Recently, Prof J. E. Hirsch at UCSD proposed a single-number measure of scientific output of a researcher.  This method has attracted a lot of attention worldwide ever since his paper about it was released as a preprint in arxiv.org. The paper has  been published in PNAS.  The h index of a particular researcher is calculated as follows. If a researcher has published N papers, among which there are  h papers each  of which has been cited for at least h times, while each of the rest N-h papers has been cited for no more than h times, then this researcher's research output can be indexed as h.  

But often the papers are coauthored by more than one person.  In his paper, Hirsch mentioned that some normalization by a factor refelecting the number of coauthors can be done in cases of large differences of numbers of coauthors .  Specifically, I would like to suggest that the h index is improved to be an normalized index, which can be calculated as follows. 

After finding the h most cited papers, each of which has been cited at least h times, the average number of coauthors, denoted as n, of these h papers is also calculated,  then the normalized index is h/n. 

This normalized  index seems to be better. 

One might argue that different coauthors of a paper contribute differently. True. But it is hard to find out and measure this difference. Furthermore, to make such a difference in evaluation universally  would  make negative impact on cooperations. Nothing is perfect. The simple normalization by the average number of coauthors is sufficient.