Chemistry papers in LaTeX: formulas, schemes, and SI units

Chemists usually show up in LaTeX for one of two reasons: the journal insists, or a collaborator already works that way. Either way the pain points look familiar: reaction schemes, isotope labels, condition tables, and a bibliography stuffed with journals and patents.

This guide covers undergrad reports through first journal submissions. It will not replace ChemDraw for complex natural products. It will keep the manuscript and the experimental section from fighting the compiler.

Reactions with mhchem

mhchem is the usual answer for formulas and reactions:

\usepackage[version=4]{mhchem}
\ce{CO2 + C -> 2CO}
\ce{^{14}C}
\ce{H2SO4}
\ce{Fe^2+ <=>[\ce{H2O}] Fe^3+}

Pin version 4 so syntax does not drift between coauthors, or between Overleaf and a local TeX install. Longer walkthrough: chemical formulas.

Multi-step schemes with reagents above and below the arrow need more than inline \ce. mhchem handles the formulas; layout often wants a journal scheme environment or an external figure. Do not cram a twenty-step total synthesis into a run of inline chemistry lines.

Structures: chemfig or external drawings

chemfig draws simple structures in pure TeX and keeps everything vector-local. Complex natural products are almost always better as ChemDraw, Marvin, or similar, exported to PDF, then \includegraphics. Prefer vector PDF over a screenshot PNG so bond lines stay sharp in print and in two-column layout.

Name files by compound ID (cmpd-12.pdf). Keep the editable ChemDraw source in the lab archive even if git only holds the PDF that goes in the paper.

Experimental sections and long tables

Condition tables (temperature, solvent, yield, catalyst loading) work well with booktabs and often siunitx for numeric columns:

\usepackage{booktabs,siunitx}
\begin{tabular}{l S S}
  \toprule
  {Entry} & {T / \degreeCelsius} & {Yield / \percent} \\
  \midrule
  1 & 25 & 72 \\
  \bottomrule
\end{tabular}

If a table runs longer than a page, switch to longtable before deadline week, not after the journal complains about 7pt fonts. The caption should say what “yield” means (isolated, NMR, GC) so readers are not guessing.

Units and spectroscopy snippets

Pick a unit style and stick to it: siunitx, or manual \mathrm if the group already agreed on that. NMR and HRMS snippets in experimental sections are often plain text with careful spacing. Do not invent a fragile macro language mid-paper unless the lab already shares one.

Report significant figures the way the notebook does. LaTeX will happily print false precision if you type it in.

Schemes vs figures

Many chemistry journals distinguish schemes (reactions) from figures (plots, ORTEP, spectra). Use the environments the journal class ships (scheme, figure) so numbering matches production. Cross-ref with \ref and stable labels (sch:reduction, fig:ortep-1).

Safety of the draft

Unpublished routes and industrial partners turn cloud-only editors into a policy question, not only a convenience question. Keeping .tex and figures as ordinary files on disk with Git is boring in a useful way.

Oleafly is a local research editor: bundled compilers, project-wide ref and cite checks, offline spell and grammar on prose only (so \ce and cite keys are not flagged as “typos”), real Git checkpoints, optional AI that only edits through approval diffs. No account. The draft does not need a third-party editor to exist. Network is for packages, citation lookup, or AI if you turn those on.

Cite the right primary literature

Chemistry bibliographies lean hard on journals and patents. Export from Zotero or EndNote into a .bib with stable keys, then validate missing fields before submission: BibTeX validator. Empty bibliographies are almost always a path or backend mistake, not missing citations.

Journal styles differ (ACS, RSC, Springer). Use the bibliography style the template ships with. Do not glue numeric and author-year together in one preamble.

Collaboration

Own sections by person when you can. One person owns the .bib. Agree on mhchem version and structure-export format up front. When several experimental sections grow in parallel, compile the full PDF every week so integration surprises show up early.

Pre-submission checklist

Worth a final pass:

• mhchem version pinned; formulas compile without warnings you have been ignoring.
• Structures are vector where possible; compound IDs match across text, schemes, and SI.
• Tables use three-rule style; units consistent; yields defined in the caption or notes.
• Scheme/figure numbering matches the journal class.
• .bib validated; style matches the template.
• Sensitive routes stay in a local folder with access control that matches lab policy.

SI and journal checklists

Many chemistry journals publish author checklists (graphics resolution, TOC graphic size, compound characterization). Skim that PDF before you draw schemes. A wrong TOC graphic aspect ratio is a dumb reason for an admin rejection.

Supporting information

SI often has its own template or heading rules. Keep SI sources in the same repo as the main paper with clear filenames. When the journal requires full characterization, check that every compound number in the main text actually shows up in the SI.

Chemistry formatting is fussy because the objects are dense. Get the reactions and tables right and the experimental section reads like careful work, not a paste from a notebook PDF.