I am fascinated by the range of conditioning approaches we all use for GC HTC reactors. Methane or hexane are the two sources of carbon but the amount varies and interestingly, the frequency varies as does apparently the reactor life. I would be very curious to hear more about peoples experiences and best practices. Our lab's experience is below.
On our system we condition with 2 x 1ul injections of hexane in straight mode, run an alkane standard such as A5 overnight and then start running samples. We do not recondition the reactor. We are using the Thermo alumina tube reactors with the welded connection to a silicosteel capillary on the oven size. A reactor typically lasts 2-3 weeks at 1420 C. The end occurs not because of leaks but because of a very large size effect and systematic offset. Reconditioning very temporarily fixes this but not long enough to make it worth our while. Our typical external precision that includes analytical uncertainty as well as uncertainty in realizing the VSMOW scale is ±3 per-mil (1s) calculated as in Polissar and D'Andrea (2014 GCA). We do not do a multi-point calibration as would be best because appropriate standards are not available. (I think to do this properly would require several standard mixtures containing many peaks, each standard having very similar dD values for all peaks and the different standards having very different dD values).
We run a lot of isotopic standards and a drift sample throughout our analysis period. These are run at multiple concentrations every day to track any size effects along with systematic offsets etc.
We find that in the first few days to week after conditioning:
* standards such as A5 'behave' and have relatively low obs-exp variations across peaks (see Polissar and D'Andrea, 2014 GCA for an example).
* size effects are small (this is the time to run those small samples). Small samples are slightly more positive than large samples (by ~a few per-mil). Cutoff for size is ~150 ng on-column which equates to ~20 V-s on our system.
After ~ a week:
* standards such as A5 do not behave and have high obs-exp variations across peaks. This may occur because Mix A5 has large peak to peak isotope differences (by design, Sessions et al., 2001 GCA) that are affected by memory in the HTC reactor (c.f. Wang and Sessions, 2008 GCA). We observe a compression of the isotope scale in Mix A5 at this time that is consistent with memory effects.
* drift samples/sample replicates behave and give identical values to those from the beginning of the reactor
* size effects become larger and samples must be run at higher concentrations. On our system the size effects become negative at this time. The effect can be large, ~10 per mil for small samples. Cutoff for size is ~250 ng on-column which equates to ~35 V-s on our system.
After ~2-3 weeks:
* size effects become unmanageable, even for very large injections (cutoff would need to be >500 ng on column)
* throw the reactor away and start again
Because the reactor is at its 'best' in the first week, we have tried running reactors for only a week and then replacing them. For this, we used empty alumina tubes connected with valco fittings inside the GC (with the transfer capillary fed into the reactor ~3-4 cm) rather than the expensive thermo reactors. This worked pretty well although requires a lot of hands on work as we changed the reactor every 4-6 days and it is more prone to leaks. If we have samples that are large it is easier to just run a thermo reactor for 3 weeks.
Marilyn, FYI we have completely removed the sled, multifunction valve cluster, and 4-port on the isolink from the flow path and simply have a valco tee in the oven for the backflush vent capillary (and replaced the multifunction valve cluster @#$% with a SGE valve for the backflush vent, a la the GCC III). To switch from H to C requires changing some fittings but the decrease in problems and increase in precision are well worth it. For C measurements we oxidize continuously with a slow bleed of 1% O2 in helium introduced through a 2nd tee on the oven side of the reactor. The increase in precision and accuracy of d13C with this bleed was substantial although it wouldn't work for d15N measurements.
Lamont Associate Research Professor
Division of Biology and Paleo Environment
Lamont-Doherty Earth Observatory of Columbia University
61 Route 9W, Palisades, NY 10964
845.365.8400 (phone) • 845.365.8150 (fax)
[log in to unmask] • www.ldeo.columbia.edu/user/polissar