No thanks! I would like to know more about CHROMacademy

 Over 3000 E-Learning topics / 5000 Articles & Applications
 

Thermo SCIENTIFIC
The CHROMacademy Essential Guide
Understanding HPLC Column Characterization and Selection

The Essential Guide from LCGC’s CHROMacademy present an educational webcast on HPLC Column Characterization and Selection for reversed phase separations.  In this session, Dr. Tony Edge (Technical Manager, Thermo Scientific) and Tony Taylor (Technical Director, Crawford Scientific), present a definitive guide to the advanced classification of HPLC stationary phases.  The session will consider the critical characteristics of HPLC supports and bonded phase ligands for reversed phase HPLC and the way in which these materials are classified in terms of the essential performance characteristics by column vendors and independent testing organizations alike.  We will also consider how classification results can be used to select columns for HPLC method development and the relationships between analyte and column characteristics. A must see for everyone using or developing methods for reversed phase HPLC.

xc Tony Taylor
Technical Director
Crawford Scientific
 

The Camtasia Studio video content presented here requires JavaScript to be enabled and the latest version of the Adobe Flash Player. If you are using a browser with JavaScript disabled please enable it now. Otherwise, please update your version of the free Adobe Flash Player by downloading here.

x Dr. Tony Edge
Technical Director
Thermo Fisher, UK
 

Topics include:

  • Essential characteristics of an HPLC stationary phase
  • Review of bonded phase chemistry
  • Why characterize HPLC columns?
  • Column classification databases
  • Common classification tests
  • Interpreting classification results
  • Using column classification to aid column selection
  • Relating column properties to analyte characteristics
  • Similar and orthogonal phases
  • Column classification – the future
 

  Key Learning Objectives:

  • Discover the critical characteristics of HPLC stationary phases and supports
  • Understand the ways in which modern HPLC phase are characterized by manufacturers
  • Gain access to the leading free column characterization databases
  • Familiarize yourself with the advanced tests used to classify columns and learn to interpret their results
  • Understand how column classification results can influence your column selection choices for HPLC method development
  • Learn how to correlate analyte properties to HPLC stationary phases to increase your understanding of separations in Reversed Phase HPLC
  • Explore the future possibilities for  HPLC column classification
The CHROMacademy Essential Guide to
Understanding HPLC Column Characterization and Selection


This Essential Guide outlines some of the various methods for column characterization and the databases produced from the results which aid in the column selection process. Some of these databases are free to access and can be useful in selecting HPLC columns for method development and helping to understand the various affects and problems we encounter with our HPLC stationary phases.

  ask the CHROMacademy experts

Share this tutorial

There are many factors which influence the performance of an HPLC stationary phase, of which the chemical nature of the bonded phase ligand is important but by no means all encompassing, in determining the important separation characteristics. As users and developers of HPLC methods, we need to improve our understanding of these factors so that we can better understand problems when they occur and learn to exploit key stationary phase characteristics to our advantage during method development or improvement.

 
 
 
 

To study this topic in detail, we should first consider what chromatographic performance characteristics are important to us. The factors by which we define ‘successful’ chromatography are often those incorporated into our system suitability tests and include:

(Note: our definitions here are for the purposes of illustrating the effects of various stationary phase characteristics. For the standard definitions of these terms please refer to the IUPAC Nomenclature for Chromatography document

 
 
 

Retention (retention factor) –

Measures the relative retention of retained versus unretained solutes and indicates the ability of the stationary phase to interact with the analyte.  Differing extents of analyte interaction will produce a separation between analytes by retarding their progress through the column to a greater or lesser extent.  The analyte interaction(s) with the stationary phase will differ depending upon the chemical nature of the analyte, stationary phase and (sometimes) the eluent system used.

 
    Figure 1: Calculation of Retention Factor in HPLC
 
 

Efficiency –

Is important as it is a measure of the band (peak) spreading as a function of time. As the analyte band within the column disperses (broadens) any chemical separation which has been derived is mitigated. Thus highly efficient (narrow) peaks are preferred as a reasonable separation can often be achieved in a shorter time and the ability of the stationary phase to chemically distinguish between the two analytes can be more easily exploited.

 
    Figure 2: Calculation of Efficiency in HPLC
 
 

Selectivity –

The ability of a stationary phase to distinguish between species (analytes and other sample components) based on their chemical and physico-chemical properties and selectivity retard them within the column so that they elute as separated bands into the detector.

 
    Figure 3: Calculation of Selectivity in HPLC
 
 

Resolution –

The spacing between successive peaks within a chromatogram, measured as the difference in retention times at the peak apices divided by the average peak width. A resolution of 1.5 of greater is known (colloquially) as a ‘baseline’ separation and represents a situation in which two successive peaks are separated at the baseline point. Obtaining optimum resolution is often important for both qualitative and quantitative analysis.

 
    Figure 4: Calculation of Resolution in HPLC
 
 

Peak Shape (Asymmetry) –

The ‘Gaussian’ nature of the peaks within the chromatogram can be affected by several factors including unwanted secondary interactions with the stationary phase. Peak fronting or tailing can lead to compromised resolution between peaks.

Figure 5: Calculation of Peak Asymmetry in HPLC
 
 
 

There are very many characteristics of a stationary phase which affect the quality of a separation, not least of which is the chemical nature of the silica (or polymer etc.) stationary phase support particles and the bonded phase ligand. Below we have listed just some of the parameters which need to be considered when choosing, for example, a C18 column.

One should note that all of the factors listed pertain to the same USP stationary phase classification (L1) and indicate the myriad of combinations which account for the marked differences in retention, selectivity and performance between C18 (L1) columns from different manufacturers:


Bonded phase

  • Standard octadecylsilyl alkyl silane
  • Polar Embedded C18 (with imide, carbamate etc. spacer) for enhanced retention / selectivity of polar analytes
  • Mixture of C18 and shorter alkyl chains to give different selectivity / water wettability etc.
  • Nature of the silane substituents (e.g di-isobutyl silane) used as bulky substituents for pH stability at lower pH
  • Carbon loading
  • Carbon loading to silica surface area ratio (governs the ‘wetability’ of the phase and its ability to retain and separate polar analytes)
 

Nature of the base silica

  • Sol or sil-gel particle
  • Type I or Type II silica (governs the surface activity of the silica and the number of available ‘acidic / lone’ silanol groups
    which ultimately cause peak tailing if not effectively end-capped)
  • Silica metal ion content (effects peak tailing)
  • Totally porous, polymeric (effects pH stability) or superficially porous (for increased efficiency) support particles
  • Pure silica or organic / inorganic hybrid (to achieve wider pH stability)
  • Spherical or irregular silica particle
  • Particle size and particle size distribution (including the newer sub 2μm materials)
  • Pore size
  • Surface area
  • Deactivation (end-capping) post bonding – nature of the end-capping reagent (volume, polar, non-polar),
    end-capping reaction conditions (determines the extent of end-capping)

Bonded phase

  • Column Length
  • Internal diameter
  • Nature of the material used for column construction (typically stainless steel)
  • Metal passivation technique
  • Interior tubing surface polishing
  • Nature of the frit and spreaders used in the column end fittings
  • Packing quality (including method of packing, packing pressure, solvents used etc.)

A dizzying array of variables to alter the characteristics of a C18 phase leading to a very convoluted set of options when selecting HPLC columns. And these are the variables that we, as end users, are aware of – each manufacturer will have their own ‘tricks of the trade’ for column packing and hardware optimization.

Many of the variables above have been explained in a previous CHROMacademy Essential Guide which, for CHROMacademy members, can be found at this link //www.chromacademy.com/resolver-may2010.asp

There are some common assumptions that we tend to make as chromatographers to reduce the choices that we need to consider. These, might include:

  • Column manufacturers have learnt to pack columns very well using good quality silica and hardware
  • The issues around silica (column) batch to batch reproducibility have been dealt with to a greater or lesser extent
  • The pH stability of the column will be documented and can be chosen depending upon the demands of our analysis
  • The efficiency and longevity (mechanical stability) of the column and packing material will be optimised by the manufacturer (otherwise they would not have an established business)

So this broadly leaves the issue of selectivity – which after all is the driving force of separations in the liquid phase. How does one screen a column for ‘selectivity’, select a column which is similar to or radically different (orthogonal) to one we have used in the past? Apart from experience and the literature, we now also have various in-silico column classification databases, of which there are several available, and which use test results based on various chemical probes, designed to highlight specific columns characteristics.

 
 
 

In order to discuss the nature of the chemical probes that we might want to use, lets first have a quick review of silica as a stationary phase support and investigate some of the categories of bonded phase that exist in modern HPLC columns. Silica consists of siloxane bridges (silylether linkages), in which silicon atoms are three-dimensionally bridged by oxygen atoms, and terminate at the surface with silanol species -

 
Figure 6: Chemical nature of the silica matrix used as a stationary phase support in HPLC
 
 

Bare silica, chemically untreated, can be used as the support material for normal phase (adsorption) chromatography and after chemical modification is the primary support material for reversed phase (partition) chromatography.

Silica is a support material with exceptional assets including:

  • High mechanical strength - required to cope with back pressures of 1000+ bar
  • High surface area - required for high efficiency separations (300m2/g is typical for a 5μm particle with 100Å pore diameter)
  • Available in a form pure enough to chromatograph polar and ionisable components when used in conjunction with bonded phases

Whilst we think of silica support materials as ‘spheres’, in reality there is less than 1% of the silica available surface on the outer surface of the sphere, the vast majority being found within the internal pore structure.

The major drawback to using silica is its susceptibility to hydrolysis at high pH (typically >pH7.5) especially in highly aqueous environments. Factors that influence the rate of silica hydrolysis include:

  • Aqueous content of the eluent (silica is more soluble in highly aqueous eluent systems)
  • Buffer type used
  • Temperature of the column / eluent
  • Pore volume (higher surface areas, achieved with smaller pore widths / volumes, will be more susceptible to faster hydrolysis especially in highly aqueous environments)
 
 

Silanols and Peak Tailing

Fully hydroxylated silica will have a Silanol surface concentration of ≈8µmol/m2. Following chemical modification > 4µmol/m2 of these silanols may remain even with optimum bonding conditions due to steric limitations of the modifying ligands. This indicates that on a molar basis there are more residual silanols remaining than actual modified ligand. In order to remove some of these residual silanols, an end-capping process may be undertaken. Short chain, less sterically hindered hydrophobic ligands, (commonly trimethyl / tri-iodo chlorosilanes or similar), are reacted with the remaining unbounded silanol species, leading to improved peak shape with polar and ionisable analytes. This is only a partial solution, however, as not all of the surface silanol groups will be reacted even using sterically small liagnds and optimised bonding conditions, also the end-capping ligand is prone to hydrolysis especially at low pH.

Silanol groups are present in numerous conformations, with some being more active than others at causing analyte peak tailing and / or irreversible retention.

Figure 7: Various surface Silanol and Metal Ion conformations
 
 

Acidic (lone) surface silanol groups give rise to the most pronounced secondary interactions with polar and ionisable analytes. These lone silanol groups can also be ‘activated’ via the presence of metal ions within the silica matrix, which can also cause further unwanted secondary retention via chelation effects with polar and ionisable analytes. Modern silica is designated as being Type I or Type II, which primarily describes the nature of the silanol surface. Type I silica is ‘high energy’ (non-homogenous) and contains a higher density of lone silanol groups, whereas Type II silica is more homogenous (vicinal / associated / inter-hydrated) and therefore gives rise to much improved peak shapes due to a lowering of the energy of interaction between the surface silanol species and analyte polar functional moieties.

In order to create a more uniform (homogeneous) silica surface, manufacturers ensure the silica surface is fully hydroxylated prior to chemical modification. The incorporation of an acid wash step and avoidance of treatments at elevated temperature renders the majority of the surface in the lower energy geminal and bridged (vicinal) confirmations, creating Type II silica. Figure 8 shows how the peak shape of a basic polar analyte is affected when analysed using Type I silica as compared to Type II silica.

Figure 8: Basic, polar analyte (*) analysed using Type I (right) and Type II (left) silica under reversed phase conditions (45% MeCN / 55% 0.1%TFA, pH 2.1, 35oC).
 
 

Mobile phase pH will affect the degree of silanol ionisation and therefore the degree with which the silica surface may potentially interact with polar and ionisable analytes, causing peak tailing. Typically the pKa of surface silanol species lies in the range pH 3.8 – 4.5 and at eluent pH ≤3 all but the most acidic will be fully protonated and therefore peak tailing will be at a minimum (Figure 9).

Figure 9: Strength and extent of unwanted secondary interactions between the analyte and silica surface is highly pH dependant. Basic analytes in a low pH (<3.5) eluent with a highly acidic silica surface will present the worst combination of factors in terms of analyte peak tailing.
 
 

As the eluent pH increases the degree of ionisation (through deprotonation) increases, and peak tailing becomes more pronounced as the silanol groups interact with charged and polar species in solution (Figure 10).

Figure 10: Effect on peak shape of eluent pH and extent of silanol / analyte ionization.
 
 

Bonded Phases

The diversity of bonded phase ligands is huge, however it is necessary to review some simple representations of ligand ‘categories’ in order to gain an insight into the results obtained from the various classification methods test probes. Here we describe some basic characteristics which influence the physico-chemical properties of the stationary phase.

‘Alkyl’ Phases

Figure 11: Schematic representation of the variable aspects of alkyl stationary phase ligands.

Standard reversed phase type stationary phase. Alkyl chains of various length used to alter the hydrophobicity of the phase. The ligand is attached via a silyl ether bridge to the silica surface. The ‘R’ groups are typically methyl although other chemistries are used for hydrolytic stability – see below.

 
 

Base Chemistry Considerations

Figure 12: Schematic representation of the variable aspects of base silica and steric protecting groups associated with the siliyl ether linkage to an alkyl reversed phase ligand.

The silyl ether linker can be protected using hydrophobic moieties (R) as shown above. This acts to restrict access to the silica surface somewhat and reduces the influence of polar / ionizable groups on the separation.

Many manufacturers ‘end-cap’ unreacted surface silanol groups to reduce secondary interactions. Again, whilst this improves peak shape, it tends to lead to a very hydrophobic surface which will suit some applications and not others. Some manufacturers use ‘polar end capping’ reagents, which will render the surface slightly polar (allowing the phase to be used in 100% aqueous conditions for example) without giving rise to significant secondary interaction effects.

Some manufacturers produce ‘hybrid’ silica particles in which bridging organic groups are used to straddle surface silanol moieties. Primarily this is designed to lend some pH stability to the silica surface, however the surface bridging groups also influence the general hydrophobicity of the surface. The chemical nature of the R, R1 and R2 substituent’s will all have an effect on the selectivity of the stationary phase.

 
 

‘Polar Embedded’ Phases

Figure 13: Schematic representation of the variable aspects of polar embedded ligands (reproduced with permission from reference 4).

These phases are used when analytes contain key polar functional moieties which can influence the selectivity of the separation or where highly polar analytes are being separated and there is a need to carry out the separation using very low (or no) amounts of organic modifier in the eluent system. The chemical nature of the spacer group, the embedded polar moiety and the length of the alkyl chain will all affect the selectivity of the phase.

 

‘Phenyl’ Phases

Figure 14: Schematic representation of the variable aspects of phenyl type ligands (reproduced with permission from reference 5).

Phenyl phases are popular when analytes contain aromatic groups or π – electron systems. They are known to interact to different extents with aromatic and non-aromatic analytes as well as, in some cases, being able to differentiate based on the degree of unsaturation / degree of substitution. The phenyl rings may also be substituted with fluorine atoms to form a highly electron deficient phase known as PFP (pentafluorophenyl).

 
 
 

There are many factors which influence the performance of an HPLC stationary phase, of which the chemical nature of the bonded phase ligand is important but by no means all encompassing in determining the important phase characteristics.

In 2005 there were some 220 C18 (L1) phases available [1]. One can only speculate on the number available today, each of which will be subtly different in separating the analytes in which we are interested.

As users and developers of HPLC methods, we need to improve our understanding of the factors which affect separations so that we can better understand problems when they occur and learn to exploit key stationary phase characteristics to our advantage during method development or improvement.

This might be achieved by testing each (every?!) stationary phase using a ‘standard’ set of chemical probes, which we know will react in a predictable way, depending upon the phase characteristics. In this way we can produce comparative data that will allow us to select phases which we suspect might be best at exploiting important chemical and physico-chemical differences between our analytes. We can then map column characteristics and group columns (even those of the same ‘nominal’ bonded phase) into those which are ‘similar’ or ‘different’ (sometimes called ‘orthogonal’ in this context’), allowing us to manipulate our analyte retention and separation selectivity accordingly.

 

Several attempts have been made to produce a ‘definitive’ set of chemical probes and tests to best characterize the huge number of stationary phases available. This number was over 600 in the latest report we could find from 2003, and we speculate that well over 1000 different types are currently available. As yet a harmonized set of test probes and methodologies has not been identified.

An early attempt at producing a generic set of probes for testing HPLC column characteristics was made by Tanka and co-workers [2] and since then work by the USP Working Group on HPLC Columns, the Impurities Working Group of the PQRI Drug Substance Technical Committee in collaboration with Dr Lloyd Snyder [1] and work carried out by Euerby and Petersson [3-5] to expand the original probes designed by Tanaka have all undertaken to identify a definitive set of probes which will allow the various important physico-chemical phase characteristics to be specified. Most of these groups have also combined their data with various chemometric approaches to produce quantitative databases based on principal component analysis (PCA) or tools to visualize the relative groupings of commercially available columns according to their key descriptors.

One such diagram is shown in Figure 15 - the terms which are used to describe the column characteristics will be described in subsequent sections.

 

Figure 15: Spider diagram representing the various characteristics of the stationary phase.

HR - hydrophobic retention
HS - hydrophobic selectivity
SS - steric selectivity
HBC - hydrogen bonding capacity
BA - base activity
C - chelation
IEX - ion exchange capacity at pH2.6 and 7.6
AI - acid integration

Several of these databases are free to access and links are given in subsequent sections.

 
 
 

The Tanaka Protocol / Euerby and Petersson Model

The Tanaka protocol [2] was the first and seminal paper in this area, and is described below in terms of the classification method of Euerby and Petersson which is an extension and development of Tanaka’s orginal method [3-5].
 
Retention Factor (kPB) - measured using the retention of pentylbenze (kPB) (methanol used as the dead time marker). Retention factor:
k = (tr – t0 ) / t0 pentylbenzene
(tr – retention time of pentylbenzene, t0 – retention time of the first baseline disturbance on injecting methanol)
Conditions: MeOH–H2O (8:2, v/v), 1.0 ml/min, 40 oC, 5 μl injection of pentylbenzene (0.6 μg/ml)
 
 
The retention factor of pentylbenzene indicates the surface area of the packing material as well as the surface coverage of the bonded phase. This gives an indication of how retentive the phase will be in the reversed phase mode – higher kpb values indicate a column is more hydrophobic and hence will retain hydrophobic analytes more effectively. As always there are caveats to this – such as the Phenyl type phases which are less hydrophobic but result in larger kpb values due to π – π interactions with the aromatic moiety within the analyte molecule. This is a better measure of phase retentivity than % Carbon Load.
 
 
Hydropbobic Selectivity αCH2 - is measured using the retention factor ratio (selectivity) between pentylbenzene and butylbenzene and reflects the ability of the phase to separate compounds which differ only by a single methylene group. This measurement is often compared to the Hydrophobicity measure (H) in the Hydrophobic Subtraction model of Snyder and Dolan [6], another popular classification system which will be studied in a subsequent section.
αCH2 = kPB / kBB butylbenzene
Pentybenzene logP (octanol/water) : 4.90
Butylbenzene logP (octanol/water) : 4.27
Conditions: MeOH–H2O (8:2, v/v), 1.0 ml/min, 40 oC, 5 μl injection of pentylbenzene (0.6 μg/ml) and butylbenzene (0.3 μg/ml)
 
This parameter describes the ability of the column to differentiate between compounds with similar LogP values– i.e. it’s ability to separate hydrophobic compounds where the non-specific hydrophobic interaction is the primary mechanism for retention. Factors affecting this parameter include the carbon loading to surface area ratio (ligand / bonding density) and the pore diameter of the phase.
 
 
Shape Selectivity αT/0 - describes the ability of the phase to discriminate between planar structures (triphenylene) and those with greater spatial volume (o-terphenyl). A similar descriptor (Steric Resistance, S*) is used within the Hydrophobic Subtraction model which uses tetra-benzonapthalene and benzo-a-pyrene as probes.
αT/0 = kT / k0  

o-terphenyl

triphenylene

Conditions: MeOH–H2O (8:2, v/v), 1.0 ml/min, 40 oC, 5 μl injection of o-terphenyl (0.05 mg/ml) and triphenylene (0.05 mg/ml)
 
Although there is lower than expected correlation between the Hydrophobic Subtraction method and the Euerby Petersson model, they both effectively describe the ability of the column to differentiate on spatial volume and the ability of the surface to differentiate between ‘bulky’ solute molecules. This property can be important when analysing compounds whose chemistry is similar but whose spatial volume differs through charge repulsion / attraction, functional group chemistry interactions etc. The ability of the phase to be shape selective is influenced by ligand (bonded phase) density, monomeric or polymeric bonding chemistry and the shape and functionality of end-capping reagents that may be used. Larger values of this term indicate greater shape selectivity.
 
 
Hydrogen bonding capacity αC/P - is a measure of the retention factor ratio (selectivity) between caffeine and phenol. It is a descriptor of the columns ability to hydrogen bond with a solute. The Hydrophobic Subtraction model further sub-classifies this parameter into hydrogen-bond acidity (A) and hydrogen bond basicity (B). Typically the value reflects the number of available silanol groups which are capable of intermolecular hydrogen bond interaction with the solute and the nature and degree of end-capping.
αC/P = kC / kP  

caffeine

phenol (pKa 9.95)

Conditions: MeOH–H O (3:7, v/v), 1.0 ml/min, 40 oC, individual 5-μl injections of phenol (1 mg/ml) and caffeine (0.5mg/ml).
 
Older, Type-A, silica will show a greater degree of hydrogen bonding capacity, as will non-endcapped silica. Some columns are designed to undergo polar-polar interactions with solute molecules including phases that are endcapped with ‘polar’ functional groups to impart some degree of ‘(100%) Aqueous’ compatibility or those with polar embedded ligands. In general these columns are able to interact with polar and ionisable solute molecules to a greater extent than purely alkyl silica columns, especially through functional group proton acceptor / donor strength differences, and hence may offer an alternative selectivity.
 
 
Total ion-exchange capacity αB/P pH7.6 - is the selectivity between benzylamine and phenol at a mobile phase pH of 7.6 and reflects the total silanol activity of the column.
αB/P = kB / kP (pH 7.6)  

Benzylamine (pKa 9.33)

phenol (pKa 9.95)

Conditions: 20 mM KH2PO4 , pH 7.6, in MeOH–H2O (3:7), 1.0 ml/min, 40 oC, individual 5-μl injections of phenol and benzylamine HCl both at 0.5 mg/ml.
 
The pKa of the silanol groups on the silica surface is dependent on many factors which include; the number of lone silanol groups, the degree to which the silica surface is in the vicinal (inter-hydrated) form, the concentration and type of metal ions within the matrix etc. As the pH is raised, the degree of surface charge increases as the surface silanol groups become ionised. Older, Type-A silicas, and those which are non-endcapped will become more extensively charged and at a lower pH than the newer Type-B, Hybrid and end capped silica’s. This test measures the Total Ion Exchange capacity of the phase as it is assumed that at pH 7.6 all ionisable silanol groups are in the ionised form. This gives an indication of the ‘ion-exchange capacity’ of the phase and the ability to interact with (and potentially be selective towards) ionised solutes as well as polar, but non-ionised, functional groups. To some extent this parameter also indicates the possibility of poor peak shape (tailing) with acidic or basic analytes due to an increase in secondary interactions with the stationary phase.
 
 
Acidic ion-exchange capacity αB/P pH2.7 - is measured using the retention factor ratio between benzylamine and phenol at pH 2.7.
αB/P = kB / kP (pH 2.7)  

Benzylamine (pKa 9.33)

phenol (pKa 9.95)

Conditions: 20 mM KH2PO4 , pH 2.7, in MeOH–H2O (3:7), 1.0 ml/min, 40 oC, individual 5-μl injections of phenol and benzylamine HCl both at 0.5 mg/ml.
 

This parameter should give a lower result than αB/P pH 7.6 as the surface silanol groups should be un-ionised at this pH, with only the most acidic groups remaining in the ionic form. High values indicate the presence of highly acidic silanol groups and therefore the likelihood of poor peak shape with acidic or basic compounds. The magnitude of the result and the difference between this result and that obtained in the previous test together give an indication of the columns ability to interact with polar analytes to give a secondary interaction, and therefore an alternative selectivity, and the extent to which peak tailing might be expected. A high result for the test at pH 7.6 and low result at pH 2.7 might indicate some polar interaction with the stationary phase is possible whilst poor peak shape is not expected – this is typical for Type B silica, or where ligand bonding density is controlled in order to render the phase 100% water compatible.

Euerby and Petersson [3] tested 135 columns under the conditions shown above and subjected the results (all variables) to Principal Component Analysis. Column types included: C18, Aqua, Cyano, Phenyl, Polymer, Polar embedded, Perfluorinated, C8, Polyethylene glycol, Mixed alkyl, Polar endcapped, Short alkyl ligand, Amino, Phenyl-hexyl, Aluminium oxide, C12, Zirconium oxide.

The variables obtained were scaled by subtracting the average from each variable and each was divided by its standard deviation (auto-scaling). This effectively allows direct comparison of the results derived from different columns using either PCA SCORE plots or, via the use of simple mathematical transformation, using a database to rank phases in terms of the magnitude of their differences for each variable and also through a combined ranking score, the column difference factor (CDF).

 
 
 

Conditions:
20 mM KH PO , pH 2.7, in MeOH–H2O
 (3.3:96.7, v/v), 1.0 ml/min, 60 oC, 5 μl injection of the hydrophilic base test mixture, detection at 210nm.

Analytes
1 Nicotine
2 Benzylamine
3 Terbutaline
4 Procainamide
5 Salbutamol
6 Phenol

Figure 16: Principal component contribution plot for two C18 columns using the methodology from reference 3 (top) and comparison of the actual performance of these phases with pharmaceutically relevant basic analytes and phenol under reversed phase conditions (reproduced with permission from reference 3).
 
 

As can be seen from Figure 16, two phases who’s scaled PCA scores show a marked difference in acidic surface silanol activity (aB/P pH 2.7) do show marked differences in their retention and peak shape for basic compounds.  For Hypersil C18 (Type A silica) analytes 1,2 and 4 are irreversibly retained and analyte 5 shows increased retention and poor peak shape.   These results are somewhat predicted by the large negative PC contributions for the both the total and acidic ion exchange capacities, indicating that the Hypersil C18 phase will heavily interact with the protonated bases under the acidic conditions of the test.

Using PCA, by plotting the scores for two principal components (PC), it is possible to graphically find similarities and differences between objects (stationary phases). 

How much of each of the original variables that are included in a PC is described by so-called loadings, one for each variable. By plotting the loadings for two PCs, it is possible to see which of the original variables are most important (distance from the origin) and if any variables are correlated (the same or opposite directions on a straight line through the origin) (Figure 16).

Euerby and Petersson were able to reveal several ‘groupings’ amongst the silica phases by studying the SCORE Plots of various principal components.

 
Figure 17: (left) PC1 and 2 score plot for all columns excluding non-silica and amino phases: A, mostly non-C18 and traditional C18 acidic (type A) silica phases; B, mostly non-acidic (type B); C, polar embedded phases. (right) PC1 and 2 loading plot for all columns excluding non-silica and amino.
 

Three ‘groupings’ are initially discernable from plotting the first and second principal components (PC1, PC2).  For column descriptions see reference 3, or the ACD Labs database described at the end of this section.  The groupings are correlated with primary characteristics as follows:

Group A – high silanol activity (high aB/P) and low retentivity (low kPB) – correlate primarily with shorter chain phases and those using older ‘Type A’ silica which has larger numbers of lone, acidic silanol moieties. 

Group B – newer C18 phases which show higher retentivity and lower silanol activity (lower aB/P values)

Group C – include the polar embedded phase which show higher shape selectivity (aT/O)

The positions of columns on the SCORE plot correlate very well with the manufacturers (sometimes limited) information on column chemistry and the loading plot, which correlates the column properties investigated against position on the SCORE plot.

These first two PC’s explained 61% of the variations noted.  Plotting the first and third principal components revealed further important differences between columns.

 
 
 
Figure 18: (left) PC1 and 3 score plot for all columns excluding non-silica and amino phases: D, non-C18 phases; E, acidic phases; F, perfluorophenyl phases; G, highly hydrophobic phases; H, cyano phases. (right) PC1 and 3 loading plot for all columns excluding non-silica and amino phases.
 

Group E – acidic phases (Type A and non-end-capped) which show highly acidic (αB/P pH 2.7) and total silanol (αB/P pH 7.6) activity

Group D – are the non C18 phases and contain a sub classification (H) which contains most of the cyano phases which were investigated

Group F – contains the fluorinated phases which are differentiated by their increased shape selectivity (αT/O)

Group G – contains the highly retentive phases which in general have higher carbon load (>22%) and which are characterised by increased retention of pentylbenzene (kPB) and increased hydrophobic selectivity (αCH2)

Group H – contains the cyano phases (see D above)

 
 

The information from the SCORE plots above is very useful to help select columns which show the same general characteristics.  For individual column selection, it is more useful to rank the columns in terms of overall similarity or difference (orthogonality).  This can be most easily done through measuring differences in the individual parameters in a database format. 
The data generated from this, and subsequent studies, has been mathematically transformed into a commercially available database (see subsequent links and references section) and typical results are shown below.

 
Figure 19: (top) most similar column to the Hypersil Gold C18 from the ACD Labs Column Selector database based on the work of Euerby and Petersson, showing a small column difference factor (bottom) most dissimilar column to Hypersil Gold C18 showing a much larger column difference factor.
 
 

The Hypersil Gold C18 (ThermoFisher Scientific, Waltham, Massachusetts, USA) column is most similar (in this database) to the XTerra MS C18 (Waters Inc., Milford, Massachusetts, USA) column with the largest difference occurring in the general retentivity of the phase (kPB).  Both phases are high purity alkyl phases with effective surface deactivation, the X-Terra phase being of the bridged silica / organic hybrid type.  All other parameters show close correlation.

The least similar phase is the Primesep B which is a mixed mode phase, containing an alkyl chain with an embedded cationic moiety.  This goes some way to explaining the reduction in silanol activity as the surface silanol species are shielded from the probes by a layer of the embedded cations.  The reduction in retentivity can be explained by the decrease in hydrophobicity of the phase, due to the bonded phase charge, and mixed mode phases of this type are also known to be more highly shape selective.

Since their original work described here, Euerby and Petersson have gone on to identify further probes to better describe certain stationary phase types [4,5]. They describe two such probes for further characterization of ‘Phenyl’ type stationary phases as such:
Aromatic selectivity (π-acidity of the phenyl phase), (αPB/O).

 

The retention factor ratio between n-pentylbenzene (PB) and o-terphenyl (O), αPB/O = kPB/kO. This descriptor is believed to be a measure of the aromatic selectivity, which is influenced by the density of aromatic character on the phase.

Aromatic selectivity (π-basicity of the phenyl base),(αTNB/NB, αDNT/NB, αTNB/DNT).
The retention factor ratios between 1,3,5-trinitrobenzene (TNB) and nitrobenzene (NB), 2,4-dinitrotoluene (DNT) and nitrobenzene (NB) and  1,3,5-trinitrobenzene (TNB) and 2,4-dinitrotoluene (DNT).
These descriptors are believed to be measures of the aromatic selectivity, which is influenced by the density of aromatic character on the phase.

Figure 20, shows the loading plots for PC1 and PC2 for both the Tanaka and phenyl test probes. The plot clearly shows correlation between the various probes and allows identification of phases with predominant characteristics important to our analysis.

 
Figure 20: Loading plot for the Tanaka and Phenyl test probes (reproduced with permission from reference 5).
 
 
Of the 21 phenyl phases analysed, several discernable characteristics were identified via the analysis of a number of lipophilic bases. Some of these differences account for the alternative selectivity of the phases shown in Figure 21.
  Figure 21: Orthogonal selectivity of several phenyl phases in the analysis of lipophilic bases, Chromatographic conditions: 20mM KH2PO4 , pH 2.7 in MeOH:H2O (45.5:54.5, v/v), 60 oC, 5μl injection of a lipophilic base test mixture and detection at 210 nm (reproduced with permission from reference 5).
A number of pentafluoropropyl (PFP) phases were also analysed and the following general conclusions about ‘phenyl’ type phases were reached:
  • Phenyl phases tend to show lower hydrophobic retention than their C18 counterparts – which is absolutely to be expected
  • The length of the alkyl spacer between the silica and phenyl moiety as well as the inclusion of an electronegative atom prior to the phenyl group strongly influences the ability of the ligands phenyl moiety to undergo π-π interaction with aromatic moieties of the analyte molecule
  • The main difference between the chromatographic selectivity of the phenyl (π-base) and pentafluorophenyl phases (π-acid) was the latter’s enhanced shape selectivity and reduced aromatic selectivity parameters
  • In general the phenyl phases appeared to have a higher hydrogen bonding capacity than their alkly counterparts (perhaps due to the use of caffeine asa probe which may also undergo some aromatic retention)
  • Typically, phenyl phases with longer alkyl linking chains, such as phenyl hexyl phases, exhibit enhanced hydrophobicity, increased shape and aromatic selectivity, decreased ion exchange and increased apparent hydrogen bonding capacity
 
 

USP Classification Model

The USP Working Group on HPLC columns consists of members of the National Institute of Standards and Technology (NIST) and the five largest manufacturers of HPLC columns in the United States [1]. This group uses the NIST Standard Reference Material SRM 870 to evaluate columns using the following test conditions (outlined in the SRM Certificate of analysis):

Mobile phase: 80 % methanol / 20 % buffer (v/v) (5 mmol/L potassium phosphate adjusted to pH 7
(final phosphate concentration in the mixed methanol/buffer mobile phase is 1 mmol/L).
Flow rate: 2 mL/min
Column temperature: 23 0C ± 2 0C
Injection volume: 5 μL

This procedure uses a mixture of five organic compounds (uracil, toluene, ethylbenzene, quinizarin, and amitriptyline) in methanol to characterize column performance and is intended primarily for the characterization of C18 columns used in reversed-phase liquid chromatography.
This group identified four parameters to be used in the characterization of the columns:

Hydrophobicity / column retentiveness (capacity factor of ethylbenzene, H or Hy)  
Chelation (tailing factor of quinizarin, C or CTF)  
Activity toward bases (silanol activity) (capacity factor CA or CTA, and tailing factor of Amitriptyline, TA or TF  
 
Figure 22: Examples of separations of SRM 870 on commercial C18 columns - highly deactivated Type B silica with polar embedded ligand (top), high silanol and metal ion activity phase (bottom).
 
 

Shape selectivity (bonding density)

The term ‘shape selectivity’ is used to denote a chromatographic quality exhibited by certain stationary phases for which enhanced separations of geometric isomers may result based on their molecular structure rather than other physical or chemical properties of the solute.  SRM 870 does not characterize shape selectivity, however this property can be assessed by use of other chromatographic test mixes, such as SRM 869a

[//ts.nist.gov/MeasurementServices/ReferenceMaterials/archived_certificates/869a.pdf], Column Selectivity Test Mixture for Liquid Chromatography.  This test uses benzo[a]pyrene (BaP), phenanthro[3,4-c]phenanthrene (PhPh) and 1,2:3,4:5,6:7,8-tetrabenzonaphthalene (TBN) as probes and the shape selectivity measurement is the selectivity between TBN and BaP.  The structures and space filling models of these compounds are shown in Figure 23.

Figure 23: Testing columns for shape selectivity according to the USP classification model.
 
 

This group have also developed a mathematical model to derive a function (F) used for a more quantitative overall column comparison.  This model also incorporates a quantitative measure of Bonding Density (BD) as shown:

   

Bonding Density (BD) in μmol/m2
In which:
X = surface coverage (mol/m2)
%C = percent carbon loading of the bonded silica
nC = number of carbon atoms in the bonded ligand
MW = molecular weight of the bonded ligand
SA = surface area of silica substrate (m2/g)

 
 

The mathematical expression is shown in Figure [24] alongside typical results from the USP database column comparison.

Figure 24: USP ranking factor F expression alongside typical results for similar and orthogonal phases.
 
 

PQRI / Hydrophobic Subtraction Model

The Impurities Working Group of the PQRI Drug Substance Technical Committee are investigating the effects of technological improvements (one of which includes improvements in stationary phase design), and have adopted the approach of Snyder and Dolan [7] and their proposed Hydrophobic Subtraction Model.

This model is very well characterized and widely reported in literature [8-10] and uses the set of test probes shown in Table 1 under reversed phase conditions: 50% acetonitrile/ buffer; pH 2.8 and 7.0; 35 0C

 

Table 1: Test solutions used in the PQRI classification system

 
Based on the selectivity measurements from the various test mixtures, many reversed phase columns have been characterized by six column-selectivity parameters: relative retention (kEB), hydrophobicity (H), steric interaction (S*), hydrogen-bond acidity (A) and basicity (B), and relative silanol ionization or cation-exchange capacity (C), the latter of which changes with eluent pH (2.8 and 7.0). These measurements have led the formulation of the general equation of the Hydrophobic Subtraction Model shown:
 
log α = log k/kEB = η’H – σ’S* + β’A + α’B + κ’C
 
 
Figure 25, taken from reference 7 shows these interactions in schematic form.
Figure 25: Schematic representations of the five interactions described by the Hydrophobic Subtraction model
(reproduced with permission from reference 7).
 

Column hydrophobicity (H) increases with an increase in total carbon. End capping (because of its low (<10%) contribution to the overall carbon load has little effect on H. H has only a minor effect on column selectivity and is a measure of the column retentivity.

Column steric interactions (S*) increase as the bonded phase ligands move closer together on the silica surface (bonding density increases). That means an increase in S* for increased chain length or concentration of the bonded phase. S* also increases for packings with narrow-pore size. S* has a significant effect on column selectivity, especially for molecules of different shape.

Column hydrogen-bond acidity (A) due to non-ionized silanols decreases when the column is end-capped and the number of accessible and unreacted silanols decreases. This parameter has a significant effect on column selectivity for nonionized basic molecules such as amines and amides.

Column hydrogen-bond basicity (B) arises from various functional groups within the bonded phase. For all type-B (high-purity silica) and some type-A (older, less pure silica) columns, it is postulated that water from the mobile phase partly dissolves in the bonded phase, and subsequently preferentially interacts with non-ionized acidic species. So, columns with larger values of B preferentially retain acidic compounds. In the case of columns with embedded polar groups, the basic polar group (urea, amide, carbamate etc.) can strongly bind both phenols and carboxylic acids. Some type-A columns have larger values of B, believed to be the result of metal impurities in the silica.

Silanol ionization (C) results in a negative charge on the column, and this charge attracts ionized (positively charged) bases and repels ionized (negatively charged) acids. For samples that contain ionized acids or (especially) bases, the column parameter C is a very important contributor to column selectivity. For samples that do not contain acids or bases, C is much less important. Column ionization and values of C increase as mobile-phase pH is increased as has been extensively explained above. End-capping results in decreased access to ionized silanols and a large decrease in C.

Once again, from the data gathered under the testing conditions described here, a more quantitative measure of column similarity of difference has been proposed as shown.

The Fs value of this column (1) relative to the one you have selected (2). Fs is essentially the distance separating two columns in a six-dimensional parameter space. In general, Fs values below 3 are considered excellent matches, Fs values below 5 are considered reasonable matches, and Fs values above 5 are considered poor matches.
 
 
Figure 26 shows similar and orthogonal phases as reported from the PQRI database. Note that the database allows some degree of refinement to take into account analyte types (acidic, basic or both) and the intended eluent pH – which is very useful for column selection during method development.
 
Figure 26: Similar and orthogonal column selections from the PQRI database.
 
 

Links to Commercially Available Databases
Tanaka / Euerby & Petersson Column Selector Database
Column selection database based on the work of Euerby and Petersson:

//www.acdlabs.com/products/adh/chrom/chromproc/index.php#colsel

Note: in order to run this application you will also need to install the Freeware version of ACD Labs ChemSketch which can be downloaded here:
//www.acdlabs.com/resources/freeware/chemsketch/

USP & PQRI Databases
Both the USP and PQRI databases can be found at the following location:

//www.usp.org/app/USPNF/columns.html

 
 
 

References

  1. Pharmacopeial Forum Vol. 31(2) [Mar.–Apr. 2005]
  2. K. Kimata, K. Iwaguchi, S. Onishi, K. Jinno, R. Eksteen, K. Hosoya, M. Arki, N. Tanaka, J. Chromatogr. Sci. 27 (1989) 721.
  3. M. R. Euerby, P. Petersson, J. Chromatogr. A 994 (2003), p. 13 - 36.
  4. M. R. Euerby, P. Petersson, J. Chromatogr. A 1088 (2005) 1–15.
  5. M. R. Euerby, P. Petersson J. Chromatogr. A 1154 (2007) 138–151.
  6. D. Visky, Y.V. Heyden, T. Iva´nyi, P. Baten, J. De Beer, B. Nosza´l, E. Roets, D.L. Massart, J. Hoogmartens, Pharmeuropa 14 (2002) 288.
  7. N.S. Wilson, M.D. Nelson, J.W. Dolan, L.R. Snyder, R.G. Wolcott, P.W. Carr, J. Chromatogr. A 961 (2002) 171-193.
  8. J.W. Dolan, A. Maule, D. Bingley, L. Wrisley, C.C. Chand, M. Angod, C. Lunte, R. Krisko, J.M. Winston, B.A. Homeier, D.V. McCalley, L.R. Snyder, J. Chromatogr. A, 1057 (2004) 59–74.
  9. L.R. Snyder, A. Maule, A. Heebsh, R. Cuellar, S. Paulson, J. Carrano L. Wrisley, C.C. Chan, N. Pearson, J.W. Dolan, J.J. Gilroy, J. Chromatogr. A 1057 (2004) 49–57.
  10. N.S. Wilson, M.D. Nelson, J.W. Dolan, L.R. Snyder, R.G. Wolcott, P.W. Carr, J. Chromatogr. A 961 (2002) 195-215.
 
 

Related Reference Materials from CHROMacademy:

Theory of HPLC – Column Chemistry Module *** CHROMacademy Registered users only ***

Column Selection for Reversed Phase HPLC Essential Guide Tutorial *** CHROMacademy Registered users only ***

Column Selection for Reversed Phase HPLC Essential Guide Webcast *** CHROMacademy Registered users only ***

 

Further reading and resources from LCGC Magazine:

Facilitated HPLC Column Selection in RPLC

A Global Approach to HPLC Column Selection Using Reversed Phase and HILIC Modes: What to Try When C18 Doesn't Work

Selectivity in Reversed-Phase LC Separations, Part III: Column-Type Selectivity

 
 
 
As a member of CHROMacademy, you will also get access to LCGC magazine articles from your favourite authors - John Dolan, John Hinshaw, Mike Balough, and Ron Majors; over 300 news items refreshed daily; vendor application notes, webcasts and podcasts; electronic laboratory tools and calculators; and access to the growing Essential Guides archive, recorded tutorials by industry experts.

Enroll online today at: www.chromacademy.com

Subscribe now for $399
and get instant access to all CHROMacademy Essential Guide Webcasts


subscribe now

  join today
 
 
     
 

The following subjects are covered in CHROMacademy.com

The Theory Of HPLC
Introduction (1.5hrs)
Chromatographic Parameters (3hrs)
Band Broadening (3hrs)
Column chemistry (4hrs)
Reverse phase (partition) chromatography (6hrs)
Ion-Pair Chromatography (3hrs)
Normal phase (absorption) chromatography (3hrs)
Gradient HPLC (3hrs)
Quantitative and Qualitative HPLC (3hrs)
FAST HPLC (4.5hrs)
HILIC (3hrs)
SFC (3hrs)
Ion Chromatography(3hrs)

Theory and Instrumentation of GC
Introduction (1.5hrs)
Chromatographic Parameters (3hrs)
Band Broadening (3hrs)
Gas Supply and Pressure Control (2hrs)
Sampling Techniques (4.5hrs)
Sample Introduction (5hrs)
GC Columns (5.5hrs)
GC Temperature Programming (3hrs)
GC Detectors (2.5hrs)
SFC (3hrs)

Instrumentation of HPLC
Mobile Phase Considerations (3.5hrs)
Solvent Pumping Systems (4hrs)
Autosamplers (4.5hrs)
Detectors (4.5hrs)

Solid Phase Extraction
Molecular Properties (4hrs)
SPE Overview (3.5hrs)
SPE Mechanisms (4.5hrs)
Method Development (5.0hrs)
Primary Sample Preparation Techniques (2hrs)
Liquid / Liquid Extraction Techniques (1.5hrs) Approaches to Automation for SPE (1.5hrs)

Fundamental GC-MS
Introduction (1.5hrs)
GC Considerations (4.5hrs)
GC -MS Interfaces (2.5hrs)

Fundamental LC-MS
Introduction (1.5hrs)
Electrospray Ionisation Theory (6hrs)
Electrospray Ionisation Instrumentation (4hrs)
Mass Analyzers (9.5hrs)
Atmospheric Pressure Chemical Ionisation (3.5hrs)
Atmospheric Pressure Photoionisation (3hrs)
Solvents, Buffers and Additives (3.5hrs)
Vacuum Systems (3hrs)
Flow Rates and Flow Splitting (3hrs)
Orbitrap Mass Analyzers (3hrs)

MS Interpretation
General Interpretation Strategies (11hrs)
Intro to MS Proteomics Research (3.5hrs)

loading data
loading data
loading data
loading data
loading data
Home | About UsContact Us | SubscribeTerms and Conditions | Advertise | Privacy Policy |

loading data

loading data

loading data

 

loading data


loading data