# Double-bar notation for second-order tensors

I want to use the double-bar notation for second-order tensors, which is common in continuum mechanics (e.g. for the strain and stress tensors).

I've searched the Comprehensive LATEX Symbol List but failed to find anything conclusive in there.

I've found this discussion, in which Stefan and Thorsten propose the alternatives below.

Is there a preferred way of typesetting this accent?

Subsidiary question: what is the difference between \bar and \Bar?

\documentclass{article}
\usepackage{amsmath}
\begin{document}
$\bar{\bar\tau} \quad \Bar{\Bar\tau} \quad \overline{\overline\tau}$
\end{document}


-
\show\Bar gives \bar so no difference (at least in amsmath) – percusse Apr 6 '13 at 15:08
amsmath.sty has \def\Bar{\bar}, so (as percusse said) there's no difference (in amsmath). – Gonzalo Medina Apr 6 '13 at 15:12
In a previous version of amsmath, in order to get double accents one had to use the uppercase variant; this is no longer necessary from amsmath version 2. The old commands have been retained for backwards compatibility. – egreg Apr 6 '13 at 16:53

egreg's comment comforted me in the thought that both \bar{\bar{...}} and \overline{\overline{...}} can be used to typeset double accents without any nasty side effects.

However, the best way to replicate the double-bar notation for rank-2 tensors, as found there, for instance, seems to be using \overline{\overline{...}}.

\overline{\overline{...}} should be preferred to \bar{\bar{...}} because the \overline produces a vinculum of the same width as its argument, whereas \bar only produces a vinculum of fixed width.

\documentclass{article}
\usepackage{amsmath}
\newcommand*{\rttensor}[1]{\overline{\overline{#1}}}
\newcommand*{\rttensortwo}[1]{\bar{\bar{#1}}}
\begin{document}
\noindent
$\rttensor{\epsilon} \quad \rttensor{\sigma} \quad \rttensor{G}$\\[1em]
$\rttensortwo{\epsilon} \quad \rttensortwo{\sigma} \quad \rttensortwo{G}$
\end{document}


-
Aesthetically, however, \bar\bar is more pleasing than \overline\overline. – JohnRos Oct 1 '15 at 12:50
@JohnRos Not to me, especially if the tensor name is long. – Jubobs Oct 1 '15 at 12:52