Measurement Uncertainty in Elemental Analysis
Lenny Lin
Introduction
I read a literature, An International Study Evaluating Elemental Analysis [1]. Due to the availability of the raw data file, I wrote this report to present my observation and provide my comments.
Th literature stated that, For elemental analysis, the majority of journals require ±0.4% of each value to confirm sufficient purity for publication, and that the ±0.4% guideline for journals would actually require that some samples be 99.6% pure, without factoring in the error associated with measurement and what the trace impurities are. Another study [2] supported their arguement with statistical study that the deviation was quite low among measurements but was still greater than the typical deviation reported in a survey of results from the literature. They suggested that many results in the literature appear to be too precise, thus raising questions about the integrity of elemental analysis data in the literature at large. They suggested that an element-agnostic deviation of 0.4% based on single replicates is not a statistically realistic journal requirement for synthetic samples.
Experiment and Data
In the literature, samples of the 5 compounds (Table I) were sent to 18 laboratories. Based on the statistical analysis of the returns on C, H, and N elemental analysis.|
Compound |
C (%) |
H (%) |
N (%) |
|
2-hydroxybenzimidazole |
62.39 (62.68) |
4.44 (4.51) |
20.79 (20.88) |
|
Bisoctrizole |
74.68 (74.74) |
7.67 (7.65) |
12.54 (12.76) |
|
Diacetyle Pyridine |
66.23 (66.25) |
5.50 (5.56) |
8.58 (8.58) |
|
DL-tryptophan |
64.66 (64.69) |
5.91 (5.92) |
13.62 (13.72) |
|
Succinimide |
48.36 (48.49) |
5.02 (5.09) |
13.97 (14.14) |
I reorganized the provided raw data and presented them in the following table (Table I). The column “Diff (%)” is the output of analysis value minus theoretical value.
Figure I: Raw data
Note: The red dashed lines are ±0.4%
guidelines. The outliers which are outside of guidelines are represented
with red dots.
Statistical Analysis
All data were presented on box-plots.
Box-plots of C (%)
Figure II: Box-plot of C (%)
Note: The red dashed lines
are ±0.4% guidelines. The outliers which are outside of guidelines are
represented with red dots.
There are 146 C (%) data. There are 24 outliers. The ratio is
16.4%.
- All results from four out of 18 labs were outliers for C(%) of DL-Tryptophan. The ratio of “failed” labs is 22%.
- Six out of 18 labs reported outliers for C(%) of all five compounds. The ratio of “failed” labs is 33%.
Box-plots of H (%)
Figure III: Box-plot of H (%)
Note: The red dashed lines
are ±0.4% guidelines. The outliers which are outside of guidelines are
represented with red dots.
There are 146 H (%) data. There are 3 outlier. The ratio is
2.1%.
- The result from Numega Lab was an outlier for H (%) of Succinimide.
- The result from Guelph Chemicals and Midwest Microlab was an outlier for H (%) of Bisoctrizole.
Box-plots of N (%)
Figure IV: Box-plot of N (%)
Note: The red dashed lines
are ±0.4% guidelines. The outliers which are outside of guidelines are
represented with red dots.
There are 144 N (%) data. There are 20 outlier. The ratio is
13.9%.
- All results from KU Leuven and UQ were outliers for N (%) of Succinimide
- All results from KU Leuven were outliers for N (%) of Bisoctrizole.
- All results from Guelph Chemicals and Midwest Microlab were outliers for N (%) of Diacetyl Pyridine.
Comments
I suggest taking into consideration of the following two aspects.It is the guideline for all three elements which causes more outliers in analysis of C (%) than those for H (%) and N (%). It would be appropriate to set up one guideline for each element.
Few labs failed all testings for N(%) of a couple of compounds. A round-robin testing program would help labs to perform analysis consistently.