Why does word-initial upsilon always have a rough breathing?

Question

How did a rough breathing develop before all words starting with an upsilon in Ancient Greek? This is a commonly noted fact about the distribution of these sounds (or rather spellings), but I’m having a hard time figuring out the etymological reason for it.

In some words, it is easily explicable as the regular outcome of a word-initial *s- in Proto-Indo-European: *súpnos > ὕπνος (húpnos)

But in other words, there is no *s reconstructed for the PIE form. Are there any historical linguistics hypotheses about the source of the rough breathing in these words (e.g. a prothetic *s- prefix that later weakened to a rough breathing, a shift [w] > [v] > [f] > [h], a sound change inserting /h/ before all word-inital /u/ at some point in Ancient Greek (that seems phonetically unmotivated to me, although I guess it's the most straightforward option), something like that).

Some example words (I’m taking these from Wiktionary):

• *wódr̥ > ὕδωρ (húdōr)
• *webʰ- > ὑφαίνω (huphaínō)
• *upo > ὑπό (hupó),
• *uperi > ὑπέρ (hypér)

Are these last two related to the prothetic s- seen in Latin sub, super? (Wiktionary gives super < *eks-uper, but ὑπέρ < *uperi, which would seem to suggest that it is just a coincidence).

A related question: a rough breathing seems to also have developed in some words that didn't start with upsilon, but that had *w in their PIE form. In most words starting with *w in PIE, like *wérǵom > ἔργον (érgon) the *w simply developed to ϝ (representing the sound /w/) and was later lost, leaving a smooth breathing behind. Why did a rough breathing develop in the following words?

• *wek(ʷ)speros > ἕσπερος (hésperos)
• *wes- > ἕννυμι (hénnumi)

Is this related to the *s that comes later in each of these roots?

I don't know if it's also relevant that the PIE word-initial cluster *wr- merged with the reflex of the cluster *sr- in Greek, giving ῥ (rh, rho with a rough breathing).

Answer 1

Greek initial /h/ develops regularly from IE *s and non-syllabic *i. But (as you mention) initial υ always has rough breathing, even where it was not preceded by etymological /s/ or /j/. One explanation that has been considered is that Greek initial /u/ developed a /j/ on-glide, very much as has happened in English (unit, use, united…), which then partook in the development of inherited /j/ to /h/.

Answer 2

As a conjecture, it seems possible that the generalized rough breathing before #u- has to do with the cross-linguistic tendency for high vowels to produce extra turbulence, which can then be phonologized. Most often this takes the form of spirantization or affrication of stops before high vowels, but aspiration can also result: in Ikalanga (Bantu), stops become aspirated before /i u/. I don't know of an exact parallel to the Greek case, though.

For your second question about the w- > h- change, this is indeed often said to be somehow due to the following -s- in most of these words. This isn't unproblematic, though, since (a) the phonetic justification for such a change seems slim, and (b) there are counterexamples of both kinds: ἑκών, ἕδνα with unexpected rough breathing, ἄστυ with no rough breathing where this account would predict it.

Answer 3

It is known that in PIE any words starting with r- were prohibited. It is conjectured that this was the case of all sonorants.

For instance, the pronouns usually rendered with initial i̯- has been recently reconstructed with an initial laryngeal e̯i̯-.

One of your roots, the u̯ebh- (wasp, web, weave etc) is known to start with a laryngeal, though what laryngeal it was is different in Mallory&Adams and Brill Latin.