OF ARTS AND SCIENCES. 115 



It is to be observeil, however, that enough of this impurity to change 

 the crystalline habit and color of the curcumia has but a very slight 

 effect on the percentage composition, and, it may be added, does not 

 lower the melting-point more than one degree. That it is formed by 

 the oxidation of curcumin by the action of the air, appears from the 

 fact that pure yellow curcumin was partially converted into red burrs 

 when moistened with alcohol and exposed to the air for a long time. 

 When once formed, the impurity can be removed only by repeated 

 crystallization, and the amount of Madras curcumin at our disposal 

 did not admit of this, nor did we take the trouble to prepare a larger 

 supply, as the above analysis with the melting-point 178° is sufficient 

 to establish the identity of the Madras and Bengal curcumins. 



The following comparison shows that our results agree tolerably 

 well with those of Daube, but are entirely at variance with those of 

 Kachler and Ivanow-Gajewsky. (As no analyses are given in the 

 abstract of the latter's article, which alone is at our disposal, we have 

 given the theory for his formula, C^gH^gO^, under his name.) 



Jackson 

 and Menke 

 Mean 

 67.92 68.30 

 5.70 5.63 



There are, then, eight analyses of curcumin, which support a percent- 

 age of carbon in the neighborhood of 68 against three (or more) in 

 favor of one near 70.* 



It is probable that the high results obtained by Ivanow-Gajewsky 

 and Kachler were due to the presence of resinous impurities, since 

 their predecessors, who analysed exceedingly impure curcumin, as 

 shown by the low melting-point, obtained the following results. 



This view is further supported by tJie fact that we obtained a higher 

 melting-point, 178°, than anyone else; Daube found 165°, Ivanow- 

 Gajewsky 172°, later 140°, while Kachler gives no melting-point, 



* Ivanow-Gajewsky obtained a percentage of lead in a plumbic salt agree- 

 mg with his combustion ; but tliis result is more than counterbalanced by the 

 analyses of derivatives of curcumin given later in this paper. 



