I am analyzing herbicides by GC using a 30m x 0.25 DB-XLB and a 30m X 0.25 DB-5.625. I have tried a split injection and splitless and am having trouble with linearity of my curvbe which rangesfrom 0.05ug/ml to 0.75ug/ml. The later eluting peaks are better than the early. My injection port temerature is 220 and the ECD is at 330. My column flow is about 2 and my total flow is 40ml/min. My starting temp. is 50 which I hold for 2min, ramp 10deg/min to 160 hold for 2min then ramp at 2deg/min to 200, then finally ramp at 15deg/min to 270.
If you can make any suggestions please reply. Thanks!
By Scott Fredrickson on Friday, June 25, 1999 - 05:05 pm:
There doesn't appear to be anything basicly wrong with your instrument set-up, except that you didn't mention your split ratio. Split/splitless injectors can be tricky to adjust. Make sure you know what ALL the flows really are. I've been surprised more than once. The last time was two days ago, and I've been at this a long, long time.
From the bench chemist's practical viewpoint, there is only one thing to consider--how close the detector is to saturation. However, many factors contribute to the 'background', including gas purity, column type (bleed), sample matrix history, split ratios, etc.
I would guess that your splitless linearity is much worse than the split runs, because you have much more compound in the detector when it elutes--probably 10-50 times more, depending on the split ratio. If you can run split mode, do so. The entire system will live longer.
Increasing the make-up gas flow to lower the amount of compound in the detector at any instant may improve linearity, but will increase your instrument detection limit.
The difference in linearity between early and late peaks demonstrates the issue of detector saturation. The earlier, taller peaks, even with less area, push the detector closer to saturation. At some point, usually long before the column is overloaded, the detector response will flatten out.
Herbicide extracts, unless you are doing clean water, will have lots of material in them that will not make the detector respond, but may still plate out over time and coat the detector cell. You can see this stuff on a mass spec.
You will have to consult your manuals, and perhaps the manufacturer, to see how clean your system is. For instance, our HP systems when clean run at about 200 picoamps. When they get up to about 5000, the linear curve is too short to be useful, and we send them in for cleaning.
You may have to live with the non-linear curve, and use more complex curve fitting algorithms for your calculations, but the curve should be reproducible. You might have to use only the range that is 'linear' and reproducible, and/or make new standards with concentrations that give similar peak heights.
Another issue is that of the analytical system irreversibly absorbing herbicides at the low to sub-nanogram levels typically analyzed. This also affects linearity, but gets less as the amount injected increases. With both of these phenomena happening at the same time, it is possible to end up with an S-shaped standard curve. System maintenance intervals (usually injector insert, septum and pre-column exchange/break-off) must be established to address the garbage build-up problem.
By Anonymous on Thursday, July 1, 1999 - 01:30 pm:
Thank You Scott for your reply to the problem I have been having with my Herbicide analysis. I am convinced that I am not saturating the ECD. I am only injecting 1ul and my peaks are not that big. I am using a HP5890 series II plus and my signal on the one ECD is 13 and the other is 14. I know that a low signal like this indicates a clean detector but I am wondering if there could be another problem with the detector that could effect the linearity. Also the water samples I am running have many components besides the materials of interest. There is something in these samples that remains in the system for some time and causes increased sensitvity of the ECD. My standard that I run after every ten samples has to be within 15%RSD to pass. After running samples my %RSD on the DB-XLB is over 100% and on the DB5.625 it is 20-30%. After letting the system sit for the weekend the standard comes back under the 15% on both sides. Also if I cut the guard and put a new Y in the system the %RSD also comes down. It appears that what ever is in these samples sits on the head of the column and gradually bleeds off. Do you know what type of compound could increase the sensitivity like this? If you have any information about these problems please respond. Thank You!!
By Scott Fredrickson on Wednesday, July 7, 1999 - 03:23 pm:
I was hoping it was a simple instrument problem. Sorry to bring you bad news, but....
The problem you have is very familiar to us, and the reason we are working toward using LC/MS to analyze these types of samples. We have seen this problem for some compounds even with laboratory water. It is probably NOT the detector acting up, but the injection system and column. As I see it, these components have 'active sites' that absorb a relatively constant amount of the standards you inject. So, the system appears to be reproducible, and probably non-linear. But when you inject samples, with LOTS of extra, unknown junk included, these active sites are saturated with the junk, and more--apparently much more--of your herbicides get through to the detector, giving a larger response. And, when the system is left hot overnight, or the injector insert and precolumn replaced, the effect goes away. Temporarily.
Here are things we have tried to help the problem. We have not solved the problem, only learned to minimize it, at great expense.
The system may become more stable as it runs longer. You may be able to 'condition' the system by running the same sequence multiple times until the system settles down and you get acceptable results. If this works, the conditioned state can be maintained for awhile by cooling the injector and column when the system isn't in use. At least, it won't take as long to re-condition. An autosampler work-out, and lots of data evaluation. This is the easiest thing to try, and the least difficult to explain and support.
If the response drifts the same way, it can be mathematically corrected to get closer to the true value (say, a 5% increase in response for each injection). However, the problem is that the standards, at the same levels, will give different responses in matrix than they do in solvent. If you do nothing else, this must be checked out, or your results may be wildly wrong, even if meet QC requirements. Your spikes should already be giving you clues about how bad this problem is. Are 130-200% recoveries common?
Standards can be made in 'clean' matrix,if that matrix is available. Lab distilled water is probably not going to give the same results.
Sometimes, samples can be diluted, minimizing the co-extractive load, and give better results. The samples are screened into high, medium, and low categories, and then each category is run separately. The results are great at high levels, and so-so at low levels. Again, lots of GC time.
Sample cleanup may be required. This is difficult, because the ECD won't detect the interfering materials, probably (they may not even elute intact from the column), so you can't 'see' what's going on. And you don't even know the class of compound giving you trouble. And of course, the cleanup adds more chances for errors.
Your DB-5 column is closest to making your RSD's. The XLB column may be a write-off, and have to be replaced with something else, maybe DB-17.
When you just don't have the resources to do any better, or are out of ideas and time, you might have to negotiate new QC requirements for this analysis. Even 30% RSD's may be adequate for the decisions that will be made from the data.
Oh, one more thing. You can use rock-solid standards to see what your ECD linearity really is, and check out the tuning tips I mentioned. I'd recommend any of the classic multiple Cl pesticides that don't break down on-column, such as heptachlor, and endosulfan I and II. Most of the CLP Pesticides (except endrin) fall into this category. I'd consider the herbicides to be an order of magnitude more difficult to analyze.
Good Luck! And start studying LC/MS.
By Scott Fredrickson on Wednesday, July 7, 1999 - 03:29 pm:
I had my train of thought broken--sorry. The second thing to try after 'conditioning' the system is simply to run fewer samples between standard sets, or use some variation on the EPA 'continuing calibration' scheme.
I had a similar problem with endosulfans on XAD air tube extracts (basicly polystyrene extracts) and ran 6 samples between 4 level standard sets to meet the 15% criteria. Thought I was never going to finish the project.
By Donna Ruszaj on Friday, July 9, 1999 - 03:37 am:
Scott thank you again for your information. You are the only contact I have found that has experienced this same problem. Have you ever tried adding trimethylchlorosilane to samples to tie up active sites or would this cause more problems. I remember reading an article where they had tried this but I haven't been able to find the article again and I cannot remember the details. Or possibly some other silanizing agent. I am using a 0.25um film thickness and I was also wondering if going to a thicker film had helped you.
One more question. With the XAD work you mentioned did you basically run a new curve between your 6 samples?
Thanks again for all your help.
By scottca on Thursday, July 15, 1999 - 04:13 pm:
Yes, I had to run a new curve every six samples.
I think there would be serious problems trying to put trimethyl CHLORO silane into an ECD with the samples. Not only with the detector, but there would be other uncharacterized reaction possibilities. If I were going to try this, I'd use a system that I could afford the time and money to re-build. It might be worth a try to cool, dis-connect and cap the detector, and silanize the system, though.
Our formulation lab has used an on-column silanization scheme, but they use FID's. The silicone coats the detector and causes problems after awhile.
A thicker film would probably help. In general, we have less trouble with difficult compounds when using 0.53mm megabore columns, where the typical film thickness is 2.65 um--ten times as thick as the typical 0.25 mm x .25 um capillaries. You could try a megabore in your ECD if you keep the helium flow low enough (<10 mL or so). ECD's won't run on 100% He. I don't know what the upper limit of He to Ar/CH4 is before they quit, but I know if you mis-plumb the detector, they don't work!
Have you been able to make any progress?
Posting is currently disabled in this topic. Contact your discussion moderator for more information.