Combat Poor Calibration Linearity by Optimizing Splitless Injection Parameters
19 Sep 2012“Help! My 8270 calibrations aren’t linear. I’ve already done inlet maintenance.” Does this sound familiar? Maybe you’ve replaced your column too. Agilent wants you to buy a new inert source with a 6mm ID drawout plate. The solution, however, may be much easier (and cheaper).
The linearity requirements for individual compounds listed in EPA Method 8270D specify that the calibration is considered linear if the relative standard deviation (RSD) of the relative response factor (RRF) is less than 20%. Failing this, one of the alternative fit methods described in EPA Method 8000 may be used, such as linear regression, provided that a minimum correlation coefficient (R) of 0.99 is achieved. Additionally, for the initial calibration (ICAL) as a whole to meet acceptance criteria, less than 10% of the individual compounds may have RSDs that exceed 20% and fail to meet the minimum R value for the alternative fit methods.
Correct instrument settings are as important as an inert system when you are looking for linear and reproducible responses over a large calibration range. Properly calculating the splitless hold time maximizes transfer of analytes to the column. This increases the raw response and enhances reproducibility, which is critical to achieving calibration linearity. Ideally, you want to flush the inlet volume at least 1.5 times before opening the purge valve and evacuating residual solvent, which can impair the chromatography of early eluting compounds. This thorough inlet flush is especially important when performing semi-volatiles analysis. The high molecular weight compounds take a while to fully volatize and the acids and bases have a long residence time due to chemical adsorption onto the various surfaces in the inlet. Deactivated wool and liner surfaces help mitigate this, but once samples are being analyzed, the buildup of non-volatile co-extracted materials plays an increasingly significant role in adsorptive loss in the inlet.
To calculate the splitless hold time, we used a free GC column pressure/flow calculator for Windows available from Agilent (Simply enter the column dimensions and instrument starting conditions as shown in Figure 1). We used a 30 m x 0.25 mm ID x 0.25 µm df Rxi-5Sil MS (cat # 13623) with an initial oven temperature of 40 °C and employed a 30 psi pressure pulsed injection in a 275 °C inlet. The resulting inlet flow was 2.78 mL per minute. A single taper Restek Premium liner with wool (cat # 23303.5) has a volume of 900 µL, so we were aiming to flush the inlet with a minimum of 1350 µL helium before opening the purge valve, which should take 0.49 minutes.
Figure 1 - HP Flow Calculator
To highlight the benefit of the extended splitless hold time, we ran two 8270 calibrations: one with the splitless hold time set to 0.59 minutes (1.82 inlet volumes, right in the middle of the suggested 1.5 to 2.0 volume range); the other was set with the splitless hold time of 0.25 min (0.77 inlet volumes, approximately half the suggested minimum). The calibrations were geometric series from 0.5 ng/µL to 32 ng/µL. The responses and relative response factors for the first internal standard group from the 0.5 ng/µL calibration point are listed in Table 1 along with calibration RSDs.
Table 1 - Calibration Summary Comparison
Looking at a selection of compounds in the first internal standard group, the benefit of the extended splitless hold time is obvious. Not only are the responses higher at 0.5 ng on column, the reproducibility between runs is improved as well (reproducibility isn’t generally associated with the variable concentrations of a calibration, but surrogates (SS) and internal standards (1,4-dichrobenzene-D4 in this case) are kept at consistent concentrations in all calibration levels).
If you plot the RRFs against the calibration levels, you should see straight lines with no slope. Since RRFs are normalized for concentration, they should be the same for each calibration level. Figure2 shows the plot relative response factors for pyridine, 2-fluorophenol, 2-nitrophenol, and 2,4-dichlorophenol for the calibration run with a splitless hold time of 0.25 min. Figure 3 is the same plot using the RRFs from the 0.59 min splitless hold calibration. As you can see, the extended splitless hold RRFs are much closer to the straight, slope-less lines predicted. 2-Fluorophenol, a surrogate, is present at the same concentration in each calibration level. This is a very good example of the lack of reproducibility that results from ending the splitless hold prematurely.
Figure 2 - RRF vs. Calibration Level, splitless hold time 0.25 min
Figure 3 - RRF vs. Calibration Level, splitless hold time 0.59 min
This lack of reproducibility makes it difficult to generate linear calibrations. If you plot the response factor (RF) against the calibration level, the linear regression lines and correlation coefficients indicate a much better fit for the extended splitless hold time calibration. While both sets of acquisition parameters yield R values greater than 0.99 for the compounds in question, the short splitless hold time exhibits exceptionally poor performance at low concentrations.
Figure 4 - RF vs. ICAL level, splitless hold time 0.25 min
Figure 4a - Poor regression line fit at low concentration
Figure 5 - RF vs. ICAL level, splitless hold time 0.59 min
Figure 5a - Excellent fit of regression lines at low concentration
Using pyridine as an example, we'll back calculate the concentration of the first calibration level (0.5 ng/µL) using the RF and the linear regression line equations obtained from the Figure 4 and 5 charts. The basic linear equation has the formula of y = mx + b. For our purposes, y is the RF, m is the slope of the regression line, x is the concentration and b is the y-intercept.
Short Splitless Hold Time:
y = 0.0751x - 0.0916 (from Excel)RF = 0.0162x = 1.435 ng/µL% deviation from 0.5 ng/µL: -187%
Extended Splitless Hold Time:
y = 0.0923x - 0.0017 (from Excel)RF = 0.0454x = 0.51 ng/µL% deviation from 0.5 ng/µL: -2%
Section 11.4.5.6 of EPA Method specifically addresses this potential for bias at the low end of the calibration when linear regression is employed. Using the instrument response and regression line equation, the calculated concentration of the low calibration point should be within 30 percent of the actual concentration (in this case, 0.5 ng/µL). The short splitless hold calibration for pyridine misses acceptance criteria by a wide margin, while the extended splitless hold calibration shows little deviation. A comparison of Figures 2 and 3 summarizes the performance difference nicely, in terms of relative response factors.
To summarize, double check your acquisition parameters and make sure you are purging the inlet sufficiently before you start making expensive and time intensive modifications to your instrument.