This newsletter is to give an insight on some of the common bias potentials of the trap fraction for Method 25 samples.

I have been asked about this a couple of times recently and thought it might be good to review it for everyone. There are three bias potentials for the trap fraction. There is a positive bias from the background of the sampling procedure, equipment, sample recovery, and analysis. This is the largest of the bias potentials and while it does vary, it is generally below 25 ppm for a 4 L sample volume. This, of course, can be lowered by taking a large volume sample, thus diluting the bias impact. This is the bias for which there is discussion of a blank subtraction to improve the lower concentration level accuracy.

The second bias potential is also a positive bias. This is the result of an amount of CO2 which is not removed from the trap during the flush portion of the recovery and is thus counted as NMOC when the trap fraction is analyzed since the NMOC is converted to CO2 during the recovery process. This is tied to the statement in the method concerning the product of the moisture and CO2 concentrations. According to a discussion with Dr. Jayanty on this some years ago, the documented bias was only about 2% of the sample concentration they tested. The assumption had been the ice crystals had formed pockets which trapped the laden sample gas which could not be released until the ice melted. There have been other mechanisms proposed, but the impact is the biggest concern. With that in mind, I developed a spreadsheet which could calculate the impact of a volume of the sample gas being retained in the trap for whatever reason. This was based on actual sample data, so I plugged in some general data for this example. For a constant 4.5 L sample volume in a 6 L sample tank, a 6 L ICV volume, variable levels of CO2, and variable volume equivalents retained the following results are determined.

8% CO2 with 0.5 ml trapped = ~12 ppmC bias

8% CO2 with 0.75 ml trapped = ~18 ppmC bias

8% CO2 with 1.0 ml trapped = ~24 ppmC bias

5% CO2 with 0.5 ml trapped = ~7 ppmC bias

5% CO2 with 0.75 ml trapped = ~11 ppmC bias

5% CO2 with 1.0 ml trapped = ~15 ppmC bias

1% CO2 with 0.5 ml trapped = ~1 ppmC bias

1% CO2 with 0.75 ml trapped = ~2 ppmC bias

1% CO2 with 1.0 ml trapped = ~3 ppmC bias

Of course, even 0.5 ml is a large amount of gas equivalent to become trapped, but it does give an indication of the impact. Again, the larger sample tanks, ICVs, and sample volumes can reduce this impact.

The final bias is the negative bias of incomplete recovery of the sample from the trap. The method defines a cut off which is sometimes not possible to meet and some of the sample remains. This is a little more common with a lower temperature recovery of 200 o C as opposed to 250 o C or 300 o C. One would assume that anything passing into the trap through a filter at 121 o C would come back out at nearly twice that temperature, but that is clearly not the case. The restriction then relates to the ability to accurately measure the pressure in the ICV, which increases rather quickly with a flow of 150 cc/min or more. A larger ICV will allow a longer period of recovery and even the use of dual ICVs will help, but at some point the problems associated outweigh the benefit of continuing the recovery. At that point the recovery will be stopped if the impact on the reported sample is not significant.

Very early on in our history, we did quite a lot of internal research on recovery processes because of this type of situation. We recovered blanks, we recovered unused traps, we recovered traps after just being recovered, we recovered internal audits, and we arranged to get duplicate samples for comparison. One of these series of inquiries involved the recovery of traps and the impact of stopping the recovery at higher effluent concentrations. We recovered the trap fraction into an ICV until the pressure was an issue and then we recovered a second and then, if necessary, a third ICV until the cutoff level was reached. We then analyzed each of the ICVs individually and determined a concentration for the sample for each ICV and the combined sample concentration. Unfortunately, that was 20 years ago and I have not been able to find the exact data for the samples, but I do have the basic information from which our SOP concerning trap recovery was created.

There was no impact, which could be determined after a second ICV was completed, on the total sample. In most cases, there was no determinable impact on the total sample after the first ICV was completed. In the case of very high concentration samples, the impact was negligible if the effluent concentration was well over 200 ppmC at the end of the first ICV. This is because the effluent concentration is slowly decreasing over time so the concentration at any given point will generally be higher than any given point afterward. This is, of course, different for the initial recovery as the trap is heated from the ~ -78o C starting point at the end of the flush process to the 200o C at the end of the recovery. This heating will at times, create “bands” of compounds eluting from the trap, which will rise and then fall as more of the compound is released and then drops off until the next compound reaches the temperature that will cause it to elute from the trap. After the bands the concentration will return to a slow drop, which may or may not be steady.

Now there have been some comparisons of sample concentration to elution concentration during cleaning. We have done little in determining what is removed form the trap during cleaning because we have used a batch cleaning process for most of the time. This makes it difficult to make any claims other than the analysis of the individual traps after cleaning have shown a lower concentration in the effluent and at a higher temperature than required by the current method. I can, however, point out the impacts of effluent concentration on the sample even at what seem to be high concentrations.

If we were to use some type of continuous monitor, it should show some type of elution curve as described in the “bands” above, but related to the increased temperatures of cleaning the trap. The curve could reach the height of a 50 ppmC or 100 ppmC calibration, but the height would not remain constant so the concentration over time would be significantly less. This is due to the curve rising to the height corresponding to that concentration, but the curve starts closer to the zero line and should also return to near that starting point. Thus, the concentration during the time for which the curve would be determined is much lower than indicated by the height or even the area under the curve. For the sake of a simple example I have used the two concentrations indicated, a flow of 150 cc/min, and an assumed time for the curve to rise and return to the previous baseline of ~ 40 seconds. This gives a volume of ~ 100 ml for the gas not included in the sample. If the effluent concentration is assumed to be for the entire period, along with the prior assumption of sample volume, tank volume, ICV volume, we can determine the possible impact. Given an ICV must be pressurized, the total volume of gas will be greater and in this case I am assuming the total to be 8 L for ease of calculation. To determine the impact, all one needs to do is determine the concentration of carbon in the ICV filled with zero air after the introduction of the 100 ml slug of “lost” carbon and then calculating the concentration for a trap fraction analysis for 4.5 L sample volume.

8 L of 0 ppmC concentration gas plus 0.1 L of 50 ppmC concentration gas is ~0.6 ppmC.

8 L of 0 ppmC concentration gas plus 0.1 L of 100 ppmC concentration gas is ~1.2 ppmC

The concentration calculation for a 4.5 L sample is simple, since we have assumed the same volume for the sample tank and ICV and constant pressure and temperature, which only leaves the ratio of the gas volumes. With a 4.5 L sample volume and an 8.1 L volume in ICV the sample volume would be divided into the ICV volume to give a multiplier for the measured concentration in the ICV of 1.8.

Effect of 50 ppmC slug on 4.5 L sample is ~1 ppmC negative bias.

Effect of 100 ppmC slug on 4.5 L sample is ~2.2 ppmC negative bias.

The bottom line on these potential biases is that they are usually not very significant in relation to the use limit guidance for the method.

Wayne Stollings

Triangle Environmental Services, Inc.