and d15N by Elemental Analysis
I] Principles of the Elemental Analyzer
II] Quantitative and Isotope analysis
III] Sample preparation
IV] Sequence Setup (how to set up the sequence based on sample
V] Data processing
I] Principles of the Elemental Analyzer
The Elemental Analyzer (EA) operates by combustion and reduction of
material to produce CO2, N2, H2O, and
SO2 (when set up for sulfur analysis). The general sequence
is like this:
- A solid sample wrapped in a tin (or silver) capsule is dropped
into a combustion reactor at 1000 °C. A pulse of oxygen is timed to
flow through the reactor at the time the sample is dropped.
Combustion products are: CO2, H2O, NxOy,
and SO2. Unreacted O2 from the initial oxygen
pulse is also an effluent.
- The effluent from the combustion reactor then flows through a
copper reduction furnace at 650 °C. Unreacted O2 reacts
with the copper to form CuO and thus is removed from the gas stream.
CO2 and H2O are uneffected by the reduction
furnace whereas the nitrogen oxides are reduced to elemental
- The gas stream then flows through a magnesium perchlorate trap.
Here water forms a hydrate with the Mg(ClO4)2
and is removed from the gas stream. CO2, N2
and SO2 continue on through the system.
- Gases then flow through a molecular sieve packed GC column at 60
°C where they are differentially retained. N2 flows
through the GC column the fastest followed by CO2 and
much later by SO2.
- The effluent from the GC column passes over a TCD (Thermal
Conductivity Detector) before it leaves the EA. In systems which
provide only quantitative results the TCD is used to quantify CHNS.
In our system, the mass spectrometer provides the quantitative data
anyway so we do not use the data feed from the TCD.
- Effluent from the EA is then split and released into an open
split in the Conflo III interface
- The Conflo III is the source of the sample gas and reference gas
for the isotope ratio mass spectrometer.
For good (good means both accurate and precise)
isotope ratio measurements it is essential that all of the carbon and
nitrogen in the sample gets converted to the final analysis gases CO2
and N2. If the combustion is not complete, or is too slow,
then the numbers will not be correct. Additionally, the amount of
nitrogen or carbon required to get good quantitative results is far
lower than that required to get good isotope measurements. For
quantitative analysis samples should have a minimum of 1ug of C or N.
For good isotope ratio measurements we require a minimum of 25ug of N
(gives a 1V peak in isodat) or 12ug of C (gives a 1.5V peak in isodat).
Samples should also not exceed 1.1mg of N or 3mg of C.
Here is an example of what a chromatogram looks like for a typical
organic compound (glycine):
The first two "square" peaks are nitrogen gas from a compressed gas
cylinder (ref gas), the following peak is sample nitrogen. The funny
drop in the blue trace at about 250 seconds is due to the magnet jump
to start monitoring CO2 isotopomers. The CO2 trace occurs
at 282 seconds here and then is followed by two ref gas peaks of CO2
from a compressed gas cylinder. It is important to note that although
we use gases from compressed gas cylinders as a reference point
throughout the sequence, we use real working standards (i.e.
calibrated materials that run through the EA just like every sample)
to rescale the measured isotope ratios to the VPDB and AIR scales. The
square peaks are only necessary for isotope analysis and play no role
in quantitative analysis.
II] Quantitative and Isotope analysis
Why do we need more material for isotope ratios than for
The answer is that one factor that limits the precision (standard
deviation) of the measurement is closely equal to the square root of
the number of counts (ions observed). At lower sample sizes the amount
of carbon or nitrogen that makes it into the ion source simply is not
enough to reduce the precision to the levels that the instrument
itself can usually achieve. Here is an example of what one sees with
varying amounts of a sample containing nitrogen:
Here is the calibration curve:
It's a great calibration curve!
And here is what the isotopes look like for the same set of
The standard deviation for the five measurements with areas
10Vs is 0.12 ‰. The standard deviation of the three points below 10Vs
is about 2 ‰. Note that the instrument will always give a number,
however, to evaluate the accuracy and precision of that number it is
necessary to perform a check like that shown in the plot above.
III] Sample preparation
Some things to think about concerning EA
It is primarily up to the user
to prepare their samples correctly for isotopic analysis. That
being said, we are very happy to discuss what should/can be done
to ensure that the samples submitted for analysis are correct
representations of what the user is after.
Homogeneity: it is important
that the sample be isotopically homogenous prior to analysis.
For soils, this requires drying of the sample at 60°C for 48 hours followed
by use of a ball mill for 10 minutes.
Here are some illustrations of
the importance of sample homogeneity using dirt collected from
outside our building and dried for 48 hours at 60 °C. The dirt was
then placed in a ball mill and aliquots were removed after specific
time segments so that the total milling time is shown in the plot
(Data collected by Dave Baysinger for his honor's thesis)
Note that the dirt we started with had some small pebbles in it as well
as lots of dust. We only weighed out about 1mg of material for the
actual measurement so it is quite easy to get a significant variability
in the amount of carbon in any given sample.
Here is a look at the relative standard deviation (stdev/mean x 100)
from the same set of samples.
(Data collected by Dave Baysinger for his honor's thesis)
The lesson is that it is important to have a homogenous sample in order
to get numbers that actually reflect the overall sample. For a real
sample with similar variability as our dirt, a single measurement after
milling it for only 2 minutes could have come out anywhere between 0.4
and 1.5 wt % C. That's a huge range for this instrument!
Inorganic vs organic carbon.
The carbon in a soil sample may be of organic and inorganic
origins. Analyzing a sample without addition of acid will give a
carbon isotope ratio that reflects a weighted average of the
inorganic and organic carbon isotope ratio of the sample. In
order to determine the isotope ratio of only the organic carbon,
it is necessary to add HCl to drive off carbonates prior to
isotopic analysis. We recommend using 5 times more HCl than is
necessary to neutralize all of the inorganic carbon in the
sample assuming that it is present as a carbonate salt (i.e. you
will need two moles of HCl to neutralize one mole of CO32-).
How much acid to add? It is necessary to
add a minimum of the stoichiometric equivalent of H+ to
neutralize the carbonate. The best is to add an excess of acid (at
least 10 times the stoichiometric equivalent). This is what happens
to the isotope ratio of a sample as carbonate is removed:
(Data collected by Dave
Baysinger for his honor's thesis)
The numbers by each data point give the
ratio of equivalents of acid needed to neutralize all of the
carbonate in the sample. (i.e. at a value of 1.5 enough H+
has been added to completely neutralize 150% of the carbonate that
was originally in the sample)
Sample size: how much can you weigh out?
The EA can efficiently combust samples up
to roughly 50 mg in size with fairly low amounts of C or N.
IV] Sequence Setup (how to set up the sequence
based on sample type)
When using samples that require dilution of
CO2 (ie. you are weighing out large amounts to access the
nitrogen, or you are running organic matter such as plant leaves),
use this sequence
as a template.
Here is an example of a sequence viewed
with Isodat Workspace:
When using samples that do not require
dilution of CO2 (usually this is for samples with very
low weight % C and/or no nitrogen isotope ratios will be measured),
sequence as a template.
V] Data processing
Although it is possible to have all of the
data processing done in an automated fashion, new users will not
understand how these results are obtained. For any data of
significance it is of most value for the owner of the results to
understand how the results were obtained. Consequently, here at the
ASU Keck Foundation Laboratory for Environmental Biogeochemistry we
require users to manually process the results for their analyses. We
provide training and then the tutorial provided here (and elsewhere
for other methods) are intended to help remind you how to get the
results. There are a lot of details that the analyst may think about
when processing results, however, it is very difficult to explain
them all. It is advised that all data sets generated in the lab be
shown to a lab supervisor (Natasha or Stan) for verification prior
Here is the data file used for the
following example of data processing.
Before starting to export your results,
make sure that no analytical sequence is currently running! Isodat
can lose data if exporting results while it is acquiring data.
1) Export your results. In Workspace, find
your data files, highlight them all, then right click:
2) Click on "reprocess" and go ahead and
then click on "add" to add the export template. Choose the
appropriate template (probably there is only one listed), After
selecting the template, change the file name for the exported
spreadsheet to something that describes what the samples are.
Typically, we recommend keeping the analysis date in the file name
as well. Allow 5-10 minutes for Isodat to export your results. Do
not use the computer while it is doing this and NEVER export while a
sequence of samples is running!!!
3) Once the export is done, you should
click on the "results" shortcut on the start bar at the bottom of
the screen and go find your results. Open your spreadsheet and it
should look like this:
4) Right click on the bottom tab and select
"move or copy", then choose "make a copy" and "move to end". Rename
the new worksheet tab "C_1":
5) Sort results by element. To do this,
highlight the "Gasconfiguration" column and choose the "A to Z" sort
6) Copy this worksheet into a new one (use
the tab at the bottom of the page) and call the new worksheet "N_1".
In this new worksheet, delete all of the data with a CO2
gas configuration, then highlight the "Rt" column and select the "Z
to A" sort option. Note that Rt stands for "retention time", which
is the time at which the peak was detected.
7) Delete the columns relating to CO2
from the nitrogen worksheet:
8) Recall from the data file shown at the
top of this page that the first two peaks to be observed are the
square nitrogen reference gas peaks. After them we will see the
sample nitrogen peak. Consequently, by sorting in reverse Rt, the
first data in the spreadsheet are for the sample gas peak and the
later peaks are for the reference gas peaks. Find the point where
the reference gas peaks start to show up and insert 5 lines to
separate them from the sample gas peaks. In the cell just above the
ref gas peaks write in "ref gas peaks":
9) Inspect the columns for the m/z 28
amplitude of the ref gas peaks. They should all be pretty close to
the same value. If you see a significant deviation (perhaps 1 or 2
volts or more) from the mean, then there may be a spike or dip
occuring on that peak. Although it is very rare, this is a quick
quality control check to find such things. Also, the first reference
gas peak is measured against the second one. Although this does add
time to the analysis, it provides an added level of quality control
in that one can see if there are baseline problems causing
significant deviations in the measurements. The ref gas peaks should
all have values that are close to zero. See below:
10) Copy the N_1 worksheet into a new
worksheet and call that one "N_2". In the new worksheet delete all
of the reference gas peak results. Then sort your results by line
11) Delete the extraneous columns (see
12) Copy the "N_2" worksheet to a new
worksheet and call it "N_quant". You will use it to calculate
quantitative results for your samples. In the "N_quant" worksheet,
delete the column for isotope ratios and add the columns shown
13) Sort the results by "identifier 1".
Then group the data (use "cut" and "insert cut cells") as shown
below with the blank capsule first, followed by the linearity
standards (spinach leaves in this case), then any other standards,
and finally the samples.
14) Subtract the "area all" of the tin
capsule blank from all other sample areas. For example, in the sheet
above for the first NIST 2710 standard (cell F27), use the formula:
"=D27-D$2". If this isn't clear, take a look at the formula used in
the attached spreadsheet. In the column for "known ug N" calculate
the weight of nitrogen in the provided standards. The spinach leaf
standard has a composition that is known to be 6.07% N by weight. It
is also 38.85% C by weight. Again, see the formulas in the linked
spreadsheet if you have any difficulties calculating this. Be sure
you understand how this is calculated!
15) Plot the known ug N as a function of
the corrected area for the linearity standards, this is your
calibration curve for N. If you are not familiar with excel, ask
someone to help you with this. Then choose "xy scatter" and go
through the options. make sure the provide a title and label the
axes appropriately. Although it may be tempting to cut corners now,
it is so much easier to figure things out one year later if they are
presented well. See image below for labels and titles. Please,
please, please delete the gridlines and remove the legend.
(Anthony's pet peeve)
16) Click on "finish", then right click on
a data point in the plot and select "add trendline". Choose the
"linear" trendline then go to options and select "set intercept to
0", "show equation", and "show R squared".
17) The slope of the line shown here tells
us how many micrograms of nitrogen produce the observed corrected
area. We will be multiplying it by the corrected area for the
samples (and standards) so it is important that we are not limited
in significant figures. In the plot shown above, the slope was
only provided with two significant figures, yet we are going to
multiply it by the corrected area, which is measured with much
better precision. To avoid limiting our final results, right click
on the equation for the line and choose "format label", then go to
"number" and choose "number" with 5 decimal values.
This changes the equation to y=0.88997x
18) Calculate the "measured ug N" by
multiplying the line slope by the sample (and standard) corrected
area. Finally, for the standards, calculate the relative error of
the measurement: =(known-measured)/known x 100
19) Plot the relative error for the
linearity standards (Spinach leaves) as a function of the measured
ug N. As extra credit, you can add extra data series to the plot
showing where your other standards fall on the plot. This one is
The plot above shows that the relative error of
your measurements (for the standards at least) gets bigger for
smaller amounts of nitrogen in the sample. This is to be expected
and we also see that the relative error of the glycine and NIST 2710
measurements fall on this curve (this part is a bit lucky, they
could just as well be biased in the opposite direction). The value
of this plot is to allow you to estimate the relative error of the
measurement of your sample (we are talking accuracy here and not
precision). For example, if the sample you are concerned with had a
measured weight of N of 25ug, then you can guess that the relative
error in that measurement is around 15%. For larger samples, it
might be less. In general we suggest that you apply an error
estimate of 10% of the measured value (so a measurement of sample
composition as 1.0% N is really 0.9 to 1.1% N by weight) but for
small amounts of N you should use the relative error plot to
estimate the error for your sample. Note that it is very difficult
to estimate the error of a sample measurement when the sample has
less nitrogen than the lowest standard that was run. In such cases,
it is better to weigh out as much sample as possible and rerun it
20) Copy the headers from the top of the
page and paste them above your sample results. You can delete the "rel
error" header for your samples and instead label that column as "wt
% N". If you are measuring samples collected on filters then this
may not be necessary, but for anything else, calculate the wt % N
for your samples. Don't forget to convert from ug N to mg N in the
You now have your quantitative results for
nitrogen, now on to carbon:
21) Copy the "N_2" worksheet to a new
worksheet called "N_isotope" and add the column headers shown below:
22) Plot the d15N/14N as a function of the
Ampl 28 (or of the area, it doesn't make much difference). Label
your title and axes as shown.
Here's a close-up of the plot:
The above plot shows what looks like very
stable isotope ratios from about 2V to 23V. There is a drop around
1V to 2V, and in particular, the first data point, at 0.3V, is way
off from the rest of the data set. Ideally this plot would show
identical isotope ratios for all peak amplitudes. The phenomenon
observed above is fairly typical. At some point when the signal gets
too low we can no longer make good isotope ratio measurements. The
value of this plot is to inform us as to where that happens and also
to allow us to make some corrections for any effects related to
sample size (linearity effects). From this data, we see that the
first point is too low in signal for a good measurement so we will
drop it from the plot and throw out results for any samples that
have peak amplitudes below 842mV (the amplitude of the second
23) Remake the plot without the first
point. You may need to rescale things so that you really fill in the
plot area with the data points:
With the plot expanded and the first data
point dropped, we see that the isotope ratio measured for the rest
of the linearity standards covers a range of about 0.6 to 0.7 ‰ over
a range of amplitudes covering 22 V. The standard deviation for the
data points observed above is 0.23 ‰, which is not bad at all. If we
fit a line to this data, then we can make a small correction for the
linearity effect we see over this range. (Note that there are more
complex ways to go about making linearity corrections, but a line
fit for this data set will be good enough and easier to explain)
Finally, since we have observed that the
isotope ratio measurement behaves poorly below 842 mV we must go
through the sample data set and flag every sample that has a
nitrogen m/z 28 amplitude of less than 842 mV. We must flag those
samples so that we know that the numbers are not reliable for them.
24) Add a trendline to this plot and choose
"show equation" and "R-squared". The correlation coefficient does
not have to be good. In general it is lower when the data is better
behaved (i.e. no change in isotope ratio with sample size).
This plot shows that as you increase sample
size, there is a change in the measured isotope ratio. Clearly for
larger sample sizes this effect disappears, but it is more
pronounced for smaller sample sizes. Our attempt is to slightly
improve the data set, not to make it perfect. The intercept value
for this equation is dependent on the specific sample, so we will
not use the intercept in the linearity correction. The appropriate
formula for the linearity correction is: CV = OV - (M x A), where CV
is the Corrected value, OV is the Old value, M is the slope of the
linearity line, and A is the amplitude of the peak for the sample.
For more clarity, see the spreadsheet provided with this tutorial.
25) Calculate the linearity corrected value
for all samples and standards. In addition, insert three lines
between the last linearity standard and the next line. Calculate the
mean and standard deviation for the linearity standards that were
used for the linearity correction. Now use the known values of the
isotope ratios for the standards (click
here for a quick list) to fill in the "known d15N" for
In the above plot we see that the standard
deviation for the linearity corrected linearity standards has gotten
smaller. This will ALWAYS be the case. At worst, the value will not
change. If the linearity corrected standard deviation increases then
you made a mistake in writing the equation or something else.
26) Plot the known d15N of the high and low
standards as a function of the linearity corrected d15N. Be sure to
label title and axes correctly as shown below:
Note: make sure that excel identifies the
data as "columns" and not "rows" in the initial plot preview. If
your normalization lines look like two data sets instead of one,
then excel probably misidentified the data series. Try again and ask
for help if you need it.
27) Clean up the plot a bit and add a
trendline: Right click on the chart area (to the left of the title
for example) and select "font" and turn off the autoscaling. Make
sure the font size is 10 point. Add a trendline (right click on a
data point to do this) and choose linear and "show equation".
28) Shrink the plot, make sure that you
have 5 significant figures in the equation (right click on the
equation text box and choose format -> number -> 4 or 5 decimals as
appropriate (or scientific with 4 decimals). Make the plot fairly
small and place it as shown below. You will be stacking plots like
this so smaller is easier to work with, you just need to see the
29) Use the normalization equation to
convert all samples measured between the two sets of high/low
standards to the international scale (AIR for nitrogen and VPDB for
Be sure to flag any problematic numbers
such as those for samples with low intensities.
30) Calculate the mean and standard
deviation for the linearity check standard. Copy the plot and paste
it next to the next set of "sandwiched" samples. Right click on the
plot and choose "source data". Now highlight the new set of data
points for this section of samples:
If this is done correctly, the equation
will update and the slope and intercept of the new line should be
very close to those in the previous plot. Rename this plot
"normalization 2" and repeat for the entire data set:
Be sure to flag any results which are known
to be incorrect.
Repeat all of the steps above starting at
step 6, but this time you are after the carbon results so remove the
nitrogen numbers from the C_1 worksheet and go from there. Note that
when you sort the results by retention time (Rt) you should choose
"A to Z" so that the sample peaks show first and the reference gas
peaks show later. Because we set the sensitivity for peak detection
rather low, one will sometimes see peak tails integrated as well.
You can tell these by their low amplitude (ampl 44) and their
retention time which will be somewhere in between that of the major
sample peak and that of the next reference gas peak. The example
data set provided with this tutorial has the steps and results shown
for Carbon. When you are done with Carbon, then move on to the next
More to come soon....
Page last updated: June 27, 2007