HILLEBRAND.] 
CIIKOMU'M. 
81 
will strengthen it, and less than 2 milligrams of chromic oxide in 1 
gram of rock can then be exactly measured. For smaller amounts it 
is best to employ from 3 to 5 grams of powder and then to concen- 
trate the chromium by precipitation by mercurous nitrate, as detailed 
in the next section under Vanadium (p. S3); otherwise it may bo diffi- 
cult or impossible, because of the large amount of alkali carbonate 
present, to obtain a filtrate of sufficiently small bulk to show a decided 
color. 
If niter has been used in the fusion and the crucible has been at all 
attacked by it, a yellow coloration of the filtrate may be due to dis- 
solved platinum, but neither the proportion of niter nor the tempera- 
ture of the blast should ever be high enough to permit the crucible to 
be attacked. 
Comparison of colors. — The final solution is transferred to a gradu- 
ated flask of such size that its color shall be weaker than that of the 
standard chromium solution. Definite amounts of the latter are then 
diluted with water from a burette until of the same strength as the 
test solution, exactly as described on page 68 for the colorimetric 
estimation of titanium. For very minute amounts it is necessary to 
use Nessler tubes, as in ammonia estimations, instead of the glasses 
and apparatus described and depicted under Titanium (p. 70). 
As with colorimetric methods in general, this one gives better results 
with small than with large percentages of chromium, yet it can be 
applied in the latter cases with satisfactory results by making a larger 
number of consecutive comparisons with the same solution. 
A FEW COMPAKATTVE DATA. 
A few comparisons between colorimetric and gravimetric determi- 
nations of chromium are here given to show the order of agreement, 
the former having* been made several months and even years after 
the latter. 
Gravimetric 
per center./).;. 
Colorimetric 
per cent Cr 2 3 . 
Trace. 
0.018 
0. 05 
.051 
.14 
.12 
.08 
.C33 
Trace 
.013 
None. 
. 0086 
None. 
. 0067 
The outcome was somewhat surprising, for it was hardly to be 
expected that the long and laborious quantitative separations should 
Bull. 17G 6 
