I have since similar problems over the years both for %N and C and for
delta N and delta C.  I have observed this on both the a 5 year old Europa
Sci. 2020 with SL analyzer (a CN analyzer very siimilar to the Carlo Erba
systems), and to an extent with a new FM Delta Plus with an 10 year old
Calro Erba NA 1500.

The Calro Erba FM combination was linear for reference gas and solid
samples for delta values (over 60 ug N, see below), but not for % N or
%C.  I ended up running 10 variable weight standards over any run I do, we
generally run 131 samples in a batch.  I then use a correction separately
in a spread sheet to calculate the % values accurately.   Below 60 ug of N
in the tin capsules, the solid standards have a progressively lower and
lower delta value.  This can be corrected for by standards by methods
explained in:-

Fry, B., et al. 1992.  Anal. Chem. 64, 288-291.

and

Finnegan MAT Flash Report No. G29.

Note the later application report reports very good results for samples
less than 10 ug N per tin capsule.  However, we have not been able to
duplicate this with our system and a service engineer told me that this
method used a smaller than usual combustion tube, GC column, etc. to
measure the very low N.

On the Europa system, on which I have probably run 80-100,000 samples,  I
got good linearity when the system was new.  However, gradually the
linearity declined, and the source had to be re-tuned to get the system
linear, then cleaned, then the collectors rebuilt.  Eventually I developed
a method where I always run the 10 variable weight standards and correct
for any non-linearity in either the delta results or the % results, and
check these against 2 other different standards. I have observed that all
of the following seem to have an effect on linearity:-

MS sources of problems:
1) ion source tuning.
2) replacing the filament.
3) cleaning the source
4) having the faraday cups re-coated with DAG (had to be returned to
factory for the work).

All these problems are straightforward and can be tested with reference gas
injection.

On the EA I have noticed the following that might affect linearity:
1) Insufficient separation of N and C peaks, and/or inappropriate
integration of peaks.
2) Too much ash in the combustion furnace.
3) Insufficient O2 for combustion.
4) The EA is aging and just isn't linear anymore, for unknown other reasons..

The last item is hard to quantify, but does appear to happen.  I have heard
rumors that both Analytical Precision Products and Elemental Micro Analysis
Ltd., will work on an EA to remove some of these linearity problems, but
have not confirmed them.  Does anyone know if this it true, and what is
commonly repaired on these EA's?

I think any analyst has simply got to be vigilant and test for linearity on
every run, make small corrections if necessary, testing the correction with
a standard other than the one used to correct for the non-linearity.

The only paper I know of that reports correction for linearity is:

Ohlsson, K.E.A. and Wallmark, P.H., Analyst 124 (1999) 571-577.

Hope this helps,

Paul.





At 12:03 PM 10/17/00 -0600, you wrote:
>Our EA (CE Elantech)-isoprime has been giving us sample size 13C linearity
>trouble too, but in the opposite direction.  I ran a cellulose standard in
>varying sizes (5 - 20 micrograms) and found the opposite relationship and
>worse than your problem (although the magnitude of the problem could just
>be a result of pushing the sample sizes to the extreme for our set up):
>our largest sample (1.2 e-8 major beam) was 8 per mil LIGHTER than our
>smallest (1.8 e-9 major beam), with the expected value being just about in
>the middle of that range. The relationship is not linear across all sample
>sizes/beam sizes, however.  There appears to be a plateau/reduced slope
>(but same direction of change) in the middle of that range, with the
>larger sample on the plateau (9.7e-9) being 1.2 per mil lighter than the
>smaller one (3.4e-9).  The ref gas beam size was 6.8 e-9 for this test.
>
>Given that our symptoms are opposite in sign, my problem may be very
>different than yours, but perhaps not entirely unrelated.  You say that
>your problem is an ea problem.  I am not sure if mine is an ea problem or
>a mass spec problem or a bit of both.  The mass spec as seen through other
>prep systems (all gas samples) attached to it is functioning fine, but
>those systems are controlling sample sizes quite closely.
>
>I am interested in any input on ways to reduce the effect of sample size
>on delta 13C (other than the obvious of controlling sample size) and
>whether the sign of the slope (positive or negative change in 13C with
>sample size) of the linearity problem can be used to diagnose the source
>of the problem.  Are there any references on this topic?
>
>Thank you,
>Francesca
>
>
> >I've been having a few problems with our EA for the last few weeks. Here
> >are the details:
> >It's a Carlo Erba NC 2500 connected up to a Micromass Optima. The Optima is
> >working fine for dual inlet work, but the EA has not been used for about 18
> >months.
> >Linearity tests for N2 and CO2 reference gases give a range of 0.015-0.020
> >permil over the full range of refgas sizes, i.e. there is nothing wrong
> >with the mass spec. A run of fifteen solid organic standards of widely
> >differing sizes (equivalent to 0.05-0.75mg N) gives a scatter of d15N
> >values with a standard deviation of 0.14 permil - in other words nitrogen
> >works OK.
> >However, the situation for carbon is problematic. A similar run of
> >standards, (attempted organic standards as well as graphite) gives a range
> >of d13C values, with the largest samples being 1.8 permil heavier than the
> >smallest samples.
> >I assumed that the problem was probably related to poor combustion - there
> >is a good flash combustion - sample drop time is 21 seconds, not dissimilar
> >from the recommended values. He pressure/flow is 150 kPa,  88 ml/min. O2
> >pressure/flow is 60 kPa / 20 ml/min - I have had the O2 up to 150 kPa / 55
> >ml/min with no effect on the d13C/sample size variation. I've also had the
> >GC oven up to 110°C for four hours to rid the column of any crud. The water
> >trap and furnace tubes have new chemicals and there are no major leaks.
> >Also there is a good (R2 = 0.999) linear relationship between the
> >integrated area of mass 44 and the mass of the sample.
> >I've also fiddled with the way the peaks are integrated, with no success.
> >Any ideas as to what could cause the d13C/samplesize variation, and what I
> >can do about it? As I say, it is evidently an EA problem, not the mass
> spec.
> >
> >cheers,
> >
> >Jason
> >
> >
> >*********************************
> >Dr. Jason Newton
> >University of California Santa Cruz
> >Department of Earth Sciences
> >Santa Cruz
> >CA 95064
> >USA
> >
> >email: [log in to unmask]
> >Tel: +1  (831) 459 5751
> >Fax: +1  (831) 459 3074
>
>
>
>~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>Francesca Smith
>
>Current address:                Department affiliation:
>
>INSTAAR                         Department of Geophysical Sciences
>Campus Box 450                  University of Chicago
>Boulder, CO 80309-0450          5734 S. Ellis Ave
>                                 Chicago, IL 60637
>phone: (303) 492-7808
>fax: (303) 492-6388
>
>email: [log in to unmask]
></blockquote></x-html>




Paul D. Brooks,
Dept. ESPM-ES,
147 Higard Hall MC 3110,
UC Berkeley, Ca. 94720-3110.

phone (510)643-1748,
FAX (510)643-5098.