As usual, it's not easy to pin down who said or did it first, but I 
think Dupuit is probably the best answer, if you absolutely have to have 
a name. He drilled the concept, but did not use the term.

I suggest looking at James R. Hines, Jr., ?Three Sides of Harberger 
Triangles,? _The Journal of Economic Perspectives_, Vol. 13, No. 2, 
(Spring, 1999), pp. 167-188.

In my forthcoming micro text, in a chapter on taxes and deadweight loss, 
I say in the references section (which I use throughout to do a little HOT :-)):

"Hines explains that the theory of deadweight loss dates back to Dupuit, 
Jenkin, and Marshall, but Harberger?s papers in the 1950s and 1960s 
?illustrated the techniques, the usefulness, and the realistic 
possibility of performing such calculations, and in so doing, ushered in 
a new generation of applied normative work.? (p. 168). For this reason, 
argues Hines, ?Welfare loss triangles are ?Harberger triangles? because 
Harberger?s papers measured them, did so in a consistent manner, and 
assisted and encouraged a host of others to do likewise.? (p. 185).

Hines offers a nice bibliography if you want to dig deeper.

Humberto Barreto