Variability problem in HPSEC [June 22, 2004]
Posted: Tue Aug 24, 2004 8:33 pm
By David Kreller on Tuesday, June 22, 2004 - 02:33 pm:
Hello,
This question will regard variability (mostly in retention time) in size exclusion chromatography (SEC).
Currently we are taking a close look at variability and are trying to understand how much variability is inevitable, and how to decide when there's too much.
For example, we run a set of molecular weight standards (polystyrene sulfonate) and from them establish a calibration curve for the relation between molecular weight and retention time. Of course the retention time depends on size/ hydrodynamic radius, which depends primarily on weight, but if you are familiar with SEC/ GPC you know that.
The Waters Empower software we are using has the GPC component that sets up a calibration curve for us and tell us the relationship (the equation for the line) and the R2 value. Our first question is ‘What is an acceptable R2 value?, More specifically, when would you say that an R2 value is too low to go ahead and use it on your samples?’
We intersperse standards throughout our sample sets. Recently a standard was analyzed as many as 10 times at periodic points throughout a long run. The injections were made from the same vial. Treating the standard as a sample, the software estimated its molecular weight. One standard deviation in that calculated molecular weight was 10-15 % of the Mn!!
In your opinion is that too much variability? It seems like it is bordering on the limit of too much. What can a person do to control variation in this technique? We have seen literature on the performance of our system (Waters 2695) that says variability in retention time should be < 1%, but we sure aren’t seeing that. Might that be the expected variability in retention time in the absence of the column? I'm sure the generalities of SEC are the same as other LC techniques in terms of good practice.
We have thought that flow rate and temperature might control retention time, with flow rate having the greatest effect. This is a pretty good Waters HPLC (2695) and I would tend to doubt that the flow rate would vary by much more than a couple of %.
Thanks for any input,
David Kreller
CE/GEOS
University of Notre Dame
-------------------------------------------------------------------------------------------------------
By Anonymous on Wednesday, June 23, 2004 - 06:44 pm:
Hi...Hope the information below can clear your doubt. Depending on which methods you use, there is indeed some variation in calculated molecular value. Also how you integrate the peak has to be consistent, ie the start and end of intergration have to stay consistent. If you start the integration at different point, your Mn values will vary.
Since you are using Empower software, there is this column under the peak table that can show the start and end integration in terms of retention time. Alternatively, some people use molecular marker to check the variation.
Accuracy of various molecular weight determination methods.
Standard Error of Estimate
Mw by LALLS molecular weight above 50,000 +/- 5 to 10 %
Mw by LALLS molecular weight below 50,000 +/- 10 to 15 %
Mn by LALLS +/ 20 to 30 %
Mw by MALLS molecular weight above 100,000 +/- 3 to 10 %
Mw by MALLS molecular weight below 100,000 +/- 10 to 15 %
Mn by MALLS +/ 20 to 30 %
Mn by Membrane Osmometry MW above 50,000 +/- 5 to 10 %
Mn by Vapor Pressure Osmometry MW 15,000 - 30,000 +/- 10 to 15 %
Mn by Vapor Pressure Osmometry Mw belo15.000 +/- 2 to 5%
Mw by GPC/SEC +/- 3 to 10 %
Mn by GPC/SEC +/- 10 to 15 %
Mz by GPC/SEC +/- 10 to 15 %
Mp by GPC/SEC +/- 3 to 10 %
Mw by GPC/DV Universal Calibration +/- 5 to 10 %
Mn by GPC/DV Universal Calibration +/- 10 to 15 %
Mz by GPC/DV Universal Calibration +/- 10 to 15 %
Mp by GPC/DV Universal Calibration +/- 5 to 10 %
Mn by End group analysis MW Below 10,000 +/- 10 to 15 %
Mz by Ultra centrifuge Mw above 100,000 +/- 10 to 15 %
Mv by Viscotek viscometer Y-501C +/- 5 to 10 %
Intrinsic Viscosity by Viscosity Y-501C +/- 5 to 10 %
-------------------------------------------------------------------------------------------------------
By Anonymous on Wednesday, June 23, 2004 - 07:11 pm:
If you are not controlling temperature (your message suggests you are not), you must. Variations in temperature affect RT, this is a critical parameter to control in GPC.
-------------------------------------------------------------------------------------------------------
By Anonymous on Wednesday, June 23, 2004 - 07:30 pm:
Just to add to the previous message, a temperature change of 2 C can result in a RT shift of 2-3%. A properly operating 2695 should give retention times of less than 0.1% RSD when running GPC.
-------------------------------------------------------------------------------------------------------
By David Kreller on Friday, June 25, 2004 - 11:44 am:
Hi folks,
Thanks for your feedback so far. I have a couple of questions following from the messages;
1: Where did all those numbers come from for the typical relative error in all those ways of determining Mn? Was it some standard analytical textbook?
2: About the effect of temperature, would it be correct to assume that elevated temperature would decrease retention times, and that this would be due to breaking bonds of interactions between the analytes and the stationary phase?
Thanks,
Dave
-------------------------------------------------------------------------------------------------------
By David Kreller on Friday, June 25, 2004 - 12:15 pm:
Hi again,
Another question that should have been included in the last post.
Q: Would you ever not include one of the standards in the calibration? And if you would leave one out, how would you make that decision? Based on getting a better R2 value?
We have noted that the Empower software lets us decide if standards will or not be included.
Dave
Hello,
This question will regard variability (mostly in retention time) in size exclusion chromatography (SEC).
Currently we are taking a close look at variability and are trying to understand how much variability is inevitable, and how to decide when there's too much.
For example, we run a set of molecular weight standards (polystyrene sulfonate) and from them establish a calibration curve for the relation between molecular weight and retention time. Of course the retention time depends on size/ hydrodynamic radius, which depends primarily on weight, but if you are familiar with SEC/ GPC you know that.
The Waters Empower software we are using has the GPC component that sets up a calibration curve for us and tell us the relationship (the equation for the line) and the R2 value. Our first question is ‘What is an acceptable R2 value?, More specifically, when would you say that an R2 value is too low to go ahead and use it on your samples?’
We intersperse standards throughout our sample sets. Recently a standard was analyzed as many as 10 times at periodic points throughout a long run. The injections were made from the same vial. Treating the standard as a sample, the software estimated its molecular weight. One standard deviation in that calculated molecular weight was 10-15 % of the Mn!!
In your opinion is that too much variability? It seems like it is bordering on the limit of too much. What can a person do to control variation in this technique? We have seen literature on the performance of our system (Waters 2695) that says variability in retention time should be < 1%, but we sure aren’t seeing that. Might that be the expected variability in retention time in the absence of the column? I'm sure the generalities of SEC are the same as other LC techniques in terms of good practice.
We have thought that flow rate and temperature might control retention time, with flow rate having the greatest effect. This is a pretty good Waters HPLC (2695) and I would tend to doubt that the flow rate would vary by much more than a couple of %.
Thanks for any input,
David Kreller
CE/GEOS
University of Notre Dame
-------------------------------------------------------------------------------------------------------
By Anonymous on Wednesday, June 23, 2004 - 06:44 pm:
Hi...Hope the information below can clear your doubt. Depending on which methods you use, there is indeed some variation in calculated molecular value. Also how you integrate the peak has to be consistent, ie the start and end of intergration have to stay consistent. If you start the integration at different point, your Mn values will vary.
Since you are using Empower software, there is this column under the peak table that can show the start and end integration in terms of retention time. Alternatively, some people use molecular marker to check the variation.
Accuracy of various molecular weight determination methods.
Standard Error of Estimate
Mw by LALLS molecular weight above 50,000 +/- 5 to 10 %
Mw by LALLS molecular weight below 50,000 +/- 10 to 15 %
Mn by LALLS +/ 20 to 30 %
Mw by MALLS molecular weight above 100,000 +/- 3 to 10 %
Mw by MALLS molecular weight below 100,000 +/- 10 to 15 %
Mn by MALLS +/ 20 to 30 %
Mn by Membrane Osmometry MW above 50,000 +/- 5 to 10 %
Mn by Vapor Pressure Osmometry MW 15,000 - 30,000 +/- 10 to 15 %
Mn by Vapor Pressure Osmometry Mw belo15.000 +/- 2 to 5%
Mw by GPC/SEC +/- 3 to 10 %
Mn by GPC/SEC +/- 10 to 15 %
Mz by GPC/SEC +/- 10 to 15 %
Mp by GPC/SEC +/- 3 to 10 %
Mw by GPC/DV Universal Calibration +/- 5 to 10 %
Mn by GPC/DV Universal Calibration +/- 10 to 15 %
Mz by GPC/DV Universal Calibration +/- 10 to 15 %
Mp by GPC/DV Universal Calibration +/- 5 to 10 %
Mn by End group analysis MW Below 10,000 +/- 10 to 15 %
Mz by Ultra centrifuge Mw above 100,000 +/- 10 to 15 %
Mv by Viscotek viscometer Y-501C +/- 5 to 10 %
Intrinsic Viscosity by Viscosity Y-501C +/- 5 to 10 %
-------------------------------------------------------------------------------------------------------
By Anonymous on Wednesday, June 23, 2004 - 07:11 pm:
If you are not controlling temperature (your message suggests you are not), you must. Variations in temperature affect RT, this is a critical parameter to control in GPC.
-------------------------------------------------------------------------------------------------------
By Anonymous on Wednesday, June 23, 2004 - 07:30 pm:
Just to add to the previous message, a temperature change of 2 C can result in a RT shift of 2-3%. A properly operating 2695 should give retention times of less than 0.1% RSD when running GPC.
-------------------------------------------------------------------------------------------------------
By David Kreller on Friday, June 25, 2004 - 11:44 am:
Hi folks,
Thanks for your feedback so far. I have a couple of questions following from the messages;
1: Where did all those numbers come from for the typical relative error in all those ways of determining Mn? Was it some standard analytical textbook?
2: About the effect of temperature, would it be correct to assume that elevated temperature would decrease retention times, and that this would be due to breaking bonds of interactions between the analytes and the stationary phase?
Thanks,
Dave
-------------------------------------------------------------------------------------------------------
By David Kreller on Friday, June 25, 2004 - 12:15 pm:
Hi again,
Another question that should have been included in the last post.
Q: Would you ever not include one of the standards in the calibration? And if you would leave one out, how would you make that decision? Based on getting a better R2 value?
We have noted that the Empower software lets us decide if standards will or not be included.
Dave