> library(ggplot2)
> ggplot(iris,aes(Sepal.Width)) + geom_density() + xlim(1,3.5)

Using xlim or ylim the plot is not cutted, ggplot subsets the data before calling the stat function (stat_density in this case). You can see it in the warning message.

Warning message:
Removed 19 rows containing non-finite values (stat_density).

ggplot(iris[iris$Species == "setosa",],aes(Sepal.Width)) + 
    geom_density()

Here, we are subsetting the dataframe before passing it to ggplot. It is a very useful tool derived from the data frame data structure.

To make the code more readable, one can also use dplyr's filter:

library(dplyr)
iris %>% filter(Species == "setosa") %>% ggplot(aes(Sepal.Width)) + 
    geom_density()