Preprints

Preprint servers like arXiv, bioRxiv, medRxiv, ChemRxiv, HAL, or SSRN are a routine part of scholarly publishing, but the original BibTeX had no notion of them: there is no @preprint entry type, and no standard field for a preprint identifier. Several competing conventions have grown up to fill the gap, and none of them works everywhere. This page surveys those conventions, explains the one bibdeskparser recommends and implements, and spells out its tradeoffs.

Encoding preprints in BibTeX

The choices for representing a preprint as an entry in a .bib file are guided by pragmatism and the limitation of particular BibTeX .bst styles. The first choice is which entry type to use for a preprint. As far as standard entry types go, the potential candidates are @unpublished, @misc, or @article. BibLaTeX has additional types, like @online, but these are not supported by more traditional tooling.

The second question is in which fields to encode the preprint information. Styles only take into account standard fields (fields they know about, that is), and silently ignore all other data. For older styles, the list of fields is rather short.

The structured eprint fields

The modern de-facto standard for machine-readable preprint information is the eprint field family: eprint holds the identifier, archiveprefix names the server (assumed to be arXiv when absent), primaryclass gives the arXiv category (e.g., quant-ph), and a fourth, archive, gives the base URL of the server’s identifier pages (see below). The convention originated with the SLAC/SPIRES database (today’s INSPIRE-HEP) and the first three fields are included in arXiv’s own “Export BibTeX citation”. It is fully supported by many slightly more modern styles, and most importantly by REVTeX’s apsrev4-x/aipnum4-x, by Elsevier’s elsarticle styles, by biblatex (which reads archiveprefix/primaryclass as aliases for its native eprinttype/eprintclass fields, on any entry type, and hyperlinks arXiv: 2205.15044 [quant-ph] by default), and by classic styles retrofitted with urlbst.

The catch is that under the plain classic styles the fields are invisible: with plain, unsrt, abbrv, alpha, the natbib variants, or IEEEtran, an entry whose identifier lives only in eprint renders as just “Author. Title, year.” – the identifier is dropped without a trace.

Thus, for use with older styles that do not support these fields, preprint information must be additionally included in one of the standard fields. For styles that do support eprint, the way the information is rendered still depends on which entry type is used.

As a concrete example, preprints.pdf renders fourteen variants of the same references – an arXiv preprint-only paper, a HAL preprint-only paper, and a published paper with a preprint – in one REVTeX document (revtex4-2 with the default apsrev4-2 style and hyperref loaded).

Using @unpublished for preprints

A preprint-only work has, by definition, not been published, so @unpublished is arguably a “semantically accurate” type. However, unlike the other possible entry types, @unpublished requires a note field in every classic style. This represents both an opportunity and a challenge. If the style does not support the eprint fields, then putting the preprint information in the note as plain text, note = {arXiv:2205.15044}, would be appropriate. The drawback is that this is unstructured text: nothing hyperlinks it, and reference managers cannot parse it back out. Moreover, a note is usually positioned at the end of the citation instead of as “venue” between title and year, which may be undesired (and @unpublished ignores howpublished).

When the style does support the eprint field and renders it independently, the “challenge” is what to put in the mandatory note field. An attractive option is to use the note for the publication status, e.g. “preprint” for preprints that have not yet been submitted, “submitted” or “submitted to Phys. Rev. A” for preprints that are known to have been submitted, and a more descriptive note like “unpublished report” or “lecture notes” for material that has been uploaded to arXiv without the intention of publishing it as a peer-reviewed article.

REVTeX in particular (see preprints.pdf) renders this combination best of any option: the hyperlinked, category-tagged eprint moves into the venue position, before the publication year, followed by the note (“arXiv:2003.10132 [quant-ph] (2020), lecture notes”). This makes @unpublished by far the most attractive and richly featured option, as long as one is okay with defining a note.

Using @misc for preprints

@misc with the eprint fields is what arXiv’s own “Export BibTeX citation” hands out. Its advantage over @unpublished is that @misc has no mandatory fields, i.e., no required note field. If a note field (and/or a howpublished field) is present, it is rendered. However (at least in REVTeX), the placement is different from @unpublished: The eprint information is not in the “venue” location, but instead after the note. Instead, howpublished takes the “venue” location. This makes @misc acceptable for styles that do not support eprint. The preprint information should be stored in howpublished.

If the style does support eprint, it would render the information redundantly with howpublished, so howpublished should be omitted in that scenario. The result will be properly linked, but the preprint information will not appear in the “venue” slot.

Using @article for preprints

The last option abandons the eprint fields and instead puts the preprint reference directly in the journal field of a regular @article, e.g. journal = {arXiv:2205.15044} (Google Scholar’s BibTeX export popularized a variant of this). No style guide documents the practice, but it buys the one thing the structured forms cannot guarantee: every BibTeX style renders an article’s journal, so the identifier is never dropped. The citation comes out essentially as

M. H. Goerz et al., “Quantum optimal control via semi-automatic differentiation”, arXiv:2205.15044 (2022)

under plain, unsrt, IEEEtran, and REVTeX alike – the preprint reference formatted exactly like a journal reference, which is how readers actually treat it.

What it gives up is structure. Because the identifier is plain text in journal, the rendering shows “arXiv:2205.15044 (2022)”, unlinked by default. Proper linking may be reproduced by having a url field pointing to the URL of the preprint. Using a “pseudo-journal” like this is a double-edged sword: In a tool like BibDesk, we can potentially filter on the arXiv prefix, but any tool that reads journal strictly as a venue name (journal-abbreviation pipelines, reference-manager imports) may be misled by it.

Like @misc with howpublished, any eprint field should be omitted, or used exclusively with styles that do not support eprint. Otherwise, the preprint information will be presented redundantly in the resulting bibliography.

Nonstandard BibLaTeX types

Under BibLaTeX, an @online entry with the structured eprint fields renders and hyperlinks natively, in the right place, with the category tag, and without a required note to leave empty.

The bibdeskparser library does not build around a non-standard type like @online because BibDesk at its heart is a classic-BibTeX program. Its preview pane runs real bibtex, not biber, and a .bib database maintained for years across collaborations and journal submission systems cannot assume that every document downstream of it will use biblatex. An @online entry that renders beautifully under biblatex degrades to “Author. Title, year.” the moment someone feeds the same file to plain or IEEEtran. The stored form therefore stays within classic BibTeX; biblatex reads the eprint fields it carries anyway.

How bibdeskparser stores preprint-only entries

bibdeskparser stores a preprint-only work as an @unpublished entry that carries the structured eprint fields, a pseudo-journal, the DOI, and a publication-status note:

@unpublished{Wilhelm2003.10132,
    Author = {Wilhelm, Frank K. and others},
    Title = {An introduction into optimal control for quantum technologies},
    Journal = {arXiv:2003.10132},
    Eprint = {2003.10132},
    Archiveprefix = {arXiv},
    Primaryclass = {quant-ph},
    Doi = {10.48550/arxiv.2003.10132},
    Note = {preprint only},
    Year = {2020},
}

The journal value has the fixed shape <Archive>:<identifier> – a pseudo-journal. The recognized archives (arXiv, bioRxiv, medRxiv, ChemRxiv, HAL, and SSRN by default) are configurable via the [preprint_archives] table. Since arXiv assigns a DOI (10.48550/arXiv.<identifier>) to every preprint, and bioRxiv/medRxiv/ChemRxiv/SSRN identifiers are DOI-based to begin with, essentially every preprint-only entry can carry a doi; HAL deposits carry a stable URL instead.

The stored form redundantly combines fields that are relevant to different choices for exported entry types. The @unpublished type with the eprint fields is both semantically accurate and the best-rendering REVTeX form, so the stored entry can be copied verbatim into an external .bib file (the extra journal field is ignored on @unpublished), and BibDesk’s preview pane, which runs real BibTeX, shows the same clean rendering. The note records the publication status – “preprint only”, “submitted to Phys. Rev. A”, “lecture notes” – which renders in every BibTeX style and satisfies @unpublished’s required-field check; bibdeskparser never fills it in automatically in the database, so a missing note (which BibDesk flags as an incomplete required field) is a deliberate signal to record the status by hand. The pseudo-journal gives the entry a venue inside the database: BibDesk’s table view shows arXiv:2003.10132 in its journal column. The doi, last, is the canonical identifier for duplicate detection and the preferred hyperlink target.

import_bibtex() (and add()) normalize incoming preprint-only entries into this form. Recognition does not depend on any single marker: an entry is preprint-only if its journal is a recognized pseudo-journal (any entry type – including the earlier @article-based form of this convention), or if it is a @misc or @unpublished entry with an eprint from a recognized archive (e.g. arXiv’s own BibTeX export). Everything derivable is derived – the pseudo-journal from the eprint or vice versa, the doi from a doi.org resolver URL or (for arXiv) from the identifier itself – and a url that merely restates the archive’s page for the identifier is dropped when the entry carries a doi. Only the note is never derived.

Exporting preprint-only entries

When export() writes a preprint-only entry (of any stored form) for use with LaTeX, one choice remains, and it depends on the bibliography style the document will use – does it render the eprint field, or silently drop it? That choice is the preprint parameter of export (--preprint on the command line), with the default set by the preprint_export setting:

• preprint="unpublished" (the built-in default): @unpublished with eprint/archiveprefix/primaryclass and the doi. Under REVTeX (apsrev4-x, aipnum4-x) this is the best-rendering form (venue-position identifier, category tag, status note); it is equally understood by elsarticle and biblatex. Minimal exports guarantee the required note: the stored one, or the synthesized text “preprint” for an entry without one (full exports never synthesize it, so that re-importing a full export cannot plant a note in a library).

• preprint="misc": the same structured fields as a @misc entry – the form arXiv itself exports. Equivalent information; the eprint renders after the year instead of in the venue position, and there is no required note.

• preprint="article": the pseudo-journal form – @article with journal = {arXiv:2003.10132} and the DOI written as its resolver address in url. Use this for classic styles (plain, unsrt, abbrv, IEEEtran, …), where the structured forms would lose the identifier entirely.

• preprint="stored": no transformation, for inspecting the entry as stored.

Every form derives whatever it needs (eprint from the pseudo-journal or vice versa), so all stored forms export identically; minimal exports reduce to the essential fields of the chosen form – always including eprint/archiveprefix for the structured forms, and always keeping a stored note. The stored entry itself never changes.

The archive field

REVTeX’s apsrev4-x/aipnum4-x styles construct the hyperlink of a rendered eprint as <archive>/<eprint>, where archive is a BibTeX field with the built-in default https://arxiv.org/abs – the SPIRES-era companion of eprint/archiveprefix that modern BibTeX exports no longer emit. For arXiv identifiers the default is right; for any other archive it produces a broken link, and no other field can compensate (@unpublished ignores url entirely).

bibdeskparser therefore emits the archive field automatically wherever the structured eprint fields of a non-arXiv preprint are exported: Archive = {https://hal.science} for a HAL identifier, Archive = {https://www.biorxiv.org/content/10.1101} for bioRxiv, and so on. The base URL is derived from the archive’s URL template in the [preprint_archives] table whenever the template has the form <base>/{id}; archives whose page URLs embed the identifier differently (SSRN, ChemRxiv) get no archive field – for those, prefer the article export form, whose link does not rely on the eprint machinery. A stored archive field is always exported as-is; conversely, import drops an archive field that matches the derivable base, since exports regenerate it.

Preprint information on published papers

Recording eprint/archiveprefix on a published paper is valuable: The journal reference and DOI of a published article often lead to a paywall, while the eprint link leads to a free copy, so a bibliography that carries both serves every reader. High-energy physics has done this for decades: INSPIRE-HEP retains the arXiv identifier on every published record, and the SPIRES-era guidance explicitly asked for identifiers “for both published and unpublished papers”. The styles cooperate, rendering both parts – REVTeX’s apsrev4-x prints the journal reference followed by the hyperlinked arXiv:2205.15044 [quant-ph], biblatex does the equivalent with eprint=true (its default), and bibdeskparser’s own render() does the same, with the DOI-linked journal segment ahead of the eprint link. And it costs almost nothing to keep up: add_preprint() searches arXiv for the preprint matching a published entry and fills in the fields automatically; see How to fill in missing arXiv identifiers.

Minimal exports of published articles therefore include the eprint/archiveprefix/primaryclass fields – and, like full exports, the archive link base when the preprint is on a non-arXiv archive, so its link renders correctly.