In January 2011 the updated version of EU GMP Annex 11 on computerized systems was published that became effective on 30 June 2011 (1). In April 2011 this column (2) reviewed Annex 11 and also Chapter 4 on documentation to see if this was Europe's answer to 21 CFR 11 on electronic records and electronic signatures (3). In this "Questions of Quality" column I want to review the technical requirements of the new version of Annex 11 to see how the software commonly used in regulated laboratories meets these requirements and where they don't. To begin, I had better explain what I mean by a technical requirement of software. 

A technical requirement is where compliance with a regulation is met through a feature or function in the software that has been designed, engineered and tested by the vendor of the product. In doing this, a vendor should seek the opinion of customers who although they work to the same regulation will tend to have different individual interpretations of how the software should comply with the regulations. The role of the vendor would be to sift through these interpretations, which will occasionally differ, to specify the software so that there is a generic approach to compliance. Where companies have different views then a vendor may use software configuration to enable individual laboratories to implement their own interpretation of a regulation. Here some features can be turned on and others off, allowing different laboratories to implement their own interpretations of a regulation but at the same time allowing a vendor to sell a standard product.  So in this column we are looking at how software should meet the technical requirements contained in the new version of Annex 11. 

What Are Annex 11 Technical Requirements?

The obvious starting point is to sift through the principle and 17 clauses of Annex 11 and identify what are the technical requirements so that we can then discuss each one individually. Table 1 presents my interpretation of Annex 11 to identify the technical requirements. In this table the technical requirement of each clause is highlighted in bold.  As I just mentioned, this is my interpretation of the technical controls, there are clauses of Annex 11 that at first glance may appear to be technical controls (for example, clause 7 on backup) but are mainly procedural in my view. 

Table 1: The technical requirements of EU GMP Annex 11.			

Of the requirements in Table 1, I am not going to discuss clause 15 release of a batch by a Qualified Person (QP) as this typically occurs outside of the laboratory and not many analytical scientists are QPs. Also, the requirement for archiving is subject for discussion in one or two "Questions of Quality" columns and therefore we will not discuss this any further.  Therefore, this leaves the following sections to consider: 					

Clause 12. Security (however, due to some similarities with clause 9, I will discuss these technical requirements under the section on audit trail)

Looking through the list you will see that many of the topics are concerned with data integrity which is complementary to the discussion in my last "Questions of Quality" column 

(4), so we will consider the discussions about data integrity we looked at there as well as what we shall cover in this column. We will therefore start in a boringly logical order with the lowest numbered clause and proceed in increasing numerical order. 

Automated Data Transfer Checks between Systems

Looking on the bright side, if all the computerized systems in your laboratory are standalone, then you don't have to consider this section as you are condemned to life imprisonment with eternal transcription error checking instead of hard labour. Actually, considering what I have just written, there is probably not much difference between the two apart from the former being physical and the latter mental hard labour. 

However, if you have seen the light and are linking your computerized systems together, e.g. a Chromatography Data System (CDS) with a Laboratory Information Management System (LIMS) then you will want to make sure that what was sent from the CDS was the same as received in the LIMS and vice-versa. This is not a regulatory requirement, it is simply good analytical science. However, as we can see in Table 1, Annex 11 pokes its nose in by requiring 

"appropriate built-in checks for the correct and secure entry and processing of data"

. Therefore, there needs to be software functions to ensure that there is no corruption or truncation of data between the sending and receiving computerized systems. Ideally, there would be an interface option available from one or other of the software vendors, typically the LIMS vendor. This should be a standard software module that transfers the data between the two systems or it could be configurable application to accommodate the requirements for any CDS. However, it should not be a oneoff module of code written specifically for your situation as this will be more difficult to support over time. 

Don't forget that the interface between the two systems needs to be specified in a user requirements specification and tested in your validation of one or both systems. 

Annex 11 requires that if critical data are manually entered into a computerized system, then a second check needs to be made. This check can either be manual or automated. The key driver of this regulation is GIGO: garbage in, garbage out. Put rubbish data in to a computerized system you will get rubbish results out at the other end. It is therefore good common sense to put checks into the system to ensure that the data are correct before you go any further – in a worst case scenario it is cheaper to do this than have a product recall of nonconforming product. 

First you need to define what critical data are. Consider some of the following situations: 					

Is the weight of an analytical reference material used to prepare a stock solution that is entered manually into a LIMS considered critical data?

Are sample weights or factors entered manually into the sequence file of a CDS considered critical data? 

The answer to this and other questions would be "yes" in my view. However, you need to define and document what are critical data for all systems in your laboratory as part of the specifications for the system to comply with Annex 11. Therefore you have choices to make: Do I check everything manually or attempt to automate the process? In my last "Questions of Quality" column (4) I discussed some of the ways of ensuring that manually entered data could be verified against predefined criteria, such as data format checks to ensure the correct number of numbers either side of the decimal point and also within the range of data anticipated. However, these verification checks cannot eliminate transcription errors just reduce them. So unless you are using automated data acquisition and data transfer, I don't think that you can eliminate manual checks of manually entered data, you'll just reduce the incidents of gross error with data field verification, but the second person check, although reduced, will still be needed. This is more reason to automate data transfer between systems rather than rely on the four eyes principle.  

Annex 11 has the requirement that printouts can be printed on request, even for fully electronic systems, but as we can see from clause 8.2 there is the requirement for printouts involving batch release that it should be possible to 

"generate printouts indicating if any of the data has been changed since the original entry"

. So, if we are analysing samples for batch release including stability samples then this section applies to laboratory systems. 

For this technical requirement, let us consider a CDS. The printout from the CDS has to identify if data have changed since the original acquisition, interpretation or data entry. At the integration level we are very well served with a CDS as this has been built into commercial integrators and data systems since the mid-1970s. When peaks are automatically integrated, the integration is identified with capital letters e.g. BB for baseline to baseline resolution as shown in Figure 1, however, if you manually reposition where the baseline is drawn on the training edge the integration becomes, say, Bb where the training baseline has been repositioned manually. This is available from all CDS now and can be printed on demand.  

Figure 1: Chromatogram illustrating that integration is automatic.			

 this is as far as it goes in most systems.  

Printouts for changes to everything in a batch analysis you may have to hunt for, depending on the CDS you have, because you will need to identify changes to the sequence file entries, instrument method and processing method. It is not always straightforward or easy to obtain a single printout about the changes made during the analysis of a set of samples analysed for batch release. So what are you going to do about it? Perhaps you might want to check out how you can meet this requirement before an inspector calls. 

To help, the European Medicines Agency has published a question and answer section on the web (5) and here question 10 provides some help: 

Question: What alternative controls are accepted in case a system is not capable to generate printouts indicating if any of the data has been changed since the original entry? 

Answer: As long as this functionality is not supported by the supplier, it may be acceptable to describe in a procedure that additionally a printout of the related audit trail report must be generated and linked manually to the record supporting batch release.

So from the inspector's perspective you can hunt to find this information. However, from your perspective should you not be asking your CDS and LIMS vendors to provide this functionality as standard? 

Most Audit Trails Don't Comply with Annex 11

I was thinking of titling this section "what the regulator gives with one hand they take away with the other" but it was a bit too long to fit on the page. Here's what I mean: section 9 on audit trail allows a company to determine if they need to have an audit trail or not (giving with one hand) based on a risk assessment. 

, clause 12.4 states that if your system holds data or documents it 

"should be designed to record the identity of operators entering, changing, confirming or deleting data including date and time"

 (taking away with the other hand). Although you do not need to have an audit trail to fulfil the requirements of 12.4, if the system holds data then there needs to be a mechanism to identify users who enter, change or delete data. For the majority of systems we use in the laboratory, an audit trail is really the only sensible solution to meet requirements in clauses 9 and 12.4 of Annex 11, especially if you want the benefits of working electronically. 

In addition, if time and date are changed there needs to be a record of this and is an added justification for ensuring that all laboratory computerized systems are linked to the network and time stamps are independently maintained by the IT department. Time and date are critical elements for ensuring data integrity in computerized systems used in the laboratory, therefore there needs to be a means of ensuring their accuracy and also preventing unauthorized people from playing about with the computer clock. The simplest way is to link systems to the network and do this automatically. Summer and winter time changes can also be automated in this way. 

Let us now focus on clause 9 for audit trails. I reiterate that if working electronically an audit trail is mandatory in my view to ensure data integrity and to identify who has done what to the data. Although clause 9 does not mention creation or entry of data, clause 12.4 does and therefore the audit trail has to encompass data creation as well as changes and deletion of already entered data. Personally, I would restrict the ability to delete data to one or two individuals or nobody so that the deletion aspect of this clause can be dealt with quickly. 

What is new, although perfectly logical, is the requirement to add a reason for change when data are changed or deleted. This brings GMP into line with Good Laboratory Practice (GLP) that does require a reason for change when altering data. Your CDS should be capable of having, at a minimum, a free text entry to allow a user to enter a reason for the change when such a change has been made. A better design for an audit trail would be to have context sensitive user predefined entries that a chromatographer can select when making a change – a faster and easier way of entering data and avoiding all those mistyped and misspelled entries. 

The final sentence of this clause has two requirements for us to consider: 					

Audit trails need to be available and convertible to a generally intelligible form and: 

The audit trail is present in the system (obviously) but it needs to be readable and understandable. This may not be the case with some CDS audit trails as some can be piles of unintelligible rubbish. Therefore, when you are considering a new CDS for a regulated laboratory, this is one area that you must pay close attention to. Therefore ask the following questions: is there a single audit trail for the whole system or are entries split into smaller units? How are the entries made: can you see who made the change, when it was made, the old and new values and reason for change? Can you search the audit trail easily with user selected criteria to obtain a summary of audit trail entries associated with a specific analysis run? 

This is the major weakness with virtually all CDS on the market in that 

 you can search the entries effectively you 

 demonstrate electronically that you have reviewed the entries. 

Now here's the issue "regularly reviewed". This brings the EU into line with what the FDA have been requiring laboratories to review audit trails: those who have not been complying have featured in a number of warning letters and 483 observations. The simple reason is that there is no point having an audit trail if it is not reviewed. However, what is meant by "regular"? This depends on the use of the CDS. If it is used for batch release, then regular means prior to signing off the batch results from the CDS. 

Ask yourself a question. Do you really want to trawl through masses of audit trail data to find out about data changes or are you lazy? Join the club.  

What is required is a review of the audit trail by exception: so the system needs to inform you if there have been any changes to the data since initial entry or acquisition for that analytical run thus allowing you to focus your attention where it is needed. If there are no changes – job done! Or is it? How are you going to demonstrate that you have reviewed the audit trail even if there are no changes to data? Hmmm, my CDS does not do this. Therefore, who are you going to call? Let me give you a clue, it is not Ghostbusters but your CDS vendor to ask them how they will implement features to enable you to comply with the law.   

We shall return to the audit trail while we review Annex 11 electronic signatures in the next section as there is a further impact on some CDS products used in the laboratory and how electronic signatures can be implemented in many applications. 

Electronically Signing Electronic Records the Annex 11 Way

Detailed interpretation of sub clause 14a can derive the same electronic signature requirements as required under 21 CFR 11 except for the horrible administrative over head of writing a formal letter to the FDA and requiring people to verify during an inspection that their electronic signature is the equivalent to their handwritten one. There is also no need to classify a computerized system as either open or closed, nor the need to specify the types of electronic signature that can be used (electronic, biometric or digital). Annex 11 does all of this in 40 words compared with 745 (excluding definitions) in Part 11. A neat and effective way to write a regulation and perhaps if the FDA get round to revising Part 11, they can learn from this simple approach. The reason is that the primary authors of Annex 11 are the inspectors themselves and not lawyers. 

However, looking at the way Annex 11 is written there are issues concerning electronic signatures that directly impact on the design of software used in GMP regulated laboratories. Clause 14 requires that electronically signed electronic records or a report of an analysis must be available outside of the computerized system that generates it but within the boundaries of your company. This is not an unreasonable requirement as the users of the information generated by an analytical laboratory typically reside outside of the laboratory in development, regulatory or manufacturing. After the laboratory data have been interpreted by an analyst, a report will be prepared with an electronic signature attached to the report.  

The electronic signature must therefore be applied to a representation of the records to which it pertains and not stuffed as an afterthought by a vendor's software engineer into the audit trail where it is difficult to see that the record has been signed. In short, the record and the signature must be together, as clause 14 makes perfectly clear.  Furthermore, the linkage between the signature and the record interpretation must be permanent as required by Annex 11. How this is to be achieved is left to the software vendor to interpret and the laboratory to see if it complies with their interpretation of the regulation. However if the signature is not attached to the record, then the system does not comply.  

Articles

Questions of Quality

The new version of EU GMP Annex 11 has been effective for about a year. There are new technical requirements in the Annex – do your laboratory computerized systems comply with them all? In this column we look at the impact of these new regulatory requirements for Good Manufacturing Practice laboratories.

Can You Meet the Technical Requirements of Annex 11?

It has been about nine years since this list was last published in 

(1). During the interim my toolbox has gained some new items, and a few others have seen little or no use. Given the modern obsession with list-making, readers may enjoy this update and perhaps learn about some tools they might like to add or replace in their own toolkits. 

Every profession has its specialized tools. Those used in chromatography are often just as specialized as those used in computer repair or automotive work. Many of the tools and accessories that gas chromatographers keep on hand for installing, maintaining and repairing their chromatographs are also found in plumbers', carpenters' and homeowners' toolkits. Wrenches, screwdrivers, pliers and metal-tubing cutters are some easily recognized examples. Other items such as dental instruments or paper correction fluid are familiar, but their use in the laboratory environment might not be immediately obvious. Still others, like a column flowmeter, septum nut wrench or a specialized fused-silica column cutter, aren't found outside the laboratory at all. 

Here, then, is the list of today's tools and accessories along with some information about their use and significance. One or more specialty manufacturers offer many of these chromatography-specific items in their catalogues or on-line offerings. I scanned through several catalogues and web sites and gleaned some new items that I have included here. 

A butane lighter is a convenient source of hydrocarbon gas for measuring an approximate unretained peak time. Butane is effectively unretained at temperatures above 75 °C on liquid-phase coated columns with phase ratios above 50. Columns at low temperatures or with lower phase ratios (thick stationary films) may retain butane and separate the traces of ethane and propane present in the butane fuel. Use the earliest observable peak for the best estimate of unretained peak time. Natural gas is mostly methane; if your laboratory has a supply of natural gas (mine doesn't) it makes a good substitute for a lighter and is less retained than butane. Just be sure to turn off the gas after you've filled a syringe with it. A lecture bottle of methane with a suitable pressure regulator is another excellent source of the unretained substance. Concentrations in the low percent range work well. 

Hydrocarbons won't work for unretained peak time measurements with electron-capture detection (ECD). Instead, try loading the syringe with a puff from a pressurized can of dust remover such as Dust-Off, which contains 100% 1,1-difluoroethane. 

Hydrogen or helium — whichever is not the same as the carrier gas — make good unretained peak markers for porous polymer or molecular sieve columns that retain hydrocarbons strongly, plus these two gases should be readily available in most gas chromatography (GC) laboratories. Flame ionization detection (FID) will not respond to hydrogen or helium, but other detection methods such as thermal conductivity detection (TCD), pulsed discharge detection (PDD) or helium ionization detection (HID), the latter with hydrogen as the unretained substance, should respond well. 

Recorder cable finally has gone the way of chart recorders: You might find some lying around but they haven't been used in years. I have one or two in the lab just in case the need arises. I keep some spare USB cables as well. 

Use white correction fluid to mark the measured position on a column that corresponds to the correct column penetration depth into an inlet or detector. Measure the depth after inserting the column into the nut and ferrule and making a fresh cut on the column end. Some regulated laboratories' policies don't permit the use of correction fluid for data integrity reasons. A septum into which a slot has been cut may be slid onto the column below the nut and ferrule to act as a positioning aid. Remove the septum before heating the column oven. A positioning gauge (see "Ruler" below) is a good alternative. I have also used a black permanent marker for this purpose; just be careful not to get it on your fingers or clothing. 

The best fused-silica column cutting tool is the one that holds the column in an adjustable chuck and cuts with a diamond chip as the operator rotates a thumb wheel. This tool also has a magnifying glass on the opposite end for inspecting the fresh cut for squareness and lack of burrs or hanging polyimide coating. A pen-like tool with a sapphire tip or ceramic scoring wafers or scribes that make a sharp cut on the column so that it may be broken cleanly in two are the best inexpensive alternatives. In any case, a fresh cut should be made and inspected just before placing the column into the inlet or detector and after sliding on the nut and ferrule.  

A large plumber's tubing cutter for ¼-in. metal tubing, and a smaller one for ⅛-in. tubing are used extensively during instrument installation and gas-supply set-up. The correct ⅛-in. size cutters are available from some chromatography supply companies — don't try to use the mini cutters found in home supply stores. Keep a supply of new cutter blades on hand. These disks wear out rapidly, and a dull blade will distort the tubing diameter and make it difficult to slide on the back and front ferrules. Power cutters with high-speed rotary abrasive cutting wheels are also available. These are mandatory for cutting 1/16-in. tubing for low-dead-volume gas valving applications, and they work well for any tubing up to ¼-in. diameter. After cutting and squaring the tubing (see "Tubing Reamer" and "Deburring Tools" below) be sure to clean out the ends with a solvent such as isopropanol so that loose particles can't get into the instrument components or interfere with the passage of peaks through column ends. 

Gas chromatographers can use 5 or 10 m of deactivated 0.53-mm i.d. fusedsilica tubing with a press-fit connector as a retention gap when necessary. Shorter pieces can act as a column-to-detector adapter when you don't want to put the coated column end into a detector to avoid column bleed from the column end at hot internal detector temperatures. 

A plastic dental mirror with a front-silvered surface makes it easy to examine the underside of an inlet fitting in the oven, or to check other inaccessible areas for loose or missing parts. The mirror can also be used to detect the flame in a flame ionization or flame photometric detector by observing condensation of emitted water vapour on the cool mirror surface. A shiny wrench is a good substitute for the mirror in this case. 

A dental pick is very handy for removing septa from septum nuts and debris such as bits of graphite ferrule from fittings. 

Diagonal cutters are used only for cutting electrical wires. They should not be used as a substitute for a tubing cutter. Don't even think about threatening your fusedsilica column with one! 

I have a box of these that is in frequent demand. I use them to place small drops of isopropanol onto fittings for leak checks. Some types have rough volume indications more like a pipette, and I have used them to make a crude dilution of qualitative test mixtures when accuracy wasn't required. 

A good assortment of graphite and graphite–Vespel capillary column ferrules is essential. Keep at least 10 of each inner diameter in the fitting sizes that match your instruments' oven fittings. In a pinch, graphite ferrules can be squeezed into sealing on smaller columns than they were designed for. It's always a good idea to install new ferrules with a new column; old ferrules can be reused on the same column if a seal can be made without over tightening the fitting. 

Ferrules for metal tubing are also essential. I used to prefer brass ferrules for copper and stainless-steel tubing, but nowadays I match the ferrule material to the tubing because some brass ferrules will bind to stainless steel upon tightening. Some instruments use 1/16-in. or 3/32-in. stainless tubing. This tubing is best connected using ⅛-in. graphite–Vespel reducing ferrules for ⅛-in. fittings, or 1/16-in. ferrules for the 1/16-in. tubing. Chromatographers should be aware of the potential for atmospheric oxygen contamination of the carrier gas from improperly installed supply tubing and ferrules. Even with the best filtration in place, a leak between the filters and the instrument will nullify the effect of the filters. 

An assortment of needle files can be used to pick out ferrules from fittings, remove burrs and shape the ends of metal tubing before it is connected to a fitting. Don't forget to clean off all traces of metal before connecting. 

A flexible 2-ft magnetic pickup comes in handy when you drop a small part inside the instrument. Another similar tool has a three-jawed "claw" operated by a plunger, and it will pick up nonmagnetic items. 

An electronic flowmeter is an expensive investment, but I believe that it will pay for itself many times over with improved accuracy and precision over bargain-priced bubble flowmeters. I prefer the type of electronic meter that senses flow directly and that allows the operator to select the type of gas in use, such as air, helium or hydrogen. The option to calculate split ratios from the measured split vent and column flows is a handy feature. 

If you use bubble flowmeters, keep two sizes on hand. The large size is good for measuring FID air or inlet split vent flows up to several hundred millilitres per minute. The smaller size is better for packed-column or hydrogen flamegas flows in the 10–50 mL/min range. Don't try to use a bubble flowmeter to measure capillary column flows below 10 mL/min. The carrier gas will diffuse out of the bubble and you will get a low reading. Measure the unretained peak time instead and calculate the flow rate from it. Note that this calculated flow rate or the rate displayed by electronic pressure control will only be as accurate as the column dimensions the operator uses. See an earlier "GC Connections" (2) for a detailed discussion. 

This item is useful for those who must install glass wool in inlet liners, or for the hardy few who pack their own columns and use glass wool to hold in the packing. These days I find little use for it. 

Inlet liners are often broken or chipped during installation or removal. They also can become contaminated with sample residue or may lose their deactivation if used for too long at high temperatures. Keep some spares on hand, both for packed inlets and for split or splitless injections. If you use deactivated liners, it is better to purchase them already deactivated than try to treat them yourself because of the chemical hazards and waste disposal problems this will create. 

GC Connections

This month's instalment presents the latest tools and accessories for the gas chromatographer. Have you got what you need in your troubleshooting toolkit?

A GC Troubleshooter's Toolkit

The use of an evaporative light scattering detector (ELSD) with high performance liquid chromatography (HPLC) has been well documented. However, because the mobile phase is nebulized to produce droplets the use of supercritical fluids as the mobile phase with an ELSD is also favourable because the fluid depressurizes before the detector entrance, which allows the formation of aerosol. This article investigates the ELSD response variation with supercritical carbon dioxide–based mobile phases and compares the response (peak area) with HPLC.

The use of an evaporative light scattering detector (ELSD) with high performance liquid chromatography (HPLC) overcomes some of the drawbacks and limitations of using ultraviolet (UV) detection: ELSD can be used for the detection of compounds lacking a chromophore group, provides no baseline deviation during gradient elution and can be used with solvents with a high degree of UV absorption (1,2).  

The principle of ELSD is based on nebulizing the mobile phase — a process where an aerosol is produced by the addition of a nebulizer gas (nitrogen) (3–8). This process takes place in a nebulizing chamber that can have various dimensions and volumes depending on the manufacturer. It can also be adapted to the type of chromatography used — HPLC, ultrahigh-pressure liquid chromatography (UHPLC) or micro-LC. The droplets, consisting of the analyte molecules and an unknown amount of solvent, are introduced into a drift evaporative tube and heated at varied temperatures. The droplet size is reduced by further solvent evaporation as the droplets pass through this tube. Finally, the droplets go through a light source and the deflected light is collected by a photomultiplier. 

The response depends on the particle (droplet) size, which seems to increase with the increasing compound concentration and which is related to numerous physical properties of the solvents composing the mobile phase: viscosity, surface tension and volatility. This explains that the response (

) is generally nonlinear for noncapillary columns and follows the relation: 

				Consequently, using a gradient elution in reversed-phase (RP) HPLC, that is, increasing the organic solvent content in the mobile phase, leads to changes in the droplet size that favour the response of the compounds in relation to their retention time (4,9,10). A compensation gradient added after the column makes it possible to limit this drawback by keeping the mobile-phase composition at the detector entrance constant (10,11).			

Other parameters induce great response changes, such as the atomization pressure (3,4), which is related to the flow rate of the nebulizer gas; the geometry of the nozzle (mainly the capillary diameter) (7,8); and the location of the mixing point of the mobile phase with this gas. The wavelength of the light source used also plays a role in the response intensity (6). 

On the other hand, because of the nebulization process required, coupling supercritical fluid chromatography (SFC) with ELSD has been successfully performed previously (12–18). The specificity of the carbon dioxide–modifier (an organic solvent) mobile phase used in SFC is the pressure required at the column outlet to ensure the dense state of the fluid. This back-pressure is achieved with a back-pressure regulator or a restrictor, located after the column. However, after the back-pressure regulator or restrictor, carbon dioxide returns back to a gaseous state, thus favouring aerosol formation in the nebulizer chamber. However, this carbon dioxide depressurization cools the outlet capillary, which can induce dry-ice formation and possibly plug the capillary. Heating, provided either by the back-pressure regulator or by an additional transfer line, avoids this drawback.  

Beyond these studies addressing this coupling of SFC with ELSD, however, few studies are available describing the varied effects of the numerous chromatographic parameters affecting SFC separations. The difficulty of interfacing SFC with a low-pressure detector, given the phase separation between the liquid solvent and gaseous carbon dioxide, has been addressed (19), but no clear comparison has been performed between SFC and HPLC to evaluate how the use of carbon dioxide instead of a liquid affects ELSD responses in terms of area or sensitivity. This comparison is the topic of this paper. In this study, we used an ELSD system with a slightly adapted nebulizer, mainly with carbon dioxide–ethanol mobile phases, with the goal of operating in "green" chromatographic conditions. 

Two model 980-PU pumps (Jasco, Tokyo, Japan) were used, one for carbon dioxide and another for the modifier. Control of the mobile-phase composition was performed by the modifier pump. The pump head used for carbon dioxide was cooled to -2 °C by a cryostat (Julabo F10c, Seelbach, Germany). When the two solvents (modifier and carbon dioxide) were mixed, the fluid was introduced into a dynamic mixing chamber (model PU 4046, Pye Unicam, Cambridge, UK) connected to a pulsation damper. The injector valve was supplied with a 20-µL loop (model 7125, Rheodyne, Cotati, California, USA).  

The columns were thermostated by an oven (Jetstream 2 Plus, Thermotechnic, Langenzersdorf, Germany). A UV detector with a high pressure–resistant cell was set to control the injection repeatability and volume (HP 1050, HewlettPackard). The detection wavelength was 210 nm. After the detector, the outlet column pressure was controlled by a Jasco 880-81 back-pressure regulator. The outlet regulator tube (internal diameter (i.d.) = 0.25 mm) was heated to 60 °C to avoid dry-ice formation during the carbon dioxide depressurization. The ELSD system [Sedex 85 (Sedere, Orleans, France)] was connected to the outlet capillary. The nebulizer tubing was slightly different from the one used with HPLC, in that the capillary diameter was 150 µm. The nebulizer gas (nitrogen) pressure was 2 bar and the drift temperature was 50 °C. Chromatograms were recorded using Azur software (Datalys, France). 

The main column used was a 250 mm × 4.6 mm, 5-µm 

 Synergi Fusion RP column (Phenomenex, Torrance, California, USA), the mobile phase was 90:10 carbon dioxide–ethanol (except for the study of the modifier percentage), the temperature was 25 °C, the outlet pressure was 15 MPa and the flow rate was 3 mL/min. Caffeine was provided by Sigma-Aldrich (St. Louis, Missouri, USA). The injected volume was 5 µL for most studies, and the concentration was 2.5 mg/mL. Other columns used included a 250 mm × 4.6 mm, 5-µm 

 Cyano column (Princeton, Cranbury, New Jersey) and a 250 mm × 4.6 mm, 5-µm 

Uptisphere HDO column (Interchim, Montluçon, France). 

A first series of experiments was performed to study basic parameters: injected volume, additional gas pressure, outlet column pressure, drift temperature and flow rate of the mobile phase. To ensure accuracy of the quantitative value studied, no split was used for these experiments. That means that the entire volume of supercritical mobile phase eluted from the column was introduced into the detector. By using a back-pressure regulator, the transfer capillary (around 10 cm in length) was heated to 60 °C. This was sufficient to avoid dryice formation and plugging of the transfer line. 

Figure 1(a) shows the variation of the caffeine peak area with the injected volume from 2 µL to 15 µL, with both the UV and ELSDs. As expected, the UV response displays a linear variation, showing the accuracy of tested volumes. The variation of the ELSD response appeared to be exponential. Figure 1(b) shows, by using double logarithmic coordinates, that the relationship between the response and the injected volume is linear. This behaviour [Figure 1(a)] is classical in comparison to the one observed in HPLC (1). The standard deviation measured for 10 injections of 10 µL was 0.5% for ELSD and 1.0% for UV detection. 

Figure 1: (a) Peak area (A) variation with injected volume (Vinj) of caffeine. Each value is calculated based on six measurements. Blue diamonds: ELSD; pink squares: UV 210 nm. (b) Log A vs. log Vinj with ELSD (from 2 ÂµL to 15 ÂµL). 			

Figure 2 shows the detector response compared to the nebulizer gas pressure, from 0 bar to 4 bar (5 µL injected). For this study, the nebulizer chamber of an earlier ELSD system (Sedex 55) was also tested with the same nebulizer because this model had been used to analyse carbohydrates by SFC in the 1990s (13). The difference between the two chambers is mainly the length, with the chamber of the later system being about twice as long as that of the system previously used. 

Figure 2: Variation of peak area with pressure of nebulizer gas. Blue diamonds: Sedex 55 nebulizing chamber; green circles: Sedex 85 nebulizing chamber. 			

The highest response values were obtained with the longer nebulizing chamber at 3 mL/min, which seemed to be the optimal value. With this longer chamber, the addition of the nebulizer gas improved the response. Finally, the optimal nebulizer gas pressure was close to the pressure usually used in HPLC, despite the gaseous carbon dioxide introduced into the chamber in SFC. As reported elsewhere (20,21), the increase in the nebulizer gas flow rate induces both the shift of the droplet size toward smaller droplets and the increase of the number of droplets. These two effects should provide opposite response changes, but the second effect must be greater to explain the observed result (20). The strong effect of gas pressure on response was also noticed with HPLC (4). In contrast to our results, other studies report a rapid response decrease with an increase in the additional gas flow rate (22,23), but the design of the nebulizing chamber, and the interface section between the chamber and the drift evaporative tube are not identical from one detector to another (24) and those differences may well be responsible for the opposite observations.  

Nevertheless, the responses obtained without additional gas (Figure 2) are not dramatically lower than those obtained with additional nebulizer gas, showing a rather good production of droplets as a result of the carbon dioxide depressurization.  

) applied to the chromatographic system was also studied from 9 MPa to 19 MPa (Figure 3). With both detection methods, but mainly for ELSD, the response was enhanced by an increase in outlet pressure. Increasing 

 enhances the fluid density before the backpressure regulator, leading to a reduction in the linear speed of the mobile phase that enhances the residence time of the compounds in the cell; the increased residence time explains the response increase, mainly from 11 MPa to 16 MPa (a refractive index change in the cell also can modify the UV response) (25). The greater amount of carbon dioxide before the back-pressure regulator also results in an enhanced fluid speed in the nebulizer (after the backpressure regulator), which seems to favour the ELSD response.  

Figure 3: Variation of peak area with the column outlet pressure. Blue diamonds: ELSD; pink squares: UV at 210 nm. 			

The temperature of the drift evaporative tube was also studied from 40 °C to 70 °C (Figure 4). With caffeine, a semivolatile compound, the increase in this evaporative temperature leads to a regular decrease in response. This classical behaviour has been reported with HPLC (3) and results from the reduction of particle size by vapourization with increasing temperature. 

Figure 4: Variation of peak area with the drift evaporative tube temperature. 			

The flow rate of the mobile phase was also studied from 2 mL/min to 5 mL/min (Figure 5). With both detectors the response decreases when flow rate increases. For UV, a concentration detection method, this behaviour is expected; the residence time in the cell decreases as the mobile-phase flow rate increases. As reported previously, the nonlinear decrease of the UV response also could be caused by varied effects related to the fluid density change. However, for ELSD, a mass detection method, these flow-rate changes would not modify the amount of compound going through the detection system.  

Figure 5: Variation of peak area with the mobile-phase flow rate. Blue diamonds: ELSD; pink squares: UV at 210 nm. 			

In HPLC, varied trends have been observed. An increase in response with the flow rate (from 0.1 mL/min to 1 mL/min) was reported for a concentric nebulizer by direct solute injection (no chromatographic column) (23), whereas the opposite was seen with different mobile phases, different compounds and drift temperatures (6,21,26).  

From the Nukiyama–Tanasawa equation (27), the increase in the mobile-phase flow rate should improve the average particle diameter. Larger droplet diameters are supposed to enhance ELSD response (up to a limit value) (6,28), but they also favour droplet condensation and precipitation on the walls of the nebulizing chamber, thus decreasing the number of droplets that can enter the drift evaporative tube. Nonetheless, this equation describes the primary droplet size distribution (the distribution obtained just after the nebulizer), whereas the measured response is that of the ternary distribution (after evaporation and loss of larger droplets by impaction) (24,29), limiting the relevance of this equation to describe the phenomena encountered. Moreover, this equation does not take into account the flow conditions, which are more often turbulent than laminar (29). Consequently, several works report the poor calculation of the droplet size distribution from this equation in some cases (30). In particular, investigating the effect of a flow rate increase on the droplet size, it was observed that the droplets were smaller than estimated based on the equation (31).  

Whatever the cause, this decrease is observed in packedcolumn SFC, where the total liquid flow rate entering the nebulizer is 0.2–0.5 mL/min (the mobile phase was 90:10 [v/v] carbon dioxide–ethanol, 2–5 mL/min). The fluid depressurization induces a cooling effect, which increases with the flow-rate increase. As a result, the condensation of droplets increases, meaning that the amount of compound going through the light beam decreases. 

Figure 6 shows the response variation for a diastereoisomer separation at two mobile-phase flow rates. As expected, for the two signals, the retention and peak width are lower, and the height of the peak is quite constant with UV. However, for the ELSD response, both the peak height and area are dramatically reduced for the higher flow rate, suggesting a strong modification in the droplet size distribution. Figure 7 shows the separation and detection of various pharmaceutical compounds in carbon dioxide–ethanol mobile phase. Theophylline, which is included in this set of compounds, displays a log P equal to –0.32, explaining why this compound is often studied by hydrophilic interaction chromatography (HILIC) (32–34). The other compounds display higher log P values (1.16 for dexamethasone, 1.20 for cortisone and 1.66 for hydrocortisone) (35). The separation of these four compounds is achieved in carbon dioxide–ethanol mobile phase without any additives (acidic or basic), with good peak symmetry. The response of at least three compounds (cortisone, hydrocortisone and dexamethasone) is higher with ELSD detection than with UV detection, regardless of the flow rate used, 3 mL/min or 5 mL/min. 

Figure 6: SFC separation of diastereoisomers. Column: Uptisphere HDO; mobile phase: 97:3 (v/v) carbon dioxideâethanol; temperature: 25 Â°C; Pout: 15 MPa. ELSD conditions: nitrogen pressure: 3 bar; drift temperature: 50 Â°C. 			

As expected, the peak intensity decreases with ELSD when increasing the mobile phase flow rate, but the ELSD response is still higher than the UV response, despite the very low wavelength used (210 nm) with the SFC mobile phase. The amount of organic solvent added to carbon dioxide was also studied (3–20% ethanol). 

Figure 7: Separation of four pharmaceutical compounds. Column: Cyano; mobile phase: 90:10 (v/v) carbon dioxideâethanol; temperature: 25 Â°C; Pout: 15 MPa. ELSD conditions: nitrogen pressure: 3 bar; drift temperature: 40 Â°C. UV detection wavelength: 210 nm. ELSD detection (a); 3 mL/min (blue signal); 7 mL/min (red signal); UV detection (b); 3 mL/min (blue signal); 7 mL/min (red signal). Peaks:  1 = theophylline; 2Â = cortisone; 3Â = hydrocortisone; 4 =Â dexamethasone. 			

				As reported previously, the mobile-phase composition induces response changes, both in HPLC (9) and SFC (10), leading to the use of an additional inverse gradient to maintain a constant response (10). The use of organic solvent is reported to shift the droplet size toward smaller values (36).			

Figure 8 shows that the increase in ethanol favours the compound elution (diastereoisomers), but dramatically reduces their ELSD response. This means that isocratic conditions should be preferred for SFC–ELSD and that the stationary phase should be carefully selected to allow good separation with the smallest modifier proportion. Classifications of stationary phases in SFC favouring this choice are available (37,38). 

Figure 8: Separation of diastereoisomers depending on the percentage of ethanol. Other conditions as in Figure 6. 			

				To compare the HPLC and SFC responses, studies were performed for caffeine with different mobile phases (methanol–water, ethanol–water, acetontrile–water and carbon dioxide–ethanol), having close eluting power to maintain identical retention factors (

With all mobile phases, the column provided symmetrical peaks. Figure 9 shows a log–log plot of peak area vs. concentration for these mobile phases. Two flow rates (3 mL/min and 1 mL/min) were studied in SFC. Satisfactory linearity is reported up to 50 ppm in all cases (log–log coordinates). The response increases from methanol–water to acetonitrile–water, and ethanol–water, then SFC at 3 mL/min. The response is multiplied by three from methanol to acetonitrile or ethanol in HPLC, and by three again from acetonitrile or ethanol in HPLC to SFC with 10% ethanol (3 mL/min). Moreover, at 1 mL/min, the signal is dramatically higher in SFC. The 

 coefficient regularly decreases in the order methanol<ethanol<acetontrile<SFC at 1 mL/min. Numerous studies report that this coefficient is related to the droplet size (39). The higher the value of 

, the lower the droplet size. Does that mean that in SFC, as a result of the lower 

value, the droplets are bigger? Another study reported the reverse effect in packed-capillary SFC by using 2-propanol as modifier (40). However, the use of capillary columns, that is, of very low flow rates and injected volumes, seems to affect the response significantly because of the direct linear relationships between the peak area and the concentration observed in such conditions (8,40–42). 

Figure 9: Log A vs log C for different mobile phases in HPLC and SFC. Column: Synergi Fusion RP. Black triangles: methanolâwater; green circles: acetonitrileâwater; red squares: ethanolâwater; blue diamonds: SFC at 1 mL/min (90:10 carbon dioxideâmethanol); pink diamonds: SFC at 3 mL/min.  			

Whatever the reasons, the peak area in SFC is dramatically improved. Lower droplet condensation is observed in SFC conditions because of the lower amount of liquid in the mobile phases (10% for our studies), apparently favouring the transfer of analyte from the nebulizer to the drift evaporative tube. 

The difference observed for the two flow rates in SFC could be a result of varied cooling effects of the nebulization chamber — greater at higher flow rates — which induces a greater droplet condensation on the wall (the overall signal decreases from 1 mL/min to 3 mL/min) and a larger droplet size of the ternary aerosol, which explains the lower 

Finally, there is an experimental linear correlation in HPLC between the two coefficients, 

 (24). Figure 10 shows this correlation for 14 varied compounds in SFC: palmitic and linolenic acid, tripalmitoyl- and trilinonenoylglycerol, cholesterol, hexadecanol, octadecanol, eicosanol, docosanol, theophylline, ketoprofen, hydrocortisone, caffeine and 4-hydroxybutylparaben. A linear relationship is observed in SFC with 90:10 (v/v) carbon dioxide–ethanol mobile phase (blue diamonds). Furthermore, other data obtained in HPLC were added to Figure 10. With a Sedex 75 detector in a nonaqueous reversed-phase LC separation (acetontrile–dichloromethane) (39), the values for one compound (tricaprylin) with different post-column addition (red squares) are very close to the values obtained in SFC with various compounds. The slopes of the tendency curves are also very close for direct injection of ceramide with various pure organic solvents (pink line) (26) and for HPLC analysis of neomycin with aqueous mobile phases with the detector mentioned previously (green line) (43). The main differences are noticed on the intercept, which appears to be strongly improved with aqueous mobile phases and lowered with direct injection in pure organic solvents. 

Figure 10: Log b vs. log a. Blue diamonds: 14 compounds in SFC; red squares: tricaprylin in nonaqueous reversed-phase HPLC with acetonitrileâdichloromethane mobile phase with different organic postcolumn additions (data from reference 39); pink crosses: ceramides by direct injection in pure organic solvents (data from reference 16); green triangles: neomycin in HPLC with different aqueous mobile phases (data from reference 43). 			

The use of ELSD with SFC is favoured by the nature of the mobile phase in SFC, which enables aerosol formation and thus improves the peak area, at least for the carbon dioxide–ethanol mixture. The 

 coefficient is smaller, possibly indicating that the particle size is greater. Conversely, the log 

 value is greater, explaining the higher response obtained in SFC conditions. From the numerous analytical parameters studied, the mobile-phase flow rate and the amount of modifier mixed with carbon dioxide are the main parameters affecting peak area. The outlet pressure also plays a role in response by affecting the speed of fluid in the nebulizer.   

The parameters acting on the response of ELSD were not well known when coupling with supercritical fluid chromatography. Moreover, they could vary depending on the ELSD used, because of the different types of nebulizer chamber set on commercial detectors. This paper investigated a lot of parameters used to optimize separation in SFC (outlet pressure, modifier percentage) at different flow rates, showing the effects of the analytical conditions on response, and that should be useful for many chromatographers in the goal to achieve the best detection possible. Moreover, the use of supercritical fluid chromatography ensures higher peak areas than HPLC.  

 are associate professors with the Institut de Chimie Organique et Analytique UMR 7311, at the Université d'Orléans in Orléans, France. 

are bachelor students with Sederein Orléans, France and 

Article

The article investigates the ELSD response variation with supercritical carbon dioxide-based mobile phases and compares the response with HPLC.

Coupling SFC with ELSD: Study of the Parameters Ruling Response

A change in peak shape is one of the most common observations of problems with a liquid chromatography (LC) method. Because of this, most system suitability tests include a measure of peak shape, so a quantitative value of peak shape can be tracked over time. Poor peak shape can compromise the results of an analysis by degrading resolution between closely eluted peaks and reducing precision and accuracy of measuring peak area, especially for small peaks. A change in peak shape is one of the first signs that the column is failing, but there are other causes of peak tailing, as well. This month we look at several aspects of peak tailing as we continue our "Troubleshooting Basics" series (1–3) of column instalments. 

The ideal chromatographic peak will have a Gaussian shape, but it is rare that a perfectly symmetric peak is observed in real chromatograms. Most peaks tail slightly, and as the column ages, tailing typically increases. However, there are several other potential causes of peak tailing (or fronting) as well, so it is a good idea to track the peak shape over time to anticipate when practical problems will occur. As a result, nearly all system suitability tests include a measurement of peak shape. 

Figure 1: Measurement of tailing factor and asymmetry factor.			

The two most popular methods of measuring peak shape are illustrated in Figure 1. Other methods to measure peak shape are used much less often. The pharmaceutical industry uses the tailing factor, TF, which is determined by measuring the entire peak width at 5% of the height and dividing it by twice the front half-width. Nonpharmaceutical laboratories often use the asymmetry factor, 

, which is calculated by measuring the back half-width of the peak at 10% of the peak height and dividing it by the front half-width. You can see that if the peak is perfectly symmetric, the front and back half-widths will be the same, no matter where they are measured relative to the peak height, so for such peaks, TF ≡ 

. As tailing increases, however, the two numbers diverge, with 

 growing faster than TF, but for peaks with a value < 2 there is not a very noticeable difference. There is no inherent value in using one technique vs. the other for measuring peak shape; rather, it is important to choose one technique and use it to look for changes in peak shape over time. Most LC peaks tail or front a bit, so column manufacturers typically set their column-release specifications of 0.9 < TF < 1.2 as normal performance. As can be seen in Figure 2, when tailing increases, several practical problems can arise. The peaks are harder to integrate because the transition from the baseline to the peak or peak to baseline is much more gradual, and on noisy or sloping baselines the peak limits are difficult to determine. Generally, the peak area stays constant, so increased peak tailing translates into shorter peaks, and peak height is the limiting factor in determining detection limits, so method limits can suffer with tailing peaks. Tailing peaks also take a larger time window to be eluted, so to achieve baseline resolution between peaks the run time must be longer. And tailing peaks are aesthetically less pleasing. You can see that all these factors favour symmetric peaks. From a practical standpoint, peak tailing is difficult to eliminate, however, for many applications peaks with TF ≤ 1.5 are acceptable. When TF ≥ 2, usually corrective action should be taken to identify and eliminate the source of tailing. 

When peak tailing occurs, it usually shows up for one or just a few peaks in the chromatogram, but sometimes all the peaks in the run tail. The appearance of peak fronting is much less common. Most often, these three behaviours are caused by three different sources. We will look at each of the three problems next. 

Usually, one or a few peaks in the chromatogram tail and the cause is most often chemical in nature. The problem may appear during method development, in which case you probably do not know if the peak shape was ever good. Or more often with an existing method, a peak that had acceptable shape in the past begins to tail, and tailing increases over time. If the onset of the tailing was sudden, as when a new batch of samples was run, look for some other change that coincided with the observation, such as preparation of a new batch of mobile phase or replacement of the column or guard column.  

Mobile-phase changes are the easiest to check. For example, the pH of the mobile phase can have a strong influence on the peak shape, so if an error was made in pH adjustment, this could be the problem source. Mobile-phase problems will also usually cause changes in retention time. If such correlated changes are observed, carefully prepare another batch of mobile phase (or repeat some other change that was made) and see if that corrects the problem. Sometimes a method will be developed that is not very robust. In such cases, even small changes in temperature or other variables can cause a change in the retention or peak shape. Another mobile-phase problem that occurs occasionally is insufficient buffer concentration. Although reversed-phase separations are not strongly affected by buffer concentration, hydrophilic interaction chromatography (HILIC) and ion-exchange are much more sensitive to buffer effects. A buffer concentration of 5–10 mM is usually adequate to buffer the mobile phase, column and injection solvent in reversed-phase separations. If buffer concentration problems are suspected, double the concentration and see if this fixes the peak shape. 

After the mobile phase is eliminated as the problem source, look to the column. If a guard column is in use, remove it and make an injection — if the peak is OK now, the guard column has failed. If the problem persists without a guard column, substitution of a new column for the old one is the simplest way to check for other column problems. Column problems are more likely after ≈500 or more samples have been run, but in some cases column problems can occur much earlier. Dirty samples and mobile phases outside the 2 < pH < 8 range are two common sources of rapid column deterioration. If replacing the column corrects the problem, consider improving the sample cleanup, changing the mobile phase pH or using a guard column to help delay the problem of column deterioration. 

The source of peak tailing is not a simple process. In Figures 3(a)–3(c), you can see the influence on peak shape of increasing amounts of mefanamic acid. Under these conditions (pH 2.8), the mefanamic acid is well below its p

, so it is not ionized. At low loading [Figure 3(a)], the tailing is dramatic. Peaks with exponentially shaped tailing are likely a result of two different retention processes going on simultaneously; for example, some molecules might be interacting with column sites that equilibrate slowly (those on the tail) and some with sites that equilibrate quickly (those on the main peak). As the mass on column increases [Figures 3(b) and 3(c)], the majority of the sample molecules are retained by a single mechanism (fast equilibration), and the peak shape improves. The slow equilibration process doesn't disappear, but it is a smaller portion of the total, so tailing is reduced. On the other hand, amitriptyline, an ionized base at pH 2.8, looks better than mefanamic acid at low concentrations [compare Figures 3(a) and 3(d)]. As the injected mass of amitriptyline increases [Figures 3(e) and 3(f)], two things are observed. First, the peak tail begins to change shape until it takes on a right-triangle appearance [Figure 3(f)]. Second, the retention time gradually gets smaller. These are two classic symptoms of column overload. To verify this problem source, reduce the sample mass on column and see if retention increases and tailing improves. The cause of tailing in this case is likely because of ion exclusion. As the amount of amitriptyline adsorbed inside the pores in the column increases, the pore takes on a net positive charge (amitriptyline carries a positive charge at pH 2.8). At some point, the positive charge inside the pore is sufficient to repel another positively charged molecule, so it must travel further down the column before it finds a compatible site on the surface. The interesting thing about these two examples is that at the same pH, one analyte (mefanamic acid) gives less tailing as the sample load increases, whereas the other (amitriptyline) gets worse. It is hard to generalize about the source of peak tailing. 

Figure 3: Peak tailing for (aâc) mefanamic acid and (dâf) amitriptyline at pH 2.8. Injection mass on column: (a,d) 3 ng; (b,e) 500 ng; and (c,f) 15 Î¼g.			

Peak tailing was attributed to problems related to chemical interactions on the column. One way of thinking about peak tailing is that the active sites — the places on the column where interactions between the analyte molecules and the chemical surface of the column occur — become saturated. It is possible to conceive of a similar case in which the mobile phase becomes saturated or overloaded, and in such cases, peak fronting would occur. This indeed happens, but with reasonable buffer concentrations (for example, ≥ 5 mM), such overload is rare in reversed-phase LC. A more common source of peak fronting is illustrated in Figure 4, where peaks from two consecutive injections are shown. The peak changes from a normal appearance in injection number 527 [Figure 4(a)] to badly fronting in injection 528 [Figure 4(b)]. The most common cause of such changes is a sudden physical change in the column, usually referred to as column collapse. In the present case, the column was operated at pH 9.6 at 70 °C, but the column was designed to be used at pH ≤ 7 and ≤ 40 °C. The aggressive mobile phase conditions gradually dissolved the silica particles inside the column until they became so fragile that the minor shock of the injection valve rotating caused the internal column structure to collapse. The proper fix for such a problem would be to modify the method so that it operated within the recommended limits of the column or replace the column with a more robust one. That is, if all else fails . . . read the directions! However, for the present example, it made more sense economically to replace the column every 500 injections rather than redevelop and revalidate the method. 

Figure 4: Peak fronting for two consecutive injections: (a) 527 and  (b) 528.			

When all peaks in the chromatogram tail or are split or doubled, as in Figure 5, this is a symptom of a problem that happens at the inlet of the column before any separation takes place. The most common cause of such problems is a partially blocked inlet frit on the column. Debris from the sample, the mobile phase or a failed pump seal or injector rotor can collect on the inlet frit. As the frit becomes partially blocked, the sample stream arriving at the column inlet is distorted, resulting in peak distortion. Because this problem happens before any sample molecules are separated, it affects all peaks in the chromatogram in the same manner. You can fix the problem about a third of the time by reversing the column and backflushing it to waste for a few minutes. Most columns can be reversed for such flushing, but it is best to check the column care-and-use instructions to make sure it is allowed for your column. If reverse-flushing corrects the problem, you can proceed as normal. If it doesn't, replace the column. In either case, it is wise to eliminate or reduce the source of particles arriving at the top of the column. Improve the sample pretreatment by adding a centrifugation or filtration step. Replace worn pump or autosampler parts. Filter the mobile phase to remove particles. And a good safety measure is to place an in-line filter between the autosampler and the column. If you use a guard column, it acts as a filter to protect the column, but the in-line filter is less expensive than the guard column and easier to service, so I recommend using one in every system, even if you are using a guard column. 

Figure 5: Peak distortion for all peaks in the chromatogram.			

Changes in peak shape over time are common in the use of LC methods, but such changes should occur gradually over hundreds or thousands of samples in a well-behaved method. Tailing of one or several peaks usually points to a problem with some chemical aspect of the system, so check the mobile phase and column for problems. When all the peaks tail or are similarly distorted, it is a sign that particulate matter is collecting at the top of the column. Better sample preparation and protection of the column will help to avoid this mode of failure. Peak fronting can result from chemical problems in the system, but is more commonly attributed to catastrophic column failure, usually resulting from using the column outside the recommended operating limits. Check the column instructions for the column limits, change to a more robust column or expect to replace the column more often. 

Although peak shape changes are a sign of problems and are difficult to avoid, if a method is developed and tested for robustness, such problems should not be a major concern for the method. A good system suitability test coupled with tracking peak shape over time should allow you to anticipate when peak shape changes will compromise the quality of the data. Take the appropriate corrective action before data are compromised, and your method should be satisfactory. 

 is vice president of LC Resources, Walnut Creek, California, USA. He is also a member of 

 editorial advisory board. Direct correspondence about this column should go to "LC Troubleshooting", 

, 4A Bridgegate Pavilion, Chester Business Park, Wrexham Road, Chester, CH4 9QH, UK, or e-mail the editor, Alasdair Matheson, at 

A calculation on page 248 of the May instalment of "LC Troubleshooting" (

[5], 246–252 [2012]) was incorrect. The calculation showed that 0.1% × 0.1% = 0.01%, when in fact it should have been 0.1% × 0.1% = 0.0001%. 

LC Troubleshooting

What do you do when the peak shape changes?

Troubleshooting Basics, Part 4: Peak Shape Problems

The separation of olefin-based elastomers such as ethylene-propylene (EP) and ethylenepropylenediene (EPDM) rubbers according to their chemical composition cannot be achieved using techniques based on crystallization, which are widely used to fractionate semi-crystalline olefin copolymers. Interactive liquid chromatography using porous graphite as a stationary phase and gradients of long chain alcohols and chlorinated aromatics offers the possibility to separate these materials according to their chemical composition. Coupling the high performance liquid chromatography (HPLC) separation according to the chemical composition in the first dimension with size exclusion chromatography (SEC) according to the molar mass in the second dimension yields the complete molecular heterogeneity of these materials for the first time. 

Ethylene-propylene-diene (EPDM) terpolymers, which consist of the three monomers:  ethylene, propylene and diene, continue to be one of the most widely used synthetic rubbers for both speciality and general purpose applications. They excel with their outstanding resistance to ozone, oxidation, ageing, weather, aggressive chemicals and high temperatures (1,2). These unique features make EPDM useful for outdoor and elevated temperature applications, such as automotive profiles, singleply roofing, electrical insulation, belts, window gaskets and water seals. Dicyclopentadiene, 5-ethylidene-2-norbornene (ENB) and 5-vinyl-2-norbornene are the dienes typically incorporated in commercial EPDM. On a molecular level the application properties are influenced by the molar mass and its distribution (MMD), the chemical composition and its distribution (CCD), and the distribution of the monomers across the molar mass distribution (MMCCD). It is therefore important to study these distributions. 

The MMD of polyolefins and the corresponding average values can be analysed by size exclusion chromatography (SEC). The CCD of semi-crystalline olefin copolymers is usually determined by fractionation techniques like crystallization analysis fractionation (CRYSTAF) (3) and temperature rising elution fractionation (TREF) (4). These discriminate between the polymer chains based on their crystallizability from a dilute solution in a temperature gradient and consequently fail to fractionate amorphous ethylene-propylene and EPDM samples.  

HPLC could be a useful alternative to fractionation techniques considering the selectivity, speed of separation and the ability to separate amorphous samples that is achievable. However, in the case of olefin-based elastomers no suitable stationary and mobile phases were known. Applications of porous graphite as stationary phase for the liquid chromatographic separation of small molecules have been known for more than a quarter of a century and have been reviewed (5). Compared to conventional reversed phases, graphite displays superior selectivity for particular structural features like aromatic units or alkyl chains. However, this potential has never been exploited for the chromatographic fractionation of synthetic polymers.  

The first application in this regard was the separation of blends of linear polyethylene, and the three stereoisomers of polypropylene using a commercially available porous graphite column as stationary phase and a gradient of 1-decanol→ 1,2,4-trichlorobenzene (TCB) as the mobile phase (6). The underlying principle is the alignment of alkanes (ethylene sequences) with their carbon skeleton parallel to the graphite surface plane. This knowledge was further extended to copolymers of ethylene and 1-olefins and it was demonstrated that the elution volume of these materials decreases with increasing incorporation of the comonomer (7–9). 

Synthetic copolymers are distributed in more than one parameter (i.e., in molar mass, chemical composition and tacticity) and these distributions are mutually dependent. Therefore unravelling these heterogeneities and their mutual dependence requires the coupling of analytical techniques. Comprehensive two-dimensional chromatography (2D-LC) with high performance liquid chromatography (HPLC) coupled to size exclusion chromatography (SEC) yields information on the relationship between the chemical composition distribution and the molar mass (10,11). In this approach the samples are first separated by their chemical composition and these fractions are further analysed by SEC. Such separations were realized recently for functionalized polyolefins (12) and ethylene/1-octene copolymers (13, 14), as well as for polyolefin blends (15). In this paper we want to demonstrate how amorphous EP copolymers and EPDM terpolymers can be separated by high temperature HPLC (HT-HPLC) using porous graphite as stationary phase and how this knowledge can be used to realize a two-dimensional separation with regard to composition and molar mass. 

High-temperature HPLC (HT-HPLC): A high-temperature chromatograph PL XT-220 (Agilent Technologies, Santa Clara, California, USA) containing a robotic sample handling system PL-XTR (Agilent Technologies) for injection of the samples was used. A high-pressure binary gradient pump is a part of this chromatograph. The temperature of the sample block, the injection needle, the injection port and the transfer line between the auto sampler and the column compartment was set at 160 °C. An evaporative light-scattering detector (ELSD, model PL-ELS 1000, Agilent Technologies) was used for detection. The following parameters were used for running the ELSD: gas flow rate = 1.5 L/min, nebulizer and evaporator temperatures = 160 °C and 260 °C, respectively. The flow rate of the mobile phase was 1 mL/min. The polymers were dissolved for 2–5 h in 1-decanol at a concentration of 1–2 mg/mL and a temperature of 160 °C. 50 µL of the polymer solution was injected into the column.  

A Hypercarb column (Thermo Scientific, Dreieich, Germany) containing particles of porous graphitic carbon (average particle size 5 µm, surface area 120 m

 /g, pore size 250 Å) was used. The dimensions of the column were 100 × 4.6 mm i.d. 

High-temperature two-dimesional liquid chromatography (HT 2D-LC): 

All experiments were performed using a prototype chromatographic system for high-temperature two-dimensional liquid chromatography constructed by Polymer Char (Valencia, Spain), comprising an autosampler, two separate ovens, valves and two pumps equipped with vacuum degassers (Agilent, Waldbronn, Germany). One oven was used for thermostating the SEC column, while the second one, where the injector and a switching valve were housed, was used to thermostat the HPLC column. A scheme of the HT 2D-LC setup is shown in Figure 1. 

Figure 1: Scheme of 2D-LC at high-temperature. 

The HT-HPLC and HT-SEC columns were coupled by an electronically controlled eight-port valve system EC8W (VICI Valco instruments, Houston, Texas, USA) equipped with two 200 µL loops. From the moment of injection into the HPLC column (100 µL injection loop), the eight-port valve was switched every 2 min in order to inject 200 µL of effluent from the HPLC into the SEC column.  

The effluent from the SEC column was monitored by an ELSD. Both ovens, the autosampler as well as the transfer line between the autosampler and the columns and the ELSD, were thermostated at 160 °C. The 2D-LC was handled by software provided by Polymer Char (Valencia, Spain). WinGPC 7.0 (Polymer Standards Service, Mainz, Germany) was used for data acquisition and evaluation. 

First dimension separations were carried out on a Hypercarb column packed with porous graphite particles with the following parameters: column size 250 × 4.6 mm i.d., average particle size diameter 5 µm, surface area of 120 m

 /g and pore size of 250 Å (Thermo Scientific, Dreieich, Germany). A PL Rapide H column, 150 × 7.5 mm (Polymer Laboratories, Church Stretton, England) was used in the second dimension (SEC). 

 1-decanol, 2-octanol and 1,2,4-trichlorobenzene (TCB) of synthesis quality (Merck, Darmstadt, Germany) were used as the mobile phases. TCB was distilled before use and 1-decanol and 2-octanol were used as delivered.  

All gradient runs in HPLC had 3 min of isocratic elution before applying a 10 min linear gradient starting with 100 vol. % of 1-decanol and ending with 100 vol. % of TCB.  

In HT 2D-LC the first dimension separations were done by applying a linear gradient 1-decanol→TCB at a flow-rate of 0.1 mL/min. Starting with 100% of 1-decanol for 40 min, the volume fraction of TCB was linearly increased to 100% within 80 min and then held constant for a further 80 min. Finally, the initial chromatographic conditions were re-established. TCB was used as the mobile phase in the second dimension (SEC) with a flow rate of 2.5 mL/min. 

Separation of ethylene-propylene rubbers according to chemical composition

: To ensure that all polymer samples are completely dissolved, temperatures as high as 160 °C were used. The polymer solutions were injected into the porous graphite column and adsorbed from long chain alcohols (1-decanol or 2-octanol) on the column packing. The retained polymers were then desorbed after 10 min linear gradient 1-decanol or 2-octanol→TCB. The elugrams of the EP copolymers, which vary in their ethylene content, are shown in Figure 2(a).  

Figure 2: (a) Overlay of chromatograms of EP copolymers. The average chemical composition (content of ethylene, C2) is indicated in the figure, and (b) correlation between the elution volume at peak maximum and the average ethylene content and weight-average molar mass (Mw). Mobile phase: gradient 1-decanolâTCB. 

While the sample with the highest ethylene content (80.8 mol %) elutes at 41 vol. % of TCB in the gradient, the sample with the lowest ethylene content (25.5 mol %) elutes early at 16 vol. % of TCB. The elution volume of the samples at peak maxima increases linearly with the content of ethylene [Figure 2(b)]. At the intercepts with the compositional 

-axis, the elution volume in Figure 2(b) corresponds to that of linear polyethylene having a molar mass above 20 kg/mol at 100 mol % ethylene and the elution volume at the other end (at 0 mol % of ethylene), indicates that the propylene units present in the EP samples are isotactic or atactic. No correlation exists between weight-average molar mass (M

) and the elution volume [Figure 2(b)] suggesting that the retention is not influenced by the molar mass in the tested range  of 200–2400 kg/mol.  

Figure 3 illustrates the excellent baseline separation of an EP blend separating each component at a concentration of 1 mg/mL into individual components using the chromatographic system porous graphite/2-octanol→TCB. The difference in peak heights, that is, the response of the ELSD, could be attributed to its dependence on the composition of the mobile phase as well as that of the analyte (16). The increase in polarity of the alcohol (2-octanol vs. 1-decanol) leads to a stronger retention of the EP copolymers on porous graphite. These chromatographic systems allow the monitoring of the CCD of synthesized copolymer samples and the separation of different EP copolymers blends, as well as identifying homopolymers in them. 

Figure 3: Overlay of chromatograms of a blend of EP copolymers at a weight ratio of 1:1:1:1. Average chemical composition (content of ethylene, C2) is indicated in the figure. Mobile phase: gradient 2-octanolâTCB. 

 The exclusive influence of ENB on the retention of EPDM samples on porous graphite is shown in Figure 4 using a set of samples which are identical with regard to their content of ethylene but vary in ENB content.  

Figure 4: (a) Overlay of chromatograms of EPDM samples and (b) elution volume at peak maximum vs. ENB content. Mobile phase: gradient 1-decanolâTCB. 

It can be clearly observed from these chromatograms that the elution volume increases linearly with the ENB content. While the concentration of ethylene determines the elution volume in EP copolymers (Figure 2), the content of both ethylene and ENB influence the retention of EPDM terpolymers (17). As a hypothesis we tentatively attribute the influence of the diene to either its geometrical arrangement or the unsaturation left in the incorporated ENB, which could promote adsorption via attractive interaction with the graphite.  

The use of interactive liquid chromatography using porous graphite as a stationary phase and gradients of long chain alcohols and chlorinated aromatics to separate olefin-based elastomers according to their chemical composition is discussed.

LCGC Europe-07-01-2012

Can You Meet the Technical Requirements of Annex 11?

A GC Troubleshooter's Toolkit

Coupling SFC with ELSD: Study of the Parameters Ruling Response

Troubleshooting Basics, Part 4: Peak Shape Problems

Interactive Liquid Chromatography of Olefin-based Elastomers

Event News