If you type in words in one of the search fields, the Front will return articles with all of the words in the arXiv metadata record, ignoring order and case. (Useful full-text search is non-trivial and we apologize that we don't yet have it.) To search for an exact phrase, put the phrase in quotes or connect the words with dashes. Quotes are also necessary for phrases with parentheses, brackets, and the words "NOT" and "OR". Here are some examples:

The Abstract+ field is for the full metadata record. In the Author field, there are various ways to search on more than just last names: "First M. Last" or last-f or f.last or last-first. The Title/ID field matches both titles and arXiv ID number of e-prints. Terms in the Category field can be arXiv category codes or words in category names, or PACS, ACM, or MSC numbers. This field is also an exception in that the Front tries to match at least one term rather than all of them.

If you use more than one search field, the Front will return results that match all fields.

Search queries can also have Boolean phrases and field operators. For example:

manifolds not 3-manifolds 4-manifolds
primary:[math.AG physics.HT]
abs:shelah

Terms in parentheses are combined using AND. (AND is just a regular search term.) Terms in square brackets are combined using OR (although x OR y also works). x NOT y has the obvious meaning, while x NOT y z means (x NOT y) NOT z. Fields are indicated with colons. Here is a summary of available fields:

Although field restrictions make the most sense in the Abstract+ search window, they are valid in every window in the search box.

You can also type search requests directly into URLs, provided that you convert spaces to pluses (+) and avoid some special URL symbols (?&%=/#). For example:

http://front.math.ucdavis.edu/search/chern-simons
http://front.math.ucdavis.edu/search/title:(kontsevich+formality)
http://front.math.ucdavis.edu/search/author:[schramm+lawler+werner]+cat:math.PR

Actually, an unfielded search does not quite cover every field. Because they are highly repetitive, pages, version numbers, and formats are omitted by default.

The search engine treats all punctuation in arXiv records as equivalent to spaces. It ignores case and tries to strip away TeX accents. It stems search words to eliminate past tense and plural endings: matrix is equivalent to matrices and quench is equivalent to quenched. German transliterations are also equivalent: tschüß, tschuss, and tschuess all match each other. (For the truly technical-minded, the stemming algorithm is a variation of stage 1 Porter stemming.)

Last but not least, the search engine assumes a word break between letters and numbers, so that SU2, SU_2, and SU(2) all match. (Note that the last form must appear in quotes in a search because of the special meaning of parentheses.)

- for questions or comments about the Front arXiv contact page - for questions about downloading and submitting e-prints