Although the accuracy required of a primary standard of light is of the order of 0.1%, there is a spread of 1% in the size of the candelas derived from the primary standards constructed at several national laboratories. In this paper, the standard has been critically appraised: Some of the suspected weaknesses have been studied experimentally, others are discussed at length.
A tantalum susceptor, used around the crucible containing the platinum to absorb the induced currents and heat the platinum indirectly, increased the photometric precision by an order of magnitude. The technique unfortunately introduced excessive contamination of the platinum.
Since none of the modifications has created the desired improvement in accuracy, other types of light standard may need to be reconsidered.
C. L. SANDERS and O. C. JONES, "Problem of Realizing the Primary Standard of Light," J. Opt. Soc. Am. 52, 731-745 (1962)