n 1872 a _provisional_ solar temperature
of 1,398 deg.[708] This is below that at which iron melts, and we know that
iron-vapour exists high up in the sun's atmosphere. The matter was taken
into consideration on the other side of the Atlantic by Ericsson in
1871. He attempted to re-establish the shaken credit of Newton's
principle, and arrived, by its means, at a temperature of 4,000,000 deg.
Fahrenheit.[709] Subsequently, an "underrated computation," based upon
observation of the quantity of heat received by his "sun motor," gave
him 3,000,000 deg. And the result, as he insisted, followed inevitably
from the principle that the temperature produced by radiant heat is
proportional to its density, or inversely as its diffusion.[710] The
principle, however, is demonstrably unsound.

In 1876 the sun's temperature was proposed as the subject of a prize by
the Paris Academy of Sciences; but although the essay of M. Jules Violle
was crowned, the problem was declared to remain unsolved. Violle (who
adhered to Dulong and Petit's formula) arrived at an _effective_
temperature of 1,500 deg. C., but considered that it might _actually_ reach
2,500 deg. C., if the emissive power of the photospheric clouds fell far
short (as seemed probable) of the lamp-black standard.[711] Experiments
made in April and May, 1881, giving a somewhat higher result, he raised
this figure to 3,000 deg. C.[712]

Appraisements so outrageously discordant as those of Waterston, Secchi,
and Ericsson on the one hand, and those of the French _savants_ on the
other, served only to show that all were based upon a vicious principle.
Professor F. Rosetti,[713] accordingly, of the Paduan University, at
last perceived the necessity for getting out of the groove of "laws"
plainly in contradiction with facts. The temperature, for instance, of
the oxy-hydrogen flame was fixed by Bunsen at 2,800 deg. C.--an estimate
certainly not very far from the truth. But if the two systems of
measurement applied to the sun be used to determine the heat of a solid
body rendered incandescent in this flame, it comes out, by Newton's mode
of calculation, 45,000 deg. C.; by Dulong and Petit's, 870 deg. C.[714]
Both, then, are justly discarded, the first as convicted of exaggeration,
the second of undervaluation. The formula substituted by Rosetti in 1878
was tested successfully up to 2,000 deg. C.; but since, like its
predecessors, it was a purely empirical rule, guaranteed by no principle