There are many annoying trade-offs that are faced when making figures for scientific publication, and these are only exacerbated by the myriad of conflicting and confusing journal guidelines. For example, a journal may only allow four or five figures, a restriction that inevitably leads to the use of multi-panel compound figures in order to try to squeeze the maximum amount of information into as smaller space as possible (sometimes as much as three or four figures worth of material). Unsurprisingly, this compaction of information can lead to significant problems with figure clarity and legibility.

An obvious solution is the use of vector graphics. Here the author can strive to make the figure as clear as possible at the print scale, safe in the knowledge that anyone reading the digital copy (and to be honest, who isn't doing this in 2013) can scale up the image to their heart's content without any loss in quality. I cannot emphasize enough just how useful vector graphics are for my own work. When I need to convey detailed information about the spatial organization of atoms or molecules in my model systems, scalability allows me to provide information regarding both the small and large length scale behavior of the system in the same image, both of which may be fundamental to the underlying physics. For example, while the entire image gives a sense of the structure of atoms and molecules at the system scale, zooming in allows the reader to discern local particle arrangements or bonding motifs. Having access to full information at all scales is truly invaluable.

One problem with such a detailed graphical representation is that file sizes can quickly become rather large. A vector representation of a molecular system requires coordinates for all of the atoms in the system (which can be on the order of hundreds to thousands) as well as instructions for drawing the primitives out of which they are built. Journals typically place restrictions on figure file sizes and will either reject your image if it is too large, or silently rasterize it leading to a marked reduction in quality. Now we get to the crux of the problem: how do you produce a beautiful, scalable, vector image of a molecular system while keeping the file size to a minimum? Enter PostScript.

Despite having used PostScript for many years, in my ignorance I wasn't fully aware of what it was. It was only when working with a collaborator that I realized that it wasn't simply a vector image format, but rather a fully fledged programming language (I guess the clue was partially in the name, alas). A PostScript image is simply a plain text file containing a list of function definitions and instructions to be processed by a PostScript interpreter, such as Ghostscript. This was a real eye opener for me. Since PostScript allowed me to define my own functions for all of the primitive objects in an image, I was effectively a granted a level of control over compression through judicious choice of function names and coordinate precision.

As an example, consider this simple PostScript function for drawing a filled circle:

/c {
0 0 3.0 0 360 arc
} def

Don't worry too much about the syntax, the point is that we now have a function to draw a filled circle of any color that can be addressed by a single letter, c. This definition comes at the top of our PostScript image file. If we wanted to draw a red circle centered at (10,10) we would add the following instruction to the PostScript file.

10 10 tr 1.0 0.0 0.0 c

The first part of the instruction translates the origin of the coordinate system to (10,10), the second part calls the circle function passing the RGB color values as operands. In this way, an entire image can be generated using a sequence of commands for drawing the shapes of each molecule in the system. While extreme precision might be needed to preserve numerical stability when saving particle configurations to restart files, for visualization purposes it should be perfectly acceptable to output atomic coordinates to two decimal places. This can dramatically reduce the size of the PostScript file without any noticeable difference to the output. If the vector primitives only require fixed colors then this can be absorbed into the function itself, e.g. we would be able to draw a circle by just writing c.

The following is an example of the type of beautiful image that is easy to generate with PostScript. It shows a configuration from a statistical-mechanical model of a grana stack, which is part of the photosynthetic light-harvesting system. Obviously, for inclusion on the web, this is a poor quality PNG raster. A PostScript version can be found here. Note that the vector version is around seven times smaller than the PNG raster!

A statistical-mechanical model of a grana layer

In summary: PostScript is fun, and is a fantastic tool for producing modestly sized, publication quality figures. Since PostScript images are plain text files they can be trivially generated from any program with no need for any external libraries or additional dependencies. Moreover, Adobe (the developer) has several excellent resources for learning the language that are freely available on its website and elsewhere. These are commonly referred to as the red and blue books. Another excellent book is Learning PostScript: A Visual Approach, by Ross Smith. The book is out of print but copies are easy to find on Amazon. Links below.

The PostScript Red book The PostScript Blue book Learning PostScript: A Visual Approach