d¹³C and d¹⁸O of Carbonates

I] Overview

II] Standards

III] Isodat setup

IV] Some method variables investigated

V] Data processing

I] Overview

We use the multiprep heating block and the gasbench to measure d¹³C and d¹⁸O of carbonates. in brief, here is the setup:

1) Aliquots of samples and working standards are weighed onto either small sheets of weighing paper or into tin capsules (used for the EA) and are then poured into 12 mL borosilicate Labco Exetainer vials (the ones with the round bottoms).

2) Vials are flushed on the multiprep unit with helium at 100mL/min for 8 minutes at 70°C. Note that the sample vials get pressurized from this process.

3) Pure phosphoric acid is added manually (amount of acid is generally about 10-15 drops) to each vial. Note that we mark the caps of vials to indicate where the acid injection was made. This helps to ensure that the sample needle does not aspirate phosphoric acid into the gasbench during analysis.

4) Vials are allowed to equilibrate for a predetermined amount of time at 70°C. Currently, we equilibrate pure calcites for 1 hour. Bone and enamel are given 6 hours to equilibrate at this temperature.

5) Vials are then sampled with the gasbench. We make 11 sample injections (see more on the isodat method below) and retain all peaks that elute after 400 seconds. We have noticed that in general the data quality of the first peaks from the sample are not as stable as for the rest.

6) Results are exported to a spreadsheet. If standards show no sign of drift then a single normalization is performed on the entire data set.

7) Final isotope ratios are reported with respect to VPDB.

Here is what a chromatogram looks like for about 2 mg of a bone sample:

Here is a chromatogram for about 0.350 mg of NBS-18:

As the headspace is flushed the concentration of CO₂ in the vial decreases. Hence each additional peak is smaller than the preceding one.

II] Standards

We still have not completed the preparation of a variety of laboratory working standards, so we do use NBS-18 and NBS-19 at this time. They will be phased out in favor of larger batches of laboratory working standards when we finish preparing them. Here are the standards that we currently use:

III] Isodat Setup

We have the gasbench set up with the helium pressure at 15 psi. In addition, the helium tank delivery pressure is set to 40 psi so as to maintain a 100 mL/min flow rate at the flush lines. The helium flush is performed with dual flush needles (flush two vials at a time) for 8 minutes.

Here is what the time events look like for the method we use:

At the moment we have not developed a suite of laboratory working standards. We have one bone standard and one marble standard as well as NBS-18 and NBS-19, which we use for our analyses. We are working on obtaining two more carbonate working standards that span a reasonably wide range. Additionally, we are ordering the IAEA standard LSVEC as it has a fairly low certified value for d13C, which we need to bracket bone and enamel samples. We also prepare a linearity check with one of our working standards. Below is an example of a sample sequence we use:

For bone samples, we have found that a 6 hour reaction time is needed to ensure complete reaction/equilibration with the phosphoric acid. The beginning of the above sequence simply runs six one-hour delays prior to starting sample analysis. The sequence starts and ends with NBS-18 and NBS-19, and intersperses them occasionally to compensate for minor instrument drift. In addition, the llama leg bone standard is prepared with a variety of sample sizes to measure (and correct for) any linearity effect on the instrument. The Beijing standard is a marble standard that is run as a laboratory check. If any drift corrections are needed (rare) then we can evaluate their effect on this standard. The autosampler position number goes along columns in the multiprep unit because we initially set this method up to use an automated acid pump. However, the finer diameter line of the acid pump easily plugs at our laboratory temperature (22°C). We would like to try to make the acid pump work for us, but it has been a low priority task. We are also very aware of the fact that many labs with talented staff have given up on the acid pump in favor of a manual acid addition so that has also discouraged us from spending much time on this problem. The manual acid addition works fine and is pretty easy.

IV] Variables

It important to check how long it will take for the CO2 from the sample to completely evolve and for the isotope ratios to stabilize. We checked the reaction time between 50 and 80 °C and found that for all of these temperatures one hour was sufficient for our marble standard (data shown below). However, when a similar experiment was performed for a bone standard, we found that a 6 hour reaction time was preferred.

Equilibration time

Marble

Bone

Here are the results of a test of required reaction time for bone at 70 °C. Note that although there appears to be a sudden drop in d¹³C after about 200 minutes, the range of measured values for the entire data set has a standard deviation of about 0.1 ‰. For oxygen isotopes, there is one point around 100 minutes where there is a significant deviation from the rest of the samples. Excluding that point, the standard deviation for the rest of the samples over the entire range of reaction times is about 0.15 ‰. Allowing a 6 hour reaction time here simply ensures a maximized signal but not necessarily a significant improvement in the accuracy of the measured isotope ratio of the sample.

Temperature

Here is a summary of the effect of reaction temperature on the measured isotope ratio of sample relative to the tank gas.

The reaction temperature did not significantly change the measured d¹³C for our carbonate working standard. However, there was a significant effect on the measured d¹⁸O of the working standard, as seen in the right plot above. The measured oxygen isotope ratio for the carbonate changed by about 1 ‰ over a 30 °C range. Hence, we can assume a variability in the value of about 0.03 ‰/°C. Since the multiprep heating block is stable to about 0.1 °C it does not seem that temperature stability of the heating block will be much of a concern for routine operations. (That is not the case for isotope ratio measurements of waters by equilibration with the multiprep.)

V] Data Processing

The following instructions give a detailed step-by-step approach to the data processing we do. A few things to note: we are still trying to obtain laboratory working standards that cover a range of isotope ratios for normalization of the data. Until that time, we will use NBS-18, NBS-19, and at least the "beijing standard" that we currently have for data processing. The following data set is for bone carbonates, hence the llama leg working standard is also used for a linearity and quality check.

1) Exporting the raw data:

In Workspace, find your data folder, sort the results by date created, highlight all of the data files of interest, right click in the highlighted area and select "reprocess".

In the reprocessing window, select "add template" and  choose the "C and O" template. Then rename the filename of the export spreadsheet. It is recommended to leave the date in the file name, and then add a portion of the name with the same name as the data results folder. Now click OK and wait for isodat to finish exporting the data (this will take about 10 minutes for 80 files).

The data presented below is all in this file. Download it and try processing the data on your own to get the idea of what is involved in processing the results.

2) Go to the results folder and find your data. Open the excel file with the exported results. (Download the raw data for this demonstration here)

3) Before doing anything else to the data, right click on the worksheet tab (bottom of page) and select "move or copy". In the window, select "move to end" and check the "make a copy" box. Rename the copied worksheet "1". Every time you make several changes to the worksheet, you should make a new copy of it and increment its name by one. This will make it enormously easier to find errors in the data processing later on, should there be any problems.

4) In worksheet "1", highlight the "isref" column and sort the data by this value (click on the "A to Z" sort button).

5) Scroll down to the point where the isotope ratios (values in far right columns) all start reading "0". This is where the reference gas peaks have been sorted. Insert 5 lines to separate the reference gas peaks from the sample peaks.

6) Scan through the column with the peak heights to see that they are all about the same. You are looking for any obvious deviations, usually more than half a volt more or less than the normal value. If you find such a deviation, open the data file corresponding to this file and try to see what happened. Although very rare, we have seen surges or dips in the reference gas peaks. This step is just a quick check for any oddities that might result in a bogus measurement for the sample or standard isotope ratio.

7) Insert a column after the "ampl 44" column  and label it "mean ampl 44". Insert two columns after the d13C column and call them "mean d13C" and stdev. After the d18O column, add headers to the following two columns: "mean d18O" and "stdev" (see picture below). Calculate the average peak amplitude for the run in the "mean ampl 44" column.

8) Calculate the mean d13C value and the standard deviation only for peaks that elute after 400 seconds. We have noticed that the peaks eluting before 400 seconds are not as stable as those eluting later. The reason for this is not clear so for the time being we only use the more reliable peaks occurring after 400 seconds. Do this also for d18O and its standard deviation. Be sure to set the displayed precision in the cells to two digits past the decimal. At the moment this is more of an esthetic thing, but later on we will want to be sure to display a realistic precision in the final results.

9) Scan the calculated standard deviations for d13C and d18O and flag any that are greater than 0.2 ‰. Note that the first "linearity check" standards will have a very low signal, which will result in poor precision so do not be shocked if these are high. For any samples or standards that have high standard deviations (use 0.2 permil as your cutoff between good and bad standard deviations), look through the results and try to see if there is one particular peak that is throwing the results off. If so, then go back to the original data file to see if the baseline was correctly determined for the peak in question. In general, if there is a d18O value that deviates significantly but the d13C value is not so significantly changed, then there is probably a noise spike on the m/z 46 trace affecting the baseline determination for that trace. If you see a poor d13C peak, but the d18O data looks good, then the faulty baseline is most likely with the m/z 45 trace. If both are very bad then there may be two spikes, or all traces may be bad. Regardless, you should find a clear and obvious reason for why the value deviated as it did (and consequently is not representative of the sample) so that you can drop that one value from the integration. If you cannot find a reason for the deviation then you are obligated to keep the data point. You should speak to the lab manager to see if he/she can help to figure out what may have gone wrong. If all of the peaks seem to vary a lot, look at the mean peak intensity. If it is relatively low (say, below 2 volts) then poor precision can be expected due to the sample size. In general, when you see seemingly odd results, you should check with the lab manager to see if they can help.

Here is an example of a data file in which a dip in the m/z 46 trace resulted in an incorrect determination of the baseline for that trace. If you look at the columns for the results, the highlighted one shows not significant difference in the determination of d¹³C for this peak, however, the measurement for d¹⁸O is high by about 0.4 to 0.5 ‰ as one would expect for a baseline that artificially increases the peak are for a single trace.

10) Now copy the worksheet to a new worksheet and call that one "3". In the new worksheet, highlight all of the data, copy it, and then select edit > paste special > values. From here, sort all of the data by the mean d13C so that you can remove all of the extra data.

11) Sort the data by "line" (column A). Delete the columns labelled "Ampl 44", "d 13C/12C", "stdev", and "d 18O/16O". Be sure that you do NOT delete the columns with headers starting with the word "mean". See the picture below for what you want:

12) For the linearity standards (the llama standards in this case), plot "d13C" as a function of "mean ampl 44" (highlight the data, click on the chart wizard, and select "xy scatter"), do the same with the d18O. As a good habit, add the title and label the axes. It is recommended to remove the gridlines and legend as well.

The linearity plots below show that for mean peak amplitudes below about 2 volts there is a significant change in the measured isotope ratio. It is best to prepare samples so that the measured peaks are within a less variable part of the linearity range.

For the data set here, all sample peaks were above 2 volts, so we can check the linearity in the range that is above 2 volts, as shown in the second set of plots below where the first two linearity points were dropped. Frequently a line correction is sufficient, but sometimes a quadratic fit is much better, as in the case for d13C below.

13) The linearity plots show that the measured isotope ratio varies as a function of the peak amplitude. There are two solutions to the problem of measurement linearity: prepare all samples and standards so that they all have the same peak height (this is the ideal but frequently impractical solution), or subtract out the linearity effect by using a best line or polynomial fit. To correct for the linearity, insert a few columns in between the d13C and d18O columns as shown below. Title the new column "linearity corrected d13C" and do the same for d18O. Now, subtract out the amplitude-dependent linearity effect from the measured value. Do not use the intercept from the line or polynomial equation (see the formula used in the sheet below).

The formula in cell G2 above is: "=F2-(8.9429e-9*E2^2-1.4332e-4*E2)". Note that we put the fomula for the linearity curve in parentheses to avoid any sign errors. You always subtract off the effect. Also note that the intercept with the y-axis is not used in the correction as it is only relevant to the particular sample used for the linearity correction.

Where F2 is the "mean d13C" and E2 is the "mean ampl 44". It is important to note that the linearity correction can only improve the precision for the linearity standards. So... to ensure that the correction was applied correctly, check that the standard deviation for the linearity correction points used improves (see next figure).

Note that without a linearity correction, the standard deviation for d¹³C and d¹⁸O of the 5 lama standards above 2 volts is 0.13 and 0.16 ‰ respectively. That is quite good precision for samples with peak heights varying over a 10 volt span. By using the linearity correction, we hope to make a relatively small (about 0.1 ‰) gain in precision and accuracy of our measurements. Notice that the standard deviation for each measurement has dropped to 0.03 ‰ with the linearity correction. Although this precision seems fantastic, it is artificial in that it is the result of fitting a curve through these data. Consequently, this standard deviation does not reflect upon the analytical uncertainty for sample measurements. The best measurement of sample uncertainty is to perform multiple measurements of the same samples. In general, the user should have a minimum of one set of triplicate runs of a sample in any given sequence. The more replicates, the better one can estimate the uncertainty of the final value assigned to each sample.

14) Normalizing the results: Copy worksheet "3" into a new worksheet and call it "4". Delete the plots for the linearity correction and create the columns shown below. Add the "known" values for the standards. There are two standards that will be used to normalize the results. In this case it is NBS-18 and NBS-19. These are great for our d¹⁸O measurement but less than great for our d¹³C measurements as they do not span a wide range of d13C values. We have ordered the LSVEC standard from IAEA which will solve that problem for us.

15) Plot (use xy scatter plot) the "known" values as a function of the linearity corrected ones for the normalization standards (the NBS-18 and NBS-19 in this case). Only use the high/low standards dedicated for this measurement. The other standards will be measured as checks. Be sure to get rid of the gridlines and legend as well as add in a title and label the axes as shown below.

Continue to normalize all of the oxygen results as well:

16) Copy worksheet 4 into a new worksheet and call the new worksheet "results". In the new worksheet, cilck on the empty box to the left of column A and just above line 1 to highlight all of the contents of the worksheet. Now do Paste Special > Values in order to remove the equations used for normalization. Delete all of the plots in the "results" worksheet.

17) From here, we want to sort all results by identifier 1, then separate out the standard and samples. To prepare for this, delete the spacer column between the C and O results (column J in the figure above). Also, copy the name of the linearity check standard down into the two columns below it so that the associated mean and standard deviation will have a value for identifier 1 (see figure below). Now highlight the column for identifier 1 and sort, using the "A to Z" option.

18) Insert 10 lines at the top of the worksheet and prepare  column headers shown in the figure below. Although we report the standard deviation for the linearity check standard, it is only to see how well the linearity fit was. This value should not be misconstrued to reflect the expected precision for the samples as it is the result of a best fit line. The best test of sample uncertainty is to run replicates. In the results summary below, the standard deviation for the triplicate runs of "spl-29" would be used as the best estimate of uncertainty for the rest of the samples that were run a single time.

Be sure to paste the values for the means and standard deviations from the standards into the appropriate cells shown below. If you copy a formula into the cell you will lose the value when you delete the extra data.

19) Clean up the results data to show only the carbon and oxygen isotope ratios as well as the analytical uncertainty estimated from the triplicate run of spl-29. Delete the extra columns and data until you have a nice clean summary like the one shown below. The values measured for the check standards (beijing and llama in this case) will give an estimate of the sample accuracy. The standard deviations obtained for the replicate sample measurements will give a better idea of the precision of the measured value for the samples.

Aside: In the data file used for this demonstration, the measured d¹⁸O for the NBS-18 standard was unusually variable (spanned a range of about 0.8 ‰ through the run). Although the precision for the oxygen isotope measurements is still reasonably good, better precision would have been obtained if the standard had been more homogenous.

Page last updated: May 14, 2007