The Physics work in this field is rigorous enough. Hawking and Ellis is a standard reference, and it is perfectly fine in terms of rigor.

### Digression on notation

If you have a tensor contraction of some sort of moderate complexity, for example:

$$ K_{rq} = F_{ij}^{kj} G_{prs}^i H^{sp}_{kq}$$

and you try to express it in an index-free notation, usually that means that you make some parenthesized expression which makes

$$ K = G(F,H)$$

Or maybe

$$ K = F(G,H) $$

Or something else. It is very easy to prove (rigorously) that there is no parentheses notation which reproduces tensor index contractions, because parentheses are parsed by a stack-language (context free grammar in Chomsky's classification) while indices cannot be parsed this way, because they include general graphs. The parentheses generate parse trees, and you always have exponentially many maximal trees inside any graph, so there is exponential redundancy in the notation.

This means that any attempt at an index free notation which uses parentheses, like mathematicians do, is bound to fail miserably: it will have exponentially many different expressions for the same tensor expression. In the mathematics literature, you often see tensor spaces defined in terms of maps, with many "natural isomorphisms" between different classes of maps. This reflects the awful match between functional notation and index notation.

### Diagrammatic Formalisms fix Exponential Growth

Because the parenthesized notation fails for tensors, and index contraction matches objects in pairs, there are many useful diagrammatic formalisms for tensorial objects. Diagrams represent contractions in a way that does not require a name for each index, because the diagram lines match up sockets to plugs with a line, without using a name.

For the Lorentz group and general relativity, Penrose introduced a diagrammatic index notation which is very useful. For the high spin representations of SU(2), and their Clebsch-Gordon and Wigner 6-j symbols, Penrose type diagrams are absolutely essential. Much of the recent literature on quantum groups and Jones polynomial, for example, is entirely dependent on Penrose notation for SU(2) indices, and sometimes SU(3).

Feynman diagrams are the most famous diagrammatic formalism, and these are also useful because the contraction structure of indices/propagators in a quantum field theory expression leads to exponential growth and non-obvious symmetries. Feynman diagrams took over from Schwinger style algebraic expressions because the algebraic expressions have the same exponential redundancy compared to the diagrams.

Within the field of theoretical biology, the same problem of exponential notation blow-up occurs. Protein interaction diagrams are exponentially redundant in Petri-net notation, or in terms of algebraic expressions. The diagrammatic notations introduced there solve the problem completely, and give a good match between the diagrammatic expression and the protein function in a model.

Within the field of semantics within philosophy (if there is anything left of it), the ideas of Frege also lead to an exponential growth of the same type. Frege considered a sentence as a composition of subject and predicate, and considered the predicate a function from the subject to meaning. The function is defined by attaching the predicate to the subject. So that "John is running" is thought of as the function "Is running"("John").

Then an adverb is a function from predicates to predicates, so "John is running quickly" means ("quickly"("Is running"))("John"), where the quickly acts on "is running" to make a new predicate, and this is applied to "John".

But now, what about adverb modifiers, like "very", as in "John is running very quickly"? You can represent these are functions from adverbs to adverbs, or as functions from predicates to predicates, depending on how you parenthesize:

(("very"("quickly"))("Is running"))("John")

vs.

(("very")(("quickly")("Is running"))("John")

Which of these two parenthetization is correct define two schools of semantic philosophy. There is endless debate on the proper Fregian representation of different parts of speech. The resolution, as always, is to identify the proper diagrammatic form, which removes the exponential ambiguity of parenthesized functional representation. The fact that philosophers have not done this in *100 years* of this type of debate on Fregian semantics shows that the field is not healthy.

## Best Answer

This may not be quite what you're after, but "Relativistic Hydrodynamics" by Rezzolla and Zanotti covers (relativistic) hydrodynamics in the language of differential geometry.

This is a graduate level textbook on hydrodynamics in the context of general relativity (hence the differential geometry). It covers kinetic theory (including quantum effects), wave propagation, numerical algorithms, and analytic work in the context of astrophysics. You should have both some relativity and hydro under your belt before you tackle it, but if you do you may find it very useful. It is also fairly recent (2013), so is very up to date.