Split Ratio 0.1 to 1

[GaryR]

We have a residual solvents method we are performing a tech transfer on. It is a headspace method which has a splitless injection. We don’t seem to be able to setup our system for splitless injections without getting carrier pressure problems which cause instrument shutdown. We have performed a trial run using a split ratio of 0.1 to 1 which seems to work ok and gives similar peak response and rt to the example chromatogram supplied.
We are looking at acetone and DMF in a pharmaceutical API. Solvent is DMA and internal std is tert-butanol. We seem to have a strange carry over problem with the DMF.
We have PE headspace and Agilent 6890 GC running Empower CDM as instrument control and data acquisition.
The DMF does not appear to be a solvent contamination.
Is it ok to run a split ratio of 0.1 to 1 or might this lead to carry-over?
We do not have any problem with acetone of tert-butanol. Peak area for these two are very consistent but the DMF response is all over the place (RSD of 6 injection around 11%). Running a 1/10 dilution of the Std gives us 1/10 the area of the acetone peak but ½ the DMF peak (with RSD of 6 = 30%). Decreasing area for the diluted std kind of point towards carry-over, but solvent blank injections between working stds and dilution do not show DMF peak.


HS conditions:
Sample oven: 120C
Needle: 165C
Transfer line: 165C
Cycle time: 30min
Thermostatting: 10min
Pressurisation: 1min
Injection: 0.1min
Withdrawal: 0.2min
Gas pressure: 15psi
Vial venting: on
HS mode: constant
Sample Shaker: on.

GC Conditions:
Column: DB-624, 30mx0.53mm, 3um film
Carrier: Helium @ 4mL/min
Oven: 40C for 5min
Ramp: 20C/min to 150C, hold 3min, 30C/min to 220C, hold 5min
Inlet temp: 165C
Detector (FID): 250C

[mbicking]

I have not used the PE system, but am validating the USP residual solvents method on an Agilent HS connected to a 6890.

Does the PE system have a separate gas flow, or does the carrier go through the HS unit before the GC? (Is there a separate transfer line or is it all connected in series?)

You can operate at a split ratio of 0.1 but the 6890 has trouble maintaining a split flow that small. You may see some variability.

[GaryR]

The carrier is plumbed through the headspace to the GC via the transfer line. The vial is pierced by the HS needle and the vial pressurised for a set time (1 min). The valve is then opened for a timed injection (0.1min) from the vial directly to the transfer line. The transfer line in fitted directly into the GC inlet. The carrier pressure is specified for 15psi at the headspace. GC flow of 4mL/min gives a pressure of around 4.5psi at the start of the run, increasing to around 6psi at the top of the temperature ramp. So there's plenty of pressure there if we need to increase the split ratio. However, I'm a bit concerned that increasing the split will lead to decreased peak area and the response for DMF is already quite low.

[Peter Apps]

If you can get a clean blank immediately after a standard the problem is not carryover.

If you have 15 psi in the vial and 4.5 psi at the inlet there will be a very sharp pressure increase in the inlet as you inject, this causes all sorts of problems with split ratio, and sends extra gas out through the septum purge. 4 ml/min is on the low side for a 530 um column, you could increase inthe inlet pressure somewhat, and decrease the vial pressure. The puzzle is that acetone and butanol are repeatable, but the DMF is not - anything to do with bulk flow should affect all of them equally. How repeatable is the repeatability ? - in other words how many batches of six have you run ?

Is the transfer line connected into the carrier gas feed to the inlet, or does it go in through the septum (or some other way). What kind of inlet liner do you have, is there any glass wool in it, and when was it last changed or cleaned ?

You have a cycle time of 30 min (incidentally why are you running the temperature up so high ?) and a sample equilibrium of only 10 min - try increasing the equilibrium time.

How are the peak shapes - any tailing ?

Is it only this method that gives a problem on this instrument, what other analyses does it run successfully ?

Peter

[GaryR]

The transfer line is connected to the inlet through a septum. The inlet liner is not packed with glass wool. RSD for 6 injections of acetone is around the 1 to 2%, as are the 1/10 dilutions. We have only run this method a couple of times, with the same outcome each time. All other analysise are performing as expected. Unfortunately, we aren't able to change too much as this is supposed to be a validated method. We have tried to talk to the people it has come to us from, with little useful success.
Peak shape is quite good, with no tailing. Not too sure why the extreme ramp, the DMA solvent peak elutes around 10 minutes. If it was a problem of equilibration I would have thought they'd have seen that during validation. (But it wouldn't be the first 'validated' method we've gotten which didn't work).
My thought is to increase the split to at least 1:1, or 1:10 to see if it helps.
We have another system that I'd like to try this on to see if it's system specific, but instrument availability is a problem - QC samples take priority, and the other headspace doesn't have a vial shaker, not that that should matter to Std injections.
We ran a series with: Solvent Blank, 3x 1/10 dilution, Solvent Blank, 3x 1/10 dilution, Solvent Blank. This gave decreasing areas for DMF throughout the run. Strangely, the solvent in the previous run (same bottle) did not have DMF peaks. This is why I'm thinking carry-over, or hold-up in the septum purge trap(?). My first thought was the glass wool in the liner, but it is an unpacked liner.
??

[Peter Apps]

Just to be sure; you injected a series of vials with solvent in between standards, and (1) you saw DMF in the injections from the solvent (2) the peak area decreased during the series, in other words the injection from the last standard had a lower peak area than the injection from the first standard, and that those in between had areas smaller than the injection from the previous standard and larger than the injection from the next one. ?

This could be due to the variability per se, its not all that unlikely to get a spurious trend over six injections.

I presume that this is a fresh vial for each injection i.e. you are not making repeat injections from a single vial.

Increasing the split will make life easier for the EPC, which should improve repeatability. A proper splitless injection will work better than a 0.1 split ratio.

Check your septum purge flow rate - you could be losing large amounts of sample through it if it is set too high.

Peter

[GaryR]

Thanks Peter.
Yes, you have read it correctly. Dimishing area of DMF throughout the series of solvent and diluted Std. Solvent injections in previous run had no DMF peaks. (?) This is what makes it so strange. All runs we have done (admittedly only 3 or 4) have show decreasing area for DMF - Both standards at limit concentration, and the 1/10 diluted standards. The RSDs for the limit standards whilst still higher than acetone were within system suitability limits (NMT 10.0%)
If it was due to variability per se, wouldn't there be a similar variation to the acetone and internal std?
Fresh vials for every injection.
I will try both increasing the split ratio, and doing a splitless run. But as I say we have experienced EPC issues with splitless injections. I'm not sure why, I will need to talk to the main analysts.
Australia day public holiday here today, so I will let you know how things go this week.

Many thanks for your suggestions.

[krickos]

Hi Gary

Can not comment much on the PE HS conditions as I have not used that much and not used them for a long time

but I can not see the logic behind the pressure spike being the issue as "only" the least volatile solvent seems effected?

However I am a bit puzzeld about the temperatures:

Provided that DMA is Dimethylacetamide the bp is about 165-166°C given the HS oven temp of 120°C you will inject a fair amout of that.

The bp pf DMF is about 155°C.

When using DMF as sample solvent a transfer line injector/temp of 165°C is typical sufficent, however in this case you want to analyze the DMF. It is just possible that DMF starts to tangle (interact) with DMA on surfaces as 165°C is exactly at the bp of DMA. Decreasing areas indicates some kind of loss.
Add the low flow and..perhaps

Increasing transferline/injector temp some 20°C could help if thats the case. Regardless if it the cause or not, temperatures in needle/sample valve/loops/tranferline/injector should exceed the bp of the highest bp of all solvents including sample solvent to minimise the risk of interaction with surfaces and to "keep it clean".

Typically Agilents types of HS are more prone to this (have cleaned a fair amout of needles and loops ) and it is not always spotted right away as it is a problem that can build up over time depending on circumstances.

[Peter Apps]

The most likely cause of a decline in areas is leaky vial caps and this will also gibe poor repeatbaility, but once again I would expect this to affect all the analytes. It is worth checking anyway - put some volatile solvent in a vial, crimp the cap as usual and then immerse it completely in hot water, a stream of bubbles shows a leak.

Carryover has to be analyte sticking somewhere in the flow stream (in which case increasing temperature should help) but with the vial pressure being so much higher than the inlet pressure it could be due to backflow into the carrier gas feed line, the lighter analytes blow straight out again but the heavy one sticks to the wall of the pipe.

It starts to look as if you have multiple problems - marginal temperatures, strange split ratios and pressures and possible leaky vials. Fixing them one at a time will not make much difference, you need to check out the whole system thoroughly.

Peter

[krickos]

Carryover has to be analyte sticking somewhere in the flow stream (in which case increasing temperature should help) but with the vial pressure being so much higher than the inlet pressure it could be due to backflow into the carrier gas feed line, the lighter analytes blow straight out again but the heavy one sticks to the wall of the pipe.
Peter

ahh forgot about that one

[Peter Apps]

I did a back of an envelope calculation, and if the feed line from the EFC to the inlet is 500 mm long (which is not far out for some GCs) and 1 mm i.d. i.e. thin walled 1/16 inch pipe, the internal volume is 392 ul, which is more than double the 157 ul in a 2 mm i.d. inlet liner 50 mm long. Apart from where it joins the inlet, that volume is at room temperature, so anythig that gets in there is going to sit around. In the Varian 3700 (those were the days) and in the really ancient Packard Bell there was a coil of carrier gas feed pipe inside the heated inlet block, presumably to pre-heat the carrier gas but with the beneficial side effect of having anything that went back up the pipe promptly evaporated out again. These days carrier gas feed lines go from room temperature to inlet in a mm or three, and people have started worrying about vapour cloud volumes and inlet liners being big enough to accommodate them. Co-incidence ?

And the reason that GC manufacturers do not put a one-way valve immediately upstream of the inlet is ????????????????????

Peter

[GaryR]

Thanks for the observations.

I don't really suspect leaking vials as the acetone and internal std areas are very consistant, and I would expect the DMF to be all over the place and not showing a fairly consistant downward trend.
I am also leaning towards pressure difference between vial and column. Unfortunately we are not able to change method parameters as this would require validation.
We have two HS systems, both 6890 GCs, one with PE Turbomatrix HS, the other with PE HS40XL HS. The HS40XL does not have a vial shaker.
We performed parallel runs on the two instruments the other day and both gave large DMF peaks in the blank and Int Std. The run using the HS40XL, without vial shaking gave more consistant DMF peak areas. The Turbomatrix, with vial shaking (as specified in method) shows decreasing DMF peak area as follows:
Solvent, 143
Int Std, 103
Solvent, 80
Std, 96 (These are LOQ Stds: 1/10 dilutions of working Stds)
Std, 96
Std, 84
Std, 81
Std, 76
Std, 74 (RSD = 11.3%)
Solvent, 61
Int Std, 60
Solvent, 59

The HS40XL DMF peak areas were:
Solvent, 93
Int Std, 88
Solvent, 88
Std, 118 (These are LOQ Stds: 1/10 dilutions of working Stds)
Std, 118
Std, 115
Std, 121
Std, 121
Std, 119 (RSD = 1.9%)
Solvent, 95
Int Std, 96
Solvent, 94

The only differences between systems are the vial shaking, and the column. RSDs are peak areas only, not ratio of area to Int Std.

We swapped columns from HS40XL system to Turbomatrix system and did a did a full run including working Stds. DMF areas:
Solvent, 56
Ins Std, 57
Solvent, 53
W Std, 226
W std, 233
W Std, 234
W Std, 217
W Std, 241
W Std, 220 (RSD = 4.0%)
Solvent, 58
LOQ Std, 73
LOQ Std, 75
LOQ Std, 75
LOQ Std, 75
LOQ Std, 76
LOQ Std, 78 (RSD = 2.2%)
Solvent, 101
Looks good except for that last solvent injection.

Doesn't make a lot of sense, as our original run used the 'other ' column and gave fairly good results.
As much as I hate the idea, we are going to use this data, and this method 'as is'. Not really my decision to make though.

One last question.
We are using a 0.32 transfer line. What effect will this make to the pressure drop going to a 0.53 column?

[Peter Apps]

Hi Gary

If you cannot change method parameters then what are you trying to do ? Your results show that the method is not robust to changes in instrument, unless you are able to re-optimise that is as far as you can take it.

If you want to understand where the problem arises I would suggest that the next step is to turn off the shaker on the Turbomatrix.

The internal diameter of the transfer line is only significant when you are running at high split ratios with all the gas suplied from the headspacer. If some of the gas to the inlet is coming from the GC volume flow through the transfer line is less of an issue, but there can be some problems due to poor mixing of the gas streams in the inlet. Since they only mix by diffusion this can give discrimination between light and heavy molecules, which is part of what you are seeing.

Peter

[GaryR]

Hi Peter

When I first posted I was hoping that there might be a simple solution to our problem which was obvoius to someone else.

We have just run this on another instrument in our R&D lab with good, reproducible results. So it looks a lot like it's instrument specific, our 6890 with Turbomatrix is giving us poor reproducibility. Now to track down why - it's not a problem for other analysis and passed a recent calibration.

I will let you know how it goes.

Many thanks for your time and suggestions.

[GaryR]

Looks like this was a carry-over problem with the headspace needle.
Replaced the needle and now we are getting reproducible results.

Still seeing large DMF peak in the solvent, but now at least we are not getting diminishing areas. Correcting our LOQ injections for the solvent DMF peak is also passing system suitability requirements.