Check the needle seat and needle for signs of wear or leaking. Note: Look for signs of leaks by the injector. Leaks always indicate a problem and should be fixed immediately. Don't run samples when you have leaks. Your method and data obtained will be invalid. Any worn parts should be replaced and the system performance tested. One of the most common causes of carry-over is due to a worn sample injector valve rotary seal. A worn seal can allow sample to be retained in its spaces and grooves in-between injections, resulting in sample peaks appearing in subsequent runs. Additionally, buffer salts can lodge between the seals causing leaks or carryover. Routine HPLC service and, if applicable, flushing of all buffers/salts every day can eliminate these issues.

That Shimadzu A/S uses a A/I design which can easily become contaminated. Using the wash vial sometimes makes the problem even worse. One of the keys to using that brand of HPLC successfully is to use the correct types of wash solutions with the mobile phase (must be 100% fully miscible and rinsed several times with mobile phase).

Another common problem we see is the use of poor quality methods which do not have proper column washing procedures after EACH analysis (failure to wash the column with a stronger mobile phase than used with the analysis method and then re-equilibrate the system are the most common reasons for peaks to show up later. Cause = Inexperienced operators/lack of training). This results in accumulation of retained sample on the column. The column becomes contaminated and the resulting contamination elutes out as part of each analysis (looks like contaminated rotor seal, but is the column). Many, many reasons for problems of this type and to find the answers you must start at the beginning. Find someone local experienced in chromatography to help you and the answer will be found quickly.

Here is an article that will assist you to troubleshoot and solve the problem; "Carry-Over (Carryover) Contamination in HPLC and LC-MS Systems".

http://hplctips.blogspot.com/2015/02/carry-over-carryover-contamination-in.html

@Ahmed  the leak will show me by the column pressure which is normal and not fluctuate. I have two instruments with the same HPLC system (4000 and 5500) and they give me the same result. I will rebuild the injector valve.

@bill  the system has a big reservoir to flush the well and I use up to 2 mL with dimethylformamide. The analyte is very well dissolved and not left over in the HPLC system as I explained above that I injected a standard and then for the second injection, I initiated the HPLC gradient without using autosampler and see nothing. That means my LC flow path is clean (not stick to the column and the mobile phase flush the column well. The third injection is water using the autosampler and see carryover. So the issue is the autosampler. It has the same issue for two "different" instrument with the same HPLC autosmapler.

That is why I included the article for you to read. It addresses these issues.

ok, I will read it now, thanks,

By the way, two instruments do not have issue with carry-over for other projects (different analytes). 

Each problem must be dealt with on a case-by-case basis. Different samples have different properties which can result in different observations and trouble. Solve one problem at a time. The HPLC flow path is contaminated (the flow path includes the A/I, A/S & column). These systems require regular cleaning and maintenance. Routine service of the A/I valve may solve it, or you may have to dig a bit deeper. Keep it simple and start with the most basic things (because most problems are very basic).

You may also find this article of interest too:

"Troubleshooting HPLC Injectors (Manual and Automated)":

http://hplctips.blogspot.com/2013/06/troubleshooting-hplc-injectors-manual.html

This is what the expert in this field told me.

POEA is a difficult compound to work with analytically.  The LC we use (Waters Aqcuity H-Class Bio) is primarily a titanium flow path and we still had carry over issues.  Unfortunately, the work around we used is what your colleague recommended:  inject low to high when you can and use excess blanks.

I'm not sure if there is a good material that will help.  In tests I ran looking at containers and filters POEA stuck to them all...glass, stainless steel, teflon, etc.

Best of luck and let me know if you find a better solution!

Cheers,

I drop 5 ul of formic acid into standard 100 ng/mL (in dimethylformamide, extracting solvent) and hope to ionize the amine group. Inject 2uL of this and follow by the first blank after this standard has 24% carryover, the second blank has 11 % carryover. Working on the 3 and 4th blank injection.

 As my colleague Dan mentioned their is a reason that the literature on POEA is sparse and that is because of the analytical difficulty associated with the characteristic complexities of  many classes of surfactants combined with its cationic character. Thus, POEA readily adsorbs to all the materials we have tested.

For example, to lengthen the time for significant amounts of POEA to adsorb to the LC vials triethylamine and formic acid needed to be added to each sample extract.  We were not successful in eliminating carryover contamination and to minimize it we analyzed several blanks between each sample and as best we could would  start with the lowest standards and samples and work to the higher ones. In addition, there is the need to "remove" the triethylamine from the flow path when converting the LC/MS over to other methods, which can take a few days and may require replacement of the tubing in the sample flow path.

If you find a method that reduces or eliminates POEA carryover we would be very interested in knowing about it.

Cheers,

Mike

It is good that you're washing the outside of your needle but I suspect the analyte is sticking to the channels in your rotor seal. I see this often and the following has been successful for me. 

Trigger a few valve turns back and forth at eluting side of your gradient to wash out the seals.  This would be at high organic in reverse phase HPLC for example. Since this now leaves the rotor seals with high organic you'll need to recondition the seals back to starting conditions or you may have reproducibility issues. I just repeat the valve turn process at the reconditioning side of my gradient to solve this. It's important to know your dwell volumes because this will effect your valve timing. You have to make sure the proper solvent conditions are at your valves when you switch them. This will also need to be done for all valves in the flow path as well, like diverting valves to your detector.

Below is a typical method with B = eluting solvent and rotor positions 1-2 to column & 2-3 loading sample loop.

0.00min %B=20 START

0.10min %B=20

5.00min %B=95

6.90min %B=95 Valve=2-3

6.92min %B=95 Valve=1-2

6.94min %B=95 Valve=2-3

6.96min %B=95 Valve=1-2

6.98min %B=95 Valve=2-3

7.0min %B=95 Valve=1-2

7.2min %B=20

9.90min %B=20 Valve=2-3

9.92min %B=20 Valve=1-2

9.94min %B=20 Valve=2-3

9.96min %B=20 Valve=1-2

9.98min %B=20 Valve=2-3

10.00min %B=20 Valve=1-2 END

I hope this helps!

I tried a few things by changing from Dimethylformamide to DMSO. My FDA friend suggested that Dimethylformamide is basic and is hard to acidify. I tested by acidify DMSO with formic acid. I can lower the carryover from 25 to 20 %. I found that DMOS or Dimethylformamide are the only solvent that I can get POEA-15 from my oatmeal samples. I spike POEA-15 for a week and get 100% out of it. But the sticky issue is carryover. I will see if I can get different material for the valve in the autosampler. I proved many times that anlayte is not sticking in the flow path, but something in the auto sampler itself. I added acid to ionize the amine group to make it more "polar", next will see if I can use hydrochloric acid. I also flush the needle wash well for 2 min with formic in methanol, this is to be certain that it is not the outside of the needle wall that causes the problem. I will keep you posted.

@david The valve switching should do the trick because I suspect that it is the valve. But now I need to know how to switch the valve during the run. My instrument is Shimadzu SIL-20A. I will look at the Analyst program to see if I can control from it.

Thanks a bunch.

I have the Shimadzu engineers here tomorrow installing a fraction collector. I can ask them. We have a SIL-20AC HT but I have not programmed the valve switching on it. It's pretty easy on an Agilent.

Found the cause will fix it tomorrow 

There are two experiments I did yesterday and today. Shimadzu guy thinks that the analyte was stick to the column and suggested me to do multiple gradient in the same run to wash the column. The column I used is Restek Raptor ARC-18. I  inject the standard and see the peak at 4-5 min interval with the response getting lower and lower (100% the first peak, 10% the second peak and 4% the third peak). It demonstrated that the analyte sticks to the system even after the gradient with 95% methanol. The most important finding is he suggested me to switch the column. So I switch to Restek Ultra Aqueous column (similar dimension) but started at 40% methanol for 1 min and ramp to 100% in 2 min with multiple gradient.  Strangely, my carryover has gone....................     Now I am dialing down the method and to see if I need multiple gradient or the issue is the type of column I am using............... Yeah.....

I asked Shimadzu if the first column is more retentive than the second one even both are C18 with the same dimension. His answer is below:

No, the explanation is simply columns are not perfect and over time become even less perfect.  The stationary phase on the particles wears off, exposes bare Si bonding sites which polar groups have a high affinity for but have low elution strength in high organic solvents therefore the pulsed method accomplishes both the column wash for the low hydrocarbon chains as well as the polar heads on the Si sites.

I have learned new things everyday after 40 years in this work, amazing.

Next, I will find the second column and test it. (same brand and dimension to see if it is not a coincidence). There is a possibility that the column that works was coated with sample matrix (I used this for two years for pesticide analysis) and change the stationary phase characteristics. Better yet, buy a new column or use a new column of similar phase (I got new Luna C8), just to prove that if you have a good column (or new), the carryover should not be an issue. I love my job.

Hi Narong,

I know what you mean about learning stuff new, that's why I love science :). I'm glad you are back on track! I would get a new column like the old one you used to make sure it's a column issue. If that's possible? Usually when the stationary phase starts to slough off your plate count gets worse. But this may not be noticed if method development was only done on an old column.

Another thing is analytes with more than one pKa can cause issues that look like a carry over b/c a portion of the analyte binds strongly to the column at a given pH. A pH gradient can help with cleaning/reconditioning the column.

Well, at this point, I have proved that it is not the autosampler issue. I would have chased the wrong trees for a while if I assume that it is the hardware issue. I will test a brand new column and see how it goes.

Today, I tested two old columns that were abused over the years (Ultra Aqueous C18 from Restek) and one new Luna C8 from Phenomenex (brand new) and used double gradient program. Inject one standard solution and see only one peak at the specific retention time. So, I was unlucky to waste the time on one column (Raptor ARC-18) but lucky enough to learn how to solve the problem. Thanks to the Shimadzu tech rep Lukas Swanson. Believe it or not, he has an M.Eng degree, not a chemist.

It is re-assuring to know that we have such an astute and systematic analytical chemist at our FDA!

Just to do my job.

A good news for public. I am working on the glyphosate in food and the hypothesis is that adjuvant added to the formulation (non-ionic surfactant like POEA-15 or MON-0818) is more toxic than glyphosate (1000x) based on the study of Seralini (tox chemist from France). The surfactant is used to spread out the formulation droplets on the leave and make herbicide stick to the leaf. I suspect that the food that contain glyphosate such as oatmeal, may contain this surfactant and make it more toxic than glyphosate alone. First, when I saw POEA-15 in oatmeal during the extraction work (about 200 ppb), I thought that my hypothesis is right and start to work on the method. However, I had the carryover issue when I tried to validate it.  Once I fixed this issue and reanalyzed the same oatmeal sample, I found nothing. So the first finding is false positive due to the carry over of the standard that I injected before the sample . This can be true because this surfactant is not necessary to be systemic like glyphosate. Glyphosate will move from the leaf to root to kill the plant. Surfactant may stay at the leaf so we are safe for now.

My experience is very old but I think still quite valid. A long time ago analysts working in pharmaceutical QC moved away from injectors in which the solution to be injected comes into contact with a rotary valve like the 6-port Rheodyne or Valco types used in many injector designs.

We wrote in LC-GC July 1996 that it was already very well known that the over-filled sample loop technique gives irreproducible results because of reversible adsorption on the loop and the valve. Another serious difficulty is that unless the sealing surfaces are in very good condition, tiny internal leaks (ie not to the outside) cause memory effects of the kind described. Note that such leaks may be significant only when the valve rotor is in motion.

The only designs that are reliable (or easily fixed) in that respect draw the sample into a needle, keeping the space between the needle and the valve filled with mobile phase, and the needle is then placed in series with the flow from the pump to the column. Contemporary examples were Waters Wisp, and HP/Agilent models like 1090 and 1100. Such equipment may have different limitations, one of which we wrote about, but avoiding carry-over and adsorption effects is vital in our field.

@Christopher   Shimadzu SIL-20A has the same design as Waters Wisp and I injected only 2 uL in the needle and the sample never reaches the loop. Again, I proved that it was not the autosampler by running double gradient in one injection and saw two peaks. Once I changed the column, the problem disappeared.  Thanks,

Mr. Lee: The Wisp and HP 1090 A/I used the older A/I designs (as the newer Waters units still do). Starting with the HP 1100-series A/I (e.g. G1313A), these were the first to use the most modern high pressure design with a high pressure pump (a single piston analytical HPLC pump) in place of the old glass syringe (the syringes had to be flushed, often by drawing liquid into it multiple times or pulling the plunger out of the barrel to flush it out, all of which proved problematic). The design that HP introduced with the 1100-series (and Agilent still uses today on the 1200, 1260 and 1290) is far superior and has almost completely done away with the need for wash vials, "syringe cleaning" and most types of cross contamination. These A/I's use a flow-through design which draws the sample into the oversized loop, away from the valve, then flushes the entire sampling path (needle, needle seat, inj pump, loop...) with mobile phase throughout the active run rinsing it clean. No need to "wash" the needle or system flow path after injection at all (it is continuously washed). As with all A/I designs, samples can still stick to the outside of the needle (to address this you can program the A/I to "dip wash" the needle in a designated uncapped, wash vial). IMHO: This was one of the most innovative and practical improvements made to liquid chromatography in the past few decades. Valve rotor maintenance is still required, but when the operators are provided with proper training in how to use these 'newer' style auto injectors, most common types of carry over are no longer seen.

Regarding Narong's original issue, I addressed this in my earlier comments as  the cause being a contaminated flow path (either the column or the injector being the likely source).

The problem is "not" the contamination. It is the imperfection of the first column I used (Resteck Raptor ARC-18). It is not the autosampler or column contamination or flow path issue. The prove is

1)  run double or triple gradient in one injection (5% methanol 1 min then step to 95% methanol in 1 min, stay for 3 min, drop to 5% methanol in 0.1 min, stay for 3 min and step to 95% methanol in 1 min, stay for 3 min and drop to 5% methanol)  with the first column and I saw two or three peaks reducing in peak response from 100 to 20 to 5%. The explanation from the Shimadzu is that the column has exposed silanol site that will retain my analyte that even 95% methanol would not elute and I have to elute with 100% buffer.  My analyte is POEA-15 that has three long chain fatty acid (16 to 18 C) and has polar head of amine.  It is a complicate structure.

2) by replacing the first column with another C18 columns and those are a) old Ultra Aqueous C18 no1, b) old Ultra Aqueous C18 no 2, and c ) new Lulna C8 and repeat the double and triple gradient using the same autosampler, I see only one peak. No carryover issue whatsoever.

Narong, I thought I'd mention valves because they were mentioned during the discussion and, as so often on RG, questions like yours provide an occasion to recall all the likely causes, particularly for the benefit of people who are not primarily analysts.

Bill gave a good account of Wisp-HP type injectors. One potential disadvantage: since the measured amount of solution to be injected is held in the tip of a needle that's in motion, there are various ways you can lose some, not only because of a leak in the valve that directs the mobile phase flow.

Memory effects inside the column may perhaps be a manifestation of the "virtual leak" problem. If there is a very narrow porosity, typically between a ferrule and the tube, that doesn't extend to the outside, analyte may get trapped in it. The pressure fluctuation that occurs at the next injection may then cause some of the trapped analyte to be ejected back into the flow stream; since the detected peak is fairly sharp, one may get the impression that it arises somewhere in the injector. This used to be  a common problem in the days when 2-part ferrules were used (Swagelok...) - these have a big dead volume between the ferrules, which plays havoc if the front one leaks a bit.

Again, if the issue is the valve, or material made to the valve or any hardware issue, I would see in the column 2, 3, and 4. I did not see it so it is not the hardware issue. It is the column issue.  This is 25% carry over on the column 1 which is "huge" and not a typical carry over that we normally would see from autosampler.  The column 1 was tested on two different instrument (Shimadzu LC-API 4000 and Shimadzu LC-API5500) and the result is exactly the same (20% carry over after injection water immediately after standard). So it is not the hardware issue.

Narong, as stated many times in the previous posts, the problem was likely to be found (and according to you was) in "either the column or the injector". The column is part of the flow path and in your case it sounds like the column was contaminated, thus causing the problem (extra peaks). Columns are consumable items and one of the very first troubleshooting steps is to replace the column with an identical, new one. It sounds like you have been using many old and contaminated columns over the years. It is good practice to dedicate specific columns to specific samples or applications, not mix them (columns are cheap, your time and instrumentation are not). Label the column for the specific application and keep track of its' use in log book. Having dedicated columns (with spares as backups) are the norm today. Once the column surface has become contaminated or chemistry changed (as in the case of fouling which does not wash off or when used with ion pairing agents), the column should not be relied upon for any important work. After all, the goal is to be able to reproduce each analysis so this is best accomplished with fresh columns, high quality methods, dedicated wash methods used after each analysis and of course regular instrument maintenance and operator training. Careful attention to these areas will insure minimal down-time and excellent, reproducible methods.

Good luck to you in your work and training.