In Ancient Quenya any initial d- became l-, a change Tolkien mentioned frequently starting in the 1930s:

d > l (r) (Comparative Tables, 1930s, PE19/18).

d became l initially. (Outline of Phonetic Development [OP1], 1940s, PE19/18).

d became l initially ... No clear case of initial r occurs. [add to margin in red ball point] A very early change? Cases of d/l variation are found in other[?] languages and may go back to C. Eld. or earlier (Outline of Phonology [OP2], early 1950s, PE19/69-70).

Quenya initial d- normally > l- (etymology notes from the late 1950s, PE17/17).

Quenya has alda since its initial voiced stops (like Greek) were all spirantalized and weakened: b > v, d > l, g > ʒ > nil (etymology notes from the late 1950s, PE17/153).

This form is due to the early change in Q of initial d > l. The change was regular in both Vanyarin and Noldorin dialects of Quenya (Quendi and Eldar, circa 1960, WJ/363).

This [d/l variation] was far older and distinct from the normal Quenya change of initial d > l; though the existence of the variation already in the language may have assisted in making l the normal change (Variation D/L in Common Eldarin, late 1960s, VT48/30).

In Quenya initial d- normally > l but the verbal forms clearly used -de (Quenya Pronominal Elements, 1968, VT49/51).

As noted above, this Quenya sound change was distinct from sporadic d/l variation in Common Eldarin, which was a variation in related roots rather than a true phonetic development (VT48/23). Unfortunately, Tolkien did not describe the mechanism whereby d became l initially in Quenya. The medial development was:

Medially between vowels it appears to have been somewhat retracted, and to have become at first a weak untrilled r or đ-like sound distinct from original r. The two sounds have distinct letters in earlier Quenya spelling, which may be represented by r (original) and r̃ (derived from d). 7 = r, 6 = ř (OP2: PE19/69-70).

The exact nature of this medial development (which Tolkien most frequently represented as ř) isn’t entirely clear, but the general consensus is that it was probably an alveolar approximant [ɹ], which is how “r” is typically pronounced in English (untrilled). Perhaps initially it further developed into the lateral approximant [l] instead of to a trilled [r] as it did medially. Alternately, the initial development may have been [d-] > [ð-] > [l-]. Analogs of this sound change are hard to find in real-world languages [@@@ look into this more].

This sound change can be seen most easily in l-/d- variations between Quenya and Sindarin/Noldorin:

• √DAN-TA > Q. lanta-, S. danna- “to fall” (PE17/62).
• ✶dōmē > Q. lómë, S. dû “night” (PE17/152; Ety/DOƷ).
• ᴹ√DAY > ᴹQ. laime, N. dae “shadow” (Ety/DAY).

Conceptual Development: In Qenya of the 1910s and 20s, initial voiced stops were unvoiced, so that initial d- > t- as can be seen in t-/d- variations between Early Quenya and Gnomish/Early Noldorin:

• ᴱ√DARA(MA) > ᴱQ. taran “bang, buffet” vs. G. drambor “thudder” (QL/89; LT2/203).
• ᴱ✶dagá > ᴱQ. tá, ᴱN./G. dâ “high” (QL/87; GL/29; PE13/141).
• ᴱ✶dagla > ᴱQ. taila, ᴱN. dail “axe (blade)” (PE13/141; PE14/66).

There does seem to be one aberrant example of d- > t- in Tolkien’s later writing: ✶dāra > Q. tāra, S. Daur “wise” (PE17/102), as in Frodo’s Sindarin name. Most likely this is a fluke.