By Vangelis on Monday, July 12, 2004 - 07:06 am:

Hi,

during analysis of major anions and cations with the use of IC, a colleague whom we've been cooperating with has been facing the issue described below:

when quantifying with the use of a calibration curve which is forced through zero, the low range values are different (with a difference as high as 25%), compared to the respective values calculated when the calibration curve is not forced through zero.

for mid and high range concentrations such a difference is not observed.

Could anyone suggest whether it would be a more appropriate method for quantifying compounds of a low range, to force the curve through the zero or not?

We are ourselves carrying out some similar experiments at the moment.

-------------------------------------------------------------------------------------------------------
By Einar Pontén - SeQuant AB on Monday, July 12, 2004 - 03:21 pm:

Are you using a suppressor or not?

-------------------------------------------------------------------------------------------------------
By Chris Pohl on Monday, July 12, 2004 - 03:53 pm:

Vangelis,

It is the effect you report only observed with carbonate eluent? Do you have a negative intercept? If the answer to both questions is: yes, then the problem is most likely due to the effect of carbonic acid on detection linearity (as described in the Analytical Chemistry article by Doury-Berthod, et. al. Yan analytical chemistry volume 57, 1985 page 852). If you are observing the effect with other analytical systems, there are a number of possible scenarios where this problem might be observed but I would need to know more about the specifics: sample, injection volume, sample concentration, column, eluent, suppressor type (if applicable), suppressor operating condition (if applicable) and detector operating conditions.

-------------------------------------------------------------------------------------------------------
By Greg Cawthray on Monday, July 12, 2004 - 10:17 pm:

I personally never fit a calibration curve thru the origin as even when there is no detectable analyte present, there will still be a very very small signal present (ie noise). It is the "size" of this signal that is used to determine the limit of detection for your system, simply by the LOD rule of 3 times background noise.

If you are quantifying samples with very low analyte concentration, then you must prepare very low conc. standards to accuratley describe the linear relationship at such low signal levels. If there is no signal (ie area o), then 3 times 0 is 0 and your LOD is thus 0, impossible, thus noise is alway present in your chromatograms.

Do you happen to know what peak area you are obtaining upon injection of a blank solution, or even mobile phase?

Regards
Greg

-------------------------------------------------------------------------------------------------------
By Vangelis on Tuesday, July 13, 2004 - 02:14 am:

Thank you all for your prompt replies,
We do use a suppressor.

Dear Chris,

it has not been tried out with any eluent other than carbonate and bicarbonate (for the fast analysis method as suggested by DIONEX manual for AS14).

As about the sample it is prepared by the DIONEX 5-anion standard (with water being of the highest purity), the injection loop is 50 uL, the suppressor is the respective ASRS (for AS14) and the level of concentrations we're discussing about is as low as 50 ppb.

Would you generally recommend that one forces the calibration curve through zero or not for such sorts of applications?

-------------------------------------------------------------------------------------------------------
By Markus Laeubli, Metrohm on Tuesday, July 13, 2004 - 02:53 am:

Calibration curves with suppressors are usually slightly unlinear. Therefore I strongly recommend to use a linear or if possible a quadratic fit (NOT THROUGH ZERO). Quantification should be done within the range of calibration levels.
For very low concentrations a single point calibration with the lowest calibration level will give pretty good results. As long as there are no blank problems.

-------------------------------------------------------------------------------------------------------
By Vangelis on Tuesday, July 13, 2004 - 03:47 am:

thanks,

as about the quantification, it is indeed done only within the range covered by the calibration standards.

-------------------------------------------------------------------------------------------------------
By Chris Pohl on Tuesday, July 13, 2004 - 12:36 pm:

Vangelis,

I wouldn't try forcing through zero with a linear fit and an extended calibration range. This will cause a high bias for mid-range concentrations. Better to either use a quadratic fit (as suggested by Markus above) or use a narrow calibration concentration range if you want to use a linear fit (i.e. keep the calibration range within an order of magnitude).