Isocratic high performance liquid chromatography (HPLC) for reversed-phase separations, while both convenient and robust, has inherent issues, especially when dealing with compounds whose polarity, hydrophobicity or extent of ionization vary widely. Common problems include: 									

increased peak width (potentially resulting in poor resolution and sensitivity) for later-eluted analytes resulting from peak broadening caused by diffusion of the analyte band,

the inability to cope effectively with analytes having a broad range of hydrophobicity (log 

), often resulting in unnecessarily long analyses, and

contamination of columns by strongly retained components.

Gradient HPLC can help overcome some of these issues and provide improved peak shape, higher efficiency and sometimes altered selectivity and improved resolution, especially for latereluted peaks.  

Gradients in reversed-phase HPLC typically involve the on-line (dynamic) mixing of solvents to achieve a steady increase in the organic solvent (typically methanol or acetonitrile) over the course of the analysis, thus increasing the elution strength of the eluent over time. Buffer strength or pH gradients, although more unusual, may also be used, especially with stationary phases capable of undergoing electrostatic interaction with the analyte. 

Three parameters are required to specify an eluotropic strength gradient in reversed-phase HPLC: initial percentage of solvent B (%B), where solvent B is usually the organic or "strong" solvent, final %B and gradient time (

) over which the transition in eluotropic strength will be achieved. In practice, there are several other very important parameters to consider: 

 A period within the gradient in which the eluent composition is held at the initial %B. This achieves a degree of analyte focusing but also crucially enables easy transfer of gradients between different instruments based on the specific instrument gradient dwell volume (

: Usually achieved using a short ballistic gradient ramp to high %B to elute highly retained components (of no analytical interest) from the column. There may be an isocratic hold at this composition to ensure elution of all analytes and strongly absorbed components of no analytical interest. 

: Returning the column to the initial gradient composition. In practice, with modern instruments, this step is programmed to occur very rapidly. 

: The time taken to ensure the whole of the analytical column is returned to the initial gradient composition. This is an important step and if not properly considered can lead to retention time and quantitative variability. Most manufacturers will recommend between 5 and 10 column void volumes for proper equilibration. A very rough approximation for the column void volume in millilitres is 

 the column diameter, both in millimetres. 

 To perform gradient HPLC analysis and to transfer methods between instruments successfully, it is important to know 

. The best gradient methods have an isocratic portion at the beginning of the profile that is at least as long as the gradient dwell volume. Refer to your manufacturer for this value or an appropriate method for empirical determination. 

In gradient HPLC, retention behaviour is somewhat different from that in isocratic HPLC. At the beginning of the analysis, when the mobile-phase strength is low, the analyte is partitioned wholly into the stationary phase (or "focused") at the head of the column and will not move through the column at all. 

As the mobile-phase strength increases, the analyte will begin to partition into the mobile phase and move along the column. As the mobile-phase strength increases continuously, the rate at which the analyte moves along the column subsequently increases and the analyte "accelerates'" through the column.  

At some point within the column, the analyte may be wholly partitioned into the mobile phase, and will be moving with the same linear velocity as the mobile phase. The point at which this occurs depends upon the nature of the analyte and its interaction with the stationary phase material.  

As the rate at which the analyte elution changes during gradient HPLC, the retention factor (

) used for isocratic separations is not applicable. Instead we use an "average" retention value, 

*, or the retention factor of the analyte as it passes the midpoint of the column. Most analytes will be moving at the same rate as they pass the midpoint of the column and hence the retention factors for all analytes in gradient HPLC are very similar.  

In gradient elution, the peaks are narrow with almost equal peak widths. All compounds accelerate through the column and thus are eluted at a high velocity. The retention time difference between compounds is largely a consequence of the percent organic modifier at which each starts to accelerate. All compounds should have approximately the same speed when they leave the column. Further, the front and tail of an analyte band reside in slightly different concentrations of organic modifier — the velocity of the tail is slightly higher than the heart of the band and vice versa for the front, which results in peak focusing. The narrow peaks obtained in gradient elution generally provide better detection limits and higher loading capacities. 

Articles

The Essentials

An excerpt from LCGC's e-learning tutorial on HPLC for reversed-phase separations at Chromacademy.com.

Gradient HPLC for Reversed-Phase Separations

Choosing conditions for a final separation ("method development") is often carried out by trial-and-error. With some experience and basic high performance liquid chromatography (HPLC) knowledge, this process can evolve into enlightened trial-and-error — in which fewer trials are required. Efficient method development aims at optimal results with minimal effort; it requires a quantitative understanding of how different experimental conditions affect the quality of a separation, and is best implemented with computer simulation. My own career, which emphasized improvements in method development, provides a background for the following short review. 

I was born in Sacramento, California, where in 1939 I received a chemistry set for Christmas. What proved fascinating was that stuff called compounds could be predictably assembled from something called atoms — nicely illustrated in the form of jigsaw cutouts (Figure 1).  

 Figure 1: An example of the jigsaw cutouts illustrating the assembly of atoms. 

The concepts of valence and balanced equations were thus revealed, if in primitive form. This initial interest in chemistry grew over time with visits to the local library and later experiments (many dangerous!) in a home laboratory. When I entered the University of California at Berkeley in 1948, I was firmly committed to a career in chemistry.  

An important refinement of this goal came during graduate research at Berkeley under Professor Donald Noyce (1952–1954). My research topic was the mechanism of an aldol condensation (1), which introduced me to physical-organic chemistry and model building. This provided an important tool for the next six decades ... the only thing missing was a research area in which to apply this tool. That would soon be supplied. 

My first job was at Shell Oil in Houston, under the direction of M.J. ("Jack") O'Neal. Jack received the ACS Award in Petroleum Chemistry in 1956 for his pioneering development of mass spectrometry (MS) for the analysis of high-molecular-weight petroleum samples. Al Zlatkis, a future "impresario of chromatography," was part of the group when I arrived. A little later we were given the job of constructing a gas chromatography (GC) system using information from the proofs for a book by Keulemans (my first exposure to chromatography).  

We began measuring retention times for various hydrocarbons, which led me to a connection between chromatography and model building in the form of the Martin equation: 

 is the retention factor for a substituted molecule (for example, 

 for the unsubstituted molecule (for example, benzene), and Δ

 is a constant for a given substitutent group (for example, methyl or ethyl). Later, we were asked to develop an assay of gasoline, which resulted in the development of "two-stage GC" (2) — an early form of column switching. By now, the power of chromatography, its connection to physical-organic chemistry, and the practical importance of method development were apparent. 

Adsorption Chromatography and Petroleum Analysis (1957–1971)

In 1957, I moved to the Union Oil Co. of California, where my starting assignment was the analysis of high-boiling petroleum samples such as lubricating oil. Pioneering work of this kind had been carried out in O'Neil's group, where the sample was initially separated by adsorption chromatography into saturated hydrocarbons, aromatic hydrocarbons, and nonhydrocarbons. This was followed by the MS analysis of the hydrocarbon fractions. My experience at Shell suggested that previous such separations were empirically derived and poorly understood — so they likely could be improved. George Lake, my new supervisor at Union Oil, agreed and gave me a month to study the problem. The month grew into a separate 14-year research program and a 1968 book (Principles of Adsorption Chromatography [3]).

 A simple equation summarized many of our findings, 

 is a solute retention factor, Ψ is the phase ratio, 

 is the relative area of the solute molecule, and ε measures mobile-phase polarity or "solvent strength." 

 is the adsorption energy of the solute, values of which can be estimated as a function of solute molecular structure via equation 1 — in turn enabling predictions of 

 for different compounds and separation conditions. 

My 1968 book would not be needed in today's HPLC laboratory, as information of present interest has since migrated into more recent books and publications. But it did describe a previously unappreciated phenomenon: 

. Localization, the noncovalent attachment of polar molecules to specific adsorbent sites, was further studied over the next four decades (4) and found to largely determine solvent selectivity for adsorption chromatography (5) — an important contribution to method development for separations using silica or other adsorbents. 

Petroleum consists mainly of hydrocarbons (alkanes, cycloalkanes, alkylbenzenes, and so on) plus lesser concentrations of related sulphur compounds (alkyl sulphides and thiophenes), and still smaller amounts of nitrogen- and oxygen-containing ("hetero") compounds. Our work on adsorption chromatography quickly led to improved compound-type separations of hydrocarbons plus sulphur compounds, for their subsequent MS analysis, as at Shell-Houston. This was accompanied by the development of complementary procedures for assaying certain other compound types in petroleum by so-called "linear elution adsorption chromatography" (6). 

By 1966 we had completed work on the total analysis of hydrocarbons and sulphur compounds in any petroleum distillate, but our knowledge of the hetero compounds in such samples was incomplete. These oxygen- and nitrogen-containing molecules include several different compound types (such as, substituted pyridines, indoles, and furans), many of which were yet to be identified. Consequently, a complete analysis of these petroleum hetero compounds represented a major challenge. Our planned approach was the separation of different hetero-compound types from each other and from interfering hydrocarbons and sulphur compounds, followed by the spectral analysis of individual hetero fractions.  

A multistep separation (7) made sequential use of three different adsorbents (alumina, silica, and charcoal) plus cation and anion exchange to provide about 50 hetero fractions. We could predict which compounds might be present in each fraction (equation 2), and this information with corresponding MS, UV, and infrared (IR) data resulted in the first comprehensive analysis of the N- and O-containing compounds in a petroleum distillate (8,9).   

So, here we have an approach to pre-HPLC method development that relied on a quantitative knowledge of separation as a function of conditions for different possible compounds present in petroleum. This anticipated "efficient" method development, but is not applicable to HPLC, which deals with the more challenging separation of individual 

The arrival of HPLC in the mid-1960s has been described (for example, see reference 10). In 1966, I began packing columns with alumina for use with a crude, homemade HPLC system (11). An example of its subsequent use in our laboratory is shown in Figure 2 for the analysis of hydrogenated quinoline samples. An initial gradient separation [Figure 2(a)] revealed major peaks 1–5 that were eluted between 10 and 25 min. This suggested the routine isocratic separation of Figure 2(b), where an additional peak (peak 3a) is seen. 

Figure 2: Analysis of hydrogenated quinoline by adsorption chromatography with alumina. (a) gradient elution (0â100% methylene chloride/n-pentane in 50 min); (b) isocratic separation (35% methylene chloride/n-pentane). Adapted from reference 12.  

By 1970, the potential of HPLC had become obvious, but clearly its application within the petroleum industry would be limited. This realization — combined with the completion of our original petroleu-manalysis project — led me in 1971 to join the Technicon Corporation (Tarrytown, New York). At that time Technicon was the largest manufacturer of automated-analysis equipment for use in clinical laboratories. 

It's worth recalling what HPLC looked like in 1971, as summarized in an early HPLC book (13). Commercial equipment and columns were available, and some of what had been learned about GC was being applied to LC. For example, the quality of a separation could be described by the resolution 

 of the two, poorest separated peaks in a chromatogram; 

 can be related (indirectly) to experimental conditions via retention factors 

, separation factors α (selectivity), and the column plate number 

In the beginning, HPLC method development emphasized the improvement of resolution by predictable changes in conditions that affect values of 

 this might involve varying column length or flow rate (for example, see reference 15). The selection of preferred values of 

 ≤  10) could be achieved by changing the proportions of a binary-solvent mobile phase; for reversed-phase chromatography an increase of organic solvent in the mobile phase by 10%v provides a predictable, 2–3-fold decrease in 

The use of reversed-phase chromatography and gradient elution was quite limited in 1971; most HPLC separations were carried out isocratically by partition, adsorption, or ion exchange. Little was known of how best to change selectivity α, which is the most important parameter for controlling resolution. As in GC, random changes in the column (for example, stationary phase) were commonly tried at first. HPLC method development by "enlightened" trial-and-error was not often practiced, and a basis for "efficient" method development did not exist. 

From 1972 to 1977, I was VP of Research and later Clinical Chemistry at Technicon, with HPLC being downgraded to a "hobby interest" — however, the hobby was pursued with vigor. In 1971, Jack Kirkland and I presented the first ACS short course on HPLC ("Introduction to Modern Liquid Chromatography"), a two-day tutorial that continued during the next three decades with the later participation of John Dolan and Joe Glajch as co-instructors. Three editions of a book with the same title resulted (1974, 1979, and 2010 [17]), as well as two editions of a related book on method development (1987 and 1997 [18]). Both the course and the books proved popular, with > 6000 total attendees and 50000 books sold. 

 (Wiley) by Karger, Snyder and Horváth also appeared in 1973. 

In 1977, I reduced my management responsibilities at Technicon and returned to hands-on R&D. Our immediate goal was an integrated HPLC system for the automated analysis of pharmaceutical drugs in serum — including sample preparation. The final equipment could be regarded as a technical success (19), but a market failure; automated immunoassay proved better suited for routine use. However, we maintained an active HPLC research program from 1977 until I left Technicon in 1982. Our group included John Dolan, Sjoerd van der Waal, Steve Bannister, and Russell Gant, aided by collaborations with Nobuo Tanaka and others from Barry Karger's group at Northeastern University.  

Lloyd R. Snyder, Inc., was founded in 1982 as a one-man company based on HPLC. For the next seven years I also served as a consultant for DuPont Instruments in Wilmington, Delaware, which included supervising HPLC projects of my choice for four PhD students from nearby universities: Marilyn Stadalius (Delaware), Mary Ann Quarry and Julie Eble (Villanova), and Barbara Ghrist (Pennsylvania). In 1984, Lloyd R. Snyder, Inc., morphed into LC Resources with John Dolan as cofounder and headquarters in California; Tom Jupille joined us in 1987 as a third partner. Method development formed a significant part of our business from the beginning, with LC Resources growing to 30 employees by 2002. I also served as an editor of the 

 from 1987 to 2000; Figure 3 shows several colleagues at my 2001 retirement party.  

Figure 3: Participants at a symposium organized by the Journal of Chromatography A at Ellecom, The Netherlands (June 13â15, 2001), on the occasion of my retirement from the journal (20). Top row: P. Schoenmakers, J. Dolan, P. Carr, L. Snyder, J. Kirkland, I. Molnar, K. Bij, J. Dorsey, R. Giese, R. Marx, S. Terabe, N. Tanaka; middle row: E. Heftmann, D. McCalley, G. Vigh, R. Kaliszan, Cs. HorvÃ¡th, S. Poole, S. Rutan, E. Soczewinski; bottom row: P. Jandera, C. Poole, H. Poppe, V. Davankov.  

In 2002, LC Resources was split and sold to Rheodyne and Bioanalytical Systems. A few years later, the company was resurrected with the original three-man staff and an emphasis on short courses and collaborative research. From 1982 to present day, John, Tom and I taught HPLC to more than 10,000 individuals (apart from the ACS course), developed and marketed DryLab software (see below), and managed a method-development contract laboratory. 

By 1977, reversed-phase chromatography had become the primary HPLC procedure, and the use of gradient elution was becoming more common. However, further improvements in method development would require a better understanding of 

 (equation 3) as a function of separation conditions, for both isocratic and gradient elution. "Efficient" method development (EMD) would not be possible until these relationships were made quantitative, so as to allow predictions (and optimization) of separation by a computer. A systematic approach to method development would also be needed. 

Over time, a better understanding developed of which separation conditions have the greatest impact on selectivity (values of α). The importance and use of mobile-phase pH was well understood (for example, see reference 21), but only ionizable compounds are affected by pH. Less was known about other conditions that determine selectivity. EMD requires a choice of which separation conditions are best varied for an improvement in resolution. This choice must then be narrowed to specific solvents or columns, followed by preliminary separations. Finally, optimal conditions can be selected by some combination of further separations, experience, and computer simulation.   

Work that eventually enabled EMD will be discussed next: advances in our understanding of selectivity as a function of separation conditions; a practical model of gradient elution; quantitative predictions of separation as a function of either isocratic or gradient conditions (computer simulation); and a method-development strategy. While many workers have contributed to our present knowledge, the following summary emphasizes those projects in which my group was involved. 

The solvent-selectivity-triangle was introduced in 1973 (Figure 4), based on GC data by Rohrschneider for six test compounds and 80 solvents. This allowed the characterization of each solvent in terms of three possible interactions with the sample: hydrogen bonding (donor or acceptor) and "polar" (22). 

Figure 4: The solvent-selectivity triangle. Adapted from reference 22.  

Three generalizations follow from Figure 4. First, as expected, solvents of similar functionality, for example, alcohols or ethers, fall within the same part of the triangle (grouped within circles). Second, a change in selectivity requires the use of a solvent from a different part of the triangle. Finally, it should be possible with mixtures of just three, well chosen solvents to approximately match the selectivity of any other solvent, for example, a reduction in the number of solvent choices from the original 80 to just three. The latter observation contributed in 1980 to the widely used four-solvent mapping system of Glajch and Kirkland for reversed-phase chromatography (23). Seven experiments allowed resolution to be predicted as a function of mobile-phase composition (mixtures with water of methanol, acetonitrile, and tetrahydrofuran), followed by the computer-selection of a "best" mobile phase. Based on this approach, the DuPont Sentinel system for "automatic" method development was introduced a few years later. Sentinel represented a big step toward "efficient" method development.  

While the solvent triangle had its virtues, its derivation from data for pure solvents made it at best a qualitative model for the aqueous mobile phases of reversed-phase chromatography — as noted over the years (24). Indeed its ininitial application was intended for nonaqueous liquid–liquid chromatography, for which it was better suited. Other studies at about the same time also contributed to the selection of preferred solvents for controlling selectivity in reversed-phase chromatography (25,26). Today, methanol or acetonitrile is usually a first choice, with solvent selectivity optimized by varying their proportions in the mobile phase. 

Solvent-Strength and Temperature Selectivity:

 Before 1985 it was widely assumed that controlling selectivity required changes in either the organic solvent, mobile-phase pH, or the column. That is, changes in either solvent strength (that is, the organic-solvent concentration, %B) or temperature were regarded as ineffective. In the mid-1980s, however, two different studies (27,28) showed that a change in isocratic %B or (equivalently) gradient steepness could in some cases provide marked changes in selectivity and successful separations. While isolated examples of temperature selectivity were known from the 1950s, its potential usefulness in reversed-phase chromatography method development was widely ignored. Some key experiments suggested by Bill Hancock in 1993 demonstrated that a change in temperature could be of great value for the separation of peptides (29). A few years later, we undertook a comprehensive study of the contribution of both %B and temperature to reversed-phase selectivity for various nonpeptide samples (30). Simultaneous changes in these two conditions were found to provide a convenient and generally effective control over selectivity (31), and for many laboratories this is now a preferred first step in method development. See the example of Figure 5 in a following section on computer simulation. 

Figure 5: Computer simulation (DryLab) for the separation of a peptide mixture. 

 While a change of stationary phase can be an effective way of varying selectivity, at the present time there are > 1000 different reversed-phase columns from which to choose. A number of procedures for characterizing column selectivity were reported prior to 2000 (for example, see references 32–34), but with certain limitations (35). In 1998, John Dolan and I organized an exploratory meeting with Jack Kirkland, Pete Carr, John Dorsey, and Uwe Neue, from which a column-selectivity research plan resulted. This was followed by a 14-year program which later included the participation of Jon Gilroy and Chris Heard of Bioanalytical Systems, Dan Marchand at the University of Wisconsin–River Falls, Uwe Neue (Waters Inc., deceased), and David McCalley of the University of the West of England.  

The latter work led to the development and application of the 

 of reversed-phase column selectivity, summarized recently in (35). By means of the H-S model, it is now possible to compare more than 500 reversed-phase columns in terms of selectivity, and to reliably choose columns whose selectivity is either similar or different using open source software (36). For routine assays, similar columns are needed as backup, while different columns can be used for orthogonal separation. Both kinds of columns are recommended for assays that fulfill "quality-by-design" (Loren Wrisley, cited in reference 37). The fundamental nature of the H-S model also supports several other uses related to method development  — for example, problems with peak tailing and stationary phase degradation (35) — as well as provides a quantitative basis for understanding and using column selectivity. 

 Ion-pair chromatography, as pioneered by John Knox and Goren Schill, provides still another opportunity to affect selectivity (see Chapter 7 of reference 17). This option is becoming less popular, however, because of the relative inconvenience of ion-pair chromatography (especially with MS detection) and the increasing replacement of ionpair chromatography by hydrophilic interaction chromatography (HILIC) (17).  

 Since the 1960s, chromatographers have been generally familiar with how isocratic separation is affected by changes in chromatographic conditions. A similar awareness for gradient elution emerged much later. Despite the derivation of the fundamental retention equation of gradient elution in 1955 by Drake, practical applications of gradient elution theory would remain dormant for the next 25 years. Our interest in gradient elution began in connection with adsorption chromatography. Starting with the Drake equation, an important concept became apparent by 1964: so-called linear-solvent-strength (LSS) gradients (38) that provide comparable peak spacing at both the start and finish of the separation. 

In 1969, Dennis Saunders and I used computer modeling to arrive at a critical feature of LSS separation (39): as gradient steepness 

)+1]. This is of the same form as a change in isocratic resolution with 

+1]; see equation 3) and suggests that gradient steepness 

 in isocratic elution. A decade later, the LSS model was brought to initial completion with John Dolan and Russell Gant (40,41). Previous workers (for example, see reference 16) found that retention 

 in reversed-phase chromatography can be represented by 

 is a solute property that varies with molecular weight, and φ is the volume-fraction of mobile-phase organic solvent. This means that a linear reversed-phase gradient corresponds to LSS separation. 

Some other practical conclusions followed for linear gradients. First, "average" values of 

*) can be calculated from the gradient time 

m, and Δφ (change in φ during the gradient):

History of Chromatography

A pioneer in high performance liquid chromatography reflects on his career and how "enlightened trial-and-error" can reduce the effort involved in method development.

Efficient HPLC Method Development and Personal Reflections

In December 2000, the European Union's Water Framework Directive (WFD) came into force with the aim of protecting and enhancing the quality of different water bodies within the community. The use of various passive sampling devices in conjunction with gas chromatography–mass spectrometry (GC–MS) and liquid chromatography–mass spectrometry (LC–MS) techniques to screen pollutants has proved invaluable for investigative monitoring purposes within the remit of the directive. Used together, these new analytical approaches offer a robust solution to address specific future monitoring needs, particularly those prompted by legislative change. 

The European Union's Water Framework Directive (WFD) became part of law in the United Kingdom in December 2003 with the objective of providing for the planning and delivery of better quality surface water, ground water and coastal waters (1). Various types of monitoring activities are described including: investigative, operational and surveillance. In particular, the WFD will help to deal with diffuse pollution that remains a serious environmental concern after most discharges from point sources have now either ceased or been remediated (2). Furthermore, the Marine Strategy Framework Directive came into force in June 2008, and its key aims are similar; to protect and enhance the environmental status of the marine waters by 2020 (3). 

Most of the strategies currently used for the identification of pollutants in a body of water focus on the measurement of the concentrations of the priority substances (currently 33 priority chemicals and eight other pollutants). These are mainly organic compounds together with the metals cadmium, lead, mercury and nickel. The measured concentrations are compared to the proscribed environmental quality standards set by the European Commission for each of these pollutants (4). Each subset of pollutants, such as polyaromatic hydrocarbons (PAHs) or specific classes of pesticides, may require a separate water sample to be taken in the field and the combined cost of these analyses can prove labour intensive and expensive 

In such surveillance monitoring campaigns, many compounds that could have a significant toxicological impact on the fauna and flora within the given aquatic environment will remain unidentified. This could result in the environmental objectives for a particular body of water not being met, and the causes of the failure need to be better understood. Further investigative monitoring may then be undertaken to gather further information on the likely reason for the failure. Additionally, within the WFD, the current monitoring practice of taking low volume (1–10 L) bottle, grab or spot samples of water followed by their laboratory analysis may not always provide a useful indication of the environmental status of a water course and alternative approaches, such as biomonitoring, sensors, passive sampling and others may be required.  

Existing Approaches to Investigative Monitoring

For several years investigative monitoring within the remit of the WFD has involved the analysis of spot samples of water for unknown (non-target) organic chemical pollutants. This is usually accomplished by low resolution gas chromatographic–mass spectrometric (GC–MS) analytical methods using simple mass spectral library searching routines. GC–MS is a powerful technique for the separation and determination of volatile and semi-volatile compounds, but even with the use of high-resolution capillary columns, it is unable to resolve the multitude of compounds that can be present in complex environmental samples.  

Screening of samples via low unit mass resolution GC–MS is also susceptible to interference from other compounds of a similar molecular mass. This implies that by using conventional single quadrupole GC–MS techniques many compounds can remain unidentified, which could have a significant aquatic toxicity (5). It is therefore desirable to introduce other instrumental techniques to improve the quality of environmental assessments and also to benefit from resource reduction as further analytical developments become available (6). 

New Analytical Technologies for Environmental Analyses

Several advanced instruments for environmental analyses have been developed recently; including high resolution GC–MS and liquid chromatography–mass spectrometry (LC–MS) systems. These techniques use higher resolution spectrometers such as time-of-flight (TOF), quadrupoletimeof-flight (Q-TOF) and some trap technologies; all have been used effectively to identify complex mixtures of pollutants in water samples (7–11). In addition, nuclear magnetic resonance (NMR) spectroscopy (to confirm structures) in conjunction with high resolution LC–MS has also been proposed (12, 13). Two-dimensional GC (GCGC) has also seen significant advances, especially with the use of fast TOF-MS detectors (14). GCGC allows for enhanced separation of complex mixtures through greater chromatographic peak capacity and allows for the detection of trace contaminants that would not have otherwise been identified through conventional single dimension GC–MS techniques (15). 

Many hazardous chemicals have impacts on the aquatic environment at extremely low concentrations and these can be challenging for analytical detection. Complex sample matrices can often make the identification of target compounds very difficult due to significant 'chemical noise', resulting in sub-standard mass spectral library match factors. In order to address these issues, and, in particular the 'chemical challenges' of the WFD, the Environment Agency commissioned the National Laboratory Service (NLS) to develop new low-cost and effective methods to screen pollutants in water samples.  

The NLS developed the GC–MS target-based multi-residue method (TBMR) which allows for the identification of virtually all GC-amenable pesticides as well as hundreds of other organic pollutants in a single sample. At the heart of the GC–MS screening capability is the de-convolution reporting software (DRS) application for target compound analysis (16). This application combines results from the GC–MS Chemstation (Agilent Technologies, Santa Clara, California, USA), the automated mass spectral de-convolution and identification software (AMDIS —

 and the mass spectral search program from the National Institute of Standards and Technology (NIST) in a single report. A website that provides an explanation of how AMDIS works can be found at 

http://chemdata.nist.gov/mass-spc/amdis/explanation.html

 Aqueous samples (including surface water, groundwater and effluent) are collected into 1 L glass bottles and stored in the dark at 5 °C ± 3°C without further additives. As a result of a wide range of compounds contained within the target database and their variety of chemical characteristics, a liquid-liquid extraction method was chosen as the initial isolation method. An internal standard (D

-phenanthrene) is added to a water sample (1 L) which is extracted using dichloromethane (50 mL). The extraction solvent is removed and the remaining aqueous layer acidified (pH ~ 1–2) using sulphuric acid. The extraction procedure is then repeated on the acidified sample. The combined extracts are then carefully evaporated to avoid significant losses of volatile compounds to 1 mL using a nitrogen 'blow down' concentrator. The resultant extract is dried using anhydrous sodium sulphate and transferred to an auto-sampler vial for analysis. Typically, a batch of twentyfour samples can be prepared in a single day. Extracts are analysed using a single quadrupole GC–MS (Agilent 7890-5975 instrument) running in full scan mode (mass range: 35–566 Da) with de-convolution reporting software (DRS) that incorporates mass spectral de-convolution with conventional library searching and quantification. Sample extracts (1.5 µL) are introduced via cold splitless injection and separated on a HP5-MS UI capillary GC column (30 m × 25 µm × 0.25 µm film). (Agilent Technologies, Santa Clara, California, USA). The initial oven temperature of 40 °C (2 min) is increased to 300 °C at 10 °C/min and held for 8 mins. Over 990 target compounds can be analysed by this method without further clean-up. Figure 1 shows a typical chromatogram for a groundwater sample extracted and analysed under the above conditions. 

 Figure 1: Total ion chromatogram of an extracted 1 L groundwater sample analysed using a GCâMS instrument running in full scan mode (35â566 Da). The extract (1.5 ÂµL) is introduced via cold splitless injection and separated on a HP5-MS UI capillary column (30 m Ã 25 Âµm Ã 0.25 Âµm film) at the conditions described in the text.  

Firstly, the GC–MS Chemstation software performs a quantitative analysis for target compounds using the target ion and up to three qualifying ions under retention time locking (RTL) conditions. RTL has the ability to very closely match chromatographic retention times in any Agilent GC system even if there are subtle differences in columns differing from the nominal length due to 'end trimming' or when there are different column outlet pressures. The DRS software then sends the data file to AMDIS which de-convolutes the component spectra and searches the NLS generated target database using the de-convoluted full spectra (provided peak apices are > half a scan apart). The NLS target database comprises the Agilent RTL Hazardous Industrial Chemicals Database (a mass spectral database of 567 pesticides, solvents and endocrine disrupting compounds), plus a further 423 compounds added by the NLS based on environmental risk assessments. These mass spectra are generated under RTL conditions and entries include the retention times from the locked method. The de-convoluted spectra for all hits identified by AMDIS are sent to the NIST mass spectral library for confirmation. The routine searches against all of the 160000 compounds contained in the NIST library. This is also a full spectrum search, but now against a different library than that used by AMDIS. This approach provides three independent complementary steps of confirmation, thereby significantly increasing the accuracy of target identification, even in the most challenging of matrices. 

Table 1: Example of a report from a GCâMS DRS analysis of a contaminated groundwater sample. Compounds identified include the commonly found molluscicide metaldehyde, the herbicide bentazone and the plant growth regulator chlorpropham. 

The amounts shown in the quantitation and DRS reports are an estimate of concentration based on target ion response compared to response of the internal standard. This new method has been essential for identifying emerging environmental pollutants. Table 1 shows a report obtained from the GC–MS DRS analysis of an extracted groundwater sample containing several pesticides. The de-convoluted spectrum obtained from AMDIS is searched against the NLS RTL target database and a match factor assigned. This is followed by the retention time comparison to that in the RTL database (R.T. Diff. Sec.). As a final compound verification, the match factor obtained from the independent NIST library is shown together with its rank (Hit Num.) in the top 100 hits. High match factors obtained from the two spectral libraries provides high confidence in the data. 

The NLS have also developed a LC–TOF-MS screening method for the more polar pollutants that are not amenable to direct GC–MS analysis. 

 HLB SPE cartridges (200 mg) (Waters Corporation, Milford, Massachusetts, USA) with an automated extraction system are used. Cartridges are conditioned with methanol (8 mL) followed by de-ionized water (8 mL) and the water sample (500 mL, flow-rate 5 mL/min) is then loaded onto the cartridge. After loading, the cartridge is washed with de-ionized water and the sorbent dried fully with high purity nitrogen. The column is then eluted (10 mL) with dichloromethane:iso-propylalcohol:trifluoroacetic acid (80:20:0.1 v/v/v). The eluate is evaporated to dryness in a vacuum centrifugal evaporator and the residue re-solvated (75 µL) in acetonitrile:methanol (1:1 v/v) and de-ionized water added (425 µL). The sample is vortexed, mixed, filtered, transferred to a screw cap silanized vial and stored (2–4 °C) until analysis. Typically, a batch of twenty samples can be prepared in a single day. The extracts are analysed by LC–MS (Agilent 1200 LC system coupled to a Bruker MicroTOF MS) (Bruker Daltonics GmbH, Bremen, Germany).  

 The SPE extract (100 µL) is injected onto a Atlantis T3 LC column (Waters Corporation, Milford, Massachusetts, USA) (150 mm × 2.1 mm i.d., 3 µm particle size) held at 45 °C. The mobile phase is 2 mM ammonium formate and 0.01% formic acid in de-ionized water (solvent A) and 2 mM ammonium formate and 0.01% formic acid in methanol (solvent B). The gradient is: 5% B (0 min) with a linear increase to 100% B over 25 min, held for 5 min. Equilibration is 10 min at 5% B. 

 The interfaced TOF-MS is equipped with an electrospray ionization source. The above LC gradient is run twice for each sample: once with the TOF in positive ionization mode and once in negative mode. The nebulizer pressure, dry gas flow, dry gas temperature and capillary voltage are held constant at 50 psig, 10 L/min, 180 °C and 4500 V (negative ionization mode), and 50 psig, 10 L/min, 180 °C and 3000 V (position ionization mode), respectively. The resolution at 

 = 316.962 is approximately 12000. Scan data is acquired across the mass range: 125–1,000 Da at a rate of 1 Hz with the optimized parameters shown in Table 2. 

 Using the Bruker target analysis software, the TOF data files for each sample are searched for compounds listed in a database. Analysis was performed for the chemicals listed in Table 3. 

Table 3: Range of substances analysed in the LCâMS screening method. 

 First a calibration is performed to ensure good mass accuracy. Next, extracted ion chromatograms (EICs) are created with a width of ± 0.005 mDa for each compound in the database. The software then looks for peaks in each chromatogram at the retention time specified in the database ± 0.5 min. If a peak is detected, an average mass spectrum (background subtracted) is taken across the peak and the mass and isotope pattern present are scored against the theoretical values. Figure 2 shows an EIC and extracted mass spectrum for the anti-convulsant drug carbamazepine found in a groundwater sample. Analytes were identified using target analysis software (Bruker) which scores compounds in the database against the following criteria:  

1. Comparison to known chromatographic retention times, 

2. Accurate mass database of environmental contaminants, 

3. Isotope patterns for compounds identified in 1 and 2 compared against the theoretical values. 

Results for a groundwater sample spiked with a mixture of pesticides, non-ionic surfactants and pharmaceuticals are shown in Figure 3. 

Figure 2: Extracted ion chromatogram and mass spectrum for the anti-convulsant drug carbamazepine found in a groundwater sample. The protonated molecular ion for carbamazepine appears in the mass spectrum at m/z=237.1022. 

The reliability of taking spot samples of water is questionable, as there is a chance that potentially harmful pollutants can be missed if a sample is taken at the wrong time over a pollution event. The 'spot check' approach is only able to collect pollutants present in the column of water the moment the sample is taken. Passive sampling is a technique where pollutants are sequestered by an

 device over extended periods of time; typically 1–6 weeks (17). Such devices have been used historically to measure time-weighted averaged concentrations of pollutants in air. This approach increases the likelihood of capturing different pollution events, whether they are point or diffuse. The technique measures the 'toxicologically relevant fraction' of contaminant mixtures and can also indirectly lower analytical detection limits for the various pollutants. Typically, the extracts obtained from passive samplers are chemically complex, especially from those deployed, for example, downstream of a wastewater treatment plant. The analytical challenge is how to cope with the huge number of potential contaminants that may be present in any given sample. 

Figure 3: An example of the data output obtained from the analysis of a spiked groundwater. Compounds identified with the symbol +++ meet all three identification criteria; chromatographic retention times and accurate masses match those contained within the data base of environmental contaminants. The isotope patterns also match theoretical values. 

A wide range of passive sampling devices is now available commercially or as laboratory prototypes. A book (18) and a number of reviews (6, 19) are available that describe their construction and use in monitoring the aquatic environment. Two of the main types of sampler used by the NLS in various trials in the United Kingdom are described below.: 

 SPMDs consist of layflat, low-density polyethylene (LDPE) tubing (approximately 98 cm × 3 cm) containing triolein lipid (1 mL) as the absorption matrix (20). SPMDs are used to monitor non-polar organic compounds, defined as those with a log Kow > 3, with maximum cross-sectional diameters of 1 nm and a molecular mass of less than 600 Da. Hydrophobic compounds such as PAHs, polychlorinated biphenyls (PCBs) and organochlorine pesticides in the dissolved state (i.e. those that are readily bio-available) will partition into the SPMD and become concentrated over time (21). Figure 4 shows a SPMD sampler before deployment alongside its protective stainless steel housing. The processing, enrichment and fractionation of SPMDs have been described in a number of publications (22–24) and involve the following steps: 

1. Removal of exterior surficial periphyton and debris, 

3. Size-exclusion chromatography and collection of the fraction that contains the chemical classes of interest e.g. PAHs, PCBs and organochlorine pesticides, 

4. Class-specific fractionation using adsorption chromatography.  

The fraction collected is carefully evaporated to avoid losses of volatile compounds and is reduced in volume to approximately 0.5 mL using a nitrogen blow-down concentrator. The SPMD extracts are analysed by GC–MS using conditions identical to those for the spot water samples. 

Table 4 : Different pollutants identified in SPMD and POCIS passive sampling devices deployed in the River Mersey, one of eighteen drinking water protected areas chosen in England and Wales for field trial in 2011. Pollutants identified comprised several WFD priority substances including: the trichlorobenzenes, fluoranthene, benzo(b)fluoranthene, benzo(k)fluoranthene, benzo(a)pyrene, isoproturon and diuron. 

Polar organic chemical integrative sampler (POCIS):

 The POCIS is designed to monitor more polar organic compounds which are not accumulated by SPMDs, i.e. compounds with a log Kow > 3 with maximum cross-sectional diameters of ~ 0.1 µm (25). The POCIS consist of specific SPE resin sandwiched between two polyethersulphone membranes clamped together by steel rings (9 cm outside diameter, 5 cm inside diameter). POCIS are manufactured in-house and the SPE material (200 mg) used is the Oasis HLB phase (Waters). Figure 5 shows a POCIS prior to deployment alongside its protective stainless steel housing. 

Figure 4: The SPMD (bottom left) is used to sequester non-polar compounds such as the PAHs, PCBs and organochlorine pesticides (with log Kow > 3), whereas the POCIS (bottom right) adsorbs polar compounds such as pharmaceuticals, some herbicides and fluorinated acids (with log Kow  

 The SPE material contained within the POCIS is transferred into an empty 6 mL glass SPE column with a PTFE frit. A further PTFE frit is then placed on top of the material to keep it in place. The SPE column is then eluted (10 mL) with dichloromethane:isopropanol:trifluoroacetic acid (80:20:0.1 v/v/v). This elution mix has been found to give good recoveries for all compounds within the LC–MS database used by the NLS. The evaporation of the eluates and their analyses by LC–TOF-MS is the same as for spot water samples, except that the injection volume is reduced to 10 µL. 

Results from the Deployment of Passive Samplers

Article

This article describes the GC–MS and LC–MS screening methods developed by the Environment Agency for England and Wales for the analysis of both low-volumn water samples and extracts obtained from various designs of passive samplers.

Screening of Pollutants in Water Samples and Extracts from Passive Samplers using LC–MS and GC–MS

Supported liquid extraction (SLE) has been around for a long time, but surprisingly it is seldom used to replace classical liquid–liquid extraction (LLE). In SLE, the same aqueous phases used in LLE are coated onto an inert diatomaceous earth support, but instead of shaking the two immiscible phases together, the organic phase is passed through the column (or cartridge) and a very efficient extraction takes place. SLE offers many advantages over LLE, including equivalent or more efficient extraction, no emulsion formation, easy automation, less organic solvent use, less labour and less glassware. 

Liquid–liquid extraction (LLE) is probably the most widely used sample preparation process. It has been around for many decades, the theory is fairly easily understood and thousands of publications have demonstrated its usefulness. Most chromatographers are first exposed to the technique in their high school or undergraduate chemistry class. I covered the basics of LLE in a previous "Sample Prep Perspectives" instalment (1), and they will not be repeated here.  

There are many advantages to the use and application of LLE. First, the operation is fairly simple and the glassware used is inexpensive. Common solvents are often used for extraction, and most applications involve an aqueous phase and an immiscible organic phase. Initially, the sample can be in either phase, and, depending on the partition coefficients of sample and matrix components in the two phases, compounds of interest often can be quantitatively transferred without the presence of interferences. Adjustments to extraction conditions such as changing pH and ionic strength of the aqueous layer or addition of a second organic solvent to the organic layer are easily made. Recoveries can be quite high and reproducible.  

On the other hand, there are also some disadvantages to traditional LLE. First, it involves a great deal of time with laborious shaking that can be user dependent. Vigorous shaking (or even non-vigorous shaking) can result in emulsion formation. Emulsions can be difficult to break, and additional time is usually involved in performing this activity. In its traditional application, the technique consumes large amounts of organic solvents and generates considerable waste. Glassware must be cleaned after each use. Automation is very difficult; the technique is performed serially rather than in parallel so throughput is reduced. 

An alternative approach that gets around some of these disadvantages has been available for some time yet is relatively unknown — supported liquid extraction (SLE, sometimes referred to as solid-supported liquid extraction or solid–liquid extraction) (2). Its principles are relatively simple: a chemically inert, high-surface-area, highly purified, graded diatomaceous earth serves as a stationary vehicle for the aqueous phase of the LLE experiment. Water very easily adsorbs onto the surface (and is absorbed by the surface) of the diatomaceous earth particles. The dry solid sorbent [Figure 1(a)] is placed into a cartridge, column or well of a 96-well plate, the same devices used for solid-phase extraction (SPE). As shown in Figure 1(a), the aqueous-based sample (for example, diluted plasma or drinking water) is added to the dry sorbent and allowed to wet (disperse by capillary action and absorption) the diatomaceous earth, a process that usually takes only 5–10 min, sometimes a bit longer. Often the aqueous sample is pretreated (that is, the pH is adjusted or an ion-pairing agent or buffer is added) such that the analyte or analytes of interest are in a suitable form to be extracted into an organic solvent, just as is performed in conventional LLE.  

Figure 1: Steps in the supported liquid extraction process: (a) before extraction, (b) apply sample and (c) extract with organic solvent. In (b), allow 10â15 min for the aqueous phase to spread out over the diatomaceous earth. In (c), multiple extractions of small volumes of organic solvent are sometimes more efficient than one extraction with a large volume of solvent. 

Next, a small volume of immiscible organic extraction solvent is added to the top of the cartridge and allowed to percolate by gravity (or sometimes with gentle vacuum or pressure) through the supported aqueous phase [Figure 1(c)]. For a more viscous liquid sample, a vacuum is used to pull liquid through the cartridges. Because the aqueous sample has been widely dispersed throughout the solid support, the organic solvent has intimate contact with the thin film of aqueous phase and rapid extraction (equilibration) occurs [Figure 1(c)]. As mentioned above, this intimate contact also occurs in LLE because of the vigorous shaking of the separatory funnel. This shaking causes the immiscible organic solvent to disperse into tiny droplets that provide closer contact to the surrounding aqueous solvent. The downside of this dispersion process in LLE is that emulsions can form and the two phases can take a long time to separate.  

The final step in the process is collection of the organic effluent containing the analytes of interest from the outlet of the SLE device [Figure 1(c), right-hand side]. The aqueous phase remains on the device. A phase-separation filter is incorporated into the outlet frit of the device to ensure that organic effluents remain uncontaminated by aqueous matrix. 

With SLE, there is no vigorous shaking and therefore emulsions cannot be formed. In addition, the intimate contact between the aqueous and organic phases allows very efficient partitioning and analyte recoveries can sometimes be higher than in conventional LLE. Furthermore, the SLE process is more technique independent than conventional LLE; this often means that the experiments are more reproducible. Obviously, the glassware involved in SLE is greatly reduced compared to that required for LLE and no cumbersome washing of the separatory funnel is required. The entire SLE process is easier to automate than traditional LLE. The 96-well plate SLE format is especially amenable to automation using 

 robotics systems. Prepacked SLE products are available in all automation formats. In addition, bulk sorbent can be purchased for those who wish to make their own customized SLE devices. 

Because SLE has been around for many years, a number of companies have developed products to serve the needs of scientists who wish to improve or automate their LLEs. Table 1 lists several companies that supply products. Basically, there are three types of products: cartridges (sometimes referred to as columns or tubes); 96-well plates (mainly for automation); and bulk sorbent for those who wish to prepare their own SLE devices. This list represents typical suppliers; any supplier wishing to provide information on its products can supply information to be included in future updates. 

Table 1: Typical commercial products for SLE applications. 

The SLE approach can duplicate any developed LLE method with a few minor adjustments. In general, 1 g of diatomaceous earth sorbent is needed for each 1 mL of aqueous sample. Cartridge size is determined by the total volume of sample that needs to be extracted. The volume extracted is dependent on the concentration of the analyte or analytes of interest, the total volume of sample available and the sensitivity of the analytical system that will make the downstream measurement. For an environmental sample (for example, for screening pesticides in water), cartridges with an aqueous capacity volume of up to 100 mL are available. For biological samples such as plasma or urine, if one is using liquid chromatography (LC)–mass spectrometry (MS)-MS (triple quadrupole) for analysis, the analysis technique's excellent sensitivity would dictate that a smaller sample volume be used. For example, when using 96-well plates with a 2-mL well volume to extract a plasma sample, as little as 200–400 µL (with a maximum volume of 500 µL) can be used to obtain picogram-per-millilitre sensitivity. For the organic extraction, the volume of organic solvent should be at least equivalent to the volume of aqueous sample. As a rule of thumb, at least two column volumes of organic extraction solvent are recommended for maximum recovery. 

Table 2: Solvent mixtures compatible with SLE*. 

Almost any nonpolar solvent can be used as the immiscible organic phase. The organic solvent should be spectroscopic grade or better because the collected fraction is often evaporated to dryness, and any solvent containing nonvolatile impurities will be unsuitable for performing clean extractions. Pure organic solvents or organic solvent mixtures can be used. Table 2 provides a list of solvent mixtures compatible with SLE. The column to the right of each solvent pair indicates the maximum percentage of polar solvent tolerated. Exceeding the maximum water-miscible solvent percentage may have a detrimental effect on clean-up because some of the aqueous phase containing impurities could be stripped from the diatomaceous earth.  

To show the ease of converting an LLE method to an SLE device, I will illustrate two examples, one fairly simple and one more complex. Parabens are chemical compounds used in many processes in the pharmaceutical, cosmetic and food industries. The most commonly used parabens are the methylparaben, ethylparaben, propylparaben and butylparaben analogs, which are sometimes used singly but also in combinations. They are most frequently used as preservatives to prevent microbial growth in the products. 

A study of these four parabens was used to compare LLE to SLE (3). A previous SLE method was used as the starting basis (4) and was slightly modified for the paraben standards. For sample pre-preparation, 1 g of body wash lotion was spiked with a mixture of paraben standards; both high (175 µg/mL) and low (20 µg/mL) levels were investigated. Then the mixture was vortexed for 30 s. Next, 2.5 mL of acetone and 1.25 mL of a sodium chloride solution (20 g/100 mL) was added (as in the original LLE method), and the mixture was vortexed again for 30 s. For the experiments, a blank body wash lotion sample as well as a blank of the paraben standards were carried through the extraction process. 

The prepared sample and blank were each transferred to separatory funnels. A 20-mL portion of ethyl acetate was added followed by 5 mL of sodium chloride solution (20 g/100 mL). The separatory funnel was stoppered and then inverted and shaken with venting several times for 1 min. The phases were allowed to separate and the organic phase was collected. A 0.2-g amount of magnesium sulphate was added to the extract to remove water. The organic solvent was evaporated to dryness and reconstituted in 200 µL of methanol, which is a solvent that is more compatible with reversedphase chromatography. An aliquot of the sample and blank were separately injected into the LC column. 

The sample and blank were separately loaded onto ChemElut SLE cartridges (Agilent Technologies, 5 mL unbuffered, Part number 1298006) packed with diatomaceous earth. A time period of 15 min was allowed for the sample to disperse into the sorbent. Two 10-mL portions of ethyl acetate were percolated through the cartridge and were collected at the exit. The organic solvent was evaporated to dryness and the residue was reconstituted with 200 µL of methanol, and an aliquot of sample and an aliquot of blank were separately injected into the LC column. 

Figure 2: Reversed-phase chromatograms of blank body wash extracts: (a) SLE blank, (b) LLE blank. Column: 150 mm Ã 4.6 mm, 5-Âµm Zorbax Eclipse Plus C18; mobile-phase A: water; mobile-phase B: acetonitrile; gradient: 30â65% B in 4 min, then 65â100% B in 1 min; flow-rate: 2.0 mL/min; detection: UV absorbance at 230 nm; injection volume: 1.7 ÂµL. 

 Figure 2 shows an LC chromatogram of the body wash blank extracted by the SLE method [Figure 2(a)] and the LLE method [Figure 2(b)]. Both techniques showed a substantial number of small extractables that originated in the body wash matrix. For some reason, a large wide peak showed up in the LLE examples at about 5.2 min [Figure 2(b)]. Blanks of the paraben standards carried through the extraction processes showed no impurities of consequence and yielded well-resolved LC peaks (data not shown). Fortunately, the high performance liquid chromatography (HPLC) conditions chosen yielded well-resolved paraben peaks as can be seen in Figure 3, which shows the high-level spike. Recoveries and reproducibilities of the paraben extractions at both high and low levels for LLE and SLE are shown in Table 3. In general, the SLE method gave similar or better recoveries and lower relative standard deviation (RSD) values compared to the LLE method at both levels. 

Figure 3: Reversed-phase chromatograms of paraben extracts from body wash high-level spiked sample: (a) SLE extract, (b) LLE extract. Conditions the same as in Figure 2. Peaks: 1 = methylparaben, 2 = ethylparaben, 3 = propylparaben, 4 = butylparaben. 

To illustrate a more complex application of SLE compared to LLE, consider the multiresidue confirmation of pesticides in honey (4). This application is very important in the food safety area. Bees that gather pollen from blossoms that have been sprayed with various pesticides can carry the contaminated pollen back to their hives. This contaminated pollen could harm bee populations or lead to the development of contaminated honey. Because bees gather pollen from many different sources, a multiresidue approach is needed to test honey for possible contamination. Pirard's study (4) investigated 17 pesticides and metabolites from a variety of chemical classes. The extraction of pesticides was performed using SLE and compared to the results obtained using traditional LLE. 

Table 3: Recoveries of SLE vs. LLE at low- and high-spiked levels of parabens in body wash. 

Initially, classical LLE was used to extract pesticides from honey. A 1-g sample of honey was dissolved in 2 mL of water. Next, 6.5 mL of acetonitrile was added and the mixture was mechanically shaken for 30 min. The organic and aqueous layers were separated by centrifugation. The organic layer was evaporated down to 100 µL and added to 100 µL of water. This extract was filtered and injected into an LC–MS-MS system. For the SLE approach, a 1-g sample of honey was spiked with a surrogate standard before mechanical agitation with 1.25 mL of water and 2.5 mL of acetone for 1 h. The mixture was then loaded onto a 5-mL ChemElut SLE cartridge. After waiting for 15 min, the analytes were eluted twice with 10 mL of ethyl acetate (gravity feed). The extract was evaporated to dryness under a gentle stream of nitrogen, and 200 µL was transferred to a 10:90 acetonitrile–water solution in vials for LC–MS-MS analysis. The extracts were analysed by LC–MS-MS in the electrospray ionization (ESI) mode without further purification. The 17 pesticides were separated by optimizing the LC gradient, and the coeluted pesticides with different masses were identified using the multiple reaction monitoring (MRM) mode. 

Figure 4: Levels and recoveries of pesticides in honey. Adapted from reference 4. 

Figure 4 shows a comparison of recoveries obtained after SLE and classical LLE. The comparison showed that the SLE extraction procedure provided similar or higher extraction efficiency than LLE for most compounds. In particular, note the surprisingly poor recovery results for chlorotoluron, rimsulfuron and 2-hydroxyterbuthylazine for the LLE. For the SLE method, linearity was demonstrated from the range of 0.1–20 ng/g of raw honey with correlation coefficients ranging from 0.921 to 0.999 depending on the pesticide. Recovery rates were well above the range specified in documents (5). Reproducibility was found to be between 8–27%, which was acceptable within the concentrations studied. More than 100 samples of raw honey were analysed to test the method robustness. More details on the method can be found in reference 6.   

Although the SLE technique has been around for at least two decades, many separation scientists are unaware of its capabilities and potential to replace traditional LLE procedures. Many sample types have been handled by SLE. The analysis of drugs in biological fluids such as plasma and whole blood is one of the most popular uses of the technique. Generic procedures are available for a wide variety of acidic, neutral and basic compounds. With the use of 96-well plates coupled to LC–MS-MS analysis, only a small amount of sample is required to perform efficient liquid–liquid type extraction in an automated fashion.  

 is the editor of "Sample Prep Perspectives" and a senior scientist in the Columns and Supplies Division at Agilent Technologies in Wilmington, Delaware, USA. He is also a member of 

's editorial advisory board. Direct correspondence about this column should go to 

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

(2) J.G. Breitenbucher K.L. Arienti and K.J. McClurel, 

(3) H.F. Rossi III, D.W. Keating, D.C. Zimmerman, J. Rizzo and K.M. Usher, Agilent Application Note, in preparation. 

(4) C. Pirard, Mass Spectrometry Laboratory, University of Liege, Belgium, published in Agilent Technologies Application Note Si-01002, August 2010. 

(6) C. Pirard, J. Widart, B.K. Nguyen, C. Deleuze, L. Heudt, E. Haubruge, E. De Pauw and J.-F. Focant, 

Sample Preparation Perspectives

This article provides examples of the use of SLE, demonstrates some of its advantages and provides commercial sources for SLE products.

Supported Liquid Extraction: The Best-Kept Secret in Sample Preparation

As many of you know, a major part of my job is teaching liquid chromatography (LC) training courses around the world. Besides getting to see some pretty fascinating places, I get to meet lots of people and have a chance to help solve some LC problems that are troubling the attendees. One thing I have found is that chromatographers everywhere tend to have the same struggles with LC problems. Sometimes I feel a bit like the detective on Garrison Keillor's 

, whose radio dramas always end with the tag line, ". . . one man is trying to find answers to life's persistent questions . . . Guy Noir, Private Eye." In this month's instalment I'd like to share a few of the questions that came up in recent classes.  

One attendee mentioned a problem related to a small but noticeable increase in retention time that continued to grow over several weeks. In an earlier instalment (1) we looked at tricks to help determine the cause of retention drift. One of the ways to differentiate between hardware (flow-rate) problems and chemical (column, mobile phase or temperature) problems was to see if the disturbance at the column dead time changed in the same proportion as the retention change. If it did, a hardware problem was indicated; otherwise it was a chemical problem. However, the current problem was observed on an LC system with a mass spectrometry (MS) detector. In contrast to ultraviolet (UV) detectors, where a solvent-front disturbance is almost always seen, unless the MS system is set to specifically look for ions at the dead time, the baseline is flat in this region. As a result, the dead time diagnostic didn't help.  

The first step taken was the easy one — replace the column. When this approach didn't help, a new batch of mobile phase was made, but it didn't solve the problem either. Next came cleaning the pump check valves and finally replacement of the pump seals. None of these changes arrested the drift.  

Then one day, the user was inspecting the autosampler and noticed a reflection from a drop of liquid on the injection valve. This led him to service the valve. He discovered that there was a scratch between two of the passages in the valve rotor, creating a situation called cross-port leakage. A tiny amount of mobile phase was leaking out the waste line from the injector, effectively lowering the flow-rate to the column and thus increasing the retention times. As the leak became worse over time, the loss of flow increased and retention times drifted to even longer times. Replacement of the rotor seal corrected the problem. Because the leak was small, likely only in the tens of microlitres per minute, and a piece of opaque tubing was used to route the injector waste to the waste container, the leaking solvent was not noticed. Replacement of the valve rotor corrected the problem. 

 Figure 1: (a) Injection valve rotor showing polymeric seal (shaded) and grooves to pass liquid (clear). (b) Typical plumbing connections: 1 and 4, injection loop connections; 2, to column; 3, to pump; 5, to waste; 6, sample inlet. (c) Lines showing scratch between ports 4 and 5 (arrow). Adapted from Figure 17.3 of reference 2. 			

The problem can be more clearly understood with the help of the valve rotor sketches in Figure 1. The rotor usually comprises a polymer disk typically 15–30 mm across and 5 mm thick, with three semicircular grooves cut into the surface as shown in Figure 1(a). In operation, a mating stationary piece (the stator) contains connections to the sample loop (ports 1 and 4), the column (port 2), pump (port 3), waste (port 5), and sample inlet (port 6) [Figure 1(b)]. The configuration of Figure 1(b) is in the inject position, where the loop is in the flow path. In the load configuration (not shown), the rotor would be rotated 60° to the right and the loop would be connected to the sample inlet and waste, while the pump would feed directly to the column. The valve rotor has a very long lifetime under normal conditions — I've had numbers of 100,000 to 500,000 cycles quoted to me by instrument manufacturers — so routine replacement of the injector rotor is not something that most of us ever encounter.  

However, if a tiny piece of hard material gets caught in the injector, it can scratch the rotor, as illustrated in Figure 1(c). Now some of the mobile phase can leak from the injection loop to waste instead of 100% of the flow going to the column. When such problems occur, they are most likely caused by a hard, insoluble particle in the sample. Centrifugation or filtration of the samples before injection should avoid this problem. I have also seen a case where a hand-cut piece of tubing was not rinsed properly, resulting in a tiny sliver of stainless steel working its way into the valve and scratching the rotor. Whenever tubing is hand-cut, whether it is stainless steel or plastic tubing, be sure to rinse it before use. The easiest way to do this is to connect the upstream end of the tubing and turn the pump on for a minute or two to flush the tubing to waste before connecting the downstream end. 

One of the topics that comes up regularly in this column is the importance of controlling the column temperature. Retention times can change by 1–2% for each 1 °C change in the column temperature, and peak spacing can change, as well. One of the illustrations I like to use to emphasize the importance of using the column oven rather than ambient temperature relates to a couple of trips I took in 2008. In January I was in central China in a laboratory that was not heated, and the ambient temperature was 10 °C; six months later I was in a temporary laboratory in Israel where, if the temperature wasn't 35 °C, it certainly felt like it! These temperatures are drastically different from the 20–25 °C that most of us would quote when asked to put a number on "ambient" temperature. Imagine trying to transfer a method between those two laboratories. In addition to control of the column temperature, it is important that the mobile phase be preheated sufficiently so that there is no more than a maximum of ≈6 °C difference between the column inlet and outlet. 

Following this discussion in one of the classes, an attendee asked about the influence of the temperature of the solvent reservoir and of the autosampler. My feeling is that the temperature of the solvent in the reservoir is not important, because by the time the solvent travels through the pump and autosampler and all the associated connecting tubing, any preheating or cooling provided in the reservoir will be canceled out by the thermal mass of the system. The control of the temperature of the autosampler is most commonly accomplished by heating or, more commonly, cooling the sample tray. The goal with autosampler cooling is to prevent sample degradation before injection, and because it only cools the sample it will have little or no influence on the mobile-phase temperature. 

So the conclusion of the discussion is that you should always use the column oven to control the column temperature, even if it is just to set it a few degrees above room temperature — for example at 30–35 °C — to ensure a constant column temperature and thus more stable retention times. You should cool the autosampler tray if your samples are likely to degrade at room temperature before they are injected. Heating or cooling the mobile-phase reservoir has little, if any, affect on the chromatography. 

Another class attendee wondered why he observed broad peaks for his sample after installing a new column. This is a good example of a time to apply the divide-and-conquer principle of troubleshooting — do a physical or mental experiment to divide the problem up into smaller parts so that you can eliminate as many possible causes as you can. Before doing anything, I would reinject the sample, and if the peaks were still broad, I would reinject a reference standard. If both of these injections are still bad, it is not a problem with the particular sample. 

My first question would be to determine if the problem occurred because the column was replaced. Were the peaks of normal width before column replacement, but broader after? If so, then either the new column is bad, a nonequivalent replacement was used or something happened in connecting the column to cause the problem. The connections at each end of the column may be the source of the problem, especially if polyetheretherketone (PEEK) tubing and fittings are used. If these are not seated properly, a small gap in the connection can result in extracolumn volume and corresponding peak broadening.  

When adjusting PEEK fittings, I recommend turning the pump flow off, loosening the fitting, pushing the tubing into the connection to ensure it hits the bottom and then tightening the nut. If you try to tighten the fitting with the flow on, it can slip during the tightening process, creating exactly the problem you are trying to correct. Once again, reinject a reference standard to see if the problem is now corrected. If the problem persists, the next easy step would be to replace the column with another new column, double-checking that the manufacturer and part number is the same as the original column. While it is rare to receive a bad column today, it can happen. No harm is done by double-checking with another new column, and if it is good, you can just put the spare column in stock for later use. 

Another check that should be made is to examine the chromatogram (or chromatograms) for other changes. Has the retention time of one or more peaks changed? Has the peak spacing (relative retention) changed? If either of these situations has occurred, it suggests a change in chemistry of the system. We've already eliminated the column as the source of problems through substitution of another new column. The remaining chemical possibilities are the mobile phase and the column temperature. If the problem origination coincided with a change in mobile phase, this is a likely source of the problem. For example, if acids or bases are present, an error in pH adjustment of the mobile phase might be the problem source. Because it is easy to check by substitution, make up a new batch of mobile phase and see if the problem is corrected. The final chemical possibility is a change in column temperature. Make sure the column oven is on and set to the proper temperature. 

The remaining items on my list fall in the extracolumn effects category. We've already looked at the column connections, but extracolumn peak broadening can occur because of poor tubing connections wherever the sample contacts, so connections at the autosampler and detector should be checked as well, especially if PEEK tubing and fittings are used. The detector flow cell, time constant (noise filter) and data system data rate also can be sources of peak broadening. I'm assuming that the detector was not changed, so the flow cell is not likely the source. The detector time constant is an electronic filter that helps to reduce the noise through signal averaging, but if it is set to too large a value, it can result in broadened peaks. The rule of thumb is that it should be set to ≈0.1 times the width of the narrowest peak of interest. Similarly, the data system sampling rate should be set so that at least 15–20 data points are collected across the peak. Although neither of these settings is likely to change without intervention by the operator, a power outage or other factor might cause these to be reset to default values that are not appropriate for the present samples. It is easy to check both the time constant and data rate to be sure they aren't the source of the problem. 

The final item on the extracolumn effects list is the possibility of the use of too strong of an injection solvent, too large of an injection or both. My rule of thumb is that you can inject ≈15% of the volume of the first peak of interest if you inject in mobile phase. So, for example, if the first peak is 0.1-min wide and the flow-rate is 1 mL/min, the peak volume is 100 µL. This translates to ≈15 µL for an injection volume if mobile phase is used as the injection solvent. If a stronger solvent is used for injection, smaller volumes should be used; if a weaker solvent is injected, larger volumes may be possible. The simplest way to check for injection problems is to inject a very small volume (for example, ≤5 µL) to see if the problem is corrected. I've seen a case where a method worked well for years, then one day started giving broad or split peaks for no apparent reason. On examining the conditions, it was discovered that too much of too strong a solvent was injected. By diluting the injection solvent and increasing the injection volume to maintain the same injection mass, the problem was solved. Unfortunately, too many methods are in use where the operating conditions are "right on the edge" of reliability — it takes only a minor, and often unidentified, change in conditions to cause them to fail. 

Retention-time drift, temperature control and broad peaks from a new column are the topics of this month's discussion.

LCGC Europe-08-01-2012

Gradient HPLC for Reversed-Phase Separations

Efficient HPLC Method Development and Personal Reflections

Screening of Pollutants in Water Samples and Extracts from Passive Samplers using LC–MS and GC–MS

Supported Liquid Extraction: The Best-Kept Secret in Sample Preparation

Questions from Afar

Event News

Can We Trust Experimental Standard Deviations?