Latin Modern Math in LaTeX

Question

It seems that Latin Modern Math font works fine only with XeTeX and LuaTeX. Is it possible to use Latin Modern Math font with LaTeX and pdfLaTeX?

---

The publisher requires that I use LaTeX/pdfLaTeX, so using XeTeX or LuaTeX is not an option. I cannot use strange packages, I have to use minimal additional packages. The article's font is Latin Modern and the encoding of the file is set to uft8 and I am not permitted to change these.

I want to use the \lBrack symbol in Latin Modern Math font. The font is defined according to Unicode Math Symbols in Latin Modern Math. However, when I use \lBrack pdfLaTeX tells me that \lBrack is not defined. The issue is not the command \lBrack itself. I don't mind defining it. I just want to have the symbol for \lBrack (Unicode 027E6) from the Latin Modern Math font appear in my output, i.e. the PDF file generated by pdfLaTeX.

I don't want to use other packages that provide similar symbols, it is not an option.

If I were using XeLaTeX I could have written something like:

\documentclass{article}

\usepackage[T1]{fontenc}
\usepackage{unicode-math}
\pagestyle{empty}

\begin{document}

$\lBrack$

\end{document}

But as I wrote I am not allowed to use XeLaTeX and have to use pdfLaTeX. If I run pdfLaTeX on the same code I get an error because unicode-math requires XeLaTeX or LuaTex engine.

Essentially, I want to has some command without add anymore packages which does what the code above does in XeLaTeX.

I have tried to directly access the charter but it doesn't seem to work. For example, I tried:

\documentclass{article}

\usepackage[T1]{fontenc}
\usepackage[utf8]{inputenc}
\usepackage{lmodern}
\usepackage{amssymb}
\pagestyle{empty}

\begin{document}

$\mathsf{\mathbf{}}$ 

${\mathchar"027E8}$

\end{document}

This does compile with pdfLaTeX but it doesn't work correctly: it produces another character, not \lBrack.

ps: I don't have a clue why I need to put \mathsf{\mathbf{}} there, neither I understand why I need the amssymb package, but without these it doesn't compile. It gives an error

\textfont 7 is undefined (character è)

Answer 1

First a side remark: With xelatex you should never load inputenc (you did it in your example).

Beside this: Theoretically it is possible to use symbols from an open type font in pdftex. But you need to create a tfm-file, you need to create enc-file(s) and a map-entry. And as pdftex can't subset open type you probably also need to split the open type and create various type1 fonts. This means a lot of work - and imho all this new files will count as "strange packages" and so your publisher will not like it. If you really don't want to use any other package you could perhaps cheat like this:

\documentclass{article}
\usepackage{lmodern}
\begin{document}
$\lbrack\!\lbrack x \rbrack\!\rbrack$
\end{document}

Answer 2

Here a solution (and solution sketch) for a fixed font size (10pt in the example):

1. A LuaTeX document prints the one character on a page with known margins around the character box. The example uses 10bp. As unit bp is used, because it is the natural unit for PostScript and PDF. That makes calculations easier.

   % lbrack.tex
   \mag=10000 % enlarge by factor 10
   \documentclass{article}
   \usepackage{unicode-math}
   \setmathfont{latinmodern-math.otf}
   \pagestyle{empty}
   
   \begin{document}
   \sbox0{$\lBrack$}
   % print the TeX dimensions of the character glyph
   \typeout{* wd = \the\wd0}
   \typeout{* ht = \the\ht0}
   \typeout{* dp = \the\dp0}
   % the rule in the margin marks the base line
   \sbox0{\vrule width 10bp height .1bp depth .1bp $\lBrack$}
   % add margin of 10bp to the other margins
   \wd0=\dimexpr\wd0+10bp\relax
   \ht0=\dimexpr\ht0+10bp\relax
   \dp0=\dimexpr\dp0+10bp\relax
   % Output the box on a PDF page with the same dimensions.
   \pdfpagewidth=\wd0
   \pdfpageheight=\dimexpr\ht0+\dp0\relax
   \pdfhorigin=0pt
   \pdfvorigin=0pt
   \shipout\copy0
   \end{document}
   

   The command to generate the PDF file:

   $ lualatex lbrack
   

2. Then ghostscript's pswrite device writes the PostScript description of the character glyph in a PostScript file. However, the size must be large enough, otherwise pswrite uses an bitmap inline graphics. Therefore I have used \mag=10000 in the TeX document in the step before. This enlarges the character by a factor 10.

   The command to generate the PostScript file:

   $ gs -sDEVICE=pswrite -sOutputFile=lbrack.ps -dBATCH -dNOPAUSE lbrack.pdf
   

   The page description of lbrack.ps:

   %%Page: 1 1
   %%PageBoundingBox: 0 0 241 300
   %%BeginPageSetup
   GS_pswrite_2_0_1001 begin
   241 300 null setpagesize
   /pagesave save store 197 dict begin
   0.1 0.1 scale
   %%EndPageSetup
   gsave mark
   Q q
   0 0 2414 0 0 2996 ^ Y
   20 w
   K
   0 1249.1 1000 0 S
   0.996264 i
   1386.55 1020 m
   0 11 -8.98 20 -19.93 20 c
   -84.62 0 0 916 ^ p
   10.95 0 19.93 9 19.93 20 c
   0 11 -8.99 20 -19.97 20 c
   -249.61 0 0 -996.02 ^ p
   10.98 0 19.97 9 19.97 20 c
   h
   1242 1956 0 -916 -85.03 0 ^ H
   f
   cleartomark end end pagesave restore
    showpage
   %%PageTrailer
   

3. This description can be transferred to PDF page description operators. However the short operators above (m, ^, p, c, ...) are user defined operators that are defined in the preamble. For example, m expands to moveto and c to rcurveto. For moveto there is a direct PDF operator m, but the PDF operator for cubic Bézier curves is c that means curveto in PostScript. Thus the relative coordinates of rcurveto needs to be converted to absolute coordinates for curveto. Example, some recalculations are shown in the comments:

   0.996264 i
   1386.55 1020 m
   % 0 11 -8.98 20 -19.93 20 c
   1386.55 1031 1377.57 1040 1366.62 1040 curveto
   % -84.62 0 0 916 ^ pstack p
   % -84.62 0 rlineto 0 916 rlineto 84.62 0 rlineto
   1282 1040 lineto 1282 1956 lineto 1366.62 1956 lineto
   % current point: 1366.62 1956
   % 10.95 0 19.93 9 19.93 20 c
   1377.57 1956 1386.55 1965 1386.55 1976 curveto
   % current point: 1386.55 1976
   % 0 11 -8.99 20 -19.97 20 c
   1386.55 1987 1387.56 1996 1366.58 1996 curveto
   % -249.61 0 0 -996.02 ^ p
   % -249.61 0 rlineto 0 -996.02 rlineto 249.61 0 rlineto
   1116.97 1996 lineto 1116.97 999.98 lineto 1366.58 999.98 lineto
   currentpoint pstack pop pop
   % current point: 1366.58 999.98
   % 10.98 0 19.97 9 19.97 20 c
   1377.56 999.98 1386.55 1008.95 1386.55 1019.98 curveto
   h
   % 1242 1956 0 -916 -85.03 0 ^ H
   % 1242 1956 moveto 0 -916 rlineto -85.03 0 rlineto 0 916 rlineto
   1242 1956 moveto 1242 1040 lineto 1156.97 1040 lineto 1156.97 1956 line
   f
   

   Also the PostScript coordinates are scaled with factor 10 by pswrite (see 0.1 0.1 scale in the page setup section) and we had scaled it by factor 10 via \mag=10000. Thus the absolute coordinates needs a correction of 1000 at the left and bottom side. The result in PDF page operations can be seen in the TeX file in the next step.

4. The TeX file puts the PDF page operations in \pdfliteral in a \hbox. The console output or the .log file of step 1 tells us the dimensions of the box. The shift for the baseline that is used for the cm operator can be seen from the vertical coordinate of the \vrule in the PostScript output.

   Then the box is put into a PDF XForm Object. Thus the PDF page operations are stored in one place only and just the XForm Object is reused (optimization, if the glyph is used several times).

   Finally, \lBrack and \rBrack are defined (if they are unknown). Because of laziness \rBrack is just the mirrored version of \lBrack. IMHO that should be close enough.

   % test.tex
   \documentclass{article}
   
   \makeatletter
   \newcommand*{\@lrBrack}[3]{%
     \setbox0=\hbox to 4.16pt{%
       #1%
       \pdfliteral{%
         q % save   
         #2.01 0 0 .01 0 -2.491 cm % set current transfer matrix
         .996264 i % flatness tolerance
         386.55 20 m % moveto
         386.55 31 377.57 40 366.62 40 c % curveto
         282 40 l % lineto
         282 956 l % lineto
         366.62 956 l % lineto
         377.57 956 386.55 965 386.55 976 c % curveto
         386.55 987 387.56 996 366.58 996 c % curveto
         116.97 996 l % lineto
         116.97 -.02 l % lineto
         366.58 -.02 l % lineto
         377.56 -.02 386.55 8.95 386.55 19.98 c % curveto
         h % closepath
         242 956 m % moveto
         242 40 l % lineto 
         156.97 040 l % lineto
         156.97 956 l % lineto
         f % fill
         Q% restore 
       }%
       #3%   
     }%
     \ht0=7.5pt %
     \dp0=2.5pt %
     \pdfxform0 %
   }
   \@ifundefined{lBrack}{%
     \@lrBrack{}{}{\hfill}%
     \edef\lBrack{\mathopen{\pdfrefxform\the\pdflastxform}}%
   }{}
   \@ifundefined{rBrack}{%
     \@lrBrack{\hfill}{-}{}%
     \edef\rBrack{\mathclose{\pdfrefxform\the\pdflastxform}}%
   }
   \makeatother
   
   \begin{document}
     \[ \lBrack x \rBrack \]
   \end{document}
   

   This file can now be compiled with pdflatex:

   $ pdflatex test
   

5. For different math styles (\scriptstyle, \scriptscript style), the procedure needs to be repeated. The symbol can then be defined via \mathchoice. However, \left and \right are not supported.

Result:

Answer 3

Maybe I better understand what you're talking about. So, I'll give it yet another stab.

You talk about symbol 027E6 which is a unicode font character. You can gain access to this character through the package unicode-math. But then you'll have to use some flavor of LaTeX other than pdfLaTeX: for example, according to the package, xetex or luatex.

Otherwise you're going to have to do a bit of hacking. If, for example, you're directly entering unicode characters into your document, then you'll have to do some kind of work around as follows:

\usepackage{stmaryrd}
\usepackage[T1]{fontenc}
\usepackage[utf8x]{inputenc}
\DeclareUnicodeCharacter{"027E6}{\llbracket}

which is basically telling pdfLaTeX that you're using a different encoding for your characters/glyphs in your source file, and that when it sees unicode character 027E6 it should treat this glyph as though it were the command \llbracket provided by stmaryrd.

At this point I'm assuming you're working with a MWE along the following lines:

\documentclass{article}

\usepackage{stmaryrd}
\usepackage[T1]{fontenc}
\usepackage[utf8x]{inputenc}
\DeclareUnicodeCharacter{"027E6}{\llbracket}

\pagestyle{empty}
\begin{document}

$⟦$

\end{document}

where I've directly entered the unicode character in question.

I should point out that this really has nothing to do with Latin Modern fonts. This is all about the use of unicode characters in your document. LaTeX (pdfLaTeX) needs a way to interpret what to do with the character.

Short of this work around, you're going to have to work with something like xelatex which better handles non-ascii characters. The package unicode-math does this to some extent, but not in a manner accessible for pdfLaTeX.

For example, the following compiles quite fine under xelatex

\documentclass{article}
\usepackage{unicode-math}
\usepackage{lipsum}
\pagestyle{empty}
\begin{document}

$⟦$

\end{document}

But it compiles because unicode-math already told xelatex how to handle the glyph ⟦.

You might want to look at Entering unicode math symbols into LaTeX, direct from keyboard, on a Mac to see more about how to implement this.

If this is still not what you want, I can not stress enough the importance of supplying a MWE that reflects what you're doing on your end. It helps us better understand what you're trying to do; otherwise, it can be like shooting in the dark --- particularly if we're not understanding each other's lingo in the same way.

Answer 4

\documentclass{article}

\usepackage[T1]{fontenc}
\usepackage{lmodern}
\usepackage{amssymb}
\usepackage{fourier}

\usepackage[utf8]{inputenc}
\usepackage{newunicodechar}

\newunicodechar{⟦}{\llbracket}

\pagestyle{empty}
\begin{document}
\Huge
$⟦ \left⟦f\right]$

\end{document}