUTOMATION ONGOING

Research Software
Astronomy Computing

▶

ABSTRACT: A suite of Python tools to automate repetitive FITS image handling tasks across all active research projects. Reduces manual file loading, standardizes visualization settings, and enables batch processing of large source catalogs.

— — — — — — — — — — — — —

TOOLS

SCRIPT	PURPOSE	STATUS
opends9.py	Open FITS files in DS9 automatically, apply visualization settings	IN DEV
xrg_pipeline.py	End-to-end XRG morphology measurement pipeline	IN DEV
flux_tools.py	Region flux extraction and wing/lobe ratio computation	IN DEV

opends9.py — PLANNED FEATURES

Current Capability

Opens a specified FITS file in DS9 and applies basic visualization settings. Reduces manual drag-and-drop file handling.

Planned Improvements

Tkinter GUI with file picker, batch loading of full source folders, automated overlay controls, and export of measurements directly to CSV.

TKINTER GUIBATCH LOADINGAUTO OVERLAYSCSV EXPORT

MISSION PROGRESS

opends9.py — basic DS9 launcher

opends9.py — Tkinter GUI

xrg_pipeline.py — core structure

flux_tools.py

Full batch automation