Trouble with ZIC HILIC

[danko]

No instrument should produce jagged peaks neither any substance should render such data. So, something’s wrong alright, but we don’t know what – as yet.

Let’s see how the chromatogram/s looks like – might help indeed.

Best Regards

[lmh]

should I be concerned by unbuffered solvents in a HILIC method?

[wan]

Hi Danko,

Sorry for the late reply.

The tailing was solved after we clean up the column. Something must be plugged.

We realised the jagged peaks can be solved by using a higher dwell time. We did not try increasing that as we had always been using a dwell time of 80 or 100 msec. After doubling that to 200 msec, the peaks appear smoother. But we need more points for quantitation, so the highest we can use is 150 msec.

I need to figure out how to post pictures to show the chromatograms. Will do so after I figure that out.

Thanks!

Regards,
Wan[/img]

[danko]

Hi Wan

Are we talking data collection points/rate (point per second)?
Here is the suggested by admin. method for embedding chromatograms:
viewtopic.php?t=2617

lmh wrote:

should I be concerned by unbuffered solvents in a HILIC method?

Unless you don’t mind mixed mode interactions, I’d say yes. Anyway in most cases – all depending on the column chemistry and the nature of the analyte (e.g. ionizable vs. not ionizable).

Best Regards

[wan]

Hi Danko,

Yes, the increased dwell time is supposed to affect the # of data points of each peak. Even though we can achieve a high number of data points at a lower dwell time, the peak is jagged and thus we suspect the high # of points is misleading. It also led to problems in quantitation as the jaggedness is rather random, making it difficult for integration.

With the current settings, we managed to get 16-19 points in each peak for quantitation. One of our scientists told us that is not sufficient, but we tried many ways to improve it and can't.

Thank you for the heads-up regarding embedding images. I will post chromatograms after we finish running this set of samples, which will be more representative.

Thank you again for all your help!

Regards,
Wan

[chris_singleton]

Hi Wan,


A quick note about data points, I usually shoot for 15-20 points across the peak and get excellent data (with triple quads e.g. Sciex, Thermo, Water). I've used the same approach when I was at another pharma company and when I worked at an instrument company, so I think this range of points is pretty well accepted for quantitative LC-MS/MS. Can you post a chromatogram? And have you tried smoothing your peaks when you build your quantitative method in Analyst? If needed I use 3 smooths when using the Intelliquan algorithm.


http://www.chem.agilent.com/Library/Sup ... points.pdf

And page 3 on this app note:
http://chromatographyonline.findanalyti ... rticle.pdf

[XL]

Here is new information on simultaneous separation of melamine and cyanuric acid.
Wadsworth Center of the New York State Department of Health, in Albany, New York, used Dionex Acclaim® Trinity™ P1 column in a new HPLC method for the simultaneous separation of melamine and cyanuric acid residues in dairy products and pet foods.
In response to the acetonitrile shortage, a successful "simple, accurate, and sensitive method" was developed on the Dionex column using methanol (replacing acetonitrile) in the mobile phase. Dr. Tran, B.N. et. al, describes this validated method in the publication titled, Use of Methanol for the Efficient Extraction and Analysis of Melamine and Cyanuric Acid Residues in Dairy Products and Pet Foods, published in J. Agric. Food Chem. 2010, 58, 101-107.

[wan]

Here's how our chromatogram looks like now:



We increased the ion spray voltage to 5000 and increased the dwell time to 150 ms to obtain a smoother peak shape as I mentioned before. In addition, this data was obtained using a new column.

Alas, I forgot to copy the data file we had obtained earlier to show the comparison. My bad.

Chris: Thanks for the info! And yes, we do smooth the data when we build the quantitation method. Typically, we use 4 smooths.

Wan

[wan]

Xiaodong: Thanks for the info! I'll look that up.

Wan

[wan]

Here's how the chromatogram looks like before we increased the dwell time. The RT is different as we're using a different column. The column was changed after that due to the large tailing observed, even after washing.



The jaggedness made it difficult to quantitate the peaks. We can smooth the peaks, but that would make the results inaccurate because we have to apply a higher # of smooths.

[Uwe Neue]

You need to select a higher sampling rate (see also the comment by Chris). You have less than 15 data points per peak. If you can increase the sampling rate 4 fold, all your headaches will go away.

Your current sampling rate is about 1 data point per 5 seconds. I am sure that you can go much faster (unless your instrument was designed in the stone age).

[wan]

Hi Uwe,

I had checked my latest data and obtained 19 data points on the average.

Are you referring to the first or second chromatogram?

-
Wan

[Uwe Neue]

You see every data point on your second chromatogram.
If you get more datapoints to start with, you will not get the rugged etches. You will get more noise. If you have then a large enough set of data, say 100 data points per peak or more, you can now use software to smoothen the peak by averaging without affecting much the peak width and the resolution.
Of course, some of this depends on the software that you have and what its capabilities are...

[wan]

Hi Uwe,

The first chromatogram is the latest data I have, while the second is for comparison.

I understand what you meant in your posts as the insufficient # of data points did affect quantitation.

However, the latest data I have do not have problem when I quantitate it. On average as I have mentioned, there're at least 19 data points in the peak.

Thank you!

-
Wan

[Uwe Neue]

Well, there is more to it, but it depends on how your MS is working with your data system. Dispite the nicer picture, it is possible that your data quality between both figures has not improved, and the error is about the same - in the order of 5%. If the system does not have an internal smoothing algorithm, you could improve the quality of the integration by increasing the sampling rate, say 2 times or 4 times. While this may look more ugly to start with, you now apply the smoothing algorithm that you have used, and you get your nice peak shape back for unproblematic AND high precision integration.

Of course, I could be wrong, since I do not know the internal works of your MS software or your setup. Plus, there are some other subtle complications with MS detection. Of course, the dwell time is another important factor. You report a dwell time of 100 ms, but the displayed sampling rate was in the order of 1 per 5 seconds. So what is happening in between? Is the MS going to sleep? Or is it doing other things, like looking at other masses? The quality of the data stream depends how any signal accumulation is happening - if at all.

If I were you, I would just try it out.