We are running CO2-water equilibrations on a Gasbench II+ Delta+XP and 
in general have good results.  We have noticed "fliers" with anomalously 
heavy values in one or two recent runs and we wonder if anyone else has 
seen this and knows a possible reason.

We equilibrate 0.5 mL water with 1%CO2-He mixture for 26 hours at room 
temperature, with 88 analyses in a run.  Every fourth analysis is for 
the same lab standard water, and we use this to control for instrument 
drift, which is about 0.5 per mille in d18O with a roughly diurnal 
cycle.  In our latest two runs we see single analyses that stand out 
from this pattern, and are 1.5 - 2 per mille heavier vs the reference 
gas than the standards before and after.   Suspecting that this might be 
a sporadic phenomenon that affected samples as well as standards we made 
up a run consisting mostly of duplicates of previous analyses and found 
about 10% of samples in the run showed similar positive anomalies when 
compared with their earlier "true" values.

Anomalous analyses have completely normal chromatograms and excellent 
internal precision.  There is no drift in ref gas peaks, and signal 
strength for both sample and refgas is normal.  There is no sign of an 
N2O peak preceding the CO2 peaks, which is normally a give-away for 
contamination by atmospheric air.  The contamination of the sample gas 
would have to be huge to produce a 2 per mille anomaly in d18O of 
headspace CO2, and this would also cause a large N2O signal produced by 
reaction of N2 and O2 in the source.

Prior to all this we had performed about 1000 water analyses with 
generally excellent precision (1 sd ~0.045 per mille for d18O, estimated 
from the method of replicates and also from repeated analysis of a 
single lab standard in every run.)  However, looking back at the lab 
notebook we notice that the same problem occurred in a single 
preparatory run right at the beginning of this batch of work, in which 
we just had repeated analyses of two waters.  We never found out what 
caused it and the problem did not repeat itself.  We had some other 
running repairs to make and forgot about it.  Now it seems to have come 
back.

Has anyone else experienced this pattern of sporadic anomalies, and 
(hopefully) found the cause?

John Hill & Tim Atkinson,
Bloomsbury Environmental Isotope Facility,
University College London.