The CHROMacademy Essential Guide to GC Troubleshooting Masterclass
Thursday 21st February 2013, 11:00 AM EST, 16:00 GMT
Dr John Hinshaw and Tony Taylor demonstrate practical GC separation troubleshooting. Using real world chromatograms submitted by CHROMacademy members from many different application areas, they illustrate how to recognize problems and explain further tests and investigations necessary to properly qualify the issue. They then offer advice on how the problem may be resolved to help you gain valuable insight into recognizing, characterizing and fixing problems in GC to improve your data quality and reduce instrument downtime..
Practical GC Troubleshooting:
Learn how experts troubleshooting GC chromatography problems
A series of real chromatograms are used to demonstrate logical ways to approach various issues in GC
Learn what problems and symptoms are important and how to further investigate issues
Learn what further tests can be done to better understand problems and fix issues quicker
Learn to recognize problems quickly and and achieve better data quality and higher throughput
Who Should Attend:
Anyone working with Gas Chromatography
Anyone wishing to reduce their instrument downtime or become more effective at troubleshooting GC separation problems
Find out more about this Month's Essential Guide Webcast »
The CHROMacademy Essential Guide Tutorial GC Troubleshooting Masterclass
CHROMacademy (www.chromacademy.com) is an e-Learning, Troubleshooting and Continuing Professional Development resource for the global chromatography and mass spectrometry communities.
The site is developed by LC-GC magazine and Crawford Scientific, to bring you world class learning and problem solving in analytical science. We combine the best in multi-media learning materials with access to leading scientists, a global forum of chromatographers and the latest analytical chemistry information from instrument vendors and academic thought leaders around the world.
In development for over 12 years, CHROMacademy builds the training, application and problem solving resources you need, so that you don’t need to waste time looking anywhere else!
Dr John Hinshaw and Tony Taylor demonstrate practical GC separation troubleshooting. Using real world chromatograms submitted by CHROMacademy members from many different application areas, they illustrate how to recognize problems and explain further tests and investigations necessary to properly qualify the issue. They then offer advice on how the problem may be resolved to help you gain valuable insight into recognizing, characterizing and fixing problems in GC to improve your data quality and reduce instrument downtime.
We get many questions at CHROMacademy about peak shapes in GC, tails, shoulders, splitting, fronting we have seen it all. Sometimes the problem is physical, and due to some system component causing flow or analyte transfer disturbance, and sometimes it’s due to unwanted secondary chemical interactions. The problem below proved an interesting challenge and serves to illustrate the various approaches one might take to troubleshooting peak shape problems.
We hope that it also demonstrates how one might consider all of the method parameters to see if these are appropriate.
Figure 1: Separation of long chain alcohols and alkanes as per conditions outlined below, showing peaks shoulders / splitting after only a few injections.
Good and bad peak shapes obtained from the analysis are shown for comparison purposes.
The long chain alkane and alcohol show shouldering / splitting peak after several injections
Poor peak shape is causing issues with integration and quantitation of components
Docosanol and Octacosane show the behaviour but benzyl alcohol does not
Sample (topical cream) prepared in dehydrated alcohol at an appropriate concentration
The issues which need to be considered include:
Is this a chemical or physical issue?
Why only later eluting peaks with the problem?
Sample is a diluted cream formulation with little sample preparation – could the matrix be a factor?
Why does the phenomenon ‘appear’ during analysis?
End user notes the problem doesn’t occur with every injection (i.e. some injections may be ‘OK’)?
Dehydrated alcohol is >98% ethanol (b.pt. 78oC) – are the splitless injection conditions suitable in this respect?
Is the splitless time appropriate?
End user wonders if high water level in the formulation may be a contributing factor?
Is 300oC too high a temperature for the inlet?
So, to begin to troubleshoot this separation, let’s break down the possible causes and examine the likelihood of each factor in turn. As we noted during the webcast, there really is no substitute for experience, however looking logically through the possible causes and method parameters can provide a logical troubleshooting routine.
The stationary phase chemistry appears to be well matched the sample. The major factor in separating the components will be dispersive separation due to analyte boiling point differences; however there is the slightly more polar phenyl functionality to help differentiate ‘chemically’ between alkanes and alcohols of similar chain length.
Figure 2: 5% Phenyl PDMS Column Repeating monomer unit.
We noted that the injection is splitless, which means that we have to consider several factors in analysing the method conditions to assess their suitability. As usual when investigating any splitless injection the following questions should be asked;
Is splitless injection really necessary?
Is the initial oven temperature properly matched to the sample solvent to give good analyte peak ‘solvent focusing’ and also good thermal focusing of higher boiling analytes?
Is there a suitable purge time after which the split line will be opened to purge any vapors which would otherwise bleed from the inlet giving solvent peak tailing and a rising baseline and/or baseline artifacts for the duration of the separation?
Is the chemistry (polarity) of the stationary phase matched with that of the sample solvent, to ensure good analyte solvent focusing
Is the injection volume low enough to prevent ‘backflash’ in the inlet, leading to carryover and peak shape effects?
If you are not familiar with the questions and considerations in the above list, there are many tutorials, webcasts and e-learning modules within CHROMacademy to help you. Click here to access the information – you will need to register as a Lite Member to access free content and as a Premium Member to access restricted content.
We do not know if splitless injection is really necessary for this analysis because we were not given information regarding the possible sample concentration range. Remember that we typically use splitless injection when analyte concentration is low and we need the entire injected sample (vapor) to enter the column in order to increase response. We will assume that a sensible decision was taken in this regard.
We usually recommend that the initial oven temperature for splitless analysis is between 10 and 20oC below the boiling point of the sample solvent. Initial oven temperature here is 70oC and if we assume a boiling point of 78oC for the sample solvent, this would seem to be inadequate for good peak shape during solvent condensation – as illustrated in the animation in Figure 3.
Figure 3: Thermal peak focusing processes in splitless GC injection.
Here we have a slight conundrum however. Poor solvent focusing (which is predominantly controlled by initial oven temperature) typically effects the lower boiling (more volatile) and therefore earlier eluting compounds to a greater extent. In this case, the problem is reversed, with later eluting (less volatile) components suffering peak shape denigration to a much greater extent than earlier eluting ones.
So – here we have a possible explanation but one which doesn’t exactly fit the troubleshooting picture we usually see. To overcome these issues, the initial oven temperature should bet set to 50 – 60oC and held isothermally for at least the same time as the splitless (purge) time within the experiment.
A purge time of 36 seconds (0.6 mins.) seems reasonable – but we should investigate how many sample volumes can be cleared from the inlet under these conditions.
Figure 4: Flow / pressure calculation for the method under investigation.
Figure 5: Solvent expansion calculations for the method under investigation.
We calculated the inlet pressure for the conditions given to be around 1.6 psi. We’ll come back to further comment on this in a moment. Using this pressure we were able to calculate a solvent expansion volume of 727 μl under the conditions of the experiment. At a flow rate of 10.6 mL/min., in splitless mode, a total of 6360 μl of gas will flow through the liner in 36 seconds, which is around 9 times the vapor volume created on injecting the sample. This should be enough to clear the inlet of our analyte, however when very high boiling analytes are concerned, this calculation may not be as straightforward as we have outlined here – again more on this later in the tutorial.
One point to note regarding the calculation in Figure 5, is that the solvent vapor volume created during injection is close to the internal volume of the inlet liner. If the liner volume is exceeded, analyte spills into the inlet and outlet lines of the injector and condenses, which gives the potential for carry-over during subsequent injections, obviously leading to poor quantitative results. In this instance it is recommended that the injection volume is reduced to 0.5 μl if the required sensitivity can still be achieved.
A further potential issue of note arises from the use of the flow calculator software, regarding the carrier gas pressure. Whilst modern EPC systems are capable of very accurate and reproducible control of gases, at lower pressures the accuracy or reproducibility of the gas flow control may be lessened. This isn’t likely to be related to the problem at hand, however is worthy of note as a general point.
The sample solvent is ethanol, with a small amount of water and the stationary phase is 5% phenyl PDMS – which may be viewed as somewhat of a mis-match in polarity. This potentially causes the solvent to condense into ‘pools’ on the inner surface of the column, rather than as a contiguous ‘film’ which can evaporate to focus the analyte in a single spot within the column. Where separate ‘pools’ of solvent form on the column inner surface, this can give rise to just the type of peak shape which is shown in the problem chromatogram. Again, this behaviour typically affects the early eluting compounds within the chromatogram, as the more volatile compounds remain within the condensed solvent, rather than focusing via thermal means. This behaviour is highlighted in Figure 6.
Figure 6: Problems with peak shape created by mis-match between stationary phase polarity and sampl solvent polarity.
So – could the issue be a ‘physical’ phenomenon? Peak shapes of this type are typically associated withissues in which the analyte entry onto the column is disturbed in some way. This might typically include:
Poorly cut column
Column installed in the wrong position within the inlet
Stationary phase stripped at the column inlet
Again, one would expect that the peak shape issues would occur with all peaks within the chromatogram if the effects were physical. However to absolutely rule out physical effects, one should trim a few centimeters of column (note with 10 meter columns, removing too much column will give a marked change in analyte retention time), and re-position the column within the inlet.
Temperatures are typically the next thing we would consider when reviewing a GC method.
275oC is very close to the upper isothermal temperature of the column (260oC in this case). This shouldn’t be a problem with better quality columns, especially if this temperature is only sustained for short periods of time. However, the inlet temperature is at 300oC, obviously above the recommended upper temperature limit for this phase. There is some debate as to whether this situation would cause stationary phase loss (given the usually sharp temperature gradients within inlets etc. etc.). However, it remains a possibility that stationary phase is lost from the inlet end of the column – so disrupting the peak focusing processes which typically occur in splitless injection.
This raises the question regarding the necessity for an inlet temperature as high at this. The boiling point of octacosane is 431oC, and whilst the absolute boiling point of analytes is meaningless in capillary GC, even under the reduced vapor pressure conditions in the GC inlet, this component will need a high inlet temperature in order to effectively volatilize. In fact, it may be that the peak shape issue is being caused by slow volatilization of the higher boiling analytes – something which we would certainly recommend that the end user investigates. Of course – if this is found to be the case, the use of an uncoated retention gap would be necessary to avoid phase loss from the portion of the column inserted into the inlet.
And so – from the protracted discussion above, it’s obvious that whilst there are no outright or ‘glaring’ causes for the peak shape issues under investigation, there are a number of possible causes. I’m sure you will agree that most separation troubleshooting is like this – a series of ‘loose bolts’ which , when tightened, can lead to a much better separation.
To summarize then on the possible issues and what needs to be done:
Investigate lower initial oven temperature with isothermal hold to match the splitless time
Consider reducing injection volume to avoid problems with backflash (unlikely to solve this issue but good practice none the less)
Consider use of a retention gap or trim the column to remove any damaged phase (retention gap may also help y slightly increasing the required pressure to operate at the desired flow rate – especially if a narrower bore retention gap is used)
Investigate inlet temperature – use retention gap for increased temperatures
Consider phase mismatch and use different solvents with intermediate polarity (ethyl acetate would be a possibility for example)
Figure 7: Differences in peaks shape caused by the use of 4mm (left) and 2mm i.d. (right) liners in splitless injections – conditions shown.
Figure 7 illustrates another problem that we regularly see with splitless injection and once again it is related to establishing appropriate splitless conditions.
Whilst the volume of sample vapor generated is relatively low - again this was calculated using the Vapor Volume calculator from Agilent technologies
the problem is more directly related to the time take to clear the liner and transfer the analytes on to the column.
The 4mm i.d. liner has an internal volume of 850 μl whilst the 2mm liner has a volume of only 250 μl. Figure 7 shows the difference this can make – the peak shape using the 2mm liner is much improved as the transfer of the analyte vapor onto the column occurs 4x faster (at the same carrier linear velocity).
2mm liners are frequently used in fast GC where narrow internal diameter columns are typically used, and any ‘extra column volume’ results in gross analyte band broadening (via sample vapor diffusion).
Of course, in the above example the 4mm i.d. liner could have been used if the ‘rules’ of splitless injection had been followed more closely.
Figure 8 demonstrates the situation obtained from splitless injection of the sample but starting with an oven temperature 25oC lower than that shown in Figure 7 (25oC v’s 50oC in the example above).
Under these conditions, even though the analyte ‘bleeds’ from the inlet onto the column more slowly, the solvent focusing effect can occur which will involve evaporation of the solvent, focusing the analyte into a tight band as explained in the previous section.
Figure 8:Conditions as per the upper figure 7 but with a lower initial oven temperature, ensuring good solvent focusing under splitless conditions.
We often see a similar phenomenon in split injection where the initial oven temperature is too low or there is a long isothermal hold at the beginning of the temperature program, which allows the analyte to diffuse within the stationary phase, giving rise to broad peaks.
Figure 9:Poor peak shapes (split / shouldered) from the analysis of a test compound mix of analytes with various polarity in methanol using splitless injection with the conditions shown below.
5% Phenyl PDMS, 30m x 0.25mm x 0.25μm
40oC, 2 mins.
40oC -175oC, 1.5oC/min., 20 mins.
Splitless (purge 2.0 mins)
1 μL of 2 ppm each (1) 1-octanol, (2) n-undecane, (3) 2,6-dimethylphenol, (4) 2,6-dimethylanaline, (5) n-dodecane, (6) n-tridecane in methanol
Again another very common problem which comes across the desks of John and the technical department at CHROMacademy!
The peaks shapes shown here are awful and there appears to be a larger number of peaks than there are sample components.
Typically we see split peaks of this form under the following circumstances;
Poor column end cut
Poor installation of column into the inlet (typically column too high in the inlet)
Issues with peak focusing with splitless injection
Damaged / occluded phase at the column inlet or ferrule / septum shards stuck in the column inlet
So, again, the issues can be due to both physical and or chemical problems with the separation.
When the problem is as severe as seen in Figure 9 and as splitless injection is being used we typically investigate the splitless conditions first, although if were in the laboratory at the time the problems we first encountered we would probably also recommend that the column is re-trimmed and re-installed in the inlet…..just for good measure.
We can see that the initial oven temperature is good – 40oC is more than 20oC below the boiling point of methanol (64.7oC).
The splitless time is a little long (2 mins.) however it matches the initial isothermal hold in the method and whilst it could be a little shorter, which would sharpen the analyte peaks and reduce peak widths, it is probably not the cause of the gross peak deformation we are seeing in the example.
However, as per our first example, we did note that the solvent (methanol) is very polar in comparison to the stationary phase (5% phenyl polydimethylsiloxane - as per figure 2 above). We suggested a change to dichloromethane as the sample solvent and the chromatogram shown in figure 10 was acquired under the same conditions as that shown in figure 8 – i.e. the only change was the sample solvent.
Figure 10: Chromatogram and conditions as per figure 8 for the analytical test mix but sample solvent changed from methanol to dichloromethane.
A beautiful chromatogram! The solvent can form a contiguous film on the inner surface of the column and the subsequent solvent evaporation gives us a single focused peak. Figure 6 above shows an animation of this process.
A beautiful chromatogram! The solvent can form a contiguous film on the inner surface of the column and the subsequent solvent evaporation gives us a single focused peak. Figure 6 above shows an animation of this process.
Figure 11: Chromatogram of a pesticide test mix showing gross peak broadening for later eluting peaks – conditions as shown below.
5% Phenyl PDMS, 30m x 0.32mm x 0.55μm
40oC, 2 mins.
40oC - 300oC, 10oC/min., 8 mins.
He at 2.0 mL/min.
Splitless (Splitless 4mm liner)
Splitless time 2.0 min. @ 50mL./min.
Source – 230oC
Quad – 150oC
Minnesota Mix A Pesticide test mix in acetone
In this example our end user tells us that the peaks in the chromatogram have been very sharp in previous experiments and that whilst retention times have not shifted, the later eluting analytes are showing a gross broadening / low efficiency. The method parameters have not been altered and the method is usually very reliable……
OK – so we have the issue that peaks are broadening rapidly within a relatively short analysis time. The issues that we would typically consider under these circumstances include;
Constant pressure operating mode chosen instead of constant flow – i.e. the flow is reducing as the oven temperature increases
Reduction in column efficiency
Increase in system dead volume due to poorly installed column into inlet or detector fittings
Cold spots within the column (causing a holdup / condensation) leading to stationary phase diffusion of the analyte
If the carrier pressure operating mode is changed to constant pressure from constant flow, one would typically see a change in retention time - retention times for later eluting peaks being longer than expected – to accompany the band broadening. This is not happening in this case and so can be eliminated.
We had the end user check the installation of the column and everything seemed to be in order and the inlet coupling was thermally insulated as expected.
As the column was relatively new – this really left the coupling into the mass spectrometer. By chance – when investigating this, it was noted that the transfer line seemed cool relative to its normal temperature. This was further investigated and it transpired that whilst the instrument was reading out the correct temperature – in fact the transfer line was not being heated. After an engineer fix the chromatography was repeated and the chromatogram in figure 12 obtained.
Figure 12:As per figure 11 but with proper MS transfer line heating.
As can be seen in figure 12, the chromatography was returned to normal and we surmise that the problem was caused by (partial) condensation of the column effluent vapors in the transfer line.
Two salutary lessons should be noted from this exercise.
Don’t always believe what the instrument tells you – manually check flows and temperatures from time to time or as part of your PQ/OQ checks
2. Any cold spots in the instrument (typically caused at junctions between metal fittings and the capillary column) should be properly heated / insulated in order to prevent condensation issues and the peak shape effects we have seen here. This also applies to static headspace analyzers using autosampling equipment – it’s important the that the sample, sample loop, transfer line and inlet are all at slightly increased temperature from each other to prevent this type of phenomenon – which we often see in our troubleshooting and technical support work.
benzaldehyde, decane (internal standard), benzyl alcohol,
benzoic acid in DCM
Here we see an example of bad peak tailing during solvent analysis (conditions shown) on an instrument which had had some problems but which had undergone some significant maintenance.
In summary the previous problems and work undertaken were:
The split lines were clogged which was causing really bad peak tailing and a lot of other issues
We later ended up replacing the entire injector unit, hoping that the persistent issues were related to the injector, but this did not resolve the issue. As you can see, the baseline and peak shapes are very poor
The benzoic acid (underivatized) does not behave ideally on this column but usually appears as a sharp fronted peak that is ok for quantification. In this case, though, the software doesn't know where the "peak" ends and baseline resumes
A new column has been installed
A quick review then on the common causes of peak tailing in GC – for further details on this topic Premium CHROMacademy members can access the following webcast and tutorial guide;
Peak tailing can be caused by both chemical and physical phenomena and these break down as follows:
Unwanted secondary interactions, typically caused by polar surface species (silanol groups) due to loss of deactivation in the column (primarily due to phase bleed) or inlet liner
Poorly cut column (jagged cut leading to high number of exposed silanols from the silica tubing or vortices caused as the carrier gas enters the column)
Poorly evacuated inlet – caused by blockages in the split line, inappropriately long splitless time during splitless injection
Poorly installed column in the inlet or detector – leading to unswept dead volumes
Blockages of the jet / orifice of the FID detector
From our correspondents information we can see that several of these issues have been dealt with already, including a full replacement of the instrument inlet and split lines.
Peak tailing caused by chemical issues tends to affect polar analytes more strongly than less polar analytes due to the strength of interaction between polar functional groups and the silanol or chelating metal surface on the inside of the inlet. However in our correspondents case the peak tailing seems to apply equally to analytes of all polarity, and perhaps most telling, the solvent peak.
In this case we would suspect a physical phenomenon. The end user should ensure that a fresh liner is installed which has been chemically deactivated by the manufacturer and that the column is re-cut at the inlet and detector ends (20 cm of column should be removed from each end) and installed according to manufacturer’s instructions. The jet of the FID detector should also be checked and verified as being free from blockages. Whilst we believe this is case in this instance it is always better to double check.
We suspect that the Electronic Pressure Control module may have also become fouled and that the split valve controller or regulator in this unit may have also become contaminated – and whilst the inlet has been replaced our correspondent should manually check the split flow rate and consistency to ensure that the inlet is being appropriately swept.
This problem remains unsolved currently – however once we have a resolution to the issue we will update the thread in CHROMacademy forum to let all our readers know the cause of the issue!
Figure 14: Ratio of derivatised impurity peaks changing with varying split ratio – conditions shown below.
5% Phenyl PDMS, 30m x 0.25mm x 0.50μm
He at 1.1 mL/min.
Split (Focus liner 4mm packed)
Split ratio 10:1
proprietary pharmacologically active molecule – multiple unsaturated rings with one hydroxyl functionality and one fluorine group + MSTFA derivatization in TFH / ACN
This example is perhaps a little less ‘mainstream’ however it is indicative of the complexity that is associated with sample introduction in gas chromatography.
The sample is a proprietary however the pharmacologically active compound and it’s impurities (compounds A & B) contain both hydroxyl and fluorinated moieties. In order to avoid deleterious peak shapes (primarily tailing) our correspondent had found it necessary to derivatise the analytes using MSTFA – a reaction which proceeds generally according to the equation shown in figure 15.
Figure 15: General schema for the siliylation reaction of a hydroxylated analyte with N-methyl –N (trimethylsilyl) trifluoroacetamide (MSTFA).
In our case this reaction is carried out in a mixture of acetonitrile / THF for 4 hours at 30oC. The reaction is known to go to completion / equilibrium under these conditions.
The issue arose when trying to transfer the method from one laboratory to another and centered around an inability to reproduce the relative ratios of impurity A and B peaks.
Both laboratories were very diligent and well informed in terms of the instrument and experimental set-up. However, after much investigation and head scratching, the effect of changing split flow was investigated and the results shown in figure 14 were obtained.
It is clear that the split ratio has an effect on the relative amounts of analytes A and B which, to the trained eye, indicates that there are chemical phenomena occurring within the inlet. Of course, as the split ratio is decreased the analytes are allowed to remain within the inlet for a longer period (even if this is just fractions of a minute) and as such, the vapor phase species are subjected to the high inlet temperatures for a longer time. This may obviously lead to the reaction shown in figure 15 proceeding further in either the forward or reverse direction, and potentially leading to changes in the relative amounts of each of the derivative species reaching the column.
After more investigation it was also noted that the amount and density of the glass wool packing within the inlet liners differed significantly - one being ‘home’ deactivated and packed and the other purchased as a pre-packed, pre-deactivated part. The density (and hence the available surface area) of the liner packing material will also affect the degree to which the analytes will remain within the liner and therefore the extent to which the analytes will derivatise or underivatise accordingly.
Using the same liner and standardizing on the split ratio helped to successfully transfer the method in this case.
One should remember that seemingly minor parameters such as the density and positioning of the glass wool within the liner may have a large effect upon the vaporization processes and that labile analytes should be moved through the high temperature environment of the inlet as quickly as is practically possible.
Table 1: Relative standard deviation of peak areas of higher boiling analytes is unacceptably high.
Irreproducible peak areas for higher boiling analytes are typically caused by inadequate volatilization or a splitless time which is too short.
Whilst the boiling point of octacosane is 438oC, an inlet temperature of 300oC is possibly adequate under the reduced vapor pressure conditions of the inlet. Of course, without experimentation this is impossible to state categorically and an increase in inlet temperature may improve the analyte transfer. One word of caution –if the inlet temperature needs to be increased above the isothermal temperature maximum of the column, the use of a guard column should be considered, to avoid phase bleed from the column inlet.
An inlet flow rate of 10.5 mL/min. will clear the liner in (850μl internal volume) in a relatively short time (4.9 seconds), and therefore theoretically the splitless time of 1 minute should be more than long enough to ensure analyte transfer.
However, in this case, we felt that a combination of a potentially low inlet temperature combined with a relatively short splitless time may be confounding the transfer of the higher boiling analytes onto the column, and some of the analytes may be being lost from the inlet as the split is initiated. This problem can often be seen in laboratories where the analysis of relatively low boiling analytes is the norm, and when analysing higher boilers, unless the splitless time or inlet temperature are adjusted then quantitative problems occur with the higher boiling (later eluting) analytes.
In the first instance, to test this theory, the splitless time was extended to 90 seconds and the following results obtained;
Std injection 1
Std injection 2
Std injection 3
Std injection 4
Std injection 5
Std injection 6
Table 2: Data acquired under the conditions shown in figure 16 with an extended splitless time (90 seconds v’s 60 seconds) which show acceptable variability.
Table 2 shows that the data obtained show much more acceptable precision. The end user may further investigate inlet temperature or a slightly longer splitless time (120 sec.) to obtain best precision, however one needs to guard against broad tailing solvent peaks and rising baselines throughout the run when using very long splitless times.
Figure 17: Separation of a number of pesticides fungicides using the conditions shown below. The Relative Response Factors of some pairs of analytes show generally good agreement but over several injections (circa. 20) some RRF deviations are unacceptably high and validation of the method is not possible under the customer protocol.
As part of their validation protocol, the client measures relative response factors of the various analytes – a snapshot of the data produced is shown in table 3;
Table 3: Relative response information from a snapshot of validation data for the above method.
Table 3 highlights that whilst the relative response ratio data is generally good, there are certain injection or analyte combination where the RRF data is borderline (0.972 RRF shows a variability of 2.8% between repeat injections).
So the question is – what is causing this ‘jitter’ within what seems like a well designed method
The first thing we looked at was the chromatogram and method parameters. To our eye, several of the analyte peaks appear to be overloaded as they are fronting and obviously have asymmetry values <1. At the analyte concentrations quoted and with a 1μl injection volume, analyte amounts on column will be around 50 ng (considering the 50:1 split), which is too high for this column which would typically have a loadability of around 5 – 20ng per component on column (Figure 18).
Figure 18: Properties of some selected column geometry combinations.
The customer was advised to increase the analyte mass on column by diluting the sample by a factor of 100 or by increasing the split ratio to 200:1.
This led to a much improved peak shape – however there were still injections where the RRF was unacceptably high.
Insecticides and fungicides are reasonably adsorptive and often end users will employ more inert materials within the liner, such as Tenax or the like, in order to avoid adsorption issues. In this particular case the use of these materials had not been explored however an empty liner had been tried and the results had looked very much worse.
One possibility for the issues observed is a lack of mixing between the analyte components and the possibility of some analytes entering the inlet in the solution phase, and therefore the sample being non-homogenous at the point at which the split is made (i.e. the column inlet). This latter of these problems can be overcome by increasing the inlet temperature – however these analytes are also somewhat thermally labile and therefore this was ruled out.
The mixing issue can be overcome by slightly overpacking the liner with densely packed deactivated glass wool (typically 1 -2 cm) and ensuring that the tip of the needle penetrates the packing material – effectively mixing the same and providing a large surface area to assist with analyte volatilization.
The work on this method is ongoing – however initial results look very encouraging and Relative Response Factors show good agreement!
In the following series of figures, John Hinshaw investigates the cause of a problem baseline which is noisy and appears to show extreme column bleed!
Gas manifold tubing
30-m x 0.53-mm x 1.0 μm
Figure 19: Blank run showing noisy / rising baseline.
Figure 20: Split injection of blank solvent after 13 bake out runs at 150oC
Baseline remains noisy with large artifact peak at circa. 12 mins, baseline rise is somewhat mitigated.
Figure 21: Results after 26 further bake out runs at 200oC – baseline artifacts are still clearly visible.
Figure 22: Splitless injection of blank solvent after replacement of gas filters – the problem is exacerbated (as expected with splitless injection), but the baseline artifacts are still clearly visible.
Clearly John was having problems with this column / instrument combination.
Bleed and baseline artifacts typically disappear with column bake out (conditioning) and this column had been extensively conditioned! It had come with flame sealed ends and so the likelihood of gross phase loss due to oxidation was low and the bake-out procedure seems mild and reasonable – remember it is good policy to have the carrier flow through the column for 15 mins prior to heating a column in order to purge the dissolved oxygen from the stationary phase prior to analysis.
The gas line filters were changed in case they were exhausted or even the source of the incoming contamination but to no avail.
Then John had the idea to remove a reasonable amount of column, and the first turn was removed and the column re-installed into the instrument – the results are shown in figure 23 below;
Figure 23: Hey presto! The baseline artifacts are very much reduced and the bleed profile is as expected for a column of this type and the temperature program used.
And the salutary lesson for today is;
No matter how much conditioning of the column you do, if the column inlet is heavily contaminated or thermally degraded or occluded with sample matrix, the quickest fix will always be to remove a significant portion of the column.
Now - what is ‘significant’ in the previous sentence? Well, when we run into problems with baseline artifacts or bleed that we can’t seem to fix, we remove around 5% of the columns total length – for a 30m column this is around 15cm and so forth.
As a final caveat – the other possibility for this type of baseline noise is septum bleed products being carried from the inlet to the column. This could only occur if the septum purge has ceased to work or if shards of septum were collecting in the inlet liner. The latter of course is easy to fix – the former world require an increase in septum purge flow or a manual flow check and engineer visit if the purge valve was compromise.
Figure 24: Underlying broad distribution in sample and placebo of a formulated product.
We have included the example above to highlight a general point which can address many of the problems of the type shown above.
The formulated product contains a light mineral oil as one of the excipient ingredients, and the broad distribution of alkanes in this material appear as a ‘hump’ on the baseline. Obviously this can compromise resolution, adds baseline artifacts to the chromatogram and presents issues with integration / quantitation etc.
Whilst it is difficult to remove all of the contributing excipient, some degree of sample preparation should be considered in this type of situation. Where one can optimize the selectivity of the sorbent to discriminate between the analyte and interfering materials, then solid phase extraction or dispersive extraction (QuEChERS) can be very useful indeed. However, supported liquid / liquid extraction or phase separation media can also be sometimes useful in this regard.
The CHROMacademy Sample preparation channel is full of useful information which can be used to decide the most appropriate means of sample preparation for a particular sample. Also search CHROMacademy for articles by Ron Major (Agilent Technologies, Santa Clara, USA) who writes on this topic and has some great practical suggestions for getting just the right amount of selectivity into the sample preparation stage of your analysis.
Figure 25: Examples of a poorly (top) and well (bottom) integrated chromatogram.
Again this last example is included to address a general issue that we see a lot in our troubleshooting work.
Integration parameters / algorithms – we just don’t care like we used too! It really is worthwhile playing around with and getting to know the effect of the various integration parameters available to you . In figure 25 – the baseline obviously isn’t drawn correctly and the perpendiculars that are being drawn to this baseline could never be expected to be reproducible – and therefore you will encounter issues with quantitation.
By increasing the (slope) sensitivity of the algorithm, one is much more accurately able to identify the start and end of the peak on the rising baseline, and therefore life will be more straightforward……
Try to get a good handle on parameters such as sensitivity, dealing with tailing split or shouldered peaks using the various electronic algorithms, forcing a baseline when necessary, deciding whether to tangent skim or force a perpendicular etc. etc. etc. You should aim to satisfy Pareto’s Law – you integration algorithm should automatically integrate at least 80% of your sample s without you having to have some manual intervention to ‘fix’ the integration.
Of course – in the problem above – we would ideally like to get rid of the rising baseline – maybe using a different (more thermally stable) phase, operating in constant flow rather than constant pressure mode or doing a little more column conditioning would have worked nicely.
Whether you work in a lab or manage a lab, you will benefit from being a member of CHROMacademy.
As a member of CHROMacademy, you will get access to our vast e-Learning archive full of great interactive content and animations.
All our Essential Guide Webcasts and tutorials and LCGCs archive of magazine articles and webcasts from your favourite authors - John Dolan, John Hinshaw, Mike Balough, and Ron Majors. Plus vendor application notes, electronic laboratory tools and calculators and with our 'Ask the Expert' function - help is always at hand.
I feel empowered to fix things
I can troubleshoot effectively
I know where to go for help
I understand my analyses
I know where to get applications
I’m up to date
I’m more employable
My career is progressing
Improved equipment utilization
Faster method development/problem solving
Flexible workforce with a common standard
Better quality data
Get up to speed quicker
Less reliant on me
I spend less time on training
Subscribe for $349 per/year and access:
The entire e-Learning archive
All Essential Guide Webcasts and Tutorials
LCGCs archive of articles and webcasts
Expert troubleshooting advice when needed
Like CHROMacademy on Facebook and keep up-to-date with the latest webcast, tutorial and eLearning release schedules.
Dr John Hinshaw (Senior Scientist, BPL Global Ltd and CHROMacademy GC Dean) and Tony Taylor (Technical Director at Crawford Scientific and CHROMacademy) demonstrate practical GC separation troubleshooting. Using real world chromatograms submitted by CHROMacademy members from many different application areas, they illustrate how to recognize problems and explain further tests and investigations necessary to properly qualify the issue. They then offer advice on how the problem may be resolved to help you gain valuable insight into recognizing, characterizing and fixing problems in GC to improve your data quality and reduce instrument downtime.
Dr John Hinshaw
Senior Scientist, BPL Global Ltd and CHROMacademy GC Dean
Key Learning Objectives:
Learn from experts how to recognize problems in GC separations
Understand the steps which can be taken to better characterize problems and investigate issues
Get advice on how to fix or avoid problems in ‘real-world’ chromatography
Become more proficient in recognizing and fixing problems with GC separations or instrumentation