correction of impurites in pure solvent

[mr3573]

In order to determine of impurities of a pure compound (more than 99%) I should prepare a calibration blend contain the same impurities. In the other hand I should have a matrix similar to sample so the calibration blend usually should be prepared in the sample matrix. For example if I want determine ethylbenzene impurities, calibration blend make by addition of impurities to pure ethyl benzene. But all ethyl benzene has low impurities themselves. How could correct that effect? I used auto sampler and injection volume is almost similar in all injection but not exactly the same. If I want to use internal normalization method for calibration, how should proceed?

[chromatographer1]

Here is one way to solve this.

For example: Using Hexane or Pentane as carrier solvent, add equal volumes of EB, Tol, p-Xy o-Xy and inject this mix in order to calculate the relative response factor for the 4 analytes (should almost be 1.000 for each of the 4 analytes given)

Once that is demonstrated, then inject the almost pure EB sample, followed to two levels of that sample of EB - spiked at 1x and 2x the approximate amount found in EB. For example: let's say the EB has about 1000 ppm V/V of each of the Tol, p+o-Xy.
Then prepare EB std-B containing 997mL of EB and 1mL each of Tol, p-Xy, and o-Xy.
Then prepare EB std-C containing 994mL of EB and 2mL each of Tol, p-Xy, and o-Xy.

Inject Std-B and Std-C.

Calculate linearity of each analyte and determine the amount of analytes in the almost pure EB.

Assuming the std regression of each analyte is linear (this should be true) then you should with 3 injections be able to calculate the amount of impurity of each analyte in EB. This EB sample can now also be a std solution, call it Std-A.

You can test other samples of EB and run a single std (chose one) to calculate the new samples content from comparison with a std. Store the stds in a tightly sealed vessel in a refrigerator.

If your EB is really pure and you wish to measure ppm levels, adjust the spikes accordingly.

best wishes,

Rod

[mr3573]

Sorry, a question again. If the relative response factor of impurities determined in other solvent such as hexane, then the GC' software could calculate amount of impurities in EB pure using above factors. As an example if consider BZ, Tol, Cumene as impurities in EB, and impurities in PPM level, I tried to make following solution:

Sol B: bz 20, tol 40, cumene 60 ppm(w/w) in EB
Sol C: bz 40, tol 80, cumene 120 ppm(w/w) in EB
For preparation I added weighted amount of impurities to EB and make desired concentration by dilution. But EB solvent for dilution have same impurities themselves and I suppose zero to calculate concentration for sol B and C.

The result from GC for Sol B and C using response factor(obtained in hexane) was higher than prepared as predictable. because solvent have same component in few PPM. I found a standard method to correct effect of impurity in solvent by internal standard method. But I looking for a way to do correction for internal normalization procedure.

[chromatographer1]

You see the error of diluting with EB that contains the impurities you are trying to measure.

You can't do that.

Once you demonstrate that Bz, Tol, Cum, EB, all have the same response (+/- 1%) on the FID (proving the known, but showing government reviewers you did your homework to justify your method) then you can add ppm of each to the EB to make stds but don't dilute with EB, dilute with another solvent that doesn't contain any of these, such as pentane. I would not do w/w but v/v. You can use density factors to correct for w/w if you must.

You can make solutions of pentane that contain known volume/volume amounts of the aromatics and then add the same amount to your samples of the lot of EB.

For example you add 1mL pentane to samples of the EB that contains 0, 1, 2, 5, and 10 µL/mL of each aromatic (each solution need not be a the same concentration for each aromatic) and then inject each EB preparation which then will give you true additions at the volume levels you desire to add to the unspiked EB. Ignore the pentane peak and you can do an area % calculation with your integrator, or ignore the main peak and do a linear regression on each of the impurities.

best wishes,

ROd