To my knowledge, no, and what you have there will be about as efficient as you can get (you have to construct a tuple in order to return the key/value pair, so the overhead is essentially a single cons cell for the list).

I’d be curious to know a bit more about the use-case, though. I’m having trouble thinking of a case where it makes sense to use a map to contain a single key/value pair.

Got it!

You’re losing constant-time lookup that way, though, as you now have to scan the map to match the first character. It’ll probably be negligible for small trees but could be more significant for trees with many character prefixes.

You might consider a structure like this:

%{
  "a" => {["p", "p", "l", "e"], %{value: ...}},
  "b" => {
    ["a", "n"], 
    %{
      "a" => {["n", "a"], %{value: ...}},
      "j" => {["o"], %{value: ...}}
    }
  }
}