guanyilun/agent-sh/ads

ADS Query Syntax & Search Strategy

Reference: <https://ui.adsabs.harvard.edu/help/search/>

Query Basics

• Space = AND, OR, NOT / - (exclude), () group, "phrase"
• Precedence: NOT > AND > OR > implicit-AND > -

Fielded Search

Text Fields

Author

Precision: author:"murray, s" (broadest) → "murray, stephen" → "murray, stephen s" (tightest)

Identifiers & Publication

bibcode:ID, doi:DOI, arXiv:ID, identifier:ID, bibstem:ApJ, volume:N, issue:N, page:N, year:YYYY or year:YYYY-YYYY, pubdate:[YYYY-MM TO YYYY-MM]

Affiliation

aff:"phrase" (raw), inst:"Harvard U" (curated, matches variants), inst:"UCLA/IGPP" (department), aff_id:ID

Object & Position

object:Andromeda (SIMBAD+NED), object:"(SMC OR LMC) AND M31" (combined), cone search: object:"RA Dec:radius" (decimal deg or sexagesimal, default 2', max 60')

Classification & Counts

database:astronomy|physics|general, doctype:type, arxiv_class:"class", property:flag, bibgroup:name, citation_count:[10 TO 100], author_count:[10 TO 100], read_count:N, facility:/regex/, orcid:ID, grant:agency

Synonyms & Case Sensitivity

• lowercase → matches variations + synonyms (e.g. title:star → star, stars, stellar…)
• UPPERCASE → acronym-only (title:STAR → only "STAR")
• =field: prefix → exact match, no synonym expansion

Wildcards & Proximity

Positional Search — pos(query, positions...)

Works on: author, aff, title. Examples: pos(author:"Oort, J", 2) (2nd author), pos(author:"Oort", 1,3) (1st–3rd), pos(author:"Oort", -1) (last), pos(aff:"harvard", 1) (1st author at Harvard)

Citation & Reference Operators

citations(query) = papers citing query results, references(query) = papers cited by query results. Nestable with other terms.

Second-Order Operators

similar(query) (textually similar), trending(query) (read by same readers), useful(query) (frequently cited by results), reviews(query) (citing papers ranked by freq), topn(N, query, sort)

Tip: similar() excludes originals; use disjoint entdate ranges to control overlap.

Properties

refereed, notrefereed, article, nonarticle, openaccess, ads_openaccess, pub_openaccess, eprint_openaccess, author_openaccess, data, esource, inspire, toc

Document Types

article, eprint, inproceedings, inbook, abstract, book, bookreview, catalog, circular, erratum, mastersthesis, newsletter, obituary, phdthesis, pressrelease, proceedings, proposal, software, talk, techreport, misc

Bibliographic Groups

Institutional: ARI, CfA, CFHT, Leiden, USNO | Telescope: ALMA, CXC, ESO, Gemini, Herschel, HST, ISO, IUE, JCMT, Keck, Magellan, NOIRLab, NRAO, ROSAT, SDO, SMA, Spitzer, Subaru, Swift, UKIRT, XMM

Daily arXiv Listings

Use entdate (day-resolution) with doctype:eprint and arxiv_class. Note: pubdate is month-resolution only.

# Today's astro-ph.CO (±2 day window)
arxiv_class:"astro-ph.CO" entdate:[2026-04-07 TO 2026-04-09] doctype:eprint
 
# All astro-ph classes
arxiv_class:("astro-ph.CO" OR "astro-ph.GA" OR "astro-ph.HE" OR "astro-ph.IM" OR "astro-ph.EP" OR "astro-ph.SR") entdate:[2026-04-07 TO 2026-04-09] doctype:eprint

Subclasses: CO (Cosmology), EP (Earth & Planetary), GA (Galaxy Astrophysics), HE (High Energy), IM (Instrumentation), SR (Solar & Stellar).

Search Strategy: Simplicity First

• Start with 2–3 core nouns. If too many results, add constraints (year, author, title:). Don't start with a sentence.
• ADS is keyword/metadata matching, not semantic search. Overloading queries kills recall.
• Use detail='full' on fewer results rather than compact on many — abstracts often contain the answer.

Common Pitfalls

1. Compound multi-concept queries → zero results. Fix: 1–2 core terms + sort_by + year_from:. Scan abstracts for specifics.

# Bad: "Population III He II helium 1640 JWST metal-poor galaxy candidate"
# Good: "Population III JWST"  sort_by:date  year_from:2024

2. Non-article doctypes polluting results. Fix: doctype:article or property:refereed.

3. Iterative reformulation without simplifying. If 2+ zero-result queries on the same topic: strip to 1–2 broadest terms, or use web_search first to find titles/authors, then look up in ADS.

4. Use `web_search` → ADS as a two-step pipeline. For complex/emerging topics, start with web_search (plural queries, 2–4 varied angles) to identify key papers/arXiv IDs, then verify via ads_paper or targeted ads_search.

When Keyword Search Fails, Use Citation Networks

For interdisciplinary connections, find a known paper then walk citations (ads_citations) or references. Citation chains beat keyword matching for cross-domain discovery.

Quick Query Recipes

# Recent refereed articles on a topic
abs:"dark energy" year:2023-2025 property:refereed doctype:article
 
# Papers by first author at an institution
author:"^smith, john" inst:"MIT"
 
# Highly-cited reviews
reviews("machine learning astronomy") property:refereed
 
# Papers citing a specific work, sorted by citations
citations(bibcode:2023ApJ...950L..12A) sort:citation_count
 
# Object in specific journal
object:Andromeda bibstem:ApJ
 
# Open access with data links
property:openaccess property:data keyword:"gravitational waves"
 
# Exclude preprints, keep only journal articles
abs:"transiting exoplanet" doctype:article -arxiv_class:*
 
# Cone search
object:"12h30m49.4s +12d29m40s:5"