>> At the IAEA consultants group meeting in Vienna just before Christmas the
problem of high precision D/H measurements was discussed. All of the direct
reduction methods, using reagents such as Zn, Mn, Cr, perhaps Mg-Pt etc.
seem only capable of producing routine precisions on the order of 1 per mille
at best. Results obtained using U furnaces seem consistently to produce much
higher precisions and I have seen data at +/- 0.2 per mille produced by
several labs using this technique.
One method that does seem to offer great promise, with the benefit of ready
automation, is the H2-H2O equilibration technique using a Pt on alumina
catalyst. It also has the benefit that it can be applied directly to brines etc.

Paul Dennis

--------------------------------------------------------------------------
	
	I have used U furnaces for many many years, with a consistent  
precision of +- 0.2 permil or 0.5% of delta, whichever is greater (there are  
percentage errors in the measurment (e.g. the uncertainty of knowing the H3+  
correction) making it necessary to specify both an absolute and a relative  
precision).  There are five tricks necessary to achieve such precision:

1.  There is a cold trap to catch water on both sides of the U furnace, so  
that any water going through the furnace is recyled back through the U.  As  
the U ages, more and more water gets through the U so the recycling becomes  
more important.  We change the U when measureable amounts of water get through  
more than twice.

2. The water gets through primarily because U-oxide builds up in the furnace.  
 My furnaces support the U vertically in a U-trap, with the U resting on Cu  
wool (and also capped by Cu to keep U from getting up above the furnace if  
there is a line break or an accidental influx of air).  Below the Cu support  
on one side of the trap is a hollow space.  Every day we tap the U-trap  
several times which knocks the U-oxide down into the hollow space and gets it  
away from the U metal.  This extends the life of the U load very greatly.   
There is also a problem with U volatilizing above the Cu wool at the tops of  
the two arms.  We lower the oven at night (it stays hot all the time) below  
the U-trap, and for running samples we raise the furnace so that the hot zone  
covers the volatilized U film on the quartz tube and keeps it hot.

3.  Commercial U can be contaminated with thorium, as some labs found years  
ago: this ruins the preparation and causes large errors.  The thorium came  
from sweeping up the U chips on the cutting table by the manufacturers, after  
they had cut some thorium and had not cleaned the table.

4.  The hydrogen coming off the U metal is different in composition from the  
small amount of H2 left ON and IN the U metal, and this can cause errors.  It  
is necessary to pump (Toepler) all the H2 into a bulb above the Toepler pump  
so that the latest H2 to come off mixes with all the other H2 already  
collected.  Then the entire amount of H2 can be pushed up into a sample tube.   
If one simply pumps the H2 into a sample tube directly, and if a total sample  
is not collected, there are errors (of course if one collects it all that is  
OK, but if you get a large sample you may not be able to push it all into the  
sample tube).

5.  Contamination of a sample during loading with moisture in room air can  
cause errors at the 0.2 permil level (e.g. with seawater samples).  We have a  
tube on a ball joint which is small and narrow, so that not much air is  
present in the tube (removable to clean out after running seawater and  
brines). We flush and fill this inlet tube with Ar (heavier than air) and then  
inject the water sample (7 microliters) with a syringe and needle, so there  
is no room moisture contamination.

6. For seawater and brines, the inlet tube contains glass wool and the sample  
is injected below the wool.  Otherwise when the samples decrepitate during  
heating to transfer the water to the reaction line, the salt flies up into the  
ball joint at the top of the tube, and reabsorbs some water during the  
transfer.  This has a large fractionation effect and causes significant  
errors.

7. We run H2 sub-standards (SMOW) on the SAME side of the line as the samples  
and plot these every day, using their average as the value of the SMOW  
standard for calculating the sample data.  This takes care of any drift of the  
standard.

8.  Our H3+ contribution is always less than 2.5% of HD.  We measure it  
frequently by inputting the machine standard on BOTH sides of the machine, and  
adjusting the two voltages to be 20 and 25 volts.  Measurement of the  
apparent delta difference, done the same way as analyzing the sample, gives a  
very precise measurment of the H3+ effect.

	Examples of the precision for samples very different from the  
standard in D/H ratio can be seen in the Horibe & Craig paper on the  
equilibrium fractionation between methane and hydrogen, in the latest  
Geochimica et Cosmochimica Acta.

	With respect to measurements by equilibrating water with H2, of  
course one is then measuring D/H activity ratios, not D/H concentration  
ratios.  This can make a significant difference with brines.

H. Craig
Isotope Laboratory
SIO/UCSD