He next attacked, first with Dittmar’s and then with Schorlemmer’s assistance, the nature of the aqueous solutions of the common volatile acids of constant boiling-point, and showed that although the ratio of acid to water is constant for a definite boiling-point under a particular pressure, this does not necessarily indicate the existence of definite hydrates. The composition of the hydrated acid on boiling is entirely dependent on the pressure under which it is heated—a strong solution losing acid, and a weak solution losing water until the residue in each case acquires a constant composition, depending upon the pressure under which it is boiled.[13] In those days Gmelin’s “Handbuch” was the chief repository of chemical knowledge—or the want of it—and many suggestions as to possible profitable fields of inquiry were to be gleaned from its pages. One such subject was perchloric acid and its compounds, concerning which but little was then known beyond the composition of potassium perchlorate, established by Stadion as far back as 1816. Roscoe, with Schorlemmer’s assistance, made a fairly complete investigation of perchloric acid and its hydrates, and a number of its salts.[14] He narrowly escaped a serious accident when working with ethyl perchlorate, first prepared in 1840 by the American chemists Hare and Boye, and known to be extremely unstable. He was engaged in filtering a few cubic centimetres of the liquid into a test-tube, when the compound exploded with great violence, and a deep hole was bored into the base of the filter-stand, and many hundreds of fragments of glass were driven into his hand. That filter-stand was long an object of interest to visitors in the private laboratory of the old Owens College.[15] The writer subsequently prepared thallium perchlorate for him in a pure state, determined its composition, and established its isomorphism with the alkaline perchlorates, the crystallographic characters of which had been previously ascertained by Kopp.[16]