#Two columns and multicols

#Two-column documents

\documentclass[twocolumn]{article}
\begin{document}
\begin{abstract} … \end{abstract}
\section{Introduction}
…
\begin{figure*} … \end{figure*}     % spans both columns
\end{document}

\documentclass[twocolumn]{…}, or \twocolumn anywhere in the source, sets the whole document in two columns. \twocolumn can't be switched off halfway: \onecolumn is removed and does nothing. The content of \twocolumn[…] is kept (ICML puts its title block there). Conference styles that are two-column by themselves are recognised too: acl, aaai…, ijcai…, the IEEEtran class and acmart with sigconf; see Papers. In a beamer document, twocolumn is ignored.

#When the columns appear

The columns need room. A two-column document is laid out in two columns only while its container is at least 50rem wide (800px at the default font size), and in one column below that, e.g. on a phone. The check is redone whenever the window is resized.

#What spans both columns

Web pages scroll, so HaTeX doesn't fill pages. It cuts the document into stretches at every item that spans the full width, and balances each stretch over two columns of about equal height:

• \section, \part and \chapter headings. \subsection and lower stay in their column;
• the abstract;
• figure* and table*;
• a multicols block inside the document.

A reader therefore reads down one short column, up to the top of the next one, and on to the next full-width item. They never have to scroll back up a whole page.

#What stays in its column

Everything else stays in its column, as in LaTeX:

#Where a column breaks

The runtime measures the stretch and picks the break closest to half its height. A column can break:

• between any two blocks (paragraphs, lists, floats, theorems, code);
• between the paragraphs or items of a theorem, proof, quote, list, bibliography or the footnotes;
• inside a paragraph, just before or just after a display equation.

A column never breaks inside a line of text, and never directly after a heading. When the break falls inside a list or theorem, the element is split into two copies (the second has the class hatex-split-after), and a numbered list continues its numbering in the second column.

#multicols

\begin{multicols}{3}[\subsection*{Glossary}]
  … \columnbreak …
\end{multicols}

• \begin{multicols}{n} sets its content in n balanced columns. n is between 1 and 4, and 2 if it can't be read.
• The optional […] after it is set across the full width above the columns, as in the multicol package.
• multicols* is treated as multicols: the columns are always balanced.
• The columns appear while the block is at least 32rem wide.
• \columnbreak ends the current column there. Without it, the breaks are chosen as described above.

Unlike a two-column document, multicols doesn't have spanning items: a heading inside it stays in its column. A multicols block inside a two-column document spans the page.

#How it works, and when to call layout

The runtime lays out the columns itself instead of using CSS multi-column layout, which misplaces KaTeX's inline maths in Safari. The HTML from parse contains the whole document in one flow (.hatex-cols-flow), and enhance splits it into columns (.hatex-cols-chunk > .hatex-col):

• when the document is rendered, and again when web fonts have loaded;
• whenever the window is resized;
• whenever a TikZ picture arrives, since that changes the heights.

If the container changes width for any other reason (a sidebar closes, a split pane is dragged, a tab becomes visible), call:

HaTeX.layout(document.querySelector('.hatex'));

Pre-rendered HTML that is never enhanced (no hatex.js on the page) shows in one column.

Previous: Cross-references, citations and footnotes · Contents · Next: Slides