[ 263 ]

sound less than the moulds, but those of ircn equal tothem, or a little bigger.

Having found on trial that solid pieces of iron rose tothe surface of melted iron, and being thus convinced ofthe truth of Reaumur’s principal experiment; I thoughtI had sufficient foundation to say, in the foregoing part ofthis work, page 67, that iron expands in its passage froma fluid to a solid ssate. Some instances I had seen ofmelted iron having closely applied and fixed itself to solidpieces by which it was confined, confirmed me in thisopinion, and induced me in page 59, in proposing amethod of obviating the common imperfections of therolls for flatting gold and silver wire, by casting a hoopof steel, and after sufficiently forging the hoop, fixing aniron axis within it, to recommend, as the readiest way thatoccurred for securing the axis, pouring melted iron intothe space between it and the hoop.

An ingenious correspondent has since observed to me,that melted iron does not expand in setting, and will notfix itself to the hoop, but shrink from it as other metalswould do : that though works of cast iron are indeed ge-nerally larger than the mould, yet this increase of bulkdoes not proceed from the expansion of the metal itself,but from its fluxing and drinking into its surface a con-siderable quantity of the sand of which the mould is com-posed, which he judges to have been the foundation ofReaumur’s mistake : that large iron works cast in openmoulds, as forge hammers and anvils, have considerablehollows on the upper fide : that large works cast in closemoulds have always a cavity somewhere in the internalpart; that cannon balls cannot be cast without such cavi-ties ; of which the workmen are so sensible, that in orderto avoid as much as possible the inconvenience of the hol-low being near the side, they turn the mould upside down

soon