Accents & special letters

Type \'e and you get é; type \"o and you get ö. LaTeX’s accent commands — acute, umlaut, tilde, háček — have names that are one character long, and that character is a punctuation mark. Nothing else in LaTeX is named this way. It is a design from a time when no keyboard produced é directly, and underneath it sits the \accent primitive Knuth wrote into plain TeX. This page is a lookup for what each command actually prints: the sixteen text accent commands, the letters that are not accents (\ss, \ae, \oe, \aa, \l), the dotless \i and \j, and the math accents \hat, \bar, \vec and \dot. It also covers what a table cannot show — that choosing T1 in fontenc changes not only which marks you can set but whether an accented word can be hyphenated at all, measured here rather than asserted.

The sixteen accent commands and what each one prints

There are sixteen accents in text mode, and every one of them takes a single letter as its argument. \'o and \'{o} produce the same character; the braces are optional. The eleven marks that sit above the letter — acute, grave, circumflex, dieresis, tilde, macron, dot, breve, caron, double acute, ring — come straight from Knuth’s plain TeX and are built with the \accent primitive. The ones that hang below (\c cedilla, \d dot below, \b bar below), the one that spans two letters (\t tie) and \k (ogonek) work differently.

% braces are optional for a single letter, required for a command
r\'esum\'e, na\"ive, se\~nor, G\"odel, \c{c}a va, \H{o}r\'an
% two-letter argument: the tie accent
t\t{oo}k

Command \k unavailable in encoding OT1: what [T1]{fontenc} changes

The error means the accent you asked for does not exist in the font encoding currently in force. pdfLaTeX still defaults to OT1, an encoding from the 1980s, and \k (ogonek) is not in it. The fix is one line in the preamble: \usepackage[T1]{fontenc}. Other commands fail the same way — compiled on TeX Live 2024, \dj and \DJ (Croatian đ), \th and \TH (Icelandic þ), \dh and \DH (ð), and \ng and \NG (ŋ) all stop with the identical message.

\documentclass{article}
\usepackage[T1]{fontenc}   % without this, \k \dj \th \dh \ng all error
\begin{document}
Polish: j\k{e}zyk. \quad Croatian: \dj\ \quad Icelandic: \th\ \dh
\end{document}

The real cost, though, is not the commands that error but the ones that do not. OT1 has no single é, so LaTeX assembles one with \accent: put an acute over an e. But to TeX, \accent is a boundary at which the ligature-and-kern chain stops. Measured: in OT1 AV is 13.8889 pt while \'AV (an Á followed by V) is 15.00003 pt — the whole 1.11113 pt kern between A and V has vanished. Under T1, where Á is a single glyph, AV and \'AV both come out at 13.8855 pt.

Hyphenation makes it starker still. Run \showhyphens{universit\'e r\'esum\'e} and OT1 answers uni-ver-sité résumé — résumé gets no break point at all. Compile the same input under T1 and it becomes uni-ver-sité ré-sumé. When a document full of accented words — French, Polish, Czech, or an English paper whose only accents are in names — keeps coming out with ragged, gappy lines, this is usually why. Put the other way round: for Western text under pdfLaTeX, \usepackage[T1]{fontenc} is not a refinement but a requirement.

\i and \j: the dotless letters are only needed for seven of the sixteen

The short answer: you almost never need to write \"{\i}. Reading ot1enc.def and t1enc.def shows that for five of the accents — acute, grave, circumflex, dieresis and dot above — LaTeX already swaps the i for a dotless one through \DeclareTextComposite / \DeclareTextCompositeCommand. Compiled on TeX Live 2024, \"i and \"{\i} both come out as the identical ï at 3.33252 pt. The seven where you must supply \i yourself are the rest: \~, \=, \u, \v, \H, \r and \k.

% the dot is removed automatically for these five
na\"ive \quad \'idem \quad \`ile \quad \^ile \quad \.i
% here you must ask for the dotless letter yourself
\v{\i} \quad \u{\i} \quad \={\i} \quad \~{\i} \quad \H{\j}

\i and \j are not parts of an accent — they are letters in their own right. They sit at slots 16 and 17 in OT1 and at 25 and 26 in T1, and \i on its own prints ı, a dotless i. In a language such as Turkish, where the dotless i is a separate letter, that is not decoration but spelling. In math mode the equivalents have different names and live in a different mode: \imath and \jmath.

\ss, \ae, \oe, \o, \aa, \l: letters, not accents

This group takes no argument, because ß, æ, ø and ł are not accented vowels but letters in their own right in the languages that use them. The capitals are formed by capitalising the name: \AE, \OE, \O, \AA, \L. \SS is the exception: on TeX Live 2024, under both OT1 and T1, it prints the two letters “SS”, not a capital ß (the kernel default in latex.ltx is literally \DeclareTextCommandDefault{\SS}{SS}). For the German capital eszett ẞ, the dependable route is a font that contains it, on a Unicode engine.

The handling of \l is the one difference in this section that reaches into the PDF itself. In OT1, \l is not a real ł but a composite drawn by laying a short stroke over an l, and running the resulting TeX Live 2024 PDF through pdftotext extracts a bare l. Under T1 the real ł sits at slot 170, so search and copy-paste behave. The same happens with the below-the-line \d and \b: \d o (ọ) comes out of the PDF as o.. For body prose, looking right is usually enough; for bibliographic data and searchable terms, it is not.

\hat{x} in text gives ! Missing $ inserted.: math accents are a separate set

The text accents and the math accents are separate in name and in implementation, and crossing the border fails in both directions — with different wording, which makes the message a useful diagnostic. Write \hat{x} in running text and you get ! Missing $ inserted.; write $\'e$ inside math and you get ! Please use \mathaccent for accents in math mode. (both compiled on TeX Live 2024). The first is TeX helpfully trying to open math mode for you and failing; the second names the problem outright — that accent belongs to text.

\usepackage{amsmath}   % \dddot, \ddddot
\[
  \dot{x}, \quad \ddot{x}, \quad \bar{x}, \quad \hat{p},
  \quad \tilde{a}, \quad \vec{v}, \quad \check{s}, \quad \mathring{r},
  \quad \hat{\imath}, \quad \vec{\jmath}
\]

Two practical cautions. First, accents do not stack cleanly: layering two, as in \hat{\vec{x}}, tends to leave the upper mark floating or off-centre. If you only want more dots, use \ddot / \dddot / \ddddot; if you genuinely want layers, build them explicitly with \overset, covered on the over/under page. Second, amsmath’s \dddot does not shrink inside a superscript. Measured, $x^{\ddot a}$ is 10.55292 pt while $x^{\dddot a}$ is 16.21565 pt — the three dots stay at text size up there. Loading the mathdots package fixes it.

Type \'e, or type a literal é?

If T1 is loaded, the output is identical either way. Measured, \'e, \'{e} and a literal é all come to 4.44336 pt and resolve to the same single glyph — and since 2018 LaTeX reads UTF-8 by default, so no inputenc line is needed. For a readable source file, typing the letter directly wins. The command notation survives mainly in three places: submission rules that demand ASCII, personal names in .bib files, and the tabbing environment. Inside tabbing, \' is taken over by the “move to a tab stop” command, so accents there use the dedicated \a' form — the same command with \a in front.

% inside tabbing, \' means "go to a tab stop" — use \a' for the accent
\begin{tabbing}
  caf\a'e \= au lait \\
  th\a'e   \> vert
\end{tabbing}

• Setting Western text with pdfLaTeX: always add \usepackage[T1]{fontenc}. Hyphenation, kerning and PDF searchability are all fixed by that one line
• Type accented letters directly in the body; match the venue’s rules in .bib. Do not mix the two notations inside one project
• You only need \i for seven of them: \~ \= \u \v \H \r \k. LaTeX handles the other five for you
• \SS is SS, not ẞ. For a capital eszett, move to a Unicode engine and a font that has it
• \hat is math, \^ is text. The error text tells you which side you are on: ! Missing $ inserted. or ! Please use \mathaccent …