Hi all,

     We have recently started experiencing poor reproducibility with respect
to d18O with our Gasbench-Delta+ setup.  I was curious if anyone had
suggestions for a possible solution.  Below is our setup:

~10-year old Delta Plus
~4-year old GasBench
95% of sample throughput is carbonates (manual acid injection) with the
remaining samples being d13Cdic or d18O of water.
typical 1-sigma reproducibility for d13C based on NBS-19 (n=15 per 80
samples) = +/- 0.05 ‰ (before recent problems)
typical 1-sigma reproducibility for d18O based on NBS-19 (n=15 per 80
samples) = +/- 0.10 ‰ (before recent problems)
Method setup for carbonates: peaks 1-3 reference, peaks 4-13 sample, peak 14
reference with peaks 3 and 14 assigned values within the method
Reactions are stable at 25.0 ºC.
It is easy to get Standard On/Offs below 0.10 for d13C and 0.08 for d18O (10
REF peaks).

Over the last several weeks our reproducibility has slipped to the 0.10 to
0.20 ‰ range for d18O (sometimes worse).  The numbers for carbon still look
very good with errors always better than 0.10 ‰. We have cleaned the source,
replaced the filament, re-tuned 4-5 times, baked out the GC column (150 ºC
overnight) multiple times without any improvement in the O errors.  There
are no indicates of problems with either the flushing or sampling needles.

We performed a test where we measured the isotopic composition of CO2 from
our He+0.3% CO2 cylinder (typically for water equilibration) and the error
for d18O was +/- 0.10‰ (n=96).

I would think that issues related to the source or tuning can be ruled out
due to the good performance of C and the good numbers from the
before-mentioned CO2 experiment.  

However, with that said we do have problems with linearity.  Within a single
sample, as the sample peaks get smaller the d18O values become more
18O-depleted by up to 0.8‰ at times. We had somewhat similar problems with
this in the past but we circumvented the issue by making the last reference
peak (#14) a second anchor point.   Between samples, although less
pronounced, there is also a weak relationship between d18O value and sample
weight/peak amplitude/peak area.  

My gut tells me we have a problem with water, but would water associated
with the carbonate samples/acid cause linearity issues no matter how
well-tuned the source is?  We’ve never dealt with significant linearity
issues in the past so I’m at a bit of a loss in how to attack the problem. 
The H2O background of the He stream is less than 300 mV (mass 18.0 centered
on cup 2). We have an old water trap in the He line but not sure if it does
any good or not. Again, baking out the GC doesn’t help.  I’ve heard problems
with the nafion traps could result in the behavior we are experiencing.  I
do know that He is flowing out the vent of the trap, although we currently
have no good means of measuring the flow rate (suggestions?).

I assume it is entirely possible that we are creating the water from each
carbonate/orthophosphoric acid reaction and it is not properly being
scrubbed from the He stream on a sample-to-sample basis.  With time the H2O
is removed from the line thus giving good background numbers and good tests
with dry CO2 experiments.

An additional thought is that the nafion traps are OK but our NBS-19 has
adsorbed too much water for the traps to handle.  A few months before our
problems started our dessicator stopped working and the southeast USA is a
very humid area even inside research buildings.   Yet, I’m not even sure if
storing within a dessicator is essential.

We’ve been dodging tornadoes left and right down here in Alabama so isotopes
have been our last concern but we’d certainly like to get back on track here
at our lab.

Thanks in advance for any comments or suggestions you might have.

Alabama Stable Isotope Laboratory