What concentration of sample do you inject when you check peak purity? I injected a sample solution of ~1mg/ml and all the peaks are "impure". However, when I dilute the same solution 10X, they become pure.

I do a lot of trace analysis so I do have to inject a fairly strong solution. In the beginning, I thought the seperation is incomplete and I try for weeks by changing mobile phase, column....and all that text book stuff. They are still impure. Now I doubt that there is something other than resolution. Is the concentration of the sample a factor?

With all the experts out there, please give me some idea. Thanks!

I try to aim for a maximum absorbance value of about 200mAU when I'm performing peak purity analysis or building a spectral library.

If your absorbance is too high at the peak apex your spectra can become distorted and will fail a purity check. This can be easily detected because the purity profile will have a characteristic "W" shape instead of the normal "U" shape. This is caused by the apex spectra being distorted by too high an absorbance value.

What is your threshold value? Look at the purity profile and if you are exceeding the threshold value in the leading or trailing edge of the peak then it is probably an impurity. Good Luck.

Also check the wavelength range over which you are performing the peak purity analysis. Often analytes will absorb very strongly at short wavelengths (<200nm).

Thanks Tom, I was told somewhere in time that a purity factor of >990 means the peak is pure. I think there must have a protocol in the USP that draws a line somewhere. Or is it different from company to company? What's your criterium? Would you mind share with me?

Also, the wavelength range I use is from 200 - 700 nm.

If possible I try to use the spectral similarity curves to perform the peak purity analysis and include the threshold curve because these take into account the background noise level.

The manual states: Generally, values above 990 indicate the spectra are similar. Values between 900 and 990 indicate there is some similarity, but the results should be interpreted with care. All values below 900 indicate the spectra are different.

I try to use the maximum wavelength range that makes sense. If there are no chromaphores that can absorb in the visible wavelengths, no need using them. You will just be trying to compare noise against noise. The low end is always a tougher call because some impurities may only have absorbance at low UV wavelengths. I would look at the apex spectra and select the lowest wavelength that doesn't exceed the linear range of the detector. If the spectrum is maxed out and flat between 200 and 210 nm then you definitely need to adjust the wavelength range, amount injected, or both. Goood luck.