Mysterious "self-sealing" leak in GC

[swedding]

Background: Agilent 6890GC w/7673 Autosampler. Used for Fatty Acid Methyl Ester Analysis. Well established method in use for > 10 years (minor adjustments made due to column changes, routine maintenance, etc.)

Problem: Operator changed the column on 5/8. (Column is DB-Wax, 15m x 0.25mm x 0.25 um) After changing the column, the first injection of a sample after the instument has been sitting idle for a while, shows reduced sensitivity of ALL peaks and a slight RT shift to higher time. A second injection, from the same vial, directly after will shows correct response and RT's. All other vials in same sequence run are also okay. The oven temp at idle conditions is 130 deg C. Method is Linear Temp Program with final temp of 230 deg C held for two minutes to bake out column.

Troubleshooting Steps Taken: 1. Syringe was changed and problem seemed to stop, only to return a day or so later. 2. Check of column nuts and underside of inlet revealed that the ferrule was not crimped down against the gold seal. Changed gold seal, inlet liner, new ferrule and recut column before replacing in inlet. Changed septum and tightened nut. Checked and tightened detector nut. Okay for a day or two 3. Problem reappeared. Detector side of column was removed and ferrule showed proper crimping and correct length inserted into FID detector. Replaced and leak checked. No detectable leaks were found 4. 5/23/12 - Service Engineer came in.........Blank inlet nuts and purge gas caps were placed on instrument by service engineer to isolate inlet for pressure check. All pressure checks and flow checks were normal.

In 30 years of doing GC, I've never seen this happen. Only the 1st sample of a sequence run is affected, so it's not related to anything that would affect every sample (dirty FID, clogged purge line, etc.). It's like a leak that only shows up when the instrument sits idle and then once everything is warmed up and "expanded" after the first run, it seals up and runs everything normally.

Has anyone else ever seen this happen before? If so, what caused it? Even our service engineer is stumped by this one. I'm fresh out of ideas.......

Thanks

[Peter Apps]

Do you cool the inlet when the instrument is on standby ?

If you put a brand new septum in at the end of a session, then go to standby, is there a "leak" on the first run when you start up again.

If I recall you have an option on the GC to set initial conditions for the next run at the end of the previous one - if you do not do this the inlet has to change e.g. split ratio as the injection is happening, which is bad for peak area but should not affect retention.

Peter

[R13]

No idea right now, but additional info might generate some:
-Injector temperature/mode (split/splitless/pulsed), septum type?
-Constant flow or constant column pressure method?
-Liner O-ring: graphite or silicone?
-Injection volume? Autosampler injection mode (air gap/solvent flush/sample in the needle)? Did the service do check-out testing, is RSD normal? Did You observe change of injection RSD over longer period of the method use?
-Sensitivity decrease - 0.5%, 1%, 2%, 10% ??
-How the instrument is "put to idle" - just "detector off" or temperatures/injector pressure and flows are changed?
-New batch/other type of septa?

You mentioned first occurence after the column change. Did you observe signifficant retention time shift as compared to old column? Do you have possibility to check injections repeatability with other column of the same type?

[swedding]

Peter and R13,
To answer your questions:
Leak of brand new septum install at end of session? Prior to this column change, No, we never saw a leak on the first injection after installing a new septum. Isolation of the inlet by the service engineer showed No leaks after changing septum.

IDLE conditions = Initial Run conditions for the method. This keeps the instrument in a "ready" state for injections. GC is run 24/7 in a process lab environment.

Injectors and detectors are NOT cooled in idle state. All Temps are kept at method settings.
Inj Temp = 210C, Det Temp = 240C, Injection Mode = Split (20:1), solvent flush mode (6 solvent flushes before and after sample to clean syringe using methanol and hexane - 3 flushes each solvent) 3 sample washes and pumps prior to filling syringe for injection. Injection RSD's have not changed over time.
Method is Constant Flow with Initial flow rate 2.0 ml/min. O-ring liner is non-stick fluorocarbon for the flip inlet system.
All O-rings, inlet liner and septum changes during the 5/8 maintenance were from same type and lot as previous changes. Column was same type, rec'd Feb, 2012.
There were NO significant retention shifts between previous column and this column. RT's were within 0.03 seconds between the two columns.

Again, this is a Well Established method that has been run for over 6 years on a 24/7 basis..........with no problems during normal routine maintenance items such as septum changes, liner changes, column changes, etc.

"Something" occurred during this column change that has caused an issue where it looks like a leak happens on first injection of a sequence, after the instrument has been sitting idle, but not on subsequent injections IN the same sequence run. We do have another brand new column that we can change to. However, I would think an issue with the column or column phase would cause ALL results to be off, not just the first injection.

I appreciate your ideas and anything you can think of to check. This one has me stumped.............

[Peter Apps]

It's a puzzle for sure !

How big is the drop in sensitivity, and how much does the retention shift ? For a leak to cause a retention shift in a back-pressure regulated system the leak has to be very large - and the peaks would get markedly smaller. Check whether you lose split flow at any stage.

What happens if you run a temperature programme with no injection as the first run of the day, are the peaks of the first real injection still smaller and shifted ?

Do you have oxygen and moisture scrubbers on the carrier gas line ?, and have they been replaced lately ? What carrier gas are you using ? I have known a wax column change in retention with wet samples and no temperature programme - if there is moisture in the carrier gas the column will pick it up overnight on standby and lose it during the first run. Are you measuring peak heights or peak areas, and is the resolution affected ?

Peter

[KM-USA]

We've ALWAYS discarded/rejected the first injection of any sequence, GC or HPLC.

[swedding]

Peter.....drop in sensitivity is approximately 0.5% using normalized area percent. Retention shift is about 0.1 to 0.2 seconds higher for the first run than subsequent runs (this to me indicates a lower flow on the first run).

Based on flow rate checks the service engineer did yesterday at different stages of a blank run, we don't appear to lose split flow at any stage. He also checked the purge vent and found no leaks or change in purge pressure/flow.

Have not tried a blank run with no injection as the first run of a sequence. Will try that and let you know what happens.

We use gas generators. Hydrogen is the carrier gas and flame gas. Zero Air and Nitrogen (make-up gas). All generators have scrubbers to remove hydrocarbons, oil, moisture, etc. Scrubbers/filters were changed last week during annual plant shutdown. No moisture should be getting to the column, either from the gases or the sample injection. Sample is filtered through sodium sulfite to remove any residual moisture that might occur from sample methylation prep.

We measure peak areas. Both height and area are affected in the first run. Resolution is not affected. Still see same separations and same amount of time between peaks.

[chromatographer1]

Doing a run without an injection is a good diagnostic test.

Also injecting a blank solvent injection following that "no injection" run is a good diagnostic test.

The problem might lie in the splitter trap. If the first injection of sample through the trap changes the flow characteristics of the trap (the first allowing more sample to be split than all the following injections) a slight reduction in sample reaching the column COULD occur.

A new trap might fix the problem. They are cheap and should be replaced regularly anyway.

Good problem and good responses (except for mine). I will enjoy hearing the end of the story.

best wishes,

Rod

[Peter Apps]

Peter.....drop in sensitivity is approximately 0.5% using normalized area percent. Retention shift is about 0.1 to 0.2 seconds higher for the first run than subsequent runs (this to me indicates a lower flow on the first run). These are really small changes, which makes the troubleshooting even more of a challenge. O.1 - 0.2 s must be close to the resolution with which you measure retentions - it could be that you are simply seeing a measurement artefact. If this was the case there is no particular reason why it would only occur onthe first run. Guessing a run time of 15 - 20 min ??, if this is due to a change in flow then the flow has decreased by 0.02 % - which is surely way smaller than the repeatbaility of any flow controller, and absolutley impossible to measure with any ordinary flow meter. If all the peaks are shifted by 0.1 or 0.2 s then my guess is that it is simply an artefact of the resolution with which the GC detects the injection time. What are you normalizing areas against ? - it if is an internal standard than its area should be affected by anything wrong with the instrument to exactly the same extent as the methyl ester peaks, which would make any change in peak area undetectable. What change is there in raw peak areas ?.

Based on flow rate checks the service engineer did yesterday at different stages of a blank run, we don't appear to lose split flow at any stage. He also checked the purge vent and found no leaks or change in purge pressure/flow.

Have not tried a blank run with no injection as the first run of a sequence. Will try that and let you know what happens.

We use gas generators. Hydrogen is the carrier gas and flame gas. Zero Air and Nitrogen (make-up gas). All generators have scrubbers to remove hydrocarbons, oil, moisture, etc. Scrubbers/filters were changed last week during annual plant shutdown so the problem occurred when the moisture scrubber was getting to the end of its life - you might see an improvement now that the scrubber is new, it will take a while for the pipes and GC plumbing to dry out. No moisture should be getting to the column, either from the gases or the sample injection. Sample is filtered through sodium sulfite to remove any residual moisture that might occur from sample methylation prep.

We measure peak areas. Both height and area are affected in the first run by how much ?. Resolution is not affected. Still see same separations and same amount of time between peaks. Which answers my question above about all the times shifting - you have an injection timing issue. The difference is so small that it could be due to something as obscure as the drive belt on the syringe plunger being a bit stiff early inthe morning. What super- precise analysis are you doing that this kind of change actually makes a practical difference ?

Peter

[swedding]

Rod........Service engineer thought of the split trap filters..........he returned yesterday with new Split Vent Assembly Trap filters. He removed the trap lines and flushed them out with hexane, then replaced the trap filters. It appeared to fix the problem............until the instrument sat idle for a few hours. I had the night techs run duplicate injections in the instrument (they run a sample every two to four hours). The 3 samples run during the night show the sample problem. First injection was lower response and calculated amount of each fatty acid than the second injection. Ex. C16:0 on first injection = 3.80, second injection = 4.45. In the world of fatty acid analysis, this is a huge difference.

Peter.....this answers your question. Total Saturated fat levels for Canola oil average 7.0%. Lower limit is 6.7, upper limit is 7.3. So if the first injection shows a Sat of 6.54 and the second injection is 6.94 of the same sample, the first is Out of Limit low. We use these values to make sure 1) the fatty acid profile matches the oil type, 2) check for commingling of different oil types that can occur during changeover (plant limit is less than 2% commingling allowed). So these values are kept within tightly controlled limits.

Typical back-to-back injections from the same vial prior to this problem show repeat analysis within 0.03 to 0.05% absolute value of each other. So seeing a difference of 0.5% between injections is a huge difference. Normalized area percent of 100% oils samples is based on ratio percent of each fatty acid peak to the total area of all peaks added together. An internal std is not used. I'm dealing with operators with NO chemistry experience or education who prep and run their own samples, so the analysis has to be kept simple. Total run time of the method is 19.83 minutes, including the bakeout and equilibration time to ready the instrument for the next injection.

Time of day of the injection makes no difference. Lab is a controlled temperature and if anything is on the warm side. Total areas and peak heights don't "appear" that much different (ex. C16 1st run peak area = 414, 2nd run peak area = 446), however, since everything is ratioed to total area even small peak area differences can then "normalize" into a much larger percentage difference.

We have purchased standards that we routinely analyze. These have certified values associated with each fatty acid peak. These values are outside the certified range on first injection and correct on the second injection. We have a daily check sample of an oil that is prepped and run first thing in the morning each day. The values are tracked and recorded on a control chart. The daily check sample shows the same problem. 1st injection is low, 2nd injection values are within the allowed range of each fatty acid chain.

When analyzing 100% fats and oils, normalized area percent is the standard method in the industry..........normalized area percent is equal to g/100g using actual FID response factors and internal standards. Again, my shift operators have only basic chemistry skills and things have to be kept simple.

I am waiting for the instrument to sit idle for a few hours, then I plan to do a blank run, followed by a solvent flush.

At this point, I am starting to suspect there is something wrong with the phase in the new column. I plan to do a column change next week to see if a different column corrects the problem.

I appreciate all the responses and ideas. It was a comment Peter made about "Moisture" that made me wonder about the split vent trap and lines.

I will keep you posted on the progress..........and if I ever figure out the problem!!

[chromatographer1]

Columns and injection ports can have active sites which can affect the amount of analyte reaching the detactor, although FAMEs are not usually associated with these problems. It could be something that was injected and deposited upon the column.

That is why an injection of just solvent may not 'fix' the problem, but it would be a bit of information that might be valuable. If it requires an actual FAME sample to 'fix' the problem that also is a bit of useful information.

Certainly a replacement column should be tested to see if that affects the outcome. Otherwise, inject a sample first whenever the GC sits for a while before you make injections that 'count'

Thanks for letting us know the progress and hopefully the final solution.

Rod

[swedding]

To all who responded..........We have corrected the problem and are now back to normal.

Back-to-back injections of the same sample after the instrument has sat idle for a few hours show the same "expected" responses and the chromatograms overlay almost perfectly (no retention time shifts).

The Solution: We removed the column and replaced it with the "old" column that we had taken out when we put in the new one. That fixed the response/time shift problem, but the old column showed a lot of "ghost" peaks due to the phase being worn out.

After we confirmed it was actually a problem with the newer column, we ordered a second new column. We replaced the old column with the 2nd new one yesterday, baked out, did a solvent flush and then analyzed our standards and recalibrated. The night shift operator did some back-to-back injections of both a standard and an actual sample (after the instrument had sat idle for over two hours) and all of the back-to-back runs of same sample were identical.

I have never seen this problem with a column before, but I guess there is a first time for everything. I apologize for the delay in posting the solution. I was out of the office for a week which delayed the repair.

I appreciate all of the great responses I rec'd and look forward to being able to help some of you with your future GC issues.

Thanks,
Susan

[Peter Apps]

Hi Susan

Thanks for the feedback. Have you asked the column manufacturers for their comments ?

Peter

[chromatographer1]

Tubing manufacturers, Agilent makes their own tubing, can sometimes make a batch which will exhibit brittleness or cracks in the tubing which the coating of phase will seal from inside the column. When coated with solvent from the injection (for only a few seconds) this 'seal' can fail due to swelling of the phase coating from the solvent and begin to leak. Sometimes it stays leaking and sometimes not.

I suspect the leak was due to a crack (or defect) in the column tubing. This is not a common problem but it can happen and when it does it can be quite irritating and difficult to locate.

I would ask for a refund with the column and return it to the vendor.

best wishes,

Rod