The method you note in your question works in the sense that it does a global reset on the font used in math mode. The documentation for the fontspec and unicode-math packages provides examples of how one may restrict the scope of a math font to various subsets of math-font applications. Section 4.1 (pp. 6f.) of the unicode-math manual starts off by noting that "[t]here will probably be few cases where a single Unicode maths font suffices. ... It will therefore be necessary to delegate specific Unicode ranges of glyphs to separate fonts". It then gives examples for selecting a math font for
- specific glyphs such as
\int, \sum, and \prod
- a collection of symbols with the same math type, such as
\mathopen
- complete math styles such as
\mathbb
- comma-separated lists of Unicode slots and ranges
Section 3.1 of the fontspec manual states that "By default, fontspec adjusts LaTeX's default maths setup in order to maintain the correct Computer Modern symbols when the roman font changes. However, it will attempt to avoid doing this if another maths font package is loaded (such as mathpazo or the unicode-math package). If you find that fontspec is incorrectly changing the maths font when it should be leaving well enough alone, apply the [no-math] package option to manually suppress its maths font." [highlighting via backticks added]
rangefacility ofunicode-mathmight help). – Alan Munn Sep 25 '11 at 19:20