Dear Paul and others interested,
the figure I gave for the size of the memory referred to the TC/EA. With
the H-Device, our memory size is closer to your values, up to 1%. Thanks
for sharing the detailed handling of the syringe, it really points to a
weak spot of water measurements when injecting with a syringe..
I wonder whether the syringe memory is different in the two cases mainly
because of the differences in injection pressure: in the H/Device, we
inject into vacuum whereas in the TC/EA, the head pressure is about 1
bar above ambient. Hence, when the syringe has a leak, some of the water
will leave the syringe to the other side and can cause fractionation.
Thus, in TC/EA systems, syringe tightness is probably more critical than
when injecting to a vacuum.
Paul Brooks wrote:
> We have been able to reduce the memory effect of the syringes to reliably
> less than 1%, and most of the time in the 0.2 to 0.5% carryover range using
> an ethyl ether rinse with waste to vacuum under a septa. Furthermore,
> there is no evidence that this memory effect extends beyond the next
> sample. I presented this at the conference in Winnipeg last August. It is
> just one of several small method improvements that I haven't had time to
> write up.
> The solution to the memory problem is to dry out the syringe between each
> sample. We now do this by rinsing with Ethyl Ether between samples and
> injecting to a waste under vacuum.
> I should acknowledge John Morrison for originally suggesting the memory
> effect was in the syringe, Gilles St-Jean and others at the 2002 Canadian
> conference who suggested rinsing the syringe and emptying to waste under a
> vacuum, and my colleague Richard Neese for suggesting the use of an
> We use Hamilton #80300 10uL syringe with the Combi PAL autosampler
> injecting to the H/Device.
> We have a 2 mL vial set in position 96 in the autosampler tray filled with
> fresh daily (from the main bottle) ethyl ether capped with a normal
> septa. It is important that the ethyl ether main bottle only be used for a
> month, and be kept dry. Dry ethyl ether will form a azeotrope with water
> up to 1.2%. This means that the ethyl ether will boil off at about 34.2
> degrees C and carry with up to 1.2% water. Therefore when the syringe is
> rinsed in ethyl ether the ether combines with the water, and when the
> syringe is injected into the vacuum the ether boils off taking the water
> with it and leaving a dry syringe.
> The waste is a tube attached to our building vacuum (about -0.8 bar) that
> ends in a septa strapped to the end of the normal waste containers. All
> waste injections go to this septa.
> The sequence of events is:
> 1) Rinse syringe twice with 1.0 uL sample and inject to waste vacuum.
> 2) Rinse syringe 10 times with 1.5 uL volume (this is where the syringe is
> in the sample vial and the plunger goes back and forth 10 times in rapid
> succession filling the syringe with 1.0 uL water. This eliminates the air
> bubbles that are often in the syringe, and produces a reliable 50-60%
> volume of the sample bellows to achieve a 4 volt signal on the mass 2
> 3) Inject a 1 uL sample into the H/Device.
> 4) Fill the syringe twice with 8 uL ethyl ether and inject to the waste
> vacuum. It is interesting to note that even before the syringe plunger
> descends, as soon as the syringe needle enters the septa with the waste
> vacuum under it, the ethyl ether disappears as the ethyl ether boils off
> under the vacuum.
> 5) Push syringe plunger back and forth 5 times in the vacuum to get rid of
> any residual ether.
> We always measure the syringe carryover on every analysis run and correct
> for carryover before any other calculations are done as you describe
> below. However, the carryover is always less than 1% and usually 0.2 to
> 0.5%, and does not affect more than the next sample.
> We have found it important to keep replacing the main bottle of ethyl ether
> with a new one every month, otherwise the results seem to get very poor for
> reasons unknown.
> The only disadvantage of the method seems to be that the syringes only work
> for about 150-300 injections. Then on the rinse cycle when the plunger is
> raised up to 8 uL and pushed out the plunger bends and the syringe is
> useless. One can take the syringe out of the autosampler between samples
> and feel the syringe plunger getting harder and harder to move back and
> forth as the number of samples injected increases. A brown residue appears
> in on the glass walls of the syringe. This might be oxides from the metal
> syringe plunger, or organic residue from the water, as we run a large
> number of water samples extracted from plants. Our solution to the problem
> so far is to replace the syringe every 212 injections, two analysis runs or
> two days analysis or 168 unknown samples. The syringes cost US$20 each,
> which is a bearable US$0.12 per sample.
> We did try a SGE #002987 10uL gas tight syringe with Teflon plunger, but it
> cost US$40 and lasted a reliable 400 injections, the same cost as the
> Hamilton per sample. Strangely the amount injected was not reliable, as
> the sample bellow would close to 35-65% with the SGE syringe. Looking at
> it working, there was an air bubble between the plunger and the sample that
> does not disappear completely when pushing the plunger up and down in the
> sample fill the syringe, not surprising since it is a gas tight
> syringe. Therefore we have kept using the Hamilton syringes.
> If anyone else tries this please let me know how it works for a potential
> methods note, especially if they discover how to make the syringes last
> At 08:24 AM 12/8/03 +0100, you wrote:
>> I suspect variable memory to be the cause of the observed raw dD
>> fluctuations. We have been struggling with memory in water analysis and
>> it seems to vary quite a bit. There is changing memory in the syringe we
>> use to inject into the TC/EA sytem. We see the same with the H-device.
>> There is also memory associated with the chromium, the quartz and,
>> probably most importantly, with the surface that the water can interact
>> with before reduction.
>> In our sequence runs we therefore always include a large isotope jump
>> (>100 per mill for hydrogen) and follow the fate of the measurements for
>> about 6 analyses. The memory correction (usually around 2-4 % of a
>> mixture of the preceding measurements) is then modelled (e.g 2% of the
>> previous analysis + 1% of the one before that) to correct the jump in an
>> optimal way. The correction is then applied to all samples.
>> The memory correction is applied prior to a drift correction that can be
>> attributed to several drift sources including depletion of reference gas
>> in the bellows (H-Device).
>> The last step is the scaling to a given difference as mentioned by Paul
>> Regards Willi
>> Paul Brooks wrote:
>>> We have see a similar effect here using a FM H/Device, and I know
>>> another researcher using a Micromass continuous flow system who has a
>>> similar effect.
>>> I assume the results you show below are then normalized to the correct
>>> value of the standards. However, that means that the difference between
>>> them should be the same. Your data shows the difference between the
>>> standards has changed from -281 to -289, an 8 delta difference when I
>>> presume the precision of your system is better than 1 delta unit.
>>> -360 -79 difference -281
>>> -354 -65 difference -289
>>> I assume that most analysts are using a two point calibration for D
>>> analysis as recommended in:
>>> Brand, W. A, T. B. Coplen. 2001. An interlaboratory study to test
>>> instrument performance of hydrogen dual-inlet isotope-ratio mass
>>> spectrometers. Fresenius J. Anal. Chem. 370:358-362.
>>> Doing a two point calibration should take care of any change in each
>>> standard if it is consistent over the course of the analysis run.
>>> However, what seems to happen is that during the course of an analysis
>>> using chromium, if one calibrates with one standard, the other standard
>>> of a different isotope ratio does not always drift or change at the same
>>> rate. I calibrate with a +3.5 standard and then, if analyzing samples
>>> from 0 to -80 which is our usual range, us a -95 standard for the two
>>> point calibration. I was originally using two standards so close
>>> together because memory effects were a problem from the syringe in the
>>> auto-sampler, a problem we have now overcome.
>>> However, just calibrating with the 3.5 the -95 standard would change
>>> during the course of a 23 hour 100 injection analysis, as shown below.
>>> run number 1 2 3 4 5 6
>>> 7 8
>>> beginning of
>>> run -96.3 -95 -96 -96.2 -96.4 -94.9 -96.2 -95
>>> end of run -93.6 -95 -94 -93.7 -93.5 -95.1 -93.8
>>> difference -2.7 0 -2 -2.5 -2.9 0.2 -2.4 0
>>> One can see here that -95 standard has a tendency to be more negative at
>>> the beginning of a run than at the end, but not consistently. I
>>> eventually created a a spreadsheet that fits a curve to both the 3.5 and
>>> -95 standards, and then for every injection does a two point fit between
>>> the 3.5 and -95 curve. Quality controls since I started this procedure
>>> have been excellent, better than plus or minus 0.4 delta units long term
>>> external precision.
>>> I was able to modify my spreadsheet to work with 30 hour analysis runs
>>> that a colleague was doing with a Micromass continuous flow system and
>>> was seeing the same effect. Therefore the effect would not seem to be
>>> from the instrument, but some effect of the chromium.
>>> I would be interested in any other researchers who have seen a similar
>>> Paul Brooks.
>>> At 10:43 AM 12/5/03 -0800, you wrote:
>>>> I use IsoPrime to run dD of water by Cr reduction method. The raw
>>>> values of the working standards sometimes fluctuate day by day. The
>>>> values of -360 and -79 for the first day can be -354 and -65 the next
>>>> day. There is no big change of the machine condition except a slight
>>>> shift of peak center. However, the calculated dD values of the
>>>> repeated samples are perfectly match from day to day. It seems not a
>>>> real big problem. I will feel better if more people telling me they
>>>> have the similar experiences.
>>>> Dachun Zhang
>>> Paul D. Brooks,
>>> Center for Stable Isotope Biogeochemistry,
>>> Dept. Integrative Biology MC3140,
>>> 3060 Valley Life Sciences Building,
>>> UC Berkeley, Ca. 94720-3140.
>>> [log in to unmask]
>>> phone (510)643-1748,
>>> FAX (510)643-1749
>> Willi A. Brand, Stable Isotope Laboratory [log in to unmask]
>> Max-Planck-Institute for Biogeochemistry
>> !!! New !!!
>> Beutenberg Campus
>> Hans-Knoell-Str. 10, 07745 Jena, Germany Tel +49-3641-576400
>> P.O.Box 100164, 07701 Jena, Germany Fax: +49-3641-57-70
> Paul D. Brooks,
> Center for Stable Isotope Biogeochemistry,
> Dept. Integrative Biology MC3140,
> 3060 Valley Life Sciences Building,
> UC Berkeley, Ca. 94720-3140.
> [log in to unmask]
> phone (510)643-1748,
> FAX (510)643-1749
Willi A. Brand, Stable Isotope Laboratory [log in to unmask]
Max-Planck-Institute for Biogeochemistry
!!! New !!!
Hans-Knoell-Str. 10, 07745 Jena, Germany Tel +49-3641-576400
P.O.Box 100164, 07701 Jena, Germany Fax: +49-3641-57-70