clarke.] META- AND DIMETASILICATES. 95 
set free quantitatively. This would give the ignited residue the com- 
position shown by the subjoined alternative formulae: 
Mg 
,Si0 4X >°\v 
Mg< >Mg and Mg<f ?Mg 
X Si 2 O s ^ X Si 3 0/ 
Mg 
and of these the first would seem to accord the better with the remark, 
able stability and insolubility of the material. A metasilicate formula, 
H 2 Mg 3 (Si0 3 ) 4 , is also admissible, and accords equally well with the evi- 
dence concerning talc. The pseudometasilicate expression, however, 
seems to be preferable in view of what is known as to the genesis of 
the. species. 
Pyrophyllite and kyanite are two aluminous metasilicates which seem 
to be related; at all events the former is not infrequently the gaugue 
or matrix of the latter. Empirically they are 
Pyrophyllite AlH!Si 2 6 
Kyanite Al 2 SiG 6 
or, structurally, 
.SiU 3 .o 
Alf and Air 
\Si0 3 — H \Si0 3 — A1=0 
To these formulae I attach very little significance, except in so far as 
they represent a single observed relation. Probably the true formulae 
should be multiples of these ; for kyanite is much denser than its isomer, 
andalusite, Al 3 (Si0 4 ) 3 (A10) 3 . Both kyanite and andalusite, at very high 
temperatures, are transformed, with disengagement of heat, into a third 
isomer, sillimanite,* a phenomenon which indicates dissociation. Silli- 
manite doubtless has the smallest molecule of the three species, and 
may be AlIZSi0 4 — AlO, but even this is so purely hypothetical that it 
would be useless at present to discuss the species farther. 
With beryl, Gl 3 Al2Si«0,a, the evidence in favor of a metasilicate 
structure is fairly good, although the composition can also be expressed 
as that of a basic trisilicate. There are thus two alternatives, 
/Si0 3 
/ >G1 
Al-SiQ 3 G1 
\s 
iO a 
I y si 3 o 8x 
Gl and Gl< >G1 
I X Si 3 0/ 
,Si0 3 II 
Al-SiO, 
\ >G1 
(AlO), 
* Vernadsky, Bull. Soc. Min., XII, p. 447, and XIII, p. 256. 
