H2OFASTFIX.BIN 
WITH AD HOC FIXES TO NUCLEAR STATISTICAL WEIGHT ERRORS IN H2OFAST.BIN

     H2O16 and H2O18 * 0.25    H2O17 * 2.0    HDO16 * 0.21
 
The Partridge and Schwenke data in H2OFAST.BIN and H2OSLOW.BIN were 
normalized incorrectly by Kurucz and have been corrected.  


Each isotopolog has its own partition function and nuclear 
statistical weights.  In HITRAN at 300K the partition functions are 

H2O16 178.305   H2O18 179.904    H2O17 1079.478     HDO16 883.847
    
Atmospheric transmission is computed in TRANSYNTHE using HITRAN 
line data and HITRAN partition functions.  

In ATLAS and SYNTHE there are no isotopolog partition functions, 
only H2O16 which is derived from JANAF and is approximately the 
same as Vidier and Tennyson (2000).  At 300K it is 44.531 which is 
1/4 the HITRAN value.  Kurucz formatted Partridge and Schwenke gf 
values to agree with the HITRAN gf values without taking into 
account the change in partition functions.  The Partridge and 
Schwenke gf values in H2OFAST and H2OSLOW are 4 times too large.

Ignoring the nuclear spin factors, the partition functions of the 
isotopologs, H2O16, H2O17, and H2O18 are nearly the same to the 
percent level.  At low telluric temperatures HDO16 is about 20 per 
cent below H2O16, but at stellar temperatures the differences drop
to the few percent level.  We make the ad hoc approximation that the
partition function for each isotopolog is that for H2O16 times a 
constant.  The correction factors for the line list are then 
determined from a few strong lines to be HIT(gf/pf)/PS(gf/pf).

     H2O16 and H2O18 * 0.25    H2O17 * 2.0    HDO16 * 0.21

Isotopolog partition functions will have to be included in ATLAS
eventually.  In order to use HITRAN laboratory data in ATLAS and
SYNTHE stellar calculations, the data must be renormalized.