I thought I knew a thing or two about post formatting, but these posts on the two Mathematics sites surprised me: Vector Convolution? (Mathematics SE) and Vector convolution? (MathOverflow). The first equation in those posts uses MathJax but without the familiar $ or $$ (or on selected sites, \$) delimiters; instead MathJax seems to detect \begin{equation} and \end{equation} now. This very old post seems to suggest that didn't work before, but I assume that was an old version of MathJax. Is this behaviour documented somewhere, and are there more delimiters like this?


1 Answer 1


Here are the delimiters that I confirmed to work on MathOverflow (it will be a bit different on sites that use \$ instead of $). I’m not sure if the list is comprehensive; feel free to fix it if anything is missing or inaccurate:

  • Inline math delimiters $ ... $.

    (Neither \( ... \) nor \\( ... \\) work.)

  • Display math delimiters $$ ... $$ and \\[ ... \\].

    (The latter is an escaped version of LaTeX’s \[ ... \].)

  • LaTeX-style environments \begin{<name>} ... \end{<name>}.

    Here, <name> is any string with balanced braces { ... }; it must be the same in \begin and in \end, otherwise it is not recognized as a delimiter. (Improperly nested environments such as \begin{equation} ... \begin{align} ... \end{equation} are recognized as MathJax delimiters, but lead to MathJax error messages.)

    MathJax predefines common equation environments such as equation, align, or matrix; moreover, it supports user-defined environments using the \newenvironment command. The \begin{<name>} ... \end{<name>} structure is in any case recognized as a MathJax delimiter even if environment <name> does not exist, or if it is invalid as an environment name (e.g., if it includes unexpandable control sequences), however such cases lead to MathJax error messages.

    For single-letter environment names, \begin X and \end X are not recognized as MathJax delimiters, even though they work fine inside already-recognized MathJax expressions (just like in actual LaTeX).

    Unlike LaTeX, MathJax allows equation environments to be enclosed in other delimiters, such as $$\begin{align} ... \end{align}$$. The outer delimiters are redundant.

    Fun fact: both MathJax and LaTeX allow defining an environment with empty name with \newenvironment{}{X}{Y}. This redefines the \end command, wreaking havoc left and right; it appears that in LaTeX, \newenvironment{<name>} only checks for the existence of \<name>, but not for \end<name>, and happily redefines it if it exists. Sounds like a bug, if you ask me.

  • 1
    In LaTeX you can't do \newcommand{\end...}, so there's no need to check whether \end<name> exists in \newenvironment{<name>} as environments in their command form are always paired up. You don't, for example, have \something end with \endsomethingelse under \newenvironment{something}. It's always \something...\endsomething. Of course, in TeX you can do that.
    – Werner
    Feb 4 at 23:43
  • @Werner The point is that \end already exists in LaTeX, it does not need to be defined with \newcommand. And then, as I wrote in the answer, LaTeX accepts \newenvironment{}{...}{...} and erroneously overwrites \end, which completely breaks further environments. Feb 5 at 7:32
  • There is absolutely no value in doing \newenvironment{}{X}{Y}. You can also do \renewenvironment{document}{X}{Y} and screw things up equally. Some basic/fundamental understanding of a language is required when using (re)definitions.
    – Werner
    Feb 5 at 16:23
  • 3
    I don’t see much value in continuing this increasingly off-topic discussion, but anyway: you are making a strawman argument. A \renewenvironment is explicitly saying “I’m redefining an existing environment. I know what I’m doing. And even if I don’t, it’s my fault if this breaks anything.” In contrast, a \newenvironment pretends to check that the environment is safe to define, and does not conflict with any preexisting macros. It turns out that the check is half-hearted, and may fail to detect some conflicts; that’s a bug. Feb 5 at 17:11

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .