IN RELATION TO QUANTITATIVE CHEMICAL ANALYSIS. 
329 
changes. Solutions of intermediate strengths yield lines of intermediate length and 
strength, proportional to the quantity of magnesium present. 
(7.) O'OOOl per cent. Mg., or 1 part per 1,000,000 of solution. 
The first and third lines of the quadruple group are much attenuated through one- 
half their length, the other portions being weakened. Copper electrodes were used 
for this and the following observations on magnesium because the graphite points 
contained that element in sufficient quantity to give the first and third lines, though 
not so strongly as this solution. 
(8.) O’OOOOl per cent. Mg., or 1 part per 10,000,000 of solution. 
The same change was noted, the first and third lines being nearly invisible through 
one-half. 
(9.) O'OOOOOl per cent. Mg., or 1 part in 100,000,000 of solution. 
The first and third lines were extinguished as regards one-half their length. 
(10.) 0‘000,0001 per cent. Mg., or 1 part in 1,000,000,000 parts of solution. 
The lines appeared as in the last solution, but they were fainter. 
(11.) 0’000,00001 per cent. Mg., or 1 partin 10,000,000,000 of solution. 
In this last instance the lines were almost invisible. 
This dilution may be regarded as the point of extinction of the most persistent 
rays of magnesium with the particular spark arrangement I have generally used, or in 
other words we may easily detect 1 part of magnesium in 10,000,000,000 of liquid. As 
the quantity of solution utilised by the spark is certainly less than 0'1 cub. centim. 
we may detect with absolute certainty less than tmoTVoA, ol» o tli of a milligram of 
magnesium. 
When the strength of the spark is increased by the use of a much larger coil, and 
the striking distance between the electrodes is left the same, the strength of the 
magnesium lines is so altered that with 1 part of the metal in 10,000,000,000 of the 
solution the two lines with wave-lengths 2801 - 6 and 2794’1 are rendered as strongly as 
in the photographs of those obtained with the previous spark arrangement from a 
solution containing 1 part in 1,000,000 of liquid. The sensitiveness was therefore 
increased ten thousandfold. We have thus sufficient evidence to show that the sen¬ 
sitiveness of the spectrum reaction in this case is practically without limit. It varies 
however with different metals. 
MDCCCLXXXIV. 
2 u 
