I am typesetting my phd thesis, and I was hoping to use Palatino as main font. For some of the headings I wanted to use a fancy TTF font, so I switched from pdfLaTeX to LuaLaTeX to compile my document. Using some trickery it seems I can still use the good old fonts from mathpazo, which look nice [*]. It seems however a good idea to avoid trickery and just use Tex Gyre Pagella, which is supposedly similar, and to include fonts properly using fontspec. However, with TG Pagella the text looks significantly different/worse. See the screenshot below:

enter image description here

  • right: the version with mathpazo,
  • top left: Tex Gyre Pagella,
  • bottom left: a version with mathpazo + tgpagella.

The version of the file typeset with Tex Gyre Pagella looks much darker than the other two, which I don't really like. It seems as if the letter spacing has changed. This seems quite a change for just 'a slightly different version of' a font.

Is there a way to avoid this ``darkening''? That is, get the overall text to appear lighter with Tex Gyre Pagella, so that it looks similar to the result of using just mathpazo or tgpagella?

[*] Of course one could say: if it works why not stick with it: of course I'm now trying to change more stuff, for which I need more of the LuaLaTeX features. This brakes this setup.

The latex source used to produce the versions above:


\usepackage{fontspec}        % used only for top left
\usepackage{unicode-math}    % used only for top left

% \usepackage[osf,sc]{mathpazo}  % used for right + bottom left
% \usepackage{tgpagella}         % used only for bottom left
% \usepackage[T1]{fontenc}       % for bottom left and right 

\setmainfont{TeX Gyre Pagella} 
\setmathfont{TG Pagella Math}


\section{General Approach}

We study the problem of segmenting a trajectory with respect to a criterion.
In the continuous case, this problem can be defined as follows.  We define a
trajectory $T$ as a function from the interval $I=[0,1]$ to $\R^2$ (or $\R^d$)
and a subtrajectory, also called \emph{segment}, $T[a,b]$ as the function
restricted to the subinterval $[a,b] \subseteq I$.
A criterion $C$ is a function $C \colon I \times I \rightarrow
\{\textsc{True}, \textsc{False}\}$, which is defined on all possible segments of $T$.
We say an interval $[a,b] \subseteq I$ \emph {satisfies} a criterion $C$ if
$C(a,b) = \textsc{True}$; in this case we call the segment \emph{valid}.  A
partitioning of $I$ (or of $T$) into non-overlapping segments
whose union covers $I$ is called a \emph{segmentation}.
A segmentation of \emph{size} $k$ can be denoted by its segments
$[\tau_0,\tau_1],[\tau_1,\tau_2],\ldots,[\tau_{k-1},\tau_k]$; $\tau_0 = 0$
and $\tau_k = 1$.
A segmentation is \emph{valid} if and only if all of its segments are valid,
and \emph{segmenting} a function refers to partitioning into valid segments.
We say a valid segmentation is \emph{minimal} (optimal)
with respect to $C$ if its size is minimum;
the segmentation problem is to compute a valid minimal segmentation.
We will often omit the word ``valid'' because only
valid segmentations are useful.
  • 1
    That will probably be a problem of the pdf viewer. – Manuel Dec 4 '14 at 17:27
  • Welcome to tex.sx! – T. Verron Dec 4 '14 at 17:33
  • @Manuel hmm you may be right. On a a different system with a different both files look equally dense. The exact spacing is a bit different though, but I'm willing to attribute that to the font variation. So I guess it is indeed a problem with Evince. Strange though, because the problem occurs at all zoom levels, also just at 100%. – Noinia Dec 4 '14 at 19:31
  • 1
    @Noina Print it! And you should use csquote \enquote-command for your quotations. – MaxNoe Dec 4 '14 at 21:36
  • @MaxNoe Only if you like unreadable code... I'm not convinced there is a 'should' here. It might be convenient but better set active quotes and keep things readable. – cfr Dec 5 '14 at 3:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.