Quantitative Systematic Errors in Gas Chromatography

are unfortunately often quite large. There are several reasons, some of them will be discussed here. One of them is REGULATION and missing INFORMATION but this differs globally in wide ranges and cannot be discussed here.

NOTE: In regulated GC methods about quantitative analyses there is normally no any hint given how to check for systematic quantity errors.

The main error sources are:

1. Taking and giving the sample not correct - see Sampling / Calibration.

2. Incorrect integration - especially false data density, false integration start and stop because of poor peak detection, false base line positioning. There are further error sources in quite some - even latest commercial - integration software like missing signal size analysis and missing or poor automatic switching of data density with growing retention time; see under “integration”.

3. Missing quantity control by statistics - see Statistics.

4. Missing quality control of sampling tools. Simply a syringe or a sampling valve, which has strongly sorbing substances on the inner surface, can easily be displaced by the just selected sample with a differing chemical composition and thus falsifies the sample. This is especially a problem in court cases about the source for impurities or non kept limits for such substances.

NOTE: If for analysis we use the same syringe or sampling valve as we used some runs before in calibration runs we may find serious sample falsification due to the always and everywhere acting displacement chromatography. Example: Chemists / analysts knowing well the high polarity of ethylene glycols will very probably clean a syringe or a sampling valve by water in order to remove the water soluble glycols. But their sorption strength is so high, that water fails for a complete and immediate clean up. If the next analytical sample has a composition water plus ethanol plus organic acids, this mix sorbs stronger than the ethylene glycols and displaces still sorbed glycols. If in trace analyses the correct glycol content counts, well reproducible data falsification of several hundred relative percent may happen falsifying also court case decisions. Chromatography acts everywhere, not only in chromatography instruments.

5. Differing substances have differing specific detector sensitivities. Thus “area-% data” differ in many cases dramatically from weight-% values. This is even true with data based on the quantitatively wide ranging flame ionization detector and a sample composition exclusively with hydrocarbons. Specific calibration is mandatory. See under Calibration. The same is true with the heat conductivity detector although at least in gas analysis it behaves quite “democratic”, that means, produces at least quite correct mole-% data without the use of substance specific correction factors. But its linearity is poor.

6. Each detector has a limited quantitative working range, which at best remains under control when the peak height is measured in standard units like volt or ampere. Thus the (correctly taken) peak height acts as alarm value telling the chromatographer, that there is no more any way to correct false data but by the combination of two quantitative methods, elegantly possible using the IfC EXPORT concept.

7. Each column has a limited working range for the nature of substances: their molecular weight, their polarity, their temperature stability. As the majority of all quantitative GC analyses is done by temperature programming we have working range limits for the low starting temperature, the upper end of the temperature program and the heating rate. For isocratic GC the selection of the column temperature is limited by working ranges. This is so important that this site shows a series of figures about the temperature working range of a GC column / capillary and how to get the values.

8. All quantity values are false for sure, if the correct identification of one compound in the sample mix failed.

9. The quantity data may be wrong by thousands of percent, if one main compound in the sample could not be quantitized, for which we know two reasons: this main compound is not detected because the detector is specifically “blind” for this substance (example water and flame ionization detector), or the main compound is not eluted, or the chromatogram has been finished too early, or the backflush
(if used) remains blind, or the highest concentration of the sample (given in gram per second for a FID as example or in gr/ml carrier gas for a HCD as example) is outside the detector working range
(see topic 6. above) .

10. Data may be wrong because the chromatographer uses his equipment only 8 hours per day, switches off energy and gases and restarts his equipment without the “early morning test”. It is too space intensive to discuss this important, fast and simple test here but it is available in Teachware of
IfC.

11. A critical source of systematic quantitative errors is the position of separation lines. Main reason for wrong separation lines - lines which theoretically separate one peak from the next - is an overloaded chromatogram, poor separation or total overlapping of two or some peaks  not be visible by a too large peak width in half height or long tailings of a sample solvent peak. If substances to be quantitized “sit” on such a tailing their quantity value can be very false. See base line and separation line problems - click “here”.

12. The mobile phase flow speed may be adjusted wrong because of the use of some classical theoretical rules which are far away from practice - click “here”

According to our results and those from our thousands of course colleagues it is very helpful to use “reduced raw data” in STANDARD format and correspondingly graphics to easily check for possible systematic quantitative errors: the chromatogram is running and and than integrated. The integrated values are stored peak by peak as EXPORT data file. As mentioned in the chapter EXPORT the reduced raw data consist of
a) the retention time tms in seconds.
b) the peak width b05 in half height in seconds.
NOTE: because it is so important and against a majority believe, it is repeated: the real peak width in half height is taken by measurement, NOT by the wrong assumption, that the peak has Gaussian shape and the peak width value can be calculated. It can NOT. We checked with Europeans largest mathematical center many peak shape mathematics. When done, we realized, that a NEXT one may follow the practice. It did NOT.
c) the peak height in absolute physical units are taken either as ampere or as volt.
d) qualitative data as retention index or - if not yet known - instead of retention index values.  Zeros are written into the EXPORT reduced raw data file.
The EXPORT file is at the end of the integration job automatically enlarged by all necessary key words and information in ASCII which later identify the analytical job and allow for auto reporting.

When calling EXPORT data into one of the EXPORT programs there is a first systematic error check done:
Alarm colors tell, if either a peak height value or a peak width value is out of scale. In this case the peak integral is shown in an alarm color and / or a warning is printed on screen. We have an elegant problem solution for such cases : click on “Chromatogram Combination”. Spikes are automatically removed and the corresponding data deleted. The false data are corrected. A report is seen on screen.
This does in no way manipulate the raw chromatogram data. Raw data (in binary form) remain untouched according to internationally accepted regulation. However the ASCII EXPORT file of reduced raw data is qualified for either a complete or a strictly reduced analytical report, for auto statistics with a graphics report or for any other job analysts and the user of analytical information need.

This is a further important fact the EXPORT concept offers:
There are quite differing demands for a final result report:
Some users never need retention index data although they are the key control values to check for qualitative systematic errors - see “Qualitative GC Errors”

Others work already a long time with an existing commercial integration software but suddenly need
add-on calculations of physical values based on the results. Normally they must wait a long time until the instrument company can do the changed programming - or it turns out this is too expensive for just only one customer.
Based on the EXPORT reduced raw data file we never had any problem to offer software add-ons as solution for any new demand quickly.

The main tests to help detecting and reducing systematic quantitative GC errors:

1. Compare the latest chromatogram with an earlier one, if existing, but compare the two results graphically. Line chromatograms can be checked quickly, quantitatively, completely. For a quick check only EXPORT reduced raw data files are needed together with one of the many EXPORT programs existing.

2. Unfortunately it costs some additional time, but it pays back when avoiding or reducing systematic quantitative errors by the “six step safety  a) to f)” concept below:

Take
 a) - more than one sample source if flexibility is given;
 b) - more than one sample pretreatment if necessary;
 c) - more than one sample injection for each of the ones mentioned above;
 d) - more than one separation system - for instance by physical selectivity changes;
 e) - or use two dimensional separation;
 f) - use more than one detector if possible in parallel with known delay time between both. The delay
       time can be adjusted to become zero seconds - this is shown in one of the Teachware programs.
 g) - check for an optimal separation temperature - see “working range” below for isothermal GC or the
       optimal heating rate for temperature programmed GC.

It is necessary to check for the cost relation - not for the absolute costs of this “six step safety improvement” in quantitative GC. If the products to analyze are of importance for the economy (bio chemistry, energy industry, environmental protection etc.), than the relation of costs of analysis versus the costs of trouble caused by errors grows easily into the range of 1 to 1 million or more.
This high effort is of course applicable only during the period of method development, or in case of new analytical demands, in case of unexpected product trouble or important research projects. However: don´t we use analyses to check for the unexpected cases ?

As the “six step safety improvement” may be in total disagreement with regulation, the latter must be replaced. Most of them insist in “comparability”.
Never comparability counts, as comparable results may be drastically false and still well comparable with incredible good precision. But only accuracy counts. Accurate results are automatically comparable.

The importance of the correct mobile phase flow speed: is seen on figure 5 in page “Qual. Error HPLC”

One quite critical Error Source is not to observe the conditions of Temperature Working Ranges
of GC columns / capillaries .

If the temperature is too low, substances may not elute during the time the instrument or the
detection / integration is on. This falsifies the analytical results drastically and may damage the separation system, at least may change the polarity or selectivity of the stationary phase.
If the temperature is too high this may cause peak overlapping of early eluting substances and reduce the life time of the separation system.
 
It is quite simple to measure the temperature working range of a separation system with a few test runs and a qualified evaluation software. The following figure will show how helpful it is, to have the temperature working range available. It is of course valid only for one given column / capillary but it allows to select its strongest control factor, the temperature correctly.

In the following figure 1 the Y-axis represents the retention index of the substances to be separated, the  XY-field contains the temperatures the column / capillary must have and then the X-axis shows the corresponding retention time.

Application of figure 1: In order to get the separation of a sample containing substances with retention indices from 1200 (C₁₂₎ up to 1800 (C₁₈) done within 15 seconds and 8 minutes the separation system must have a temperature of 150 degr. centigrade. If it is adjusted to only 100 degr. centigrade, the substances of retention index C₁₈ will elute only after 5 hours.

The corresponding software package “T-Workingrange-GC” allows of course to calculate for each substance characterized by its retention index the retention time for each temperature value entered.

Figure 2 shows which measurements had to be done to get the data in figure 1: just three isothermal runs of a retention index test mixture had to be done at three differing temperatures. This provides all data for the temperature over time per retention index information. The measurements are packed in three EXPORT data sets containing the used temperature in one of the two text lines. The evaluation is completely automatic after entering the file name and the run number. As example for figure 1 and 2: the names were T.REK600.1, .2, .3.
 

for well adjusted TGC values. It does not guarantee to assure always complete substance elution in GC if one uses the temperature programmed mode, but BACKFLUSHING at least helps to keep the stationary phase alive.
See “Teachware”.

The reason why the temperature working range above in figures 1 and 2 look so perfectly linear is the very good linearity of the correlation of log(tms-tm) values with the absolute temperature T in the function   log(ts) = A * 1/T + B. NOTE: the net retention time ts and the absolute temperature are
taken. The data above have been measured with homologues from C₁₁ up to C₁₆. The linearity of the above given function is controlled by the determination coefficient (see under Statistics). The following values have been found: