Chromatography Forum

I just checked some old data and fluoride works great with a quadratic fit. That's the most non-linear anion from what I see. Unfortunately we can't use it at the lab because the method states you must use linear. So the new calibration I am trying to get implemented has fluoride from 0.1 ppm to 4.0 ppm. Weighted 1/X so our reporting limit will be more likely to pass. 1/X squared worked great for the reporting limit but at 4 ppm recovery is only 94%. That's good enough but I like things within 5%.

I did learn that the EPA will accept a quadratic curve modification for method 300.0, and I've also heard from a few industry experts that quadratic curves much better describe the data when running suppressed anion chromatography. I've also weighted my curves to 1/conc.2, which gives really good statistics for R2, %RSD, and %RSE.
I also like the well defined procedure in 300.0 for finding method detection limits, so I think I'll try that and then set my lowest calibration standard at that value, with my upper standard being 2 orders of magnitude greater. I'm still finding confusing and sometimes contradictory information on whether the reportable range starts at the method detection limit, or the lowest calibration standard, so making them one in the same seems to be easiest.

Weighted 1/X so our reporting limit will be more likely to pass.

Can you explain what you mean by this?

Do you have a preference for calibrating using multi-ion or single ion standards?

EPA will accept a quadratic curve modification for method 300.0? Is that referenced someplace? Even if it was I doubt we could use it because we have to go by TNI 2016 and ISO 17025. With quadratic curve I didn't have to weight my fluoride. We have a ± 15% acceptance criteria for our reporting limit standard. We have to run the reporting limit standard in every run so getting it to pass is important. For fluoride it's 0.1 ppm and weighting it 1/X works well without dragging down the recovery at the high end too much. 1/ x squared works really well for the reporting limit but then the high end gets only 94% recovery. Our high end is 4 ppm fluoride and we do see samples at that level so I don't want to underestimate it. We have to run proficiency samples twice per year and if we fail one it's a LOT more work doing a corrective action.

We also report chloride, nitrate, nitrite and sulfate. Nitrite is the only other one that has a slight non-linear problem. It was very linear until they changed their suppressor type. They insisted it couldn't be the suppressor so I gave up and just cut the high end of the calibration from 10 ppm to 5 ppm. Now it is looking more linear than before after yet another suppressor change. We run bromide but don't report it. It's important to have some Br in some standard because when the column ages Br and sulfate run together. Then the column has to be replaced because we have to know if we are reporting sulfate whether there is any Br in the sample. Sometimes we see 0.5 - 1 ppm Bromide.

The listed MDLs are pretty low. We don't really care about levels that low. It's just a detection limit, not a quantitation limit. Any bit of noise in the baseline would throw things off significantly. I have never even tested it at 0.002 mg/L nitrate. I do the MDL study using 0.01 ppm nitrate. At that level even weighted 1/X I get about 65% recovery out of the MDL solution. We calibrate nitrate from 0.5 - 20 ppm and that works well.

We get a multi-anion standard and make dilutions of it for calibrators. I made a multi-anion standard out of dry chemicals but the auditors like seeing a certificate of analysis apparently. The guy who audited the organic section last year was a real beast and he wanted to decertify that area so everyone is scared. It's also easier to have a company make a custom standard. And we have to keep things easy because so many people don't like to work or think. I should start my own company to make custom standards. A company can charge $200 for something that can be made with $1 worth of chemicals. When I was doing atomic absorption I ordered $30 worth of iron. That's enough to last 1000 years and it's easy peasy. If you want 100 ppm iron - just weigh 100 mg of it and dissolve in some nitric.

There was an approval letter to a specific lab published in 2007 that accepted a modified method with quadratic calibration curves. It wasn't a blanket approval, but I think it would have set a precedent for other labs. https://assets.thermofisher.com/TFS-Ass ... ion-en.pdf
I'm not familiar with TNI but my lab is currently undergoing the ISO 17025 process and they don't seem to care, as long as the method is validated or is an approved standard.

You mentioned ± 15% acceptance criteria. Is this something out of the TNI standard? I'm going through the EPA method and all of the control limits are ± 10%. Am I misunderstanding what is meant by reporting limit standard? I was assuming it would be similar to what the EPA refers to as a LFB. And on that note, what is the difference between the LFB and the IPC solutions?

I was planning on determining MDLs using my lowest cal standards, which are currently 0.05, 0.1, 0.2, 0.2, 0.3, 0.2 ppm for F, Cl, Br, NO3, PO4, and SO4, respectively. As you mentioned, I really don't care about the detection limit as it's going to be lower than my quantitation limit anyway.

± 15% for the reporting limit is just a number the lab picked. It happened before I worked there so I don't know how they arrived at the number. It seems reasonable. ± 10 % is for the mid range continuing calibration check and quality control samples, but since the reporting limit is much lower the ± 15% is better. ± 10 % can be hard to hit for fluoride and nitrite sometimes at the low levels. Acetate and formate can interfere with fluoride and there is no way to get wide separation. Our auditors have let us get away with the ± 15%. Until recently reporting limit standards were not even required, but the lab where I work has been doing them since before I worked there.

We have our reporting limit the same as our low calibrator. That way our reporting limit standard is the same solution we use to calibrate the instrument. The LFB is made from the same solution we use to spike samples to determine if there is something interfering. It is another multi anion standard that we buy. Usually the main interference is from chlorine because it oxidizes the nitrite to nitrate. IPC is the instrument performance check to make sure the instrument is still in calibration. We use a mid range calibration standard for that.

The EPA Method 300.0 doesn't seem to specify what concentration the LFB should be at, but it does specify that the recovery should be within 90-110%. I've been running an LFB at my second-lowest calibration standard concentration, but I'm having issues falling within the recovery limits, especially for nitrate and phosphate. If I run the LFB with mid-range known concentrations, then I get better recovery, but this makes me question the accuracy of lower values.
Is it acceptable for the LFB and the IPC to be the same thing? It seems like both are defined the same - reagent water with a known amount of analyte. The only difference that I see is that the method specifies the concentration of the IPC. Both need to have 90-110% recovery of the known amount...

The guidance on spiking from EPA is to do it at the regulatory compliance limit or one to five times the background concentration of unspiked field sample if the method does not specify a level. Our radiochemisry guys just found out that they had been spiking at too high a level a while ago. An auditor referenced the EPA certification manual. I spike fluoride at 2 ppm, chloride at 50 ppm, nitrite and nitrate at 3 ppm and sulfate at 15 ppm. The auditors have been accepting those levels and most of my spike pass within 10%. Sulfate is the biggest problem partly because 15 ppm is a bit low. The other problem is that some samples don;t have nitrate and without that the sulfate peak tails and is a bit bigger than it should be. The spike has nitrate and that cuts the sulfate peak short since it comes out a minute later. I fixed that with a tailing sensitivity factor that does not allow sulfate to tail too much. If I did report phosphate it would be a problem. The AS-18 column doesn't handle phosphate well all the time. It hates Mondays because the instrument is off all weekend. When it has been running for a day phosphate works better.

From my reading of the method the spike solution can be an IPC solution also. For the fluoride electrode method I run I used the calibration stock solution as a spike solution. I do you a different solution for spikes on the IC, but I don't think it is necessary. As long as you have one solution that is not your calibration solution for a quality control sample you should be OK.