Unilever R&D Vlaardingen, Vlaardingen, The Netherlands, 

Van ‘t Hoff Institute for Molecular Sciences, University of Amsterdam, Amsterdam, The Netherlands

Monitoring lipid oxidation during the shelf life of lipid-containing food emulsions, such as mayonnaise, is challenging. It is, however, essential for the development of improved, consumer-preferred products. Determining the nonvolatile lipid oxidation products (NONVOLLOPS), the precursor compounds for rancidity, is required to determine the effectiveness of product stabilization technologies. A method based on normal-phase liquid chromatography with atmospheric pressure photo ionization-mass spectrometry (LC–APPI-MS) was developed for this purpose. The inclusion of a size-exclusion chromatography (SEC) step was needed to remove interfering diacylglycerides and free fatty acids from the samples. The combined SEC and normal-phase LC–APPI-MS method allowed the identification of a wide range of oxidized species including hydroperoxides, oxo-2½ glycerides, epoxides, and other oxidized species. The method was found to be more suitable for the analysis of large sample sets. The relative levels of NONVOLLOPS from both ambient and accelerated stability tests could be determined. The results were compared to hexanal measurements. The data showed that NONVOLLOPS predict the rancidity of different formulations in a much earlier stage during shelf-life tests, providing valuable information for future product development. 

	The taste, flavour, and texture of lipid-rich foods, such as mayonnaises and dressings, strongly depends on the composition of the oil used and the emulsification methods applied during the preparation of the product. Making consumer-preferred products with healthier, unsaturated edible oils is a key challenge for the food industry. The unsaturated fatty acids in the triacylglycerols (TAGs) can undergo oxidation and hydrolysis reactions during shelf life, forming volatile and nonvolatile lipid oxidation products. Even at very low levels the volatile lipid oxidation products can cause the typical rancid off-smell of aged oil. High amounts of nonvolatile lipid oxidation products could even adversely affect the consumer’s health (1,2). Being able to accurately monitor lipid oxidation products is therefore crucial for safeguarding consumer safety and to study novel, natural stabilization technologies that suppress lipid oxidation reactions in lipid-rich foods.

	A clear drawback of the healthier unsaturated fatty acids oleic, linoleic, and linolenic acid is their much higher amenability for oxidation reactions. The processes occurring during lipid oxidation so far are not fully understood, but it is believed that most reactions proceed through radical initiation (3). First, peroxidized and hydroperoxidized TAGs are formed that are precursors for the later volatiles. Dissociation reactions can then result in the loss of volatile compounds, such as aldehydes and ketones, and the formation of so-called “2½ glycerides”. Other reaction pathways can lead to the formation of numerous nonvolatile oxidation products, such as hydroxy-TAG epoxides, TAGs with aldehyde functionalities, and polymers. Understanding the formation of these precursors is crucial for developing new stabilization technologies. The broad range of TAGs in edible oils combined with the wide variety of possible oxidation reactions makes the analysis of the nonvolatile precursors of the rancid smell very complex. 

	A number of fast and simple methods have been applied to monitor oxidation in oils. These include measuring the peroxide value (4,5) or the 2-thiobarituric acid value (6) of the oil. Although important and widely used, these measurements are highly empirical and unspecific (7). In addition, they do not really provide information down to the molecular level, which is needed to build a true understanding of the complex interplay of oxidation reactions. Chromatography combined with mass spectrometric detection is the method of choice for sensitive and selective analysis of both the volatile and nonvolatile intermediate and final lipid oxidation products. Volatile components are commonly measured by gas chromatography–mass spectrometry (GC–MS) headspace analysis (8). In the chromatographic analysis of nonvolatile oxidation products two factors must be considered. First, it should be realized that many of these products are not end products of the oxidation reaction, but are actually unstable intermediates. The analytical protocols applied for sample preparation and analysis should hence be very mild-and even then decomposition of labile intermediates might still occur. A second parameter to consider is the complexity of the mixtures obtained. Different classes of oxidized species are present, each consisting of a wide range of positional isomers and covering the entire chain length distribution.

	We have demonstrated in previous work that normal-phase high performance liquid chromatography with electrospray MS detection gives detailed information on the nonvolatile lipid oxidation products present in oxidized oils (9). The method performs a polarity-based class-separation of the oxidized TAGs, providing information on groups such as the hydroxyperoxy-, hydroxyl-, epoxy-, or oxo (ketone or aldehyde)-TAGs, or TAGs with combinations of these chemical functionalities. This method remains challenging, primarily because of limitations in the repeatability and reproducibility. Sensitive electrospray ionization of the nonpolar compounds requires the use of a post-column additive to facilitate ionization, which unfortunately rapidly pollutes the ion source. Another limitation is the sample preparation. Normal phase chromatography requires the samples to be free of water. Food products, such as mayonnaise and spreads, however, are emulsions containing up to 75% water. The products from the lipid oxidation reaction can accumulate in either the oil phase, the water phase, or at the oil-water interface. Freeze-drying effectively removes water, allowing extraction with organic solvents, but it is time consuming. Taking this all together, the normal-phase LC–electrospray MS method and the sample preparation required are far from ideal for studying the initial stages of lipid oxidation, especially if large sample sets have to be analyzed.

	In this article we describe a novel analytical approach for determining the nonvolatile oxidation products (NONVOLLOPS) in mayonnaise samples from (accelerated) shelf-life tests. The method uses atmospheric pressure photo ionization (APPI) MS. The interpretation of the mass spectra is discussed and methods to distinguish MS fragmentation from overlap of nonreacted species are evaluated. Eventually, the nonvolatile precursors of the off-smell compounds of aged mayonnaise are determined. Understanding offâsmell formation is built by correlating the results of the new NONVOLLOPS methods with those of classical hexanal measurements.

: Stable isotope labelled Cholesterol D6 (purity 97–98%) was purchased from Cambridge Isotope Laboratories for use as an internal standard. Acetonitrile, isopropyl alcohol (IPA), and n-hexane (ULC–MS-grade) were obtained from Biosolve and tetrahydrofuran (HPLC-grade) was purchased from Merck.

Normal-Phase Liquid Chromatography–Atmospheric Pressure Photoionization-Mass Spectrometry:

 The HPLC system consisted of an Agilent 1200SL binary pump equipped with an HTC PAL auto sampler (CTC Analytics), an Agilent 1200 series micro vacuum degasser, and an Agilent 1200 series thermostatted column compartment. The mass spectrometer used was an Agilent 6410 triple quad equipped with an Agilent G1971A APPI source. The ion source was equipped with a krypton discharge lamp and was operated in positive ionization mode. The optimized gas and vaporizer temperatures were 250 °C and 275 °C, respectively. The gas flow was 7 L/min, nebulizer pressure was 40 psi, capillary voltage was 2 kV, and the fragmentor voltage was 80 V. Full scan mass spectra were acquired from 

	A 250 × 3.0 mm, 3-µm pursuit XRs 2 Diol column (Varian) was used for separation. The analytes were eluted at a flow rate of 0.3 mL/min using a linear gradient of n-hexane (Solvent A) and 9:1 (

) n-hexane–IPA (solvent B) programmed as follows: 5 min at 3% B, in 25 min to 8% B, in 10 min to 25% B, 10 min at 25% B, in 2 min to 50% B, 5 min at 50% B, in 1 min to 3% B, and finally 32 min at 3% B. The total run time was 90 min. The column temperature was maintained at 25 °C and the injection volume was 5 µL. The LC–MS method is not quantitative because of the lack of pure standards. In our method we used Cholesterol-D6 as the internal standard enabling us to accurately establish (relative) concentration values. 

The HPLC system consisted of a Shimadzu LC-10 isocratic pump equipped with an Optimas autosampler (Separations/Spark Holland), a GT-103 solvent degassing unit (Separations/Spark Holland), a Mistral column oven (Separations/Spark Holland), and a Shodex RI-71 refractive index detector (Shodex). Four 300 × 7.5 mm, 5-µm PLgel polystyrene–divinylbenzene sizeâexclusion chromatography (SEC) columns (Agilent) connected in series were used for the separation. The compounds were eluted at a flow rate of 0.8 mL/min using tetrahydrofuran as eluent at isocratic pump conditions. The column temperature was maintained at 30 °C and the injection volume was 10 µL. Chromeleon Chromatography software (Interscience) was used for data acquisition. 

Volatile Compounds by Gas Chromatography:

 Hexanal measurements were performed according to AOCS recommended practice Cg 4-94 (10) on an Agilent 7890A GC–MS equipped with a Gerstel MPS2XL multipurpose sampler.

	The stability of various mayonnaise formulations was studied in a shelf-life test at ambient conditions over 200 days. The samples studied were mayonnaises obtained from local supermarkets and prototypes containing natural antioxidants (confidential stabilization technology 1 and 2). For each formulation tested and each time-point, one millilitre of sample was put in a 20-mL headspace vial. The vial was filled with air, capped, and placed in a temperatureâcontrolled room (20 °C) in the dark. Samples were taken at seven-day intervals and were immediately stored at -80 °C to stop any further oxidation and awaited analysis as one large batch at the end of the shelf-life study. Preparation of the samples consisted of freeze-drying for 48 h. Water was removed under mild conditions and all species, irrespective of whether they were present in the water- or lipid-phase, or at the oilâdroplet interface, were retained. After freezeâdrying, 4 mL of 

-hexane was added to the vials to dissolve the lipids and their oxidation products. Upon dissolution the proteins precipitated and clear sample solutions were obtained that could be directly injected into the normalâphase liquid chromatography–atmospheric pressure photo ionizationâmass spectrometry (normal-phase LC–APPI-MS) instrument. 

The normal-phase LC separation is largely based on polarity, with retention of the lipids and their oxidation products being determined by the presence of polar groups resulting from the oxidation reaction. The retention order of the oxidation products was established in our earlier studies (9). As an illustration, Figure 1 shows the normal-phase LC–APPI total ion current chromatogram of a 154-day aged mayonnaise after freezeâdrying. The first eluting group are the unreacted TAG, which elute almost unretained. The next group is that of the epoxy- or oxo-TAG, oxidation products containing one additional oxygen as a ketone or aldehyde functionality. 2½ glycerides are slightly more polar and elute between 25 min and 40 min. In the 40-min to 50-min time interval triglycerides with a single hydroperoxide group elute. More oxidized lipids are retained longer, with lipids containing two or more hydroperoxide groups eluting at retention times above 50 min. 

APPI is known for its efficient ionization of nonpolar or low-polarity compounds, such as sterols and their oxidation products (11). In APPI the ion source is less susceptible for pollution because no additives are needed to facilitate ionization as would be the case in electrospray ionization. In photo ionization, two ionization pathways can be exploited: direct photo ionization or dopant-assisted photo ionization (12). In this work ionization of the compounds was achieved by APPI using the self-doping effect of the hexane-based mobile phase (13). The photons emitted by the krypton discharge lamp have energies in the range of 10 to 10.6 eV, which is above the ionization potential of the n-hexane eluent (10.18 eV). Via charge exchange and proton transfer, the primary hexane ions cause ionization of the analytes in the gas phase. The effectiveness of this approach is shown in Figure 2. This figure shows the mass spectra for some of the more abundant peaks in the time window from 40 to 50 min in Figure 1. All mass spectra show intense signals, especially in the mass range between m/z 850 and 900. From the LC elution positions it is likely that these signals represent hydroperoxides. The molecular ions of hydroperoxide-TAGs with three C18 fatty acyl chains would be expected at 

 values of approximately 920 Da, depending on the number of double bonds in the molecule. However, these molecular ions do not seem to be present. Mass losses of 18 and 34 Da are seen, indicating the loss of water or hydroperoxide groups and resulting in the fragment ions [M+H-H

]+, respectively. Smaller ions with m/z values in the 400 Da to 600 Da range are also seen in this retention time window. These could result from in-source fragmentation, but they could also represent other species of similar polarity and lower molecular weight. Experiments at other ionization settings (fragmentor voltage 50 V to 100 V) did not change the abundance of the 400–600 Da ions. It was concluded that these signals are not caused by fragmentation, but most likely represent other compounds. Clearly the chromatograms and MS spectra obtained are too complex for further interpretation. Additional separation of these complex samples could greatly aid the interpretation of the results obtained.

In the initial normal-phase LC separations of the aged mayonnaise multiple peaks were seen with polarities corresponding to one or more hydroperoxy-groups, yet molecular weights far below that of the HOO-TAG. SEC was applied to determine the molecular weight distribution of the sample, focusing in particular on the polymerized TAG, the TAG, as well as the di- and mono-acylglycerols and the free fatty acids. By using SEC with refractive index detection the relative content of each of these individual groups could be determined. A representative SEC chromatogram is shown in Figure 3. The largest peak represents the TAGs, or, more correctly phrased, TAGs and compounds with a molecular size similar to that of a TAG. This group accounts for 95.8% of the total signal. Primary oxidation products are just one or a few oxygen atoms larger than the TAG and hence in SEC coelute with the TAGs. Significant levels of diglycerides (DAGs) and free fatty acids (FFAs) were also seen in the SEC separation. These species are present at concentration levels of 2.4% and 0.7%, respectively, and could explain the low-molecular-weight polar species seen in normal-phase LC–APPI-MS. Furthermore, polymerized triglycerides are present at a concentration level of 1.1%. These are secondary oxidation products known to be formed from the fusion of two hydroperoxides via different pathways (14). The levels of DAGs and FFAs are high relative to those of the oxidized TAGs. They interfere with the TAG-oxide identification and quantification, especially in the analysis of samples from the initial stages of oxidation where the levels of the oxidation products are still low.

	To eliminate overlap of DAGs and FFAs with the oxidized TAG in the normal-phase LC–APPI analysis, samples were prefractionated based on size by SEC prior to normal-phase LC analysis. Fractions of the eluent were collected in amber coloured vials in four time windows: 29–33.5 min, poly-TAG window, 33.5–35.5 min, (oxidized-)TAG, 35.7–37.5 min, DAG, and 37.8–40 min, FFA. To obtain enough material, the procedure was repeated twice and the fractions were combined. After evaporation of the eluent under a stream of nitrogen, the fractions were re-dissolved in 100 µL n-hexane containing cholesterol D6 as internal standard prior to normalâphase LC–APPI-MS analysis. 

	The total ion current chromatograms of the size fractions containing the (oxidized-)TAGs and the DAGs are shown in Figure 4. The majority of the polar compounds in both fractions elute between 40 min and 50 min. Note that the non-oxidized TAGs elute almost unretained in normal-phase LC and are not shown in these chromatograms. Hence, the peaks seen in the size fraction of the TAG are oxidized compounds with a size comparable to that of the TAG. The normal-phase LC–MS profiles of the two SEC fractions containing the (oxidized-)TAG or the smaller DAG are very different. Clearly however, without SEC pre-separation, these two compound groups of widely different size yet identical polarity would overlap. Six main peaks can be distinguished in the DAG fraction. The first eluting peak at 42 min has an 

 of 601.5 (43 min), and the third peak has an m/z 599.5 (44 min). These masses correspond to the [M+H-H

O]+ ions of the diacylglycerols OO, OL, and LL/OLn, respectively, where O represents oleic acid, L linoleic acid, and Ln linolenic acid. Positional isomers of these diacylglycerols also elute in the same retention order between 47 min and 49 min. From this it can be concluded that the low-molecular-weight peaks seen in the initial normalâphase LC separation without prior SEC separation were not representing oxidized TAGs, but rather partial hydrolysis products formed upon oxidation. The retention time differences between the different DAGs can be explained by the number of double bonds and the position of the free hydroxide group on the glycerol backbone. The last eluting peaks have a vacant outer position on the glycerol backbone.

The “triglyceride” SEC fraction contains the triglycerides next to the oxidized TAGs. The unreacted TAGs are unretained in normal-phase LC and are not shown in Figure 4. More polar species, such as the epoxy-TAGs and the hydroperoxy-TAGs, elute in the retention time windows of 10–25 min and 40–50 min. The interferences of the unreacted DAG are now removed meaning that the mass spectra are easier to interpret. In Figure 5, MS spectra are shown of the epoxy–oxo lipids, hydroperoxides, 2½ glycerides, and the highly oxidized lipids. The MS spectra of some important classes of lipid oxidation products are discussed in detail below. 

For the hydroperoxides, protonated molecules [M+H]+ are visible, for example at m/z 913.7, at a relatively low abundance. The most prominent peaks are 

 897.7 and 883.7, and are formed by the loss of water giving [M+H-H

O]+, or the loss of a hydroperoxide group, leading to [M+H-H

]+ ions. The DAG fragment ion m/z 603.5 is formed after loss of the fatty acyl chain with the hydroperoxide functionality. Another characteristic DAG fragment ion is visible at 

619.6. During ionization water is lost and an epoxide group is formed across the site of unsaturation converting it to a chain with no unsaturation (OS-epoxide) (14). All TAGs with two adjacent O fatty acids can form this fragment ion. In the rape seed oil used to prepare the mayonnaises analyzed here the primary TAGs showing this behaviour are OOLn, OOO, and POO. The fragment ion at 

 493.4 is formed from breakage of this epoxide bond, resulting in the loss of a C

	The earlier eluting epoxides yield mainly protonated molecular ions, with an example of this being 

 897.7. The mass spectrum is nearly identical to that of the hydroperoxides, except that there is no mass loss of 34 Da visible (loss of H

	The 2½ glycerides oxidation products, the remaining nonvolatile part of the triglyceride after chain-scission and release of the volatile off-smell, are secondary oxidation products formed from breakdown of the hydroperoxides. They are present in the DAG fraction, but not clearly visible in the TIC chromatogram. Their abundance is low in comparison to the signal of the DAGs. The structures with 2½ fatty acid chains are actually oxo-2½ glycerides, with molecular masses corresponding to that of the parent triglyceride (for example, LLO) after the loss of a hexanal molecule (C6H12O mass 100 Da). It is important to note that the oxo-2½ glycerides group is very complex. The original molecules can contain different fatty acids from which different volatiles can be split off. Moreover, the remaining fatty acid chains can be oxidized at one or more additional positions. In the extracted ion chromatograms, the oxoâ2½ glycerides are detectable in the samples. The oxo-2½ glycerides with an additional oxidation group elute later from the column and do not interfere with the analysis.

	As the degree of oxidation increases and TAG with multiple oxygens incorporated in their structure are formed, the mass spectra become even more complex. Possible structures are epoxyhydroperoxides, dihydroperoxides, dihydroxyperoxides, and cyclic peroxides. In the mass spectra multiple losses of 18 and 34 Da are visible, suggesting the loss of water and hydroperoxide groups upon ionization. Full interpretation of these compounds remains difficult however.

The SEC fractionation was applied on a limited number of samples to build understanding of the MS spectra by eliminating overlapping compounds with similar polarity but different sizes. The combined results of the chromatographic and MS spectral evaluation of the different lipid oxidation groups are presented in Table 1. The retention time windows and the m/z values used to generate extracted ion chromatograms to zoom in on specific oxidation products can be seen. Examples of the molecular structures of the oxidized species are also given in the table. Specific positions of the polar oxidation groups on the fatty acid chains can vary depending on the original location of the double bonds. Note that the combination of retention times and spectral information is unique. This means that not all samples have to go through the time-consuming SEC prefractionation. Analyzing a few samples after DAG removal is sufficient to establish the time windows and MS traces of interest. 

To meet the clear trend in society for more natural foods, the food industry is trying to replace non-natural stabilization strategies by natural preservation routes. To this end formulations with differences in the applied stabilization technology were tested in a large-scale storage test. The main goal of this study was to find suitable replacements for EDTA (E385, E386) and understand how the different stabilization routes determine the oxidation reactions and the levels and types of oxidation products. Figure 6 shows the response plots of the nonvolatile species determined by normal-phase LC–APPI-MS. Over 150 samples could be analyzed without maintenance. From this perspective the newly developed APPI-MS method performed much better than our previously described ESI approach (9). Parallel to the LC–MS measurements sensory testing was performed. Hexanal measurements were also performed, but the levels were too low for the classical static headspace method. Clear trends are visible in the plotted data and the levels of all monitored oxidation markers increase with storage time. Moreover, differences between the applied stabilization technologies are visible as well. The formulation with EDTA (positive control) appears to function best in delaying oxidation by showing the lowest increase in NONVOLLOPS over time. When no stabilization technique was applied (negative control), the oxidation products increased the most rapidly. The proprietary stabilization technology 2 performs better in this shelf-life test than technology 1: formation of the secondary oxidation products oxo-2½ glycerides and highly oxidized lipids occurs at a later time-point in the shelf-life test and the levels are lower. Interestingly, hydroperoxides seem to reach a constant plateau level, more or less identical for each of the formulations tested. The possibility that this phenomena was caused by a nonlinear response of the MS detector was ruled out by measurement of several diluted samples yielding similar curves. The occurrence of a plateau suggests that the formation of hydroperoxides at some point comes to a halt, or that the rate of breakdown and formation become equal. The latter explanation here is much more logical. The hydroperoxides are unstable intermediates that can be converted into other volatile and nonvolatile products. This is supported by the steady increase of oxoâ2½ glycerides, degradation products of the hydroperoxides, over time. Attempts to correlate the occurrence of nonvolatile oxidation products in this ambient shelf-life trial with the release of hexanal failed because of an insufficient sensitivity of the classical GC–MS headspace method. 

To enable correlation of NONVOLLOPS levels with hexanal release, an additional (accelerated) shelf-life test was performed at 50 °C with a sampling interval of three days. The effect of the increased temperature on the formation of volatile and nonvolatile oxidation products is shown in Figure 7. GC–MS was applied to measure hexanal (10). The SEC method originally set up to remove interfering species was applied to monitor the formation of polymerized TAG.

At the 50 °C elevated shelf life temperature substantial oxidation occurred. After 40 days of storage, the SEC measurements show poly-triglycerides at up to 10% of the total lipid content. In the GC–MS experiments hexanal was clearly detectable. When looking at the time curves of the hydroperoxides and oxo-2½ glycerides a trend similar to that seen in the ambient study is observed, albeit that the oxo-2½ glycerides are detectable much earlier in the 50 °C storage test. Already after 11 days this marker is detectable and clear differences between the formulations become apparent. This differentiation is not so clear from the results of the hexanal and poly-TAG analyses in the initial stages of the oxidation process. Both hexanal and poly-TAGs show a clear lag phase. Initially they are hardly formed, but after a certain lag phase their levels start to increase at a much faster rate than those of the oxo-2½ glycerides. From this observation we postulate that the oxoâ2½ glycerides may undergo further breakdown and are, like the hydroperoxides, not end products of the oxidation reaction. Whether they are lost by conversion to poly-TAG or oxidize further by the additional formation of hydroperoxide or epoxide groups on the remaining unreacted fatty acid chains cannot be derived from the current data. This concept of continuous oxidation with intermediates and end products is schematically illustrated in Figure 8. In this “oxidation engine”, triglycerides and oxygen are the starting products that form intermediate oxidation products. These products may undergo a continuous process of oxidation and breakdown and eventually end up as volatiles and polymerized lipids in the samples. Clearly the reactions occurring in lipid oxidation are more complex than the “oxidation engine” suggests (3), yet the principle of the engine clearly illustrates the processes occurring. Stabilization technologies control the initial rate of formation of hydroperoxides, as well as the speed of the engine in a way still largely not understood. The new methods presented here shed some light on the oxidation reactions and products forming, and can help in making the search for alternative stabilization technologies to protect lipidârich foods from oxidation less of a trialâandâerror exercise.

	The newly developed normal-phase LC–APPI-MS method allows the detection of nonvolatile lipid oxidation products in the initial and later stages of lipid oxidation. With the detailed information on the NONVOLLOPS, we were able to determine the efficiency of different stabilization technologies in complex food emulsions. This was achieved by successful ionization of NONVOLLOPS using the self-dopant effect of the mobile phase. Detailed interpretation of MS spectra was possible using pre-separation by SEC. The relative levels of important NONVOLLOPS such as hydroperoxides, oxoâ2½ glycerides, and epoxides could be determined in large sample sets from ambient and accelerated stability tests. It is envisaged that the developed method will play an important role in developing new, natural stabilization technologies to suppress lipid oxidation in lipid-rich foods. 

	The authors thank Justin van ‘t Veer for performing the GC–MS hexanal analysis.

		A. Lips, R.A. Chapman, and W.D. McFarlane, 

		B. Barriuso, I. Astiasarán, and D. Ansorena, 

		S. Panseri, S. Soncin, L.M. Chiesa, and P.A. Biondi, 

		L. Steenhorst-Slikkerveer, A. Louter, H.-G. Janssen, and C. BauerâPlank, 

		AOCS Recommended Practice Cg 4-94. Volatiles (VOC) in Fats and Oils by GLC. Official Methods and Recommended Practices of the AOCS, (American Oil Chemists Society, Champaign, Illinois, USA, 1998).	

		D.B. Robb, T.R. Covey, and A.P. Bruins, 

		T. Ghislain, P. Faure, and R. Michels, 

, MSc, completed his master study in analytical sciences at the University of Amsterdam in 2013. Currently, he is a research scientist in the analytical department of Unilever R&D Vlaardingen, the Netherlands. His work focuses on the development of hyphenated chromatographic techniques and highâresolution mass spectrometry methods for food and nutritional analysis.

 is science leader of chromatography and mass spectrometry at Unilever R&D Vlaardingen. He also is a part-time professor in biomacromolecular analysis at the University of Amsterdam, Amsterdam, the Netherlands. Professor Janssen’s research interests include the development of multidimensional separation systems for food, biomedical, and environmental analysis.

Articles

Cover Story

Monitoring lipid oxidation during the shelf life of lipid-containing food emulsions, such as mayonnaise, is challenging. It is, however, essential for the development of improved, consumer-preferred products. Determining the nonvolatile lipid oxidation products (NONVOLLOPS), the precursor compounds for rancidity, is required to determine the effectiveness of product stabilization technologies. A method based on normal-phase liquid chromatography with atmospheric pressure photo ionization-mass spectrometry (LC–APPI-MS) was developed for this purpose. The inclusion of a size-exclusion chromatography (SEC) step was needed to remove interfering diacylglycerides and free fatty acids from the samples. The combined SEC and normal-phase LC–APPI-MS method allowed the identification of a wide range of oxidized species including hydroperoxides, oxo-2½ glycerides, epoxides, and other oxidized species. The method was found to be more suitable for the analysis of large sample sets.

Liquid Chromatography–Atmospheric Pressure Photo Ionization-Mass Spectrometry Analysis of the Nonvolatile Precursors of Rancid Smell in Mayonnaise

Crawford Scientific, Strathaven, Scotland, UK

So, what’s not in your standard operating procedure? Documenting details can prevent headaches associated with method transfers between laboratories.

	About a year ago, I began considering the handoff of the “LC Troubleshooting” column from John Dolan. As I started discussing potential topics to address with people in the liquid chromatography (LC) community, the topic that came up most frequently was method transfer, loosely defined. And based on further discussion recently, it has become clear to me that challenges associated with this topic are only going to become more prominent in the near future. Also, the landscape around this topic is highly varied, and detail and nuance are extremely important. This means that the topic cannot be adequately addressed in a single instalment of “LC Troubleshooting”. Indeed, I expect to see the topic addressed frequently in this magazine in the coming months and years, both by myself and other experts. With this context in mind, I am viewing this month’s column as setting the stage for subsequent discussions-giving some background on how we have come to this point, and discussing the scope and nature of the problem by way of example. For this purpose, I have asked Tony Taylor of Crawford Scientific and LCGC’s ChromAcademy to join me this month. Tony’s deep experience in chromatography training and consultation gives him a rich and unique perspective to draw upon that cuts across both manufacturers (that is, instrument and consumables vendors) and consumers (for example, scientists at pharmaceutical companies) of chromatography. We hope you enjoy the discussion. – Dwight Stoll

What Exactly Do We Mean By Method Transfer?

	In today’s world of method development and life cycle management, the idea of method transfer can mean different things to different people. For example, with all of the recent changes in particle and column technology there has been increasing discussion about strategies for “transferring” a method based on high performance liquid chromatography (HPLC) technology (for example, 5-μm fully porous particles) to an ultrahighâpressure liquid chromatography (UHPLC) platform (for example, subâ2âµm superficially porous particles) (1,2). This concept is sometimes referred to as method translation. There is no question that this is a very important topic, but it is beyond the scope of what we will address here. Instead, the focus of our discussion is on scenarios that involve first the development of a method in one laboratory (typically, called the originator laboratory or the transferring or sending laboratory), and second, the implementation of that method in a physically different laboratory (typically, called the receiving laboratory). In the best case scenario, the receiving laboratory would be able to reproduce results obtained by the originator laboratory to within the uncertainty of the measurement. However, there are many factors that can lead down a path where this ability to reproduce results cannot be demonstrated. In this instalment, we highlight some of these factors and suggest ways to avoid
	the problem.

	We seriously doubt that the generation of scientists that developed HPLC in the 1960s and 1970s ever imagined that a method developed on an instrument in Boston might be transferred to a laboratory in Mumbai with the expectation that the same results would be obtained in both locations-at least not in the early days. In those days, most methods were executed for years within the same company that developed the method, if not the same building. However, since the early 2000s it has become increasingly common for a method developed in one location to be executed in a different location for the remainder of the lifetime of the method. A significant driver for this has been the trend towards outsourcing of manufacturing and even method development in certain industries and application spaces (3). In some cases, the fact that the location is different is the least of concerns-the originator and receiving laboratories may be in different companies, running instrumentation from different vendors, and using reagents from different suppliers. It is not hard to imagine how this situation can quickly lead to problems. There is not too much literature on the topic of method transfer before about 2010. However, in just the past year or so several substantive articles on the topic have appeared in chromatography, pharmaceutical, and life science journals, which reflects the increasing importance of this topic to the community (4–8).

	One needs to carefully consider the “stage of development” at which transfer occurs. These may include preclinical methods that are transferred to a different laboratory for analysis of clinical samples, methods initially developed with a focus on an active pharmaceutical ingredient (API) that subsequently need to be implemented in stability studies, and finally methods that have been developed and validated in the originator laboratory but then need to be implemented in the receiving laboratory for routine use. Of course, some method adjustments may be required in these situations. However, while transfer of methods that have been fully validated and in routine use should be straightforward, these can be the cause of as much consternation as those in which some adjustments are expected from the start.

	Further, the regulatory requirements of the receiving laboratory must be borne in mind if regulatory submission or the use of methods for product quality control (QC) is intended. Where regulatory requirements differ, this difference should be a principal concern during method transfer.

	When transferring methods, it is typical to develop a method transfer protocol, which includes details of the method (including sample preparation and data handling), the test articles that will be used including the number required to give statistically meaningful results, and any special requirements (forced degradation lots, placebo lots formulated with known impurities, and so forth), any special sample handling requirements, the number and type of tests being undertaken, and, critically, the acceptance criteria for a successful transfer. 

	The transfer protocol is often primarily focused on the acceptance criteria to use when considering the resulting data from both laboratories, and there are many different ways in which the accuracy, precision, sensitivity, ruggedness, and robustness of the methods may be assessed, and the statistical analysis of the data that will highlight any interlaboratory variability in the results. Most regulatory guidance will refer to “acceptable” performance evaluation, but United States Pharmacopeia (USP) Chapter <1224> “Transfer of Analytical Procedures” (9) does give more helpful guidance. Helpful testing regimes, statistical data analysis approaches, and acceptance criteria have been proposed by several authors. A few that are particularly noteworthy are the procedures and equivalence tests recommended by the International Society for Pharmaceutical Engineering (ISPE) (10), a further treatment of these procedures to reduce the required effort and further develop acceptance criteria by Kaminski, Schepers, and colleagues (11), and the use of ratio-based testing using methods such as Bland-Altman plots with limits of agreement (12).

	All of the above being said, the devil is always in the details, and it is invariably true that increasing the level of detail supplied to the receiving laboratory will always improve the chances of a successful method transfer (7,9). This additional information includes the detail supplied regarding the actual analytical method, where the standard operating procedure (SOP) or standard method of test (SMT) is usually not enough to capture the “vagaries” of the originator laboratory. It is far better, where possible, to include details on the method development (even in a short summary document), any validation protocols and reports as well as the control documents for methods in routine use that will identify trends and patterns in calibration samples, QC samples, and the results from “live” test articles. 

	In discussing method transfer with scientists on both the originator and receiving ends of a method, we have heard very different perspectives. From the receiving end we frequently hear scientists exclaim, “We did exactly what they told us to, and the method didn’t work!” And, from the originator we hear, “They didn’t do what we told them to do!” Upon further discussion it becomes clear that both parties are partially right, and the two views can be reconciled by realizing that not enough detail was specified in the method transfer protocol. This is akin to the oft-cited idea in computer science that computers only do what we tell them to do:

“The Analytical Engine has no pretentions whatever to originate anything. It can do whatever we know how to order it to perform.”

 - Ada Lovelace, Notes on the Sketch of the Analytical Engine, 1842 (13)

	So, it is also with an analytical method-if we do not specify all of the details necessary to reproduce the results of the analysis, we should not expect them to be reproduced faithfully. However, specifying all of the details is often not a trivial task. For example, details relevant to the preparation of buffered eluents may be such a well-established part of the culture of the originator laboratory that they are not included in the method. The originator laboratory may prepare a 0.1% solution of trifluoroacetic acid in water on a w/w basis, while the receiving lab may prepare the solution on a v/v basis-because that is the way they usually do it. In many cases this difference may not cause a problem, but we have certainly seen situations where it can cause a problem with the method transfer. So, this brings us to the following question: What’s not in your SOP (that should be)?

	The following list is a set of examples of such details that have caused problems in our own work, or have come up in discussion with scientists from different companies. It most certainly is not an exhaustive list of things to focus on. Rather we hope it motivates readers to include more detail in methods that will be transferred-again, with the aim to increase the likelihood of a successful method transfer. 

 As chromatographers, we would prefer to focus our efforts on optimizing the chromatography itself, rather than having to worry about such details as the make and model of the sample container. However, these details can be very important indeed (3). Different types and grades of septa have different extractable profiles and can affect injection carryover, because some injector designs rely on the vial septum to wipe down the outer surface of the injector needle. The vial material itself (for example, glass versus plastic) and how the material is treated at the time of manufacture (for example, washing or deactivation of glass) are also important. These details can affect adsorption of certain analytes, particularly at low concentration, and contamination at the time of manufacture can lead to the appearance of “ghost peaks” (14–16) in blank samples later on. Thus, where possible the same make, model, and grade of vial and septum should be used in both the originator and receiving laboratories.

 There are many detection variables that can affect the success of method transfer. One that has become more important in recent years is the data acquisition rate. As the resolving power and speed of LC have improved with advances in instrument and column technology, peaks have decreased in width on average, both in terms of time and volume. Methods that produce peaks that are very narrow in time require the detector signal to be recorded more frequently compared to past experience. Acquisition that is too slow can compromise resolution by making peaks appear broader than they are upon elution from the column. On the other hand, using the highest acquisition rate available all the time can lead to signals that are too noisy. Most modern ultraviolet (UV) absorbance detectors support these acquisition rates very nicely, and there is good guidance in the literature on what acquisition rate should be used for a particular peak width (17). Nevertheless, we need to be diligent about making sure this setting is specified in the method transfer protocol. Also important are factors that affect detection sensitivity and signal-to-noise ratio. In the case of UV absorbance detection, these include reference wavelength, detection bandwidth, and detection pathlength. Again, these parameters should all be clearly specified to increase the likelihood of a successful method transfer.

 One of the aspects of chromatography that makes it so powerful as an analytical tool is that very subtle changes in analyte or stationary-phase chemistry can result in selectivity differences big enough to resolve two peaks easily. However, this aspect has a dark side too. Subtle changes in column chemistry as the column ages over its lifetime can lead to changes in selectivity that can turn good resolution into bad resolution. In other words, this type of change can cause a method to drift out of specification or fail system suitability tests. Although it would be convenient if a column worked as it should with the very first injections out of the box, this is often not the case. We have first-hand experience with this and have heard several anecdotes about situations where a well-defined column conditioning step was required to “break-in” a column before using it for the intended analysis. This step may simply involve exposure to the mobile phase for a specified number of hours (18) (or even days in our experience!), or perhaps several injections of the target analytical samples before using the column to collect data on samples of interest. Sometimes this process results in in situ modification of the stationary phase by constituents of injected samples; indeed, this is why John Dolan has recommended for years that columns be dedicated to specific projects (19). On average, these problems are less serious today than they were in the early days of HPLC because column technologies have improved dramatically, but they still deserve attention.

Variation in Tubing Length and Diameter: Most modern LC instruments provide users with opportunities to change the plumbing of the system, particularly the connections that are made between the pump, injector, column, and detector. This capability, too, can be both a blessing and a curse. For example, in recent years there has been a lot of discussion about reducing the diameter of connecting tubing to minimize peak dispersion outside of the column (17). Minimizing extracolumn peak dispersion is very important from a conceptual point of view, but we have to be careful to make sure we do not end up operating too close to performance margins, and allow some room for variation in the tubing itself. The pressure drop across a piece of tubing under laminar flow conditions scales with the fourth power of the tube radius, so even very small variations that might arise from different manufacturing batches of tubing, or between different manufacturers, can lead to a significant change in system pressure, and could even exceed the pressure limit of the pumping system (20).

	As the chromatography community acquires more and more experience with method transfer, best practices will emerge and the frequency of good experiences with method transfer should increase as a result. Indeed, these best practices are beginning to emerge as experts publish guidance that can be applied broadly across different industries and companies (3–5,8). The following short list describes best practices that will support successful transfer of chromatographic methods.

 When in doubt, include more details in a method transfer protocol, even if they seem irrelevant at the time. Even though “one-to-one” implementation of the method in the receiving laboratory using equipment and materials identical to those in the originator laboratory is not always possible for a variety of reasons, including more detail in the transfer protocol at least makes it more likely from the outset. Examples include specifying the source of vials, pipette tips, centrifuge tubes, and wellplates. Also, including the many details associated with complex operations-for example, ionization conditions for electrospray mass spectrometry-could save a tremendous amount of time and effort in the transfer process.

	Use Available Data and Tools to Prevent Problems Before They Occur: As more data and software tools relevant to the method transfer process become available, there will be more opportunity to anticipate problems before they occur. For example, the recent publication by Ahmad, Blasko, and colleagues (6) describes extensive comparative data on five state-of-the-art UHPLC systems, ranging from gradient delay volume of the pumps to the time required for thermal equilibration of samples in thermostated autosamplers. Especially when it is known that the method will be transferred to a receiving laboratory with different HPLC equipment, these comparative data may be useful in anticipating differences in the performance of the method in the originator and receiving laboratories. Also, instrument control software is becoming more sophisticated in ways that support method transfer. For example, some systems now have the ability to mimic the behaviour of systems from other manufacturers. In principle this approach can simplify method transfer in cases where the originator and receiving laboratories are using systems from different manufacturers.

 The value of communication in the method transfer process cannot be overstated. Hilhorst and coworkers, writing about the transfer process from the perspective of a contract research organization (CRO) (3), emphasized the importance of mutual respect for the work of the originator and receiving laboratories. They also suggested that training receiving laboratory scientists in the originator laboratory is very helpful in the long run, even if it is expensive on the front end of the process. If travel to the receiving laboratory is not possible, then webâbased communication tools are very helpful in the process of identifying problems during the facile processing of samples and instrument setup. We can perhaps imagine a day in the not too distant future where tools such as the GoPro camera or Google Glass can be used to truly “follow” workers at a different site undertaking an analysis to spot “obvious” issues that are otherwise difficult and time consuming to solve using only email and phone conversations.

 One of the most frequently cited causes of problems in HPLC method transfers is differences in volumes of different parts of the instrument in the originator and receiving laboratories. As HPLC equipment has evolved over the past three decades, gradient delay and extracolumn volumes have changed significantly, and differences in these volumes can have significant effects on method performance. It is fair to say that these effects are well understood at this point, with detailed guidance available to help understand the challenges and potential solutions (8,21). Now, the challenge is to leverage this knowledge in practice, and to anticipate potential problems before they occur.

	Minimize Variation Where Possible: The circumstances of some method transfers will prevent one-to-one transfer for reasons beyond the control of the scientists involved (for example, HPLC equipment from different vendors in the two laboratories). For other aspects there is more control, however, and for these, every effort should be made to minimize variation between laboratories. For example, one laboratory may routinely filter their mobile phases, whereas the other laboratory may not (see the February 2017 “LC Troubleshooting” for details on this point in particular [22]). This case is an example of a detail that is highly dependent on laboratory culture and habits, and may not normally be specified in the method. Also, the quality of reagents used in a method can significantly impact results. For example, the use of lowâgrade mobileâphase additives can lead to unexplained “ghost peaks,” particularly for methods involving gradient elution. If these ghost peaks show up on analyses of blank samples, they will have to be investigated; where possible it is best to simply avoid these problems by using reagents of the same quality in both laboratories. Again, as discussed, include as many details as possible, even if they do not seem important at the time (23). 

	The topic of method transfer encompasses a vast space that includes many concepts in laboratory practice and skill, chromatography, and instrumentation, and is affected by a variety of organizations and people including instrument, consumable, and reagent vendors, chromatography training organizations, originator laboratories, CROs, and contract manufacturing organizations (CMOs). If the last two years are any indication, we will see this topic being addressed in the analytical and chromatography literature more frequently in the near future. This literature promises to provide both guidance (for example, see reference 4) and data (for example, see reference 6) that users can draw on to increase the likelihood of a successful method transfer from, or to, their laboratories. Out of curiosity we searched the technical programmes from the 2016 meetings of AAPS and Pittcon and found very little evidence of discussion of the method transfer topic in an explicit and organized way. Perhaps going forwards, it would be helpful to organize one or more technical sessions at Pittcon and other conferences that would bring together parties affected by method transfer to discuss challenges and develop a shared sense for how to move forwards as a community. In the coming months, we and others will be tackling more specific aspects of method transfer in LCGC. 

	We thank Todd Maloney, Kelly Zhang, Peter Johnson, Lianjia Ma, and Bill Long for their helpful discussions leading to the preparation of this instalment.

		General Chapter <621> “Chromatography” in United States Pharmacopeia 40 National Formulary 35 (USP 40–NF 35, United States Pharmacopeial Convention, Rockville, Maryland, USA, 2017), pp. 508–520.	

		P. Petersson, M.R. Euerby, and M. James, 

		M.J. Hilhorst, B.W. van der Strate, C.J. Briscoe, and N.C. van de Merbel, 

, 1131–1134 (2017). doi:10.4155/bio-2017-0078.	

, 595–600 (2017). doi:10.4155/bio-2017-0015.	

 (2017). doi:10.1016/j.jpba.2017.06.020.	

		I.A.H. Ahmad, F. Hrovat, A. Soliven, A. Clarke, P. Boswell, T. Tarara, and A. Blasko, 

, 1143–1159 (2017). doi:10.1007/s10337-017-3337-8.	

, 1409–1413 (2016). doi:10.4155/bio.16.34.	

		General Chapter <1224> “Transfer of Analytical Procedures” in United States Pharmacopoeia 40 National Formulary 35 (USP 40-NF 35, United States Pharmacopoeial Convention, Rockville, Maryland, USA, 2017). https://hmc.usp.org/sites/default/files/documents/HMC/GCs-Pdfs/c1224.pdf (accessed 24 August 2017).	

, 310–314 (2005). doi:10.1016/j.jpba.2005.03.015.	

		L. Kaminski, U. Schepers, and H. Wätzig, 

, 1124–1129 (2010). doi:10.1016/j.jpba.2010.04.034.	

		C. Frömke, L.A. Hothorn, F. Sczesny, J. Onken, and M. Schneider, 

, 186–193 (2013). doi:10.1016/j.jpba.2012.10.032.	

		A. Lovelace, Notes on the Sketch of the Analytical Engine (1842). http://www.fourmilab.ch/babbage/sketch.html (accessed 1 September 2017).	

, 1–11 (2004). doi:10.1016/j.chroma.2004.07.110.	

		M.R. Taylor, F. Harvey-Doyle, and K. Patel,

		T. Taylor, “The LCGC Blog” (2017). http://www.chromatographyonline.com/lcgc-blog-ghost-peaks-gradient-hplc?pageID=1 (accessed 1 September 2017).	

		J. De Vos, K. Broeckhoven, and S. Eeltink, 

, 262–278 (2016). doi:10.1021/acs.analchem.5b04381.	

, 104–110 (2008). doi:10.1016/j.chroma.2008.08.122.	

		J. Halvorson, A.M. Lenhoff, M. Dittmann, and D.R. Stoll, 

 (2016). doi:10.1016/j.chroma.2016.12.084.	

, 110–119 (2005). doi:10.1016/j.chroma.2005.04.088.	

 is the technical director of Crawford Scientific and ChromAcademy. He comes from a pharmaceutical background and has many years of research and development experience in small molecule analysis and bioanalysis using LC, GC, and hyphenated MS techniques.

 is the editor of “LC Troubleshooting”. Stoll is an associate professor and co-chair of chemistry at Gustavus Adolphus College in St. Peter, Minnesota, USA. His primary research focus is on the development of 2D-LC for both targeted and untargeted analyses. He is also a member of LCGC’s editorial advisory board. Direct correspondence to: 

LC Troubleshooting

Liquid Chromatography in the Global Economy—Preparing for Successful Method Transfer

Fast gas chromatography (GC) has received new attention recently in the form of available enhanced instrument capabilities. This instalment reviews important characteristics and requirements of fast GC: What can fast GC do for separations, and how can laboratories take advantage of enhanced separation speeds?

	Fast gas chromatography (GC) has received renewed attention in the past two years. New GC instruments and accessories in 2015–2017 include more capabilities for fast GC than their predecessors, such as rapid oven heating and cooling, extended inlet pressures, support for hydrogen carrier gas, and faster detector response times. Beyond these basics, workflow features such as ease of changing columns, energy efficiency, reduced maintenance, and improved user interfaces have also garnered significant attention. As older instruments age out and replacements are installed, laboratories may access new abilities to obtain faster separations and sample throughput. On the surface, these changes to the GC laboratory landscape resemble the long-coming transition from packed to capillary columns of the 1980s and 1990s. However, this time users may have the choice to run their separations faster with their new instrumentation, or to keep going as before and leave their capillary column methods unchanged. No doubt most will opt for the second scenario at first, but, as with a new automobile, sooner or later operators will explore how fast the new equipment can perform.

	As applied to GC, “fast” and “highâspeed” are not well defined. These terms are relative and are meaningful only when compared to an existing state. “Optimized” is a related comparator that implies a limited change. For the purposes of this article, optimized separations would be achieved without modification of the column dimensions or instrumentation, whereas fast or high-speed separations imply selection of specific column dimensions and instrument options that can produce significantly faster separations. So for example, if an existing separation’s last peak of interest is eluted in 25 min, it might be optimized by increasing the carrier flow and temperature ramp rate, thereby causing the run time to decrease to 15 min. The same separation might be upgraded to high speed by reducing the column inner diameter, changing to hydrogen carrier gas, and ramping the column oven at higher speeds. Achieving this higher speed might require a high-power oven heating arrangement or an add-on high ramp-rate module, plus faster detector response times and injection speeds to match.

	Pushing a separation to higher speeds has both positive and negative impacts. Reducing the total run-cycle time-the interval from one sequential injection to the next-directly increases sample throughput, which can be a significant advantage in laboratories with high sample loads. But changing the separation method to run at higher speeds can be a daunting and time consuming effort. Careful method optimization followed by required validation is still necessary. Having newer instrumentation already in the laboratory that is capable of high speed removes some of the expense, and this barrier will become lower over time. Deciding whether to optimize an existing separation or to proceed further into high-speed operation is fast becoming more of a business decision than a choice to implement a new and untried technology in the lab.

	After the decision is made to migrate a separation to higher speeds, the process of getting there is fairly well defined, although there are some pitfalls on the way. The first step is to pick a goal for how much faster the separation could be to realize significant gains. For an example, we will use the synthetic chromatogram of pesticides in Figure 1. Chromatogram modelling software makes it easy to test out scenarios for optimizing and speeding up complex mixture separations; in this case I used the Pro EZGC Chromatogram Modeler (Restek Corporation) (1).

	GC simulation software comes in two flavours. First there is flow calculator software that uses the same calculations as builtâin electronic pressure control to compute column flows, pressures, and velocities. Several programs are available, and most include a simple method translation mode that attempts to predict or maintain peak elution order as changes to columns and conditions are made. The second type of simulation software includes the flow calculations but goes beyond that by accessing a database of compound retention behaviour on specific stationary phases. Typically, the data are obtained by measuring two or three separations under different thermal conditions for each compound on each stationary phase in the database. These thermodynamic data are then used to estimate each peak’s retention behaviour under a wider range of conditions and column dimensions. Consideration of the influence of flow and column dimensions on peak widths is mixed in with retention information to produce an estimate of peak resolution for each peak pair.

As we will see, going from just optimizing the separation to highâspeed operation using a narrower-bore column may be quite beneficial. The chosen test compounds listed in Table 1 correspond to a pesticide sample that I separated in the 1990s as part of a temperature program optimization exercise, work that was subsequently described in an early “GC Troubleshooting” instalment (2). For those readers who are familiar with this type of sample, bear in mind that the conditions we will arrive at won’t necessarily resolve the entire set of analytes. This is only an illustration: As always, carefully check and validate methods as they are optimized or converted to the high-speed realm.

	Initially, the optimization exercise focused on the last 10 peaks of the original sample. Then, for conversion to a column with a smaller internal diameter, I included all 16 of the compounds from my original experiments, changed to a smallerâinner-diameter column with hydrogen carrier gas, and ran the simulation program’s temperature refinement option. See the end of the instalment for the results. 

The starting and optimized GC conditions are given in Table 2, and the corresponding chromatograms are shown in Figure 1. The original chromatogram was obtained at the slowest simulated ramp rate of 4 °C/min. The 10 selected peaks are well resolved in a total run time of 62 min. Increasing the ramp rate stepwise from 4 °C/min to 40 °C/min causes the retention time of the last peak to decrease from 52.2 min to 8.7 min, as illustrated in Figure 2. At first glance this is a very significant speed-up, but it is not all that it seems to be.

First of all, peaks 5 and 6 merge together as the ramp rate increases through 14 °C/min. At higher ramp rates they are separated again but their elution order reverses. This well known effect stems from chemical differences between the peaks that produce differential retention effects as temperatures shift. It is often observed on moderately polar columns such as the 35% phenylmethylsilicone phase used here. Peaks 3 and 8 exhibit similar behaviour as they migrate between the surrounding peaks with increasing ramp rates.

	The resolution of peaks 2 and 3 reaches a maximum at 12 °C/min and then decreases as the ramp rate heads out towards 40 °C/min. Figure 3 shows the predicted resolution of two peak pairs across the entire range of programming rates. An important goal of an optimization exercise is to maintain a minimum resolution of 1.5, which is indicated by the horizontal red line in Figure 3. Imposing this limitation, ramp rates between 12 °C/min and 20 °C/min have to be excluded.

A second problem arises when the ramp rate reaches 25 °C/min or higher: All of the peaks are eluted within a few degrees of the final oven temperature. Illustrated in Figure 1(c), this situation indicates that the oven ramp rate is too fast for the carrierâgas flow rate. It has gotten ahead of the peaks’ progress through the column, and the peaks no longer interact as much as they could with the stationary phase. The highest oven temperature is determined by the column’s upper programming limit, which in this case is 290 °C. Effectively, the optimal programming rate for this separation is 10 °C/min, as shown in Figure 1(b). Elution at the final oven temperature plus loss of resolution make higher ramp rates less desirable. Thus, we can realize about a factor of two increase in separation speed for this sample without changing the column or modifying the carrier-gas flow.

	A different set of conditions or different column dimensions should be considered when most of the peaks are bunched at the maximum temperature. In this case, increasing the carrier-gas flow or velocity would move the selected peaks to lower retention times. At the same time, the programming rates at which resolution for all peaks exceeds 1.5 will shift. The ramp-rate optimization exercise can be repeated at higher flows to find potential new values and perhaps get to even higher speeds, possibly further improving the separation of this 10-component subset of the original pesticide sample.

	I decided instead to jump to a higher-speed column to see just how fast the separation might be performed. The result is shown in Figure 4, where a 10 m × 0.18 mm column with a 0.2-µm film of the same stationary phase accomplishes the separation of the 10 selected compounds in under 2 min! The ramp rate was increased to 70 °C/min and the initial carrier velocity to 80 cm/s. With a GC system capable of running this faster method and column, a speed increase of about 25 times might be achieved. Significantly, the initial temperature was increased to 150 °C, which also saved some time. This last change, however, is not compatible with separating the entire 16-component mixture because the first six peaks would be crammed together at the start and poorly resolved, if at all.

In addition, this type of sample usually occurs at trace levels, and splitless or on-column injection may be required. That in turn needs a lower initial oven temperature to realize a solvent effect and solute trapping of the earlier peaks, so the higher initial temperature is not very realistic. Still, the achievable speed is impressive.

For a final try, I included all 16 compounds and optimized for speed with the same narrow-bore column except for a thicker film of 0.4 µm, and a lower initial temperature, a necessary change to resolve all 16 peaks. Figure 5 illustrates the result. The run time needed is about 6.25 min to resolve all of the peaks. A two-ramp temperature program was used to split the separation into a first section for the additional six peaks, and a second section at a higher ramp rate to resolve the 10 peaks of interest from before.

	At this point, going any further would require a trip to the laboratory to measure how real separations stack up to the predictions of the GC simulations. I can say that the simulated results given here match fairly well with the measured chromatograms from 1991, although those were performed with different column inner diameter, length, and film thickness. At the time, I found that the critical peak pair 5 and 6 were coeluted at between 8 and 10 °C/min, and peaks 2 and 3 also showed behaviour similar to the simulation. Given the observed peak reversals and the range of conditions in which peak resolution can drop below 1.5, no conclusions about optimization or conversion to highâspeed separation can be taken away from the kind of exercises presented here. Such a journey through simulation is very informative, but there is no substitute for injecting and recording real chromatograms.

		Restek Corporation (State College, PA, USA), Pro EZGC Chromatogram Modeler, http://www.restek.com/proezgc, October, 2017.	

 is a Senior Scientist at Serveron Corporation in Beaverton, Oregon, USA, and a member of LCGC Europe’s editorial advisory board. Direct correspondence about this column to the author via e-mail: 

GC Connections

Fast Gas Chromatography

Issue PDF

Click the title above to open the LCGC Asia Pacific November 2017 regular issue, Vol 20, No 4, in an interactive PDF format. 

LCGC Asia Pacific-11-01-2017

Liquid Chromatography–Atmospheric Pressure Photo Ionization-Mass Spectrometry Analysis of the Nonvolatile Precursors of Rancid Smell in Mayonnaise

Liquid Chromatography in the Global Economy—Preparing for Successful Method Transfer

Fast Gas Chromatography

Vol 20 No 4 LCGC Asia Pacific November 2017 Regular Issue PDF