I'd like to start a discussion with the latex community about the lack of a Helvetica-like font package that has full math support. From the research I've done, such a package does not exist, but many people would benefit from it. There are many questions about this in the main site (I've linked a few below), but I want to ask the more "meta" question about whether a different solution could or should be created. Here are the relevant points that I've thought of.

Why I think it does not exist yet

Helvetica makes a poor choice for setting the main text of a scientific document. In paragraph form, it can be hard to read, and the fact that some characters are indistinguishable (capital I (i) and lower case l (L) makes it dangerous for math, where meaning is carried by individual characters rather than words.

Why I still think the package should exist (and why it's important)

As the main text of a scientific document, Helvetica fails. But as the font choice for graphics in scientific papers, it is the de facto standard! Some journals (Nature and Science) prefer all the text in figures be set in Helvetica or a close relative. For graphics, readability is not as crucial because text is limited to single words or phrases. Also, any ambiguity introduced by identical characters in the figures should be obvious to those who read the the figure captions or main text. What seems to matter most for figure text is having the right style, and the neo-grotesque fonts like Helvetica have become the standard for that neutral, familiar, scientific look.

Scientists use latex as a tool to produce good-looking documents without having to be specialists in typesetting or design. However, while scientists rarely have to typeset documents, they are often required to typeset figures. Figures are both the only parts of published articles that scientists are expected to typeset themselves and the parts for which latex does not provide an adequate solution.

To see the state of scientific graphics in 2017, browse a well-respected journal, like PRX (it's open-access for y'all), and notice how figures mix serif math with sans-serif text, and even randomly use different sans-serif fonts in the same figure as if they are identical (these mixed solutions are even recommended by tex.stackexchange in the links below). I believe TeX can help solve these problem!

What options exist for typesetting graphics (and why they are inadequate)

First off, my experience in creating figures for scientific purposes is mostly from using python's matplotlib library, which allows you to use latex to typeset all or some of the text. When choosing fontsets, the easiest to work with are those that come in a latex package that includes math support. Most of such packages are for serifed fonts, but for the most part, people avoid serifed fonts in figures. Usually the body text of a paper is set in a serif font, and unless you match it exactly in typeface and size, you can create some typographical dissonance. It is much more common to use a san-serif font so that contrast with the main text is explicit and so that the smaller sizes needed in the figures can still be rendered and easily read.

This leads us to search for sans-serif font latex packages with math. One option is the package Arev, which provides the font DejaVu Sans with math support (this is the default choice in matplotlib 2.0). Another option is Computer Modern Bright. Browsing the latex font catalog, those two are really the only options for consistent sans-serif math. Other sans options on the list are either Frankensteins (one font for text, another for math, like FiraSans with newtxsf, which uses Stix Sans for math) or are just not suited for scientific figures (in my opinion) due to their unique and potentially unprofessional looks (iwona, kurier, ...).

So are Arev and CMBright good enough for professional-looking scientific graphics? I think they could be, but they do bring their own unique style to the figure, and many scientists just want their figure text to look good, standard, and unremarkable. DejaVu (through Arev) and Computer Modern bright are good fonts, but they are quite different from Helvetica (see here and here). To my eye, both fontsets give a more humanist, optimistic feel than the more neutral Helvetica. While either choice may be better than Helvetica for typesetting a full paper, for the figures of a paper using a Times-like serif font, I think Helvetica is the more appropriate choice.

There are ways to get fake math support for Helvetica using packages like sansmathfonts, mathastext, and sfmath. In my experience, using these packages can work, but it is a slippery slope to the world of endless hacking and tweaking, making it difficult to get a truly professional-looking result. What we need is something as easy as \usepackage{helvetica-with-math} that includes a Helvetica-looking version of all the greek letters in italics and upright, as well all the mathematical symbols that might be reasonably needed.

Searching around leads to many threads in both the latex and matplotlib communities of people looking for a solution to this problem. Here are just a few I came across:


I'm a latex and python user trying to make figures for my scientific papers. I've realized how hard it is to make what I thought should have been the simplest choice of using a Helvetica-like font for both text and math. It seems to me that a quality Helvetica clone package with full math support would solve most of these problems, and my feeling is that such a package could quickly become the standard package used in scientific plotting engines.

I see two good future scenarios. 1) The e-Gust foundry (creators of the Tex Gyre Heros font) develops math support for Heros and releases it as a latex package. 2) The Stix project, which solved a similar problem by giving us a Times-like font with full math support, produce a Helvetica-like font. The Stix project was funded by a group of publishers who use the fonts, and it would be great for them and us if they released a font package for figures. The figures in their journals would improve, and the scientific community would finally have an adequate answer to this recurring question.

I'm curious to see if anyone agrees with me. Is this a real need? Is anyone working on it? Am I missing a simple solution? How do we commission the package I'm suggesting? What challenges am I unaware of?

| |
  • 1
    I would say Arev and CM are well-suited for all needed purposes if you are fixed to pdflatex. For one who really needs a font like this I would recommend to find a sans font which contains all necessary symbols and use that with xelatex and mathspec. – TeXnician Apr 29 '17 at 16:27
  • 3
    I think Helvetica is not a well-suited font for scientific text with lots of symbols and with math it is even worse. It works best and it is used so for well defined and carefully crafted block paragraphs. Sans-serif fonts are in general do not fit the bill in any context. There are too many symbols which are very difficult to pull off such as capital gamma and Pi and so on. – percusse Apr 29 '17 at 17:11
  • The xelatex with mathspec solution is not the best because the symbols need to come from somewhere else. This leads to the kind of typographical dissonance that should drive any typesetting enthusiast crazy. – Gabe Apr 29 '17 at 20:08
  • 2
    Meta is about questions concerning TeX.Stackexchange. This question is off-topic. – Johannes_B Apr 30 '17 at 5:10
  • A font is designed by someone. A font contains the latin alphabet, a few numbers, a few puctuation marks. Some fonts contain cyrillic letters, or greek letters. Some contain more letters. Only few contain maths symbols. – Johannes_B Apr 30 '17 at 5:12
  • Is there any scientific study showing that sans serif figures/diagrams are standard? Every time i see one, the only reason for doing so is the author (supposedly) not knowing how to properly typeset the diagram. – Johannes_B Apr 30 '17 at 5:14
  • As it happens, the anual TUG meeting is taking place alongside BachoTeX. I bet a few GUST members will be there as well. Maybe TUG can spend a few thousands of dollars to finance the project. (Making a font takes time, it is costly) – Johannes_B Apr 30 '17 at 5:20
  • I'm afraid this is off topic both here (not about TeX-sx) and on the main site (not a question with an answer likely to need a couple of paragraphs) but a discussion point. – Joseph Wright Apr 30 '17 at 5:44
  • 5
    That said: this is not a TeX issue, it's a font/design one. Whilst some scientific publications nowadays use sanserif for figures, there is a long tradition that really such fonts should match the document body (look at most journals before authors rather than design professionals made graphics, say pre-1980). Moreover, 'serious' mathematicians use serif/sans serif to impart meaning (not design) and so are very unlikely to push for a publication-quality all-sans serif font. – Joseph Wright Apr 30 '17 at 5:47
  • I don't know, I'm browsing old Physical Review, Physical Review Letters, and Science Magazine papers from the 50s and 60s, and I can't find examples of them using a serif font for the figures. Most use something like this font. – Gabe Apr 30 '17 at 17:34
  • 2
    That's probably old technical drawing standard. That's due to using stencils in the absence of a better tool and serifs are difficult to put the pen inside them. They are drawn manually that's why you might think old stuff uses sans serif fonts. As the name implies they are old don't get stuck to them same goes for math notation. Old style is often overpompous. We know better by now. – percusse Apr 30 '17 at 20:40
  • Ah, good points. – Gabe Apr 30 '17 at 21:14

Browse other questions tagged .