What type in c# is most like an expression in R and why?

Technically, a formula object is an unevaluated function call, along with an associated environment:

f <- as.formula(y ~ x1 + x2)
> str(f)
Class 'formula'  language y ~ x1 + x2
  ..- attr(*, ".Environment")=<environment: R_GlobalEnv> 
> is.call(f)
[1] TRUE

Thinking of the y ~ x1 + x2 part as an unevaluated function call is a bit misleading, because we wouldn't typically really use it like we would other functions. But you can see it really is stored that way:

> f[[1]]
`~`
> f[[2]]
y
> f[[3]]
x1 + x2

i.e. it's a quoted string and two arguments, matching the description of a call from ?call. Of course, we wouldn't really think of ~ as a function in R, but it is one if you check help("~") you'll see it's a primitive. Using an unevaluated call in this way is one of R's more unique features, in that it is leveraging this ability to compute on the language itself to do something nominally unrelated: create a new "symbolic" object to store the conceptual definition of a model.

You can see how the formula object is actually used in this sense in model.frame.default, although that code is....fairly dense itself. The basic idea is that R is piggybacking on this ability to create functions with associated environments to allow the formula to be "evaluated" in the context of that environment, like a function would. But instead of "calling a function" it's more a vehicle for transporting model information that is then used to build the matrices you need for the model fitting process.