Workflow 10 min read 21 October 2026

PDF in Scientific Publishing: Accessibility, Data Embedding, MathML, and the Preprint Workflow

Q: How do I make a LaTeX paper PDF accessible?

The most complete approach for LaTeX accessibility is using the accessibility LaTeX package or the newer tagpdf package (currently in development as part of the LaTeX3 project) to enable tagged PDF output from LuaLaTeX. For mathematical accessibility specifically, LaTeXML converts LaTeX source to HTML+MathML, from which accessible PDF can be generated. For practical near-term use: generate your PDF normally from LaTeX, then open in Acrobat Pro to add tags via the Accessibility wizard, set the document language, add Alt text to figures, and verify with PAC 2024. This hybrid approach is less automated but achieves PDF/UA-1 compliance for most scientific papers. PDF/UA-2 with embedded MathML requires the LaTeXML or tagpdf route.

Q: Do major journals require accessible PDFs?

As of 2024-2025, accessibility requirements for journal submissions are in transition. Most major publishers (Elsevier, Springer Nature, Wiley, ACS, APS) have accessibility guidelines and encourage accessible PDF submission, but few have made it a hard submission requirement. The trend is toward requirements driven by: funder mandates (NIH, Wellcome Trust), institutional accessibility policies at universities, and regulatory frameworks (EU Web Accessibility Directive applies to academic institutional repositories). arXiv generates experimental accessible HTML alongside PDF. PLOS has accessibility guidelines for submissions. The strongest current requirement is for NIH-funded research: NIH's revised Public Access Policy increasingly addresses accessibility of published outputs. Expect accessibility to become a hard submission requirement at major publishers within 3–5 years.

Q: How does PDF/A-3 handle research data files embedded in a paper PDF?

PDF/A-3 uses the /AF (Associated Files) mechanism to embed data files as associated file streams within the PDF container. Each embedded file has an AFRelationship value: "Data" means the embedded file contains data represented in the PDF (the appropriate value for embedded research datasets); "Alternative" means an alternative representation of the same content; "Source" means the embedded file was the source used to create the PDF. The embedded file appears in the PDF's /EmbeddedFiles name tree and can be extracted by any PDF viewer that supports attachments. File size is a practical constraint — large datasets (>10MB) should be deposited separately in a data repository and linked via DOI, rather than embedded directly in the paper PDF.

Academic and scientific publishers face unique PDF challenges: complex mathematical notation, embedded datasets, figure accessibility, preprint workflows, and evolving journal submission requirements. This guide covers MathML in PDF, tagged figures, PDF/A-3 for embedded research data, and how arXiv, bioRxiv, and major journal publishers handle PDF.

Scientific and academic PDF documents occupy a demanding niche: they contain complex mathematical notation, embedded datasets, high-resolution figures with captions, citation networks, and must meet journal submission requirements that vary by publisher. At the same time, accessibility requirements are increasingly being applied to academic content — funding agencies, universities, and disability law are all converging on the expectation that scientific literature should be readable by screen reader users. The good news is that the technical machinery to produce high-quality, accessible scientific PDFs exists and has been available for years. The challenge is connecting that machinery to the workflows that most researchers actually use: primarily LaTeX and Microsoft Word, with output destined for arXiv, institutional repositories, or journal submission systems. This guide covers the specific PDF challenges of scientific publishing: mathematical content, data embedding, figure accessibility, preprint workflows, and how major publishers handle PDF accessibility today.

Mathematical content: from TeX to accessible PDF

LaTeX remains the dominant authoring tool for mathematics-heavy scientific literature. The standard LaTeX-to-PDF pipeline (pdflatex, XeLaTeX, or LuaLaTeX) produces visually excellent PDF — but historically inaccessible PDF: mathematical expressions are rendered as sequences of individual glyph drawing operations with no semantic structure, and the resulting text extraction is garbled. The LaTeXML + MathML route — LaTeXML converts LaTeX source to XML, including MathML for mathematical content. Combined with a PDF rendering pipeline that preserves the MathML in the output, this can produce PDF/UA-2 compliant scientific PDFs. The workflow is more complex than simple pdflatex compilation and requires post-processing steps. arXiv's accessibility project — arXiv has worked on accessible HTML output (via latexml) alongside the traditional PDF. As of 2024, arXiv generates experimental HTML versions of submitted papers when possible, providing a MathML-accessible alternative to the PDF. This is the direction of travel for the preprint ecosystem. Word + MathType route — Microsoft Word with MathType or the built-in Office Math (OMML) equation editor can produce equations that round-trip to MathML. Word's PDF export with accessibility settings can embed MathML in the output via the /Formula structure element. This pathway is more accessible than LaTeX for non-specialist authors producing quantitative research. The Alt text fallback — For documents that cannot be regenerated with MathML, adding Alt text descriptions to equation images in the /Alt attribute of /Figure elements is the fallback. The description should be in plain English readable form of the equation: "integral from zero to infinity of e to the negative x squared dx equals square root of pi over 2". This is inadequate for complex equations but better than nothing.

Embedded research data: PDF/A-3 for scientific datasets

PDF/A-3 allows arbitrary file types to be embedded as associated files alongside the human-readable PDF content. This makes PDF/A-3 an attractive format for research data publications that need to be simultaneously human-readable and machine-processable: Data file embedding patterns: - A research article with embedded CSV/TSV data files for all figures — readers can access the raw data directly from the PDF without a separate download. - A scientific protocol document with embedded XML data conforming to a domain standard (HL7 for clinical trials, Darwin Core for biological data, DataCite XML for dataset metadata). - An engineering report with embedded CAD files (STEP format) that CAD software can open directly from the PDF. FAIR data principles and PDF/A-3 — The FAIR data principles (Findable, Accessible, Interoperable, Reusable) advocate for machine-readable data alongside human-readable publications. PDF/A-3 with associated data files addresses the Accessible and Reusable dimensions — the data is bundled with the publication and survives as a single archival unit. Limitations — Embedded files in PDF/A-3 increase file size and are not automatically indexed by publication databases. The embedding adds value for direct distribution but does not replace formal data repositories (Zenodo, Figshare, Dryad) for data discovery. Best practice: embed data in the PDF/A-3 file and deposit in a domain data repository with a DOI.

Figure accessibility and tagged scientific PDFs

Scientific figures — plots, diagrams, microscopy images, flow charts — are among the most information-dense elements of a research paper. Making them accessible is both technically required (under PDF/UA) and genuinely challenging: Alt text for figures — Every /Figure element in a tagged PDF must have an /Alt attribute with alternative text. For scientific figures, a good Alt text is not "Figure 2" but a description that conveys the scientific content: "Bar chart showing mean reaction time in milliseconds for three experimental conditions: control (450ms), low-dose (380ms, p<0.01), and high-dose (320ms, p<0.001). Error bars show standard error of the mean." Extended descriptions — For complex figures where the full information cannot be conveyed in a short Alt text, PDF/UA-2 recommends using /ActualText or an extended description in the document body. Some publishers are experimenting with structured data captions alongside figures that provide full data tables equivalent to the chart. Colour contrast for figures — WCAG 2.1 non-text contrast requirements apply: graphical information conveying meaning must have at least 3:1 contrast against adjacent colours. Many scientific plotting tools default to colour schemes that fail contrast requirements. The viridis and cividis colormaps were specifically designed to remain distinguishable for colour-blind readers and meet contrast requirements. Figure caption tagging — Figure captions should be tagged as /Caption elements within the /Figure element, not as /P paragraphs. This allows assistive technologies to associate captions with their figures programmatically.

Frequently asked questions

How do I make a LaTeX paper PDF accessible?

The most complete approach for LaTeX accessibility is using the accessibility LaTeX package or the newer tagpdf package (currently in development as part of the LaTeX3 project) to enable tagged PDF output from LuaLaTeX. For mathematical accessibility specifically, LaTeXML converts LaTeX source to HTML+MathML, from which accessible PDF can be generated. For practical near-term use: generate your PDF normally from LaTeX, then open in Acrobat Pro to add tags via the Accessibility wizard, set the document language, add Alt text to figures, and verify with PAC 2024. This hybrid approach is less automated but achieves PDF/UA-1 compliance for most scientific papers. PDF/UA-2 with embedded MathML requires the LaTeXML or tagpdf route.

Do major journals require accessible PDFs?

As of 2024-2025, accessibility requirements for journal submissions are in transition. Most major publishers (Elsevier, Springer Nature, Wiley, ACS, APS) have accessibility guidelines and encourage accessible PDF submission, but few have made it a hard submission requirement. The trend is toward requirements driven by: funder mandates (NIH, Wellcome Trust), institutional accessibility policies at universities, and regulatory frameworks (EU Web Accessibility Directive applies to academic institutional repositories). arXiv generates experimental accessible HTML alongside PDF. PLOS has accessibility guidelines for submissions. The strongest current requirement is for NIH-funded research: NIH's revised Public Access Policy increasingly addresses accessibility of published outputs. Expect accessibility to become a hard submission requirement at major publishers within 3–5 years.

How does PDF/A-3 handle research data files embedded in a paper PDF?

PDF/A-3 uses the /AF (Associated Files) mechanism to embed data files as associated file streams within the PDF container. Each embedded file has an AFRelationship value: "Data" means the embedded file contains data represented in the PDF (the appropriate value for embedded research datasets); "Alternative" means an alternative representation of the same content; "Source" means the embedded file was the source used to create the PDF. The embedded file appears in the PDF's /EmbeddedFiles name tree and can be extracted by any PDF viewer that supports attachments. File size is a practical constraint — large datasets (>10MB) should be deposited separately in a data repository and linked via DOI, rather than embedded directly in the paper PDF.

Try it yourself

Everything in this article can be done directly in the free PDFTheGuide editor — no account required.

Edit PDF Annotate PDF Open editor