Florida State University Libraries

Electronic Theses, Treatises and Dissertations The Graduate School

2011 Innovating Two-Dimensional Liquid Bradley J. (Bradley James) Vanmiddlesworth

Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected] THE FLORIDA STATE UNIVERSITY

COLLEGE OF ARTS AND SCIENCES

INNOVATING TWO-DIMENSIONAL LIQUID CHROMATOGRAPHY

By

BRADLEY J. VANMIDDLESWORTH

A Dissertation submitted to the Department of and Biochemistry in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Degree Awarded: Fall Semester, 2011 Bradley J. VanMiddlesworth defended this dissertation on November 2, 2011.

The members of the supervisory committee were:

John G. Dorsey Professor Directing Dissertation

Michael Ruse University Representative

William T. Cooper Committee Member

Michael G. Roper Committee Member

Hong Li Committee Member

The Graduate School has verified and approved the above-named committee members, and certifies that the dissertation has been approved in accordance with university requirements.

ii

TABLE OF CONTENTS

List of Tables v List of Figures vi Abstract ix

1. Fundamentals of Chromatographic Theory 1 1.1 Introduction to chromatography 1 1.2 The stationary phase 2 1.3 Mobile phase flow 4 1.4 Separation theory 6 1.5 The need for two-dimensional chromatography 8 1.6 Research goals 11

2. Reduction of Reequilibration Time in Gradient Reversed 12 Phase Liquid Chromatography 2.1 Introduction 12 2.2 Experimental 14 2.2.1 Reagents 14 2.2.2 Liquid chromatograph instrumentation 15 2.2.3 Liquid chromatographic method 16 2.2.4 Gas chromatograph conditions 19 2.3 Results and discussion 20 2.3.1 Column reequilibration 20 2.3.2 Offline LC-GC determination of %MeCN 24 2.3.3 Pure organic-highly aqueous interface injections 24 2.4 Conclusions 30

3. Quantifying Injection Effects in Reversed-Phase Liquid 32 Chromatography 3.1 Introduction 32 3.2 Theory 34 3.2.1 Hydrophobic-subtraction model 34 3.2.2 Acetonitrile excess absorption isotherm 35 3.3 Experimental 37 3.3.1 Reagents 37 3.3.2 Liquid chromatograph instrumentation 37 3.3.3 Liquid chromatography methods 38

iii

3.4 Results and discussion 41 3.4.1 Injection solvent sensitivity, s, of methyl ketones 41 3.4.2 Comparison to the hydrophobic-subtraction model 52 3.4.3 Comparison to the acetonitrile excess absorption isotherm 55 3.4.4 Injection solvent sensitivity of lidocaine 58 3.5 Conclusions 61

4. Simultaneous Two-dimensional Planar Chromatography 62 4.1 Introduction 62 4.2 Instrumentation 65 4.3 Reagents 69 4.4 Results and discussion 70 4.4.1 Serial vs. simultaneous study 70 4.4.2 Simultaneous separation study 72 4.4.3 Amino acid separation 77 4.5 Conclusions 78

5. Summary, Significance, and Beyond 79

References 83 Biographical sketch 90

iv

LIST OF TABLES

Table 1.1 Maximum peak capacities of a given chromatographic 9 technique.

Table 2.1 Measured void volumes of the columns studied. 16

Table 2.2 Calculated capillary pressure to wet stationary phase in 27 bar (psi) from equation 2.2.

Table 3.1 Parameters of columns used in this work, as reported by 38 manufacturer.

Table 3.2 Measured sensitivities for all columns, conditions, and analytes. 50

Table 3.3 Column and solute parameters measured by the 51 hydrophobic-subtraction model.

Table 3.4 Fitting parameters for equation 3.7. 55

Table 4.1 Migration of FD&C blue 1 from initial offline spot to 69 center-of-mass.

Table 4.2 Comparison of migration distance of methylene blue between 72 single dimension runs and simultaneous two-dimensional run.

Table 4.3 Reproducibility of FD&C blue 1 in three runs by simultaneous 76 method.

v

LIST OF FIGURES

Figure 2.1 Graphic representation of flow path (colored) and control 17 signals (lined arrows) at initial setup.

Figure 2.2 Chromatogram comparison of acetonitrile front (black trace 18 at 190nm) after switching from 100% MeCN to 100% H2O at Valve A (tA = 0.50 min) and an acetone injection (cyan trace at 254 nm) at Valve B (tinj = 0.53 min).

Figure 2.3 Truncated chromatograms depicting the effect of injecting 19 simultaneously (red trace) with mobile phase switching (tinj = tA) versus injecting at the pure organic:highly aqueous interface (black trace) (tinj = tA + 0.03 min).

Figure 2.4 Representation of the run-to-run reproducibility of injection 19 series (red, green, and blue trace).

Figure 2.5 Plot of acetone retention time vs. injection number for the 21 100% MeCN to 100% H2O runs.

Figure 2.6 Plot of acetone retention time vs. injection number for the 22 100% MeCN to 90% H2O:10% MeCN runs.

Figure 2.7 Plot of acetone retention time vs. injection number for the 22 100% MeCN to 97% H2O:3% 1-PrOH runs.

Figure 2.8 Equilibration volumes required for each column at each solvent 23 system.

Figure 2.9 Equilibration volumes required for each column at each solvent 23 system.

Figure 2.10 %MeCN of eluent as determined by GC after Valve A switches 24 the solvent.

vi

Figure 2.11 Comparison of truncated chromatograms of acetone injections 27 on the Kinetex column with ~14bar priming pressure (red) and ~360bar priming pressure (black).

Figure 3.1 Chromatograms showing the effects of injection solvent on 39 peak shape of homologous series of methyl ketones, C3-7.

Figure 3.2 Change in retention time with injection solvent strength. 41

Figure 3.3 Change in peak width at 10% height with injection solvent 42 strength with (A) constant γ0μg mass, varied volume and (B) constant 1ημδ volume, varied mass.

Figure 3.4 Change in asymmetry at 10% height with injection solvent 44 strength with constant γ0μg injection mass for the (A) Poroshell 120 EC-C18, (B) Zorbax Stablebond-C18 (C) Zorbax 300Extend-C18, and (D) Zorbax Bonus-RP.

Figure 3.5 As figure 3.4, but with constant 15μδ injection volume for the 47 (A) Poroshell 120 EC-C18, (B) Zorbax Stablebond-C18 (C) Zorbax 300Extend-C18, and (D) Zorbax Bonus-RP.

Figure 3.6 Plot of efficiency vectors showing effect of percent gradient 48 on injection solvent sensitivity, s.

Figure 3.7 Sensitivity of all columns tested as a function of (A) injection 49 volume, with a constant γ0μg mass or (B) injection mass, with a constant 15uL volume.

Figure 3.8 Comparison of sensitivity values for 30μg of 2-heptanone in 54 30uL to (A) bonding density (μmol/m2), number of sorbed acetonitrile layers, and column parameters for the hydrophobic-subtraction model, and (B) to ratios of column parameters, H/A.

Figure 3.9 Measured acetonitrile excess isotherms from equation 3.4a 55 and 3.4b.

Figure 3.10 Correlation between K(C18)/K(OH) and sensitivity for values 57 of γ0μg of β-heptanone.

Figure 3.11 δidocaine peak shape with changes in the 1ημδ of 80% εeCζ 59 injection solvent onto the Poroshell column.

Figure 3.12 Lidocaine peak shape with changes in the injection volume 60 of 10% MeCN injection solvent onto the Bonus-RP column.

vii

Figure 4.1 Instrumentation of the 2D planar chromatography setup. 64

Figure 4.2 Qualification of simultaneous 2D planar chromatography 68 instrumentation.

Figure 4.3 Single-dimensional migrations of methylene blue in 75% MeOH 72 and 25% 10mM acetate buffer on reversed phase plates are comparable to the component migrations in simultaneous mode.

Figure 4.4 MATLAB graph of pixel data using the described simultaneous 74 2DPC technique.

Figure 4.5 Spots remain unresolved using A) and 75 B) conventional TLC.

Figure 4.6 Migration distance of blue 1 as a function of run time. 76

Figure 4.7 4.0 minute separation of the four amino acids.1 – histidine, 77 2 – arginine, 3 – lysine, and 4 – alanine.

viii

ABSTRACT

Liquid chromatography is ubiquitous, but two-dimensional liquid chromatography is rare. The difference is in the difficulty of method development, as once floated variables have increased pertinence. Three issues for method development in two-dimensional liquid chromatography have been described previously: 1) To preserve the resolution produced in the first dimension, an eluting analyte peak from the first dimension must be sampled at least four times across the band width. 2) The mobile phase of the first dimension becomes the injection solvent for the second dimension. 3) Each additional dimension adds further dilution of the analyte band. It is the goal of this research to fundamentally describe and mitigate the hurdles of two- dimensional liquid chromatography while concurrently adding to the overall practice of chromatography. Three investigations have been undertaken. The first describes the elucidation of pressure as the limiting factor in column reequilibration post-gradient for the goal of reducing the time of analysis of a second dimension. The second quantifies the effect of band shape distortion due to injection solvent mismatch with the goal of making predictions for column selection and organic modifier choice. The third is a proof-of-concept for a novel approach to separations where two force vectors are applied simultaneously to produce a two dimensional separation in less time and therefore, less dilution. Taken separately, these three are of general use to fundamental separation science, but collected they are of specific use to reduce the difficulty of method development of two-dimensional liquid chromatography.

ix

CHAPTER ONE

FUNDAMENTALS OF CHROMATOGRAPHIC THEORY

1.1 Introduction to chromatography It is of general interest to quantify or qualify a property of a substance of a mixture in the realm of science. Numerous detectors have been designed to target specific properties – mass detectors, spectroscopic detectors, conductivity detectors, etc. When more than one component of a mixture produces a detector response, it is desirable to separate the components to ensure the validity of the measurement. Separation techniques are characterized by the property that defines the mass transfer process. Distillation separates liquids via differing boiling points, centrifugation via density, electrophoresis via the electrophoretic mobility, and extraction via the solvation energy difference of the substance in two immiscible phases. Chromatography is form of continuous extraction that separates components via the thermodynamic solvation or affinity of a substance between two immiscible phases and the kinetics of the mobility through a retarding medium. Given the kinetic component of the separation mechanism, the two immiscible phases in chromatography must have differing velocities relative to each other, defining one as stationary and the other mobile, each with a relative affinity for the substances to be separated, here after called the analytes. The state of matter of the mobile phases defines the first segmentation of chromatography, and so one may discuss gas, supercritical fluid, or liquid chromatography. Theoretically, these three techniques are similar. The instrumentation required for these techniques are sufficiently different and thus the usefulness of the terms. Stationary phases are generally solids or liquids on a

1 solid support, and define the second useful segmentation of chromatography from the mechanism of reducing the velocity of the analyte, here after called the retention mechanism. Ion exchange stationary phases interact through electrostatic interactions with ionized analytes, chiral stationary phases interact through the handedness of the analytes, and normal and reversed phase both retain analytes via the polarity though from opposite ends of a polarity spectrum. Normal phase chromatography interacts with polar analytes through dipole-dipole and hydrogen bonding interactions, whereas reversed phase interacts through dispersion forces or more generally called hydrophobicity. Normal and reversed phase separations are the most frequent modes of chromatography due to the ubiquity of the forces of retention, the ease of tuning the separation, and the relatively low strength of the fundamental retention interactions. For separations of small molecules, reversed phase liquid chromatography is the most commonly used technique.

1.2 The stationary phase Physically, a stationary phase is generally a ligand with the desired retention interaction covalently bonded to a solid support, most frequently a polysiloxane. The polysiloxane can be a tube of crystalline quartz or a particle of amorphous silica, whether porous or non-porous. The type of polysiloxane chosen is a function of the flow properties of the mobile phase. For reproducible migrations of an analyte within a fluid, the flow must be laminar and not turbulent. The Reynolds equation describes a correlation between velocity and the onset of turbulence for a non-compressible, Newtonian fluid forced through a tube [1]: uD Re  (1.1a)  where Re is the unitless Reynolds number, u is the linear velocity of the fluid in cm/s, ρ is the density of the fluid in g/mL, D is the tube diameter through which the fluid is flowed, and is the viscosity in poise or g/(cm-s). Alternatively, for a fluid forced through a packed bed [1], ud Re  p (1.1b)  1( )

2 where dp is the Sauter mean particle diameter and is the porosity of the bed. Turbulence occurs when the Reynolds number is above some threshold value defined by the resistance to flow. For open tubes, the threshold is high, ~2100, whereas in a packed bed the onset of turbulence generally occurs between 1 and 100, as a function of the regularity of the packing order and sphericity of the packed particles. A frequent goal is to maximize the surface area of the stationary phase for which the intermolecular interaction is to take place, and can be accomplished by extending the length of a tube, or by using a packed bed of porous substrate. With low-density, low-viscosity fluids it is advantageous to use a greater length of an open tube to allow for greater flow rates. This is the case for which frequently has a column length in the range of 10 – 30m. Conversely, liquid will use a packed bed of porous silica with column lengths in the range of 3 – 25cm. Supports other than silica are commercially available as packing, such as zirconia, polystyrene-divinylbenzene, or a monolithic polymer. Silica was chosen as the usual support due to the abundance, chemical stability, physical robustness, and the well understood reaction mechanism for ligand bonding [2]. Covalently bonding the ligand to the support requires a derivatization reaction. In reversed phase liquid chromatography, the most frequent ligand is a carbon chain of some length that is terminated in a siloxane bond to the solid support. The ligand-attached silicon generally has two additional functional groups of small carbon number, termed side-groups. Methyl, isopropyl, or isobutyl groups are frequent, as well as bridging two ligands via an ethylene. However, the long carbon chain is responsible for the majority of the stationary phase-analyte interaction in comparison to the effects of the side group. Within the stationary phase synthesis reaction, ligand chlorosilanes are base-catalyzed silanized to surface silanols. The concentration of surface silanols has been measured to be ~8μmol/m2 and is considerably heterogeneous in regards to isolated, vicinal, and geminal silanols [3]. Due to the bulk of the carbon-chain ligand, the available surface silanols that can be derivatized is ~4μmol/m2, leaving 50% as a remainder. These unreacted silanols lead to a mixed chromatographic retention mechanism, and it has been posited that reduction of these silanols could improve the resulting chemical separation. Further derivatization of these remaining silanols can be done with a small molar volume

3 ligand, trimethyl chlorosilane, however only about half of the remaining silanols can be reacted, leaving ~25%. Analytes with strong interactions to these unreacted silanols, namely acids and bases, will show a mixed retention mechanism. An additional method to reduce surface silanols involves using ethylene linked silicates within the polycondensation reaction to form the support to extend the distance between surface silanols, and thus reduce the steric hindrance to increased bonding density of the ligand. However, the drawbacks to using silica as a support are the acid-catalyzed hydrolysis of the ligand siloxane, base-catalyzed dissolution of the substrate, and the thermal stability of the ligand-silane bond [2]. These limit the three useful chromatographic parameters in separations of ionizable compounds.

1.3 Mobile phase flow Mobile phases are fluids that can flow by a force and will solvate the analyte. The three flow methods used are electroosmotic flow via an electric gradient, capillary action via intermolecular attraction, and mechanical displacement via a pump or compressed gas cylinder. Within liquid chromatography, the most common flow method uses a positive displacement, dual-piston pump to deliver a constant velocity of fluid. The pressure required to force a fluid through an open tube with a velocity v is described by the Hagen-Poiseuille equation [1]: Pr 4 v  (1.2) 8L where v is the volume flow rate in mδ/min, ΔP is the pressure gradient required, π is the ratio of the circumference and the diameter of a circle, r is the radius of the open tube, and L is the length through which the fluid is flowed. Through a packed bed, the pressure required is a function of the packing order, as described by Darcy‘s law [4]: KA P v   (1.3)  L where A is the cross sectional area of the packed bed and K is the permeability in m2. Permeability has no physical meaning and is solely a proportionality constant to describe the resistance to flow of a packed bed. A correlation to packing structures is found in the Kozeny-Carman equation [4]:

4 2 3 d p  K  2 2 (1.4) kC 1( )  where Φ is the sphericity of the packing and is equal to unity for packed spheres. kC is the Kozeny-Carman coefficient and is approximately 180 for a randomly packed bed. Regarding the remaining two flow methods, electroosmotic flow and capillary action do not use a dynamic pressure gradient to migrate the mobile phase. Electroosmotic linear flow rate is modeled by the Smoluchowski equation:  E   ro (1.5) EOF  where μEOF is the linear velocity in cm/s, o is the permittivity of a vacuum in F/cm, r is the unitless relative permittivity of the mobile phase, ξ is the zeta potential in V, and E is the applied electric field in V/cm. In contrast to pressurized flow, electroosmotic flow through a packed bed is not a function of the particle diameter or packing structure. As well, in electroseparations, the analyte has an intrinsic velocity ue relative to the mobile phase: qE u   E  (1.6) e e 6r where μe is the electrophoretic mobility, q is the analyte charge in coulombs and r is the ionic Stokes radius in . These electrophoretic velocities of the bulk mobile phase and the analyte are additive, advancing the velocity of one charge and retarding the other dependent on the wall charge. Generally, with silica, the wall charge is negative and μEOF is in the direction toward the negative electrode. Regarding capillary action, the linear velocity u can be described as [5]:  K d cosθ  p0 u  (1.7) Zf where is the surface tension in ζ/m, K0 is the specific permeability of the thin layer in 2 m , Zf is the distance traveled by the solvent front in cm, and θ is the unitless contact angle between the solvent and stationary phase. Contact angle is a measure of the surface energy at the liquid-solid-air interface. In the event where a constant flow velocity is desired, capillary action is not an optimum flow process as it is a function of time and

5 solvent parameters. However, the dynamic flow rate is reproducible and is useful in conventional thin-layer chromatography as a quick test that requires little instrumentation or prior handling.

1.4 Separation theory To separate two analytes within a chromatographic system, the migration velocities must be different within the system. With differing velocities, the analytes can be separated in distance, such as in thin-layer chromatography, or time, such as in elution chromatography. In both, the analyte velocities are retarded by the retention mechanism such that we can define a unitless RF [5]:

da R F  (1.8) Zf where da is the migration distance of the center of maximum concentration of the analyte in cm. With elution chromatography analytes travel the same distance in a differing amount of time, thus it is useful to consider a unitless retention factor k‘ [4]: t  t k' R o (1.9) to where tR is the time of elution of the center of maximum concentration in minutes and t0 is the void time of the column in minutes. The two factors are relatable through [5]: 1 R  (1.10) F 1 k' Retention is a function of the solvation energies of the analyte in the mobile and stationary phase, related though the van‘t Hoff equation: - ΔH ΔS ln(k')  lnK     ln (1.11) RT R where K is the unitless for the sorption of the analyte to the stationary phase, ΔH and ΔS are the enthalpic and entropic contribution to retention, respectively, in J/mol and J/(mol-K), respectively, R is the gas constant in J/(mol-K), T is the absolute temperature in K, and Φ is the ratio of the volume of the stationary phase to the volume of the mobile phase. Partition chromatography is primarily an enthalpic driven process and therefore the temperature must be controlled for reproducible retention.

6 If the retention factors of two analytes are different, then the concentration maxima are separated and a selectivity of the stationary phase α can be described as the ratio of the retention factor of the later eluting analyte to the retention factor of the earlier eluting analyte; greater or equal to unity as defined. However, this suggests nothing of the separation of the concentration band of each analyte, termed resolution. The concentration band of the analyte is commonly called a peak in reference to the detector output graph. A series of peaks from a single run is a chromatogram. A resolution RS can be described as [4]:

2(tR,2  t R,1) RS  (1.12) W2  W1 where W is the band width in mins, and the subscripts 2 and 1 describe the parameters for the later eluting and earlier eluting peaks, respectively. Plate theory describes the relative widening of analyte peaks as a function of time spent within the system [4]: (t )2 N  R (1.13) σ2 where N is the unitless theoretical plate number describing the relative widening, and σ2 is the variance of the peak. Numerous equations exist to estimate the statistical variance of the distribution as a function of the width [6]. As tR is a function of the length of the column L, it is often desirable to normalize plate count via a plate height H in μm [4]: L H  (1.14) N Plate height can be shown as a function of the linear flow rate via the [4]: B H  A   Cu (1.15) u where A, B, and C are constants describing the independent, indirect, and direct contributions of the linear flow rate to the plate height, respectively. Using dynamic rate theory and assuming the underlying process variances are independent, these constants can be used to describe physical phenomena within a packed column. A can be modeled as a function of the packing structure of the bed and the varying flow streamlet velocities from column inlet to outlet. B is modeled as a function of the longitudinal diffusion rate of the analyte. C is modeled as a function of the kinetics of the transfer of the analyte

7 from one phase to the other. The kinetics of transfer can be described as a function of the particle diameter dp and distance penetrated in the porous support df [4]:

2 d d 2 C  p  f (1.16) Dm Ds where Dm and Ds are diffusion coefficients of the analyte within the mobile phase and stationary phase, respectively. Given equations 1.10, 1.12, and assuming a constant distribution width between two similarly retained analytes, the Purnell equation describes the methods that can increase resolution [5]:

 N  α 1 k'  R     2  (1.17) S      4  α 1 k'2  Plate count can be increased in a variety of methods, but the three most common changes are to lengthen the column, reducing the particle diameter, or reducing the diffusion distance into the particle via a superficially porous silica-based column. Selectivity is a function of the retention mechanism and is generally adjustable by changing the stationary phase ligand, though is the most effective method to increase resolution.

Increasing k‘2 involves weakening the mobile phase relative to the stationary phase. Adjusting the mobile phase strength is only effective for values where 0 < k‘ < η, but is the simplest adjustment to a method. Alternative to reducing the strength of the mobile phase, it is advantageous in time to adjust the strength dynamically during the separation, called gradient elution chromatography. The disadvantage to running a solvent gradient is the requirement to return the stationary phase to initial conditions, as discussed in chapter 2.

1.5 The need for two-dimensional chromatography In 1970, Eli Grushka [7] derived an equation to model the number of analyte peaks that can be physically resolved with a resolution of unity as a function of theoretical plate count and retention faction, called the peak capacity PC: N P 1 ln1 k'  (1.18) C 4 last

8 With optimum separation conditions, the maximum peak capacity of a given technique is listed in table 1.1.

Table 1.1. Maximum peak capacities of a given chromatographic technique. Isocratic Liquid Chromatography ~100 Gradient Elution Liquid Chromatography ~200 Temperature Programmed Gas Chromatography ~1000 High Performance Thin-layer Chromatography ~25

In 1983, Davis and Giddings described the statistical theory of component overlap which described the peak capacity required to separate n number of components randomly distributed across the chromatographic run time [8]. It was suggested that to have a 90% chance of separating all components of a random mixture, the peak capacity must be 19 times greater than the number of components. For simple, non-random samples, this requirement had not been necessary, and thus the growing ubiquity of chromatography to isolate components. However, for sufficiently complex mixtures such as a protein digest or oil sample, it is desirable to increase the peak capacity of the system. The most powerful method of increasing peak capacity is to add a second, orthogonal retention mechanism at the exit of the first, called two-dimensional (2D). If the entirety of the effluent from the first dimension is modulated to a second column, it is called comprehensive two-dimensional chromatography. In comprehensive two- dimensional chromatography, the resulting peak capacity of the system is a product of the individual dimensions. Sander and coworkers have published a recent review of the practical and theoretical considerations within comprehensive two-dimensional liquid chromatography [9]. Cohen and Schure [10] have written a book discussing sample specific method development in multidimensional chromatography. For two-dimensional liquid chromatography, the instrumentation requires two pumps, two columns, two valves, and one detector. The first dimension pump flows through the first injection valve with the sample to the first column. The eluate from the first column is collected in one of two injection loops on a modulation valve prior to the second column. Once filled, this second valve is switched to send a plug of first column‘s

9 eluate to the second column. The eluate now coming from the first column is directed to the second of the two injection loops. The second pump flows through the modulation valve with injection loops to the second column, then to the detector.

1 Injmax vmax  2 (1.19) tA The three method variables of equation 1.19 are the total run time of the second 2 dimension ( tA), the injection loops‘ volume (Injmax), and the maximum flow rate of the 1 first dimension ( vmax). The time required for the analyte with the largest retention in the second column to migrate from the modulation valve to the detector defines the run time of the second dimension. It is common to use gradient elution chromatography to reduce this time, but the column must be reequilibrated to the initial gradient conditions. The sum of the run time and the time required for the reequilibration is the total time for the second dimension and the modulation time of the modulation valve. The largest volume of first dimension eluate that can be loaded onto the second column without overloading gives the injection loops‘ volume. The injection loop volume divided by the total time for the second dimension gives the maximum flow rate of the first dimension. The migration time of the analyte with the largest retention on the first column and second column now defines the total time of analysis for the two-dimensional run. If the goal of the analyst is to maximize the peak capacity of the two dimensional system, and given equation 1.18, then the goal is to maximize efficiency of both dimensions holistically. Equation 1.15 describes the change in efficiency with the flow rates of each dimension. There has been no fundamental, quantitative study describing the injection loops‘ volumes effect on efficiency. Three significant difficulties arise in the method development of two-dimensional liquid chromatography – 1) To preserve the resolution produced in the first dimension, an eluting analyte peak must be sampled at least four times across the band width, called the Murphy-Schure-Foley criterion [11]. 2) The mobile phase of the first dimension becomes the injection solvent for the second dimension. 3) Each additional dimension adds further dilution of the analyte band, which is of concern for analytes near the limit of detection. Regarding (1), the second dimension is frequently a short, thin column running a steep solvent gradient to maintain retention times of less than a few minutes. Regarding

10 (2), injection solvent effects produce wider analyte concentration bands due to a dynamic strength of the mobile phase, as discussed in detail in chapter 3. Additionally, this has precluded the use of gradient elution in the first dimension. Regarding (3), the dilution factor of a two-dimensional technique is a product of the dilution factors of the individual dimensions plus the dilution within the connecting tubing.

1.6 Research Goals It is the goal of the research herein to fundamentally describe and mitigate the hurdles of two-dimensional liquid chromatography while concurrently adding to the overall practice of chromatography. Chapter 2 describes the reduction of reequilibration time after using a steep gradient elution in a second dimension column by using pressure of the reequilibration solvent as a variable. Chapter 3 describes the effect of injection solvent mismatch and how column selection can minimize the mixed retention that ensues. Chapter 4 describes a novel attempt at two-dimensional chromatography that applies two orthogonal force vectors to analytes simultaneously, removing all three listed difficulties above.

11

CHAPTER TWO

REDUCTION OF REEQUILIBRATION TIME IN GRADIENT ELUTION REVERSED PHASE LIQUID CHROMATOGRAPHY

2.1 Introduction In liquid chromatography, gradient elution has two main advantages over isocratic elution: sharper peak shape and reduced run time. With sharper peak shape, peak overlap is reduced and the limit of detection and the limit of quantitation are lowered (improved). Shorter run times allow for more analyses in a set time frame, which is of particular interest within comprehensive two-dimensional chromatography where the speed of the second dimension is generally the limiting factor for the total analysis time [12]. The disadvantage of gradient elution is the required time post-gradient to flush initial mobile phase composition through the column to ensure reproducible retention times of analytes in the subsequent injection [13]. The lack of reproducibility of retention times post-gradient is due to the extent of solvation of the alkyl chains of the stationary phase by the organic modifier component in the mobile phase. In order for retention to be equivalent run-to-run, the alkyl chains must be returned to the initial solvation pre- gradient. Ideally, the initial mobile phase would remove the residual gradient solvent and after some reequilibration time the eluent exiting the column would be of the same composition as the mobile phase entering the column. In 1982, Gilpin et al. [14,15] used offline LC-GC to quantitate the release of alcohol from a C-10 phase when flushed with pure water. It was shown that a 1 – 10ppm concentration of organic modifier was still detectable after flushing the column with ~600mL of pure water, approximately

12 ~850 column volumes, with a column dimension of 25cm x 2.4mm i.d. and assuming a porosity of 0.6. Ideal conditions for gradient elution cannot be attained with real world analysis times, though a minimal amount of organic modifier can remain on column and still achieve reasonable run-to-run reproducibility. This is analogous to the proposed definition of Schellinger, Stoll, and Carr [16] that full equilibration occurs when a column provides reproducible retention times for all solutes independent of the reequilibration time. However, most chromatographic analyses do not require this state of full equilibration; rather there need only be an acceptable run-to-run reproducibility. Run- to-run reproducibility is of particular importance to comprehensive two-dimensional chromatography when aligning the sampling phase [17] of sequential chromatograms to visualize a contour plot of the separation space. To determine a necessary reequilibration time, two methods have been devised. Cole and Dorsey [18] injected a weakly retained analyte (acetone, 0 < k' < 1) each minute, post-gradient, once the mobile phase composition was adjusted at the proportioning valve. Because acetone is weakly retained, its retention time should vary measurably with stationary phase solvation. Once the retention of acetone reached a constant value the column was considered equilibrated. "Constant value" or "equilibrated value" was defined later by Coym and Roe [19] as within 0.01 minutes of the average value of retention times for injections where the column is assumed to be equilibrated. Cole and Dorsey found that the inclusion of a set amount of n-propanol (3%) significantly reduced the reequilibration time by robustly solvating the stationary phase. The second method of determining necessary reequilibration time was described by Schellinger, Stoll, and Carr [16] wherein a series of gradient runs were performed sequentially with varying reequilibration times interspersed. The retention times of the analyte mixture were compared to a control run with a 15 minute reequilibration time. With the use of a second switching valve before the injection valve, the dwell volume of the pump could be bypassed, ensuring that the variation of the retention times was a function of stationary phase solvation and not flushing out the volume of the pump. It was shown that a very small amount of mobile phase (1 – 2 column volumes) was necessary to produce acceptable reproducibility of retention time (<±0.002min) for neutral analytes on the fully porous columns studied [16]. Notably, acetone was the

13 mixture component that consistently had the greatest variability with reequilibration time. For basic analytes, it was later found that 5 column volumes of initial buffered eluent were enough for full equilibration [20]. It was suggested that a new rule-of-thumb for reequilibration volume is the sum of dwell volume of the system and 1-2 column volumes, or 5 column volumes if using buffered mobile phase. Recently, columns packed with superficially-porous silica with a particle diameter of < γμm have become commercially available and provide significant performance within the pressure limitations of conventional instrumentation (400bar) [21-24]. The media within these columns have a solid, non-porous silica core with a particle diameter of 1.7μm (Agilent Poroshell, AεT Halo) or 1.9μm (Phenomenex Kinetex). Surrounding the solid core is a porous silica layer with a thickness of 0.ημm (Agilent Poroshell, AεT Halo) or 0.γημm (Phenomenex Kinetex). The enhanced performance is due to decreased resistance to mass transfer kinetics from the lesser diffusion distance into the particle as well as a narrower particle size distribution when compared to fully porous silica [25,26]. In this work, we use the Cole-Dorsey method to determine necessary reequilibration time of superficially-porous silica with a reduced dwell volume system. Given the smaller diffusion distance for residual acetonitrile (MeCN) sorbed to the stationary phase, we would expect the columns to reequilibrate faster than fully porous columns if the rate-limiting step is the diffusion of acetonitrile from the stationary phase. With offline LC-GC, we quantify the acetonitrile content of the eluent and compare superficially-porous with fully porous silica.

2.2 Experimental 2.2.1 Reagents All water used was purified to a resistance of approximately 18εΩ-cm using a Barnstead (Dubuque, IA, USA) NANOPure Diamond water purification system. HPLC grade acetonitrile (MeCN) and reagent grade 1-propanol (1-PrOH) were obtained from Sigma-Aldrich (St. Louis, MO, USA). HPLC grade acetone and sodium nitrate were obtained from Fisher Chemicals (Fair Lawn, NJ). Mobile phases were prepared by mixing the appropriate volumes of MeCN or 1-PrOH, and H2O then vacuum filtered

14 through 0.45m filters prior to use. 0.5% (v/v) acetone samples were prepared by mixing

5mL with 1L of H2O then filtered. Sodium nitrate samples were prepared mixing an amount with 300mL of water, then diluted until the detector signal was ~70mV without a column. All retention times obtained were the average of three series of repetitive acetone injections. System peaks are identified from a blank run, where no acetone injections were made.

2.2.2 Liquid chromatograph instrumentation Three Shimadzu pumps (Kyoto, Japan) were used for this study: two LC-10ATVP isocratic pumps and one LC-10ADVP gradient pump outfitted with a DGU-14A inline degasser and an FCV-10ALVP quaternary proportioning valve. One Valco (Houston, TX, USA) E90 four-port automated switching valve and a Valco E60 six-port automated switching valve were used to reduce system dwell volume and make well-timed, repetitive injections. System volumes measured were Valve A to detector (9θ.0±0.0μδ, n=θ), injection loop to detector (7θ.β±0.4μδ, n=θ), and injection loop to column (14.40±0.0βμδ, n=γ). εeasurement of the injection loop to column volume was done by acetic acid titration. First, the volume was filled with glacial acetic acid using a 60 mL plastic syringe. Then, with air from another, empty syringe, the acetic acid was expelled from the volume into 10mL of water. This solution was then titrated to endpoint with 0.010 M NaOH. Acetic acid volume was calculated from the inflection point of the titration curve, the density of glacial acetic acid (1.049 g/mL), and the molar mass (60.05g/mol). Pumps and valves were synchronized by an SCL-10AVP System controller. A Waters (Milford, MA, USA) 486 tunable wavelength UV-Vis detector was used with a wavelength of 254nm for acetone or 190nm for acetonitrile. All data were collected using a Perkin Elmer Nelson (Waltham, MA, USA) 970A integrator set to 25Hz per channel and TotalChrom 6.2.1 software for analysis. Four stationary phases were evaluated: Halo β.7μm 90Å (AεT, Inc.), Kinetex β.θμm 100Å (Phenomenex, Inc.), Poroshell β.7μm 1β0Å (Agilent Technologies), and Zorbax γ.ημm γ00Å (Agilent Technologies). The first three columns listed are superficially-porous, whereas the last column is fully porous. Column dimensions were each 4.6mm i.d. x 100mm length. Column void volumes are

15 reported in table 2.1. Each method of void volume measurement has advantages and disadvantages, as outlined and discussed in reference [27]. Static void volume was determined by pycnometry using acetonitrile and chloroform as the solvents. Kinetic void volumes were determined by the average of triplicate injections of uracil. For 100% H2O runs, the column was conditioned for 4 hours at 1.00 mL/min prior to injection. For all other solvents, the column was conditioned for 1 hour at 1.00 mL/min. With exception for the Kinetex column, the static void volumes agree well with the 100% MeCN kinetic void volumes, but the high aqueous solvents produced values that are unreasonably large, assuming well packed columns. For the Kinetex column, all kinetic voids were unreasonably large.

Table 2.1. Measured void volumes of the columns studied. Acetonitrile and chloroform were used for pycnometry. Kinetic voids were calculated from the elution time of the peak maximum for an injection of uracil. For each standard error, n = 3. Static Void Kinetic Void 90%:10% Pycnometry 100% H2O H2O-MeCN 100% MeCN

V0 (mL) σ (mδ) V0 (mL) σ (mδ) V0 (mL) σ (mδ) V0 (mL) σ (mδ) Zorbax 300 EC-18 1.025 0.016 1.468 0.008 1.065 0.018 0.953 0.006 Poroshell 120 0.997 0.018 1.653 0.010 1.084 0.011 1.001 0.010 Kinetex 2.6um 0.907 0.020 3.227 0.012 1.767 0.006 1.802 0.015 Halo 0.848 0.025 1.367 0.007 1.091 0.011 0.831 0.007

2.2.3 Liquid chromatographic method Figure 2.1 graphically outlines the mobile phase flow path and control signals of the instrument. The valve positions shown are an initial setup with Valve A in position 1 and Valve B in the load position. The Final Eluent pump (LC-10ATVP) equilibrates the column with a high organic solvent (100% MeCN or 97% MeCN:3% 1-PrOH) at 1.00 mL/min for 29.97mins, simulating the end of a gradient run. At t = 29.97mins, the System Controller signals Valve A to switch to position 2 (switching time = 0.0093mins) directing a highly aqueous solvent (100% H2O, 90% H2O:10% MeCN, or 97% H2O:3% 1-PrOH) from the Reequilibration pump (LC-10ADVP) to the column at 1.00 mL/min. Concurrently, the Injection pump (LC-10ATVP) refills the injection loop with sample continuously at 0.20 mL/min while Valve B is operating (switching time = 0.0093mins).

16 11.9 ± 0.ημδ injections start at t = γ0.00mins and continue on the minute mark until t =

49.00mins for 90% H2O:10% MeCN and 97% H2O:3% 1-PrOH (20 injections) or until t

= 79.00mins for 100% H2O (50 injections). Injections are marked relative to the beginning of injections, i.e., the 0th injection occurs at t = 30.00mins and the 1st injection occurs at t = 31.00mins, 2nd at t = 32.00mins, etc. The 0.03mins between Valve A switching and Valve B injecting allows for the volume between the two valves (19.94 ± 0.50μδ) to fill with the aqueous reequilibration solvent and make the injection where the center of mass occurs at the interface of the two mobile phases: pure organic and highly aqueous. Figure 2.2 shows the center of mass of the acetone injection with the 0.03min offset aligns with the inflection point of the acetonitrile front when the column is replaced with a zero-dead-volume fitting. Figure 2.3 shows the effect the 0.03min offset injection has on the retention and peak shape of the 0th and 1st injections of acetone following column equilibration with the pure organic mobile phase on the Zorbax column. Figure 2.4 depicts the typical reproducibility of the three injection series.

Figure 2.1. Graphic representation of flow path (colored) and control signals (lined arrows) at initial setup. Dark blue - tubing containing pure organic phase. Cyan - tubing containing highly aqueous phase. Red - tubing containing 0.5% acetone sample.

17 The accurate timing of the pumps and valves is controlled through the system controller, but the detector output with time is plotted through the integrator. The time difference between the two clocks is small, but variable. The position signal of Valve B is output to the integrator in channel B to correct for the time difference. With the ground wires of the event cable for both valves intertwined, the proper switching time of each valve can be read in the single channel: positive voltage for Valve B and negative voltage for Valve A. The time of mobile phase switching (Valve A) as read by the integrator was compared to the programmed time within the system controller, and the difference added or subtracted as necessary. All analyte retention times were corrected for the extra- column volume, the time delay/advance of the integrator, and the minute-increment injection number.

Figure 2.2. Chromatogram comparison of acetonitrile front (black trace at 190nm) after switching from 100% MeCN to 100% H2O at Valve A (tA = 0.50 min) and an acetone injection (cyan trace at 254 nm) at Valve B (tinj = 0.53 min). Column replaced with zero- dead volume fitting.

18

Figure 2.3. Truncated chromatograms depicting the effect of injecting simultaneously (red trace) with mobile phase switching (tinj = tA) versus injecting at the pure th organic:highly aqueous interface (black trace) (tinj = tA + 0.03 min). 0 injection is denoted as 00 Acetone, and 1st as 01 Acetone. Note the difference in peak shape and retention time. Mobile phase: 100% MeCN stepped to 100% H2O, Column: Zorbax 300Extend-C18.

Figure 2.4. Representation of the run-to-run reproducibility of injection series (red, green, and blue trace). System peaks are identified from the blank run (black trace) where no acetone injections were made. Mobile phases: 100% MeCN stepped to 100% H2O, Column: Zorbax 300Extend-C18.

2.2.4 Gas chromatograph conditions A Varian model CP-3800 gas chromatograph (Foster City, CA) with a flame ionization detector (FID) was used without modification. Carrier gas was set to 1.2psi and the make up gas for the FID was air. The column used was a Varian CP-SIL 5

19 CB (model CP8η10) with dimensions of γ0m x 0.βηmm (ID) x 0.βημm (film thickness). Temperatures were set at 200°C for the injection port, 220°C for the FID, and an isothermal 40°C for the column oven. Samples were collected in ~1mL increments from the outlet of the LC system via a modified ISCO Combiflash Companion (Lincoln, NE). The internal metering pumps, column, and UV detector of the Companion were bypassed and eluent passed directly to the fraction collector via a Valco manually operated six-port valve. Control of the fraction collector was done by the PeakTrak software, version 2.2.35 (ISCO). An internal standard of 25μδ of HPδC grade methanol was added to an 0.800mδ aliquot of each collected fraction. Calibration standards for the acetonitrile quantitation ranged from 0.βημδ to 10μδ acetonitrile diluted to 0.800mδ, then βημδ εeOH added. The coefficient of linearity (R2) value for the calibration curve was 0.9997 for the entire range of standards, and 0.9980 for the range between 0.βημδ and βμδ acetonitrile. Acetonitrile content of eluent samples are an average of three injections.

2.3. Results and Discussion 2.3.1 Column equilibration Figures 2.5 – 2.7 show the retention times of the acetone injections for the four columns and the three solvent systems studied. To determine whether a column could be considered equilibrated, for each injection number, n, the slope of a best fit line

 y  connecting that point and the subsequent two points,   , is compared to the 99%  x n confidence interval of the slopes calculated for the final 10 points, 1017  .2 5758 , hereafter referred to as ‗slope cut-off‘. So, if

 y     1017  2.5758  (2.1)  x n then the column is considered equilibrated. Here, the 2.5758 value is half the number of standard deviations that cover 99% of the area under a normal distribution, called Z99% in statistics. For an equilibrated column the slope should equal zero. The actual values differ for each column, and therefore the reequilibration volumes reported in Figure 2.8 compare the slope cut-off the column achieved given the conditions. Figure 2.9 gives the

20 reequilibration volumes using the largest slope cut-off from the four columns within the given reequilibration solvent composition to compare performance between columns. In agreement to the reported rule-of-thumb [16,20], equilibration occurs quickly (< 3mL) for each column and solvent system. There is no obvious correlation between superficially-porous silica and a reduced reequilibration volume, which is somewhat surprising, but may be due to the intrinsic 1-minute resolution of the reequilibration test.

8.0

7.0 Poroshell Kinetex 6.0 Zorbax Halo 5.0

4.0 Retention time (min) 3.0

2.0 0 5 10 15 20 25 30 35 40 45 50 Injection number

Figure 2.5. Plot of acetone retention time vs. injection number for the 100% MeCN to 100% H2O runs. Each point is an average of 3 injections with the standard deviation as error bars. For the Kinetex column the 0th injection is split. Both retention times are plotted. Closed symbols – ~14 bar priming pressure. Open symbols – ~300bar priming pressure.

21 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 Retention time (min) 1.0 Poroshell Kinetex 0.5 Halo Zorbax 0.0 0 5 10 15 20 Injection number

Figure 2.6. Plot of acetone retention time vs. injection number for the 100% MeCN to 90% H2O:10% MeCN runs. Each point is an average of 3 injections with the standard deviation as error bars.

5.0 Poroshell 4.5 Kinetex 4.0 Zorbax 3.5 Halo 3.0 2.5 2.0 1.5 Retention time (min) 1.0 0.5 0.0 0 5 10 15 20 Injection number

Figure 2.7. Plot of acetone retention time vs. injection number for the 100% MeCN to 97% H2O:3% 1-PrOH runs. Each point is an average of 3 injections with the standard deviation as error bars. For the Kinetex and Poroshell columns the 0th and 1st injections co-elute. Closed symbols – ~14 bar priming pressure. Open symbols – ~300bar priming pressure.

22

4 Halo Poroshell Kinetex Zorbax

3.3 2.9 3

7.0 3.0 8.8 4.6 18.3 10.6 2

22.7 11.3 25.5 21.9 1 Reequilibration volume (mL)

0 100% H2O 90% H2O 3% 1-PrOH Reequilibration eluent

Figure 2.8. Equilibration volumes required for each column at each solvent system. The slope cut-off values above the bars are reported and have been reduced by a factor of 10- 3. ~14 bar priming pressure.

4 Halo Poroshell Kinetex Zorbax

3

22.7 21.9 21.9 2

22.7 22.7 22.7 25.1 25.1 25.1 25.1 21.9 21.9 1 Reequilibration volume (mL)

0 100% H2O 90% H2O 3% 1-PrOH Reequilibration eluent

Figure 2.9. Equilibration volumes required for each column at each solvent system. The slope cut-off used is of the largest any column achieved within the solvent system and have been reduced by a factor of 10-3. ~14 bar priming pressure.

23

1

0.1 %MeCN

0.01 4 9 14 19 24 Volume from valve A (mL)

Figure 2.10. %MeCN of eluent as determined by GC after Valve A switches the solvent. The main graph is truncated to show only data within the calibration range with a logarithmic scale; the insert plots the entirety of the data linearly.

2.3.2 Offline LC-GC determination of %MeCN Figure 2.10 shows the %MeCN of the eluent per column once Valve A has been switched, as measured by GC. Between 0 – 7mL the superficially-porous columns retain the largest amount of residual organic modifier, but at 8mL the fully-porous column retains a slightly, yet statistically significant, greater amount. The internal plumbing of the fraction collector is not subtracted, making the interface occur at ~4mL in Figure 2.10. Of the columns, Poroshell and Kinetex have the fastest decrease in eluent %MeCN, whereas Halo and Zorbax gradually release the organic. Yet all four elute effectively the same amount once the initial washing of the phase is complete. The elution of MeCN continues for many column volumes, as was shown by Gilpin with alcohols [15]. Within practical reequilibration times, columns cannot be equilibrated to the extent that the mobile phase exiting the column is the same composition as what is entering.

2.3.3 Pure organic–highly aqueous interface injections

24 For all four columns and all three solvent systems, two system peaks occur shortly after the interface; the first is due to the change in refractive index and the intensity is only slightly variable with solvent system. This peak can be seen in Figures 2 – 4 when the solvent system is changed by operating Valve A. The peak does not occur when using the FCV-10ALVP quaternary proportioning valve of the LC-10ADVP gradient pump to switch between pure organic and highly aqueous mobile phases as the composition gradient between the two phases is sufficiently dispersed before reaching the detector. The second system peak is a plug of acetonitrile, verified by GC, which varies in elution time with each column but not with solvent system. The second system peak does not occur without a column on the system. These same system peaks can be found in Figure 4 of reference [16] around the 12 minute mark after the solvent system is switched to the initial gradient conditions for reequilibration. C-18 bonded phases often perform peculiarly in highly aqueous mobile phases, particularly at the interface between pure organic and highly aqueous solvents. For all columns and solvent systems studied, the 0th injection of acetone (at the interface) had a higher retention than any subsequent injection. One would expect lower retention due to a higher concentration of organic modifier. When equilibrated with 100% MeCN the retention of acetone was indeed lower (data not shown). However at the interface (all columns), and 1mL post-interface (Halo – 100% H2O, Poroshell and Kinetex – 97%

H2O:3% 1-PrOH), the retention was higher. th For the 100% MeCN switched to 100% H2O mobile phase, the 0 acetone injection on the Kinetex column was reproducibly split with 70% of the area under a first, broad peak, and the remainder under the second peak. This second peak‘s width was only slightly broader than the subsequent injections. For the 100% MeCN switched to 97% th st H2O:3% 1-PrOH mobile phase, the 0 and 1 injections on both the Kinetex and Poroshell columns co-eluted into a single peak with twice the area count and a quarter of the base peak width of the subsequent injections; see the red trace of Figure 2.11. The co- eluted 0th and 1st injections are not only sharper than the subsequent injections, but also focused, having a base peak width less than a NaNO3 void marker. The acetone injections‘ versus ζaζO3 base peak widths were 5.009±0.039 seconds vs. 5.440±0.089 seconds (Kinetex) or 5.631±0.119 seconds vs. 6.120±0.093 seconds (Poroshell). The co-

25 elution is invariant with acetone concentration (0.5% vs 0.05% by volume) or pure organic phase composition (100% MeCN vs. 97% MeCN:3% 1-PrOH). Walter et al. [28] noted similar anomalies with highly aqueous mobile phases where there would be a significant loss of retention after flow had been stopped for a length of time. The retention loss is reportedly due to a dewetting of the stationary phase when the column pressure dropped to atmospheric pressure, well below the critical pressure required to force an aqueous solvent into the pores as calculated by the Young- Laplace equation [28], 4 P  cos (2.2) c d where Pc is the capillary pressure, is the surface tension, d is the capillary (pore) diameter, and θ is the contact angle. After dewetting, if sufficient pressure was applied, the stationary phase would wet and the retention would return. Intermediate pressures would affect the stationary phase in a manner dependent on the binary wetting state of the phase, showing a hysteresis for wetting of bonded hydrocarbons. The Young-Laplace equation assumes equilibrium (hydrostatic) conditions where there is no moving phase. The hydrodynamic conditions are unsolved in theory, though at velocities with low Reynolds number (i.e., laminar flow) the expression is a reasonable approximation [29]. The approximate, calculated capillary pressures to wet phases similar to this work with a given solvent are shown in table 2.2. Note that the contact angles are not for the phases studied in this work and each critical pressure is in reference to air at atmospheric pressure, as the values were calculated from capillary flow through a plate for thin-layer chromatography with C-18 bonded phases [30,31]. Therefore, the proper capillary pressures are Pc = Pw – Po, where Pw and Po are the capillary pressure of the high aqueous phase and pure organic phase, respectively. There are three pressure readings of interest in this work: the pressure reading of the Final Eluent pump (pure organic) through column (1), and the pressure readings of the Reequilibration pump (high aqueous) when directed to waste (2) or when directed to the column (3). When Valve A switches between pumps, the column pressure drops from the pressure of the Final Eluent pump (~175 bar) to the pressure of the Reequilibration pump as it was directed to waste (~14bar). It then increases logarithmically over a few

26 minutes to the pressure required to pump the high aqueous phase (~300 bar). This occurs while the 0th, 1st, and 2nd injections are within the column, though only the 0th and 1st injections have shown any anomalous affects.

Table 2.2. Calculated capillary pressure to wet stationary phase in bar (psi) from equation 2.2. Surface tension values used for 100% H2O, 10% MeCN, and 3% 1-PrOH solvents are 71.66, 52.8, 49.5, and 29.29 dyn/cm (actual value for 5% 2-PrOH), respectively, from reference [31]. (–Cos θ) values used for 100% H2O, 10% MeCN, 3% 1-PrOH, and 100% MeCN solvents are 0.80, 0.50 (actual value for 10% MeOH), 0.78 (actual value for 4% EtOH), and -0.75, respectively, from reference [30]. Pore diameters are as listed in LC instrumentation. 90% H20: 97% H2O: 100% 100% H20 10% MeCN 3% 1-PrOH MeCN Halo 255 (3690) 117 (1700) 172 (2490) -98 (-1420) Kinetex 229 (3320) 106 (1530) 154(2240) -88 (-1280) Poroshell 191 (2770) 88 (1280) 129 (1870) -73 (-1060) Zorbax 76 (1110) 35 (510) 51 (750) -29 (-425)

Figure 2.11. Comparison of truncated chromatograms of acetone injections on the Kinetex column with ~14bar priming pressure (red) and ~360bar priming pressure (black). Coelution of the 0th and 1st injections (t = 34.3min; red) are separated by priming the pressure (t = 31.6min, 32.3min; black). All subsequent injections have reduced retention, decreased peak widths, and asymmetry factors closer to unity.

27 Figure 2.11 shows the effect of column pressure on the co-elution of the 0th and 1st injections on the Poroshell and Kinetex. Backpressure was added to the waste flow of Valve A by adding a packed column to prime the Reequilibration pump. Thus, the pressure difference seen by the column was minimized when the valve was switched. With the waste backpressure equivalent to the column pressure of the Final Eluent pump (~175 bar), the detection time of the system peaks and the still co-eluted 0th and 1st injections were reduced. However, the retention time of the 1st injection was reduced to the retention of the subsequent injections. With the waste backpressure equivalent to the column pressure required to pump the aqueous phase (>300 bar), the detection time of the system peaks are further reduced, and the 0th and 1st injections are separated. Furthermore, with the Reequilibration pump primed, the retention of the subsequent injections (2nd – 19th) is decreased, the peak widths are reduced, and the slope cut-off is reduced, as shown in Figure 2.11. With pressure priming the Reequilibration pump the reequilibration time for the Poroshell and Kinetex columns are equivalent (1 min) to the

Zorbax and Halo columns with 97% H2O:3% 1-PrOH. This experiment was applied to the Halo column with 100% H2O reequilibration solvent primed to ~300bar. Once again, the reequilibration time was reduced to 1 minute, the retention of all injections decreased, and the peak widths were reduced. With added backpressure and equivalent slope cut-off values all columns studied were reequilibrated within 1 minute. This anomalous behavior can be explained by the Buckley-Leverett equation [32] and Darcy‘s law. The Buckley-Leverett theory describes two-phase flow in porous media and, to our knowledge, has not been previously applied to chromatography. In the late 1930s, an equation was derived to describe the mechanism of water displacing petroleum in saturated sands. An excellent review can be found in the introduction of reference [29]. The system Buckley and Leverett studied is similar to post-gradient reequilibration, and can be modified to apply to chromatographic conditions. The Buckley-Leverett equation as described for chromatography: S S  US  (2.3) t x where S is the saturation of water of the packed bed, t is time, x is the column length, and U(S) is the front velocity of a given water saturation given by:

28 v  f  US    w  (2.4) A  S t where v is the volume flow rate of the reequilibration solvent, φ is the porosity, A is the cross-sectional area of the column, and fw is the fractional flow rate of the reequilibration solvent. The fractional flow rate is defined as fw ≡ vw/v , where vw is the volume flow rate of reequilibration solvent and v is the bulk volume flow rate, vw + vo. To reduce post- gradient reequilibration time in chromatography, one wishes to maximize U(Š), the shock front velocity where f(S) and δf/δS intersect. Since the term φA is constant for a given column, and columns are equilibrated when S ≈ 1, the volume flow rates are the only variables. Darcy‘s law describes the pressure required to force a fluid at a given flow rate through a porous medium, given by: KA P v   (2.5)  L where, for chromatography, K is the permeability of the column, η is the dynamic viscosity, ΔP is the pressure drop across the column of length L and cross-sectional area A. With two mobile phases, Darcy‘s equation for each is:

kw Pw vw  KA  (2.6a) w L

ko Po vo  KA  (2.6b) o L where kw and ko are the relative permeabilities of the mobile phases and ΔPw and ΔPo are KA the pressure drop of the high aqueous and pure organic, respectively. The term for L each fluid is constant, assuming Newtonian fluids and that the column dimensions are invariant with pressure. kw and ko vary non-linearly with S and sum to unity when using degassed solvents and correcting the function for the critical saturation for the front velocity, U(S→0) to reach the end of the column. If it is assumed that δPw/δL and δPo/δL are equivalent, then fw varies only with S and minimum column reequilibration time is only a function of the minimum allowable water saturation of the phase to give reproducible chromatography. However, δPw/δL and δPo/δL are only equivalent when the

29 dynamic viscosities of the two fluids are equivalent and δPc/δS = 0, as described above.

This is not the case in this work nor frequent in chromatography, so fw must also vary with total pressure, defined as:

PTotal  Pw  Po  Pc (2.7) It follows that:

vw 1 1 fw    (2.8) v  v vo   k  P  w o 1 1  w  o  o  v     w o  kw  Pw  Therefore with increased total pressure from priming the Reequilibration pump the fractional volume flow rate of high aqueous phase in the system approaches the bulk volume flow rate faster by overcoming the capillary pressure between the two solvents, giving a greater shock front velocity. This phenomenon has been noted previously in Figure 9 of reference [16] when an increase in bulk volume flow rate significantly decreased the necessary reequilibration time and volume by increasing the total pressure well above a static capillary pressure. In this work, injections made before the shock front will elute when the shock front reaches the end of the column, explaining both the increased retention at the interface and the coelution of two injections on phases (Kinetex and Poroshell with 97% H2O:3% 1-PrOH) where the initial total pressure is less than the capillary pressure.

2.4. Conclusions With proper timing, injections at the pure organic – highly aqueous interface retain peculiarly and can elucidate the limiting factors for fast reequilibration of the phase. Post-gradient reequilibration with a reduced dwell volume system can occur in very few column volumes, i.e., ≤ γ column volumes. With sufficient pressure, the time required for reequilibration is governed mostly by flushing out the system dwell volume and the column‘s pore volume. Pressures lower than the capillary pressure cannot force the aqueous phase into the pores and must depend on diffusion of the acetonitrile out of the pores to reequilibrate. Pressures primed well above the capillary pressure can serve to flush the pure organic solvents out of the pores.

30 Within a practical timeframe, the organic modifier cannot be entirely flushed from the column, however once the majority of the modifier has been flushed from the inter- and intra-particle volume, the column can retain reproducibly. An acceptable reproducibility must be determined for each analysis, and the reequilibration time adjusted to provide such. Of the stationary phases studied, more retentive phases will have a greater variation in retention post-gradient and will release sorbed organic modifier in a sharper decrease. Of the three parameters in equation 2.2, column manufacturers can adjust the pore diameter and aqueous contact angle of the phase. Greater pore diameter and lower contact angle will reduce the necessary pressure for a highly aqueous mobile phase to wet the stationary phase, such as for the polar endcapped and polar embedded phases[33]. However, the greater pore diameter decreases the internal surface area of the silica [4], and a lower contact angle reduces the retention and selectivity of nonpolar analytes. Practical chromatographers choose a column and initial gradient mobile phase composition with a selectivity as needed by the sample, then decide on a reasonable reequilibration time. It follows that using pressure to reequilibrate columns faster is the simplest method, assuming that the reequilibration solvent is set by the separation method and already contains a small amount of n-alcohol modifier. The limiting variable with column equilibration is not the distance the organic modifier must diffuse to desorb from the stationary phase, but rather the pressure required to force the aqueous phases into the pore.

31

CHAPTER THREE

QUANTIFYING INJECTION SOLVENT EFFECTS IN REVERSED- PHASE LIQUID CHROMATOGRAPHY

3.1. Introduction All samples must be dissolved in a diluent prior to injecting onto a liquid chromatographic column. For most methods, the diluent of choice is the mobile phase, and has been noted to give the best results [13]. Any other solvent can be modeled as a step gradient with strength relative to the mobile phase. Increased strength of the injection solvent has been reported to cause distorted band shapes, shoulders on main peaks, peak splitting and peak tailing, particularly for early eluting peaks. Conversely, a decreased strength of the injection solvent is described to focus analytes at the head of the column [13,34-38]. Negative effects can be mitigated by reducing the sample mass or sample volume injected. However, there are cases when methods must use a high mass or volume load, such as in impurity profile analysis where the main peak is of significantly higher concentration than trace impurities. Additionally, in two-dimensional chromatography it is advantageous to have a large modulation volume between the dimensions to increase the available time of analysis to the second dimension. This becomes a significant problem when the modes of operation have sufficiently different mobile phases, [10] or when a gradient is applied in the first dimension. Strong injection solvents are more common in RPLC as they tend to solubilize analytes readily. Expectedly, with a strong injection solvent, analytes will be retained less for the duration spent within the injection plug. Loeser et al. [39,40] reported that peak distortion occurs for analytes that elute after the injection plug in highly aqueous mobile

32 phases. When the injection solvent is strongly retained at the head of the column, less retained analytes migrate past the solvent plug and are retained by an axially homogeneous mobile phase-stationary phase system. Analytes that are more retained than the injection plug will spend a larger amount of time within the plug. If the analyte has different thermodynamics of retention in the injection plug when compared to the mobile phase-stationary phase system, then peak distortion will result in a case similar to the resistance to mass transfer from the stationary phase. With sufficiently different retention, the multiple of zones will cause a peak to shoulder or split. Since injection solvent effects are fundamentally a retention effect, all variables affecting retention can arguably affect the sensitivity an analyte-column pair has to the injection solvent. Current understanding of a hydrophobic retention mechanism considers a three phase system: 1) mobile phase 2) sorbed organic layer and 3) stationary phase ligands. Kazakevich et al. [41] derived an equation to model the change in retention volume (VR) as a function of mobile phase composition (Cel):

VR (Cel )  V0  VS  KP (Cel )[VS SK H ] (3.1a) where V0 is the volume of liquid phase in the column (mL), VS is the volume of the adsorbed layer of organic modifier (mL), KP(Cel) is the distribution coefficient of the analyte between the mobile phase and the adsorbed phase, S is the surface area of the 2 adsorbent (m ), and KH is the Henry constant for the analyte adsorption from the pure organic adsorbed layer onto the surface of the bonded phase. KH can be calculated from the retention volume of the analyte using neat organic modifier: V (100)  V K  R 0 (3.1b) H S

KP(Cel) can be calculated by the distribution of the analyte in a vapor-liquid system using headspace GC, where the liquid system is either the neat organic liquid (C100) or a mixture of water:organic, (Cel). Then KP(Cel) is K(org-vap)/K(el-vap). Alkylbenzene homologues and 2-butanone were modeled and fit well to the chromatographic measurements. For charged analytes, retention is also a function of the ionization state, which is a function of the pKa of the analyte and the pH of the solvent [42]. To date, there has been no systematic investigation to the injection solvent sensitivity of ionizable analytes in

33 comparison to non-polar compounds. Modeling the retention of ionizable analytes as a function of pH, organic modifier, and temperature is a far more difficult calculation, but has been recently described [43]. As shown by Hoffman et al. [35,36] using both simulation and experimental results, with slight changes in retention, the concentration profile of an analyte will tend to widen and become asymmetric. This suggests that efficiency calculations are a useful measure of injection solvent effects on the band profile of analytes. The Foley-Dorsey equation [44] for efficiency (N) quantitatively accounts for both a change in peak width and asymmetry :

2  t  41.7 R   W  N   10%  (3.2) 1.25  As10% where tR is the retention time at peak maximum, and W10% and As10% are the width and asymmetry of the peak at 10% height, respectively. As10% is defined as the ratio of the half-widths of the peak A and B, measured from the leading edge of the peak to a normal drawn from peak height to baseline, then from the normal to the trailing edge, respectively. As10% is defined as greater than unity, using either B/A or A/B as appropriate. In general, chromatographic peaks tail, but in the case of injection solvent effect it is more common for peaks to front. The aim of this work is to develop and apply a method on multiple columns to quantify the sensitivity to injection solvent strength for non-polar and basic analytes, with specific regard to water-acetonitrile mixtures as a solvent. Using the Foley-Dorsey equation for theoretical plate count as a desirability function, we investigate the effects of injection solvent on retention time (tR), peak width at 10% peak height (W10%), and asymmetry at 10% peak height (As10%). We show a correlation to the Hydrophobic- Subtraction model coefficients and acetonitrile adsorption isotherms to aid in column- injection solvent pair selection.

3.2 Theory 3.2.1 Hydrophobic-subtraction model Reversed-phase column selectivity can be modeled by:

34  k'  log  log   H'  S' *   A'  B'  C' (3.3)    k'ref  where the separation factor, α, is the logarithm of the ratio of the retention factors of test analytes, k’, to a reference analyte, k’ref, which is generally ethyl benzene [45]. The solute parameters describe the relative properties that contribute to retention in reversed- phase chromatography, namely hydrophobicity (η’), molecular ―bulkiness‖ (σ’), hydrogen-bond basicity (β’), hydrogen-bond acidity (α’), and approximate charge (κ’). These solute parameters are a function of the separation conditions and have been calculated for the set of test analytes in a mobile phase of 50% 60mM phosphate buffer (pH 2.8):50% MeCN (v/v) at 35°C for alkyl and polar-embedded stationary phases as reported in references [46] and [47], respectively. The parameters (H, S*, A, B, C) are the complementary interactions of the column and describe the hydrophobicity, steric hindrance, hydrogen-bond acidity, hydrogen-bond basicity, and cation exchange activity at a given pH, respectively. Average C18 columns of type-B silica will have H equal to unity, and remaining parameters equal to zero. These values should be separation- condition independent, with exception to the cation-exchange activity being dependent on the pH of the mobile phase.[48] Column parameters for a large set of columns are accessible in the USP-PQRI database (http://www.usp.org/USPNF/columnsDB.html). The physical and chemical origins of these column parameters has been described in detail in reference [49]. Parameters H and S are functions of stationary phase ligand length (C8 vs C18), bonding density, side group (dimethyl vs diisopropyl), and pore diameter. Parameters A and C are a function of the number of protonated and ionized silanols, respectively. Parameter B is considered to be a function of sorbed water in the phase, and is significantly higher for polar-embedded and polar-endcapped phases than alkyl phases.

3.2.2 Acetonitrile excess adsorption isotherm The excess adsorption isotherm of organic modifiers can be measured by the minor disturbance method where a plug of organic-modifier-rich injection solvent is made onto a column that has been equilibrated with a mobile phase of organic modifier concentration, Cel.[50,51] From the retention volume of a minor disturbance VR(Cel) in

35 the baseline and the thermodynamic void volume Vm, the excess organic amount e n org(Cel) (mols) can be written as:

C e norg C( el )  VR  VM Cel (3.4a) 0 where VM can be calculated by integration of the retention times, VR:

1 Cel Cmax VM  VR (Cel )Cel (3.4b) 0 Cmax

Alternatively, if the step values for Cel are constant, then VM is the mean of VR. From e a n org, the fraction absorbed at each mobile phase concentration x org can be calculated by:

l * e a Stx  aaqnorg xorg  * * e (3.5) St  (aaq  aorg )norg where S is the absorbent surface area (m2), t is the number of sorbed monolayers of organic modifier, xl is the molar fraction of organic in the mobile phase, and a* is the molar surface area of the pure components of the mobile phase as noted organic (org) or aqueous (aq). At 25°, a* is 0.0776m2/μmol and 0.1η9m2/μmol for water and acetonitrile, respectively, estimated by comparison to the space requirement of nitrogen [52]. The number of monolayers of organic modifier is chosen by a convention where a l l the value gives x org/x org= 0 at the inflection point, x org=I, of the negative portion of e l n org/x org, given by:

 e  1 norg  t    xl a*  1 xl  a*  a*  a* ne   (3.6) S  xl org I org org I aq aq org org I   org  I  a l Both t and I are found simultaneously by fitting x org/x org with an arbitrarily high order a l polynomial, adjusting I, and solving x org/x org= 0 while constraining limx→I- a l a l x org/x org < 0 and limx→I+ x org/x org > 0. For reversed phase modified silica with acetonitrile as a modifier, t≈γ and is a function of ligand length [41], ligand density [53], e l temperature, and pressure [54]. Knowing t, a function can be fit to the data of n org(x org):

 (K  )1 xe 1(  xe ) (K  )1 xe 1(  xe )  ne (xl )  St C18 org org  1[ ] OH org org  (3.7) org org  * l * l * l * l   KC18aorgxorg  aaq 1(  xorg ) KOHaorgxorg  aaq 1(  xorg )  where is the surface heterogeneity and is roughly the fraction of surface area covered by the modifier ligand, and KC18 and KOH are the equilibrium constants of the absorption of the organic modifier to the ligand and silanols, respectively, for a special case in which

36 the molecular sizes of the modifier and water are considered equivalent. These 2 parameters are found simultaneously by maximizing r of the fit in a regression. Extensive discussion regarding the minor disturbance method and the underlying theory can be found in reference [52].

3.3 Experimental 3.3.1 Reagents All water used was purified to a resistance of approximately 18εΩ-cm using a Barnstead (Dubuque, IA, USA) NANOPure Diamond water purification system. HPLC grade acetonitrile (MeCN) was obtained from Sigma-Aldrich (St. Louis, MO, USA). Methyl ketones C3 – C7 (acetone, butanone, 2-pentanone, 2-hexanone, and 2 heptanone) and lidocaine were obtained from Fisher Chemicals (Fair Lawn, NJ). Non-buffered mobile phase for the methyl ketone study was prepared by mixing the appropriate volumes of MeCN and H2O then vacuum filtered through 0.45m nylon filter prior to use. Buffered mobile phase for the lidocaine study was prepared by mixing citric acid and sodium citrate monobasic in proportion to give 13.5mM citrate concentration and a pH of ~2.8 prior to mixing with MeCN. Injection solvents were prepared by first adding analytes to the appropriate amount of MeCN, then diluted with water. All hydro-organic fractions are reported as volume-to-volume percents. All figures of merit (retention time, width, asymmetry, and plate count) reported were the average of three repetitive injections.

3.3.2 Liquid chromatography instrumentation A Shimadzu LC-10 stack (Kyoto, Japan) was used for this study, including LC- 10ADVP gradient pump outfitted with a DGU-14A inline degasser and an FCV-10ALVP quaternary proportioning valve, an SIL-10A auto injector with a 50μδ injection loop, an SPD-10A UV-Vis detector set to 254nm and 10Hz data rate, and an SCL-10AVP system controller. All data were collected and analyzed using Class-VP version 5.032 software. Four stationary phases from Agilent Technologies were evaluated: Poroshell EC-C18, Zorbax 300Extend-C18, Zorbax SB-C18, and Zorbax Bonus-RP. The first column listed is packed with superficially-porous silica, the remaining are fully porous silica. The last

37 column listed has a polar embedded amide linker. Column parameters are listed in table 3.1. Temperature was kept constant via column jacket and circulator, though mobile phase was not preheated due to the ΔT from ambient being ≤ η°C for the injection solvent study and adsorption isotherms.

Table 3.1. Parameters of columns used in this work, as reported by manufacturer. Bonding density calculated by reference [55]. Poroshell 120 Zorbax Zorbax Bonus- Column Zorbax SB-C18 EC-C18 300Extend-C18 RP dp β.7μm ημm γ.ημm γ.ημm Pore diameter 120 Å 80 Å 300 Å 80 Å Surface area 120 m2/g 180 m2/g 45 m2/g 180 m2/g Carbon load 8% 10% 4% 9.5% Bonding density 3.02 2.04 3.81 2.41 (μmol/m2) Dimensions 4.6x100mm 4.6x150mm 4.6x100mm 4.6x75mm propylene- propylene- diisobutyl diisopropyl Ligand bridged methyl bridged methyl octadecylsilane butadecylsilane octadecylsilane octadecylsilane Encapped? Double N Double triple Polar embedded? N N N amide linker

3.3.3 Liquid chromatography methods

Injection solvent - Methyl ketones Samples of three concentrations, 1.0mg/mL, 2.0mg/mL, and 4.0mg/mL were diluted in regularly varied solvent strengths of water:acetonitrile mixtures from 10% MeCN to 100% MeCN in steps of 10%, for a total of 30 samples. Injection volumes varied for each concentration to compare equal and changing sample masses. Combinations were 1) 30μδ of 1.0mg/mL, 2) 15μδ of 1mg/mL, 3) 15μδ of 2 mg/mL, 4) 15μδ of 4mg/mL, 5) 7.5μδ of 4mg/mL, and 6) 1.25μδ of 4mg/mL for a total of 180 injections per column. Figure 3.1 shows a typical series of chromatograms as the injection solvent is strengthened from 10% - 100% εeCζ. ζear ideal sensitivity (0.97 ≤ s ≤ 0.99, vide infra) is obtained from the small mass-small volume injections, 6. The remaining results are compared by static mass and changing volume (1, 3, 5) or by static

volume and changing mass (2, 3, 4). The mobile phase was a premixed 60% H2O:40%

38 MeCN. Volume flow rate was set to 1.00mL/min. Temperature was kept constant at 25°C.

Figure 3.1. Chromatograms showing the effects of injection solvent on peak shape of homologous series of methyl ketones, C3-7. From top to bottom the injection solvent strength is varied from 10%-100% MeCN, in steps of 10% increments. Column, injection volume, injection mass: Zorbax Bonus-RP, 30µL, 30µg.

Injection solvent - Lidocaine Similar to the methyl ketone study, three concentrations were used: 0.1mg/mL, 0.2mg/mL, and 0.4mg/mL. All other conditions were kept the same, i.e., injection volume and injection solvent choice. This ensured that no sample overloading occurred on any of the four columns studied. The mobile phase was a premixed 75% 13.5mM citrate buffer (pH ~2.8):25% MeCN. Volume flow rate was set to 1.00mL/min. Temperature was kept constant at 25°C.

Hydrophobic-subtraction model

39 Conditions were as described in [56] for all columns studied. The solute parameters from reference [46] were used for 17 analytes on the alkyl phases (Poroshell EC-C18, Extend300-C18, and Stablebond-C18): thiourea, ethylbenzene, acetophenone, benzonitrile, anisole, toluene, 4-nitrophenol, 5-phenylpentanol, 5,5-diphenylhydantoin, cis-chalcone, trans-chalcone, N,N-dimethylacetamide, N,N-diethylacetamide, 4-n- butylbenzoic acid, mefenamic acid, nortriptyline, and amitriptyline. For the polar- embedded column (Bonus-RP), the solute parameters from reference [47] were used with addition of cis-4-nitro-chalcone, trans-4-nitro-chalcone, and p-chlorophenol. 5,5- diphenylhydantoin was not used to calculate the column parameters of the Bonus-RP phase. Amitriptyline and thiourea were injected as single components on the Bonus-RP due to co-elution. Premixed 50% 60mM Phosphate buffer (pH=2.8):50% MeCN mobile phase was used as the injection solvent, with exception of the chalcones and mefenamic acid, which were first dissolved in 100% MeCN then added to the separation mixtures. The chalcones were irradiated with UV for at least 30 minutes prior to injection. Injection amount for each analyte was 500ng, with exception of 4-nitro-chalcone, which was only slightly soluble in MeCN. Each retention time was an average of triplicate injections detected at 215nm. Temperature was kept constant at 35°C via column jacket.

Acetonitrile excess absorption isotherm For the minor disturbance method, each column was equilibrated in regularly stepped mobile phase (10%, 20%, 30%... MeCN) for at least 30mL at 1.00mL/min. Mobile phases were dynamically mixed via low-pressure quaternary proportioning valve with 100% H2O in reservoir A and 100% MeCN in reservoir B. Mean retention times of triplicate injections of 10μδ and 1μδ of 100%MeCN were measured for VR in equations 3.4 and 3.5. Detector wavelength was set to 195nm. Because of the very low analyte signal, and relatively high background signal, the waveform of the gradient mixer was evident in the chromatograms, and affected the precision of the retention time. For all columns at all mobile phase concentrations, the maximum and minimum %RSD of any triplicate injection was 2.3% and 0.2%. Temperature was kept constant at 25°C via column jacket.

40

3.4 Results and Discussion 3.4.1 Injection solvent sensitivity, s, of methyl ketones Methyl ketones are ideal analytes for studying injection solvent effects due to the in a range of aqueous-MeCN mixtures. Figures 3.2 – 3.5 show the effect of sample injection solvent strength on chromatographic figures of merit from equation 3.2 for three different injection volumes or injection masses. It can be seen from figure 3.2 that changing the injection solvent strength does not significantly change the retention time of the methyl ketones in any volume or mass measured on the Poroshell column, and holds true for the three other columns. There is a slight decrease in the values, amounting to 0.006 mins on the Poroshell column from 10% - 100% MeCN, but the effect on efficiency is minor. This is expected, as the thermodynamics of retention for the peak maxima are affected by the injection solvent for only a small fraction of the total time on column [36].

7.0

6.0

5.0

4.0

3.0 Retention time (min) time Retention 2.0

1.0

0.0 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

Figure 3.2. Change in retention time with injection solvent strength. Analytes: acetone – dark blue, 2-butanone – pink, 2-pentanone – green, 2-hexanone – red, 2-heptanone – violet. Injection volumes: solid line – γ0μδ, dashed line – 1ημδ, dotted line – 7.ημδ. Injection mass: γ0μg. Column: Poroshell 1β0 EC-C18.

41 0.7 A 0.6

0.5

0.4

0.3

Width at 10% Width (min) height 0.2

0.1

0.0 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

0.6 B

0.5

0.4

0.3

0.2 Width at 10% Width (min) height

0.1

0.0 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

Figure 3.3. Change in peak width at 10% height with injection solvent strength with (A) constant γ0μg mass, varied volume and (B) constant 1ημδ volume, varied mass. Analytes as figure 3.2. Injection volumes or mass: solid line – γ0μδ, θ0 μg; dashed line – 1ημδ, γ0 μg; dotted line – 7.ημδ, γ0 μg. Injection mass: γ0μg. Column: Poroshell 1β0 EC-C18.

The change in width at 10% height as a function of injection solvent is shown in figures 3.3A with injection volume and 3.3B with injection mass. In all four columns studied, injection mass has no significant effect on the width of the resulting peak, and

42 reducing the injection volume reduces the increase in width with injection solvent strength. The change in width is mostly responsible for the reduction of efficiency, and therefore sensitivity to injection solvent strength. The major difference between the reductions of efficiency due to the injection solvent among the four columns tested is the magnitude of the change in width as injection solvent is strengthened.

2 A

1.8

1.6

1.4

1.2 Asymmetry Asymmetry at 10% height

1

0.8 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

2.4 B

2.2

2

1.8

1.6

1.4 Asymmetry at 10% height 1.2

1

0.8 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

43 2.8 C 2.6

2.4

2.2

2

1.8

1.6

Asymmetry at 10% height 1.4

1.2

1

0.8 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

2.6 D

2.4

2.2

2

1.8

1.6

1.4 Asymmetry at 10% height

1.2

1

0.8 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

Figure 3.4. Change in asymmetry at 10% height with injection solvent strength with constant γ0μg injection mass for the (A) Poroshell 120 EC-C18, (B) Zorbax Stablebond- C18 (C) Zorbax 300Extend-C18, and (D) Zorbax Bonus-RP. The data have been adjusted in the y-direction for spacing, with horizontal lines denoting asymmetry of unity. Analytes: acetone – dark blue, 2-butanone – pink, 2-pentanone – green, 2-hexanone – red, 2-heptanone – violet. Injection volume: solid line – 30μδ, dashed line – 15μδ, dotted line – 7.5μδ.

44 In figures 3.4A-D and 3.5A-D, the asymmetries of the analytes on the column studied are plotted as a function of the injection solvent strength, volume, and mass. The asymmetry functions are significantly different for each column. The only similarity among the columns is that asymmetry near the mobile phase concentration is constant despite changes in volume or mass. Any changes in measured asymmetry from changing the injection volume or mass are only evident with injection solvent strengths that are stronger or weaker than the 40%MeCN mobile phase and the change is greater the further the injection solvent strength is from the mobile phase. In other words, the asymmetry of the 60%MeCN injections depend less on volume and mass than the 90%MeCN injections. Notably, when reducing the volume from γ0μδ to 7.ημδ, the curve of the plot of asymmetry vs. injection solvent of the early eluting acetone is flattened to depend less on the injection solvent strength, and the later eluting 2-heptanone is mostly unaffected. However, when reducing the mass, acetone is mostly unaffected and there is a gradient of flattening as retention factor is increased. The curve of asymmetry vs. injection solvent of 2-Butanone is less dependent on mass than the curve of 2-heptanone. The relationship between asymmetry and volume holds for all four columns tested, however the effect of injection mass is only evident on the Poroshell and the Stablebond phase.

2.4 A

2.2

2

1.8

1.6

1.4 Asymmetry at 10% height 1.2

1

0.8 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

45 2.6 B

2.4

2.2

2

1.8

1.6

1.4 Asymmetry at 10% height

1.2

1

0.8 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

2.8 C 2.6

2.4

2.2

2

1.8

1.6

Asymmetry Asymmetry at 10% height 1.4

1.2

1

0.8 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

46 2.6 D

2.4

2.2

2

1.8

1.6

1.4 Asymmetry at 10% height

1.2

1

0.8 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

Figure 3.5. As figure 3.4, but with constant 15μδ injection volume for the (A) Poroshell 120 EC-C18, (B) Zorbax Stablebond-C18 (C) Zorbax 300Extend-C18, and (D) Zorbax Bonus-RP. Injection mass: solid line – θ0μg, dashed line – γ0μg, dotted line – 1ημg.

To quantify and compare the effect of injection solvent strength on each column,  we first define the efficiency curve with the vector of efficiency, N , versus injection solvent compositions, Ni, where i is the ordered value of acetonitrile percentage.  N  N1, N2, N3,...Ni (3.8a) For each column, injection volume, and mass, we can construct measured and ideal vectors. For this work, measured efficiency vectors are constructed as above where the i value ranges from 10% to 100% MeCN in steps of 10%. Ideal efficiency vectors consist of i components equal to the maximum efficiency measured for a given injection volume and mass.  Nideal  Nmax , Nmax , Nmax ,...Ni (3.8b) We may project the measured vector onto the ideal vector, then normalize the magnitudes, such that 0 < s < 1, and represents the cumulative percentage loss of efficiency: the sensitivity, s

47 i   i  (Nideal,n  Nmeasured,n ) Nmeasured  Nideal n1 s    (3.9) 2 i Nideal

By this calculation, s is independent of the maximum efficiency of the column, allowing columns of different plate counts to be directly compared. s is dependent on the percentage change of the plate count, and not the value of the change, as can be seen in figure 3.6. To our knowledge, it has not been suggested in the literature that particle diameter nor column length affect the injection solvent sensitivity of a column, though it is reasonable to expect a dependency on column internal diameter. If the efficiency vector cannot be modeled by a straight line, the sensitivity is then a function of the percentage step change in injection solvent.

6000

5500 y = -12x + 6120 5000 1.000 0.915 4500 0.875 0.838 0.803

Theoretical plate count Theoretical plate 4000 0.910 0.910 y = -10x + 5100 3500 0 10 20 30 40 50 60 70 80 90 100 %MeCN diluent

Figure 3.6. Plot of efficiency vectors showing effect of percent gradient on injection solvent sensitivity, s. Ideal efficiency vector with s = 1 is plotted with solid, black circles, ●. Solid shapes with dashed lines represent different percentage loss of efficiency per 10% MeCN added to injection solvent – 2% ♦; 3% ■; 4% ▲; 5% X. Open shapes with solid lines, ○ and □ represent two vectors of the same sensitivity by with differing efficiency maxima.

48 Table 3.2 lists the sensitivity of each column at each injection volume and mass, and figures 3.7A and 3.7B plot those values as a function of injection volume and injection mass, respectively, for 2-hexanone and 2-heptanone. The Stablebond-C18 phase is the most ideal column tested, having very little sensitivity to the injection solvent, whereas the 300Extend-C18 and Bonus-RP are the most sensitive to injection solvent. To explain the differences of the sensitivity, we have characterized these four columns via the different contributions to retention of the hydrophobic-subtraction model.

A 1.00 0.95 0.90 0.85 0.80 0.75 0.70 Sensitivty 0.65 0.60 0.55 0.50 5 10 15 20 25 30 Injection volume (uL)

1.00 B 0.95 0.90 0.85 0.80 0.75 0.70 Sensitivity 0.65 0.60 0.55 0.50 10 20 30 40 50 60 Injection mass (ug) Poroshell - 2-heptanone Poroshell - 2-hexanone Zorbax300 - 2-heptanone Zorbax300 - 2-hexanone Zorbax SB - 2-heptanone Zorbax SB - 2-hexanone Zorbax Bonus - 2-heptanone Zorbax Bonus - 2-hexanone

Figure 3.7. Sensitivity of all columns tested as a function of (A) injection volume, with a constant γ0μg mass or (B) injection mass, with a constant 15uL volume. Values calculated from equation 3.2 for 2-hexanone and 2-heptanone.

49 Table 3.2. Measured sensitivities for all columns, conditions, and analytes. Some values not reported due to peak overlap.

Conc mass volume (mg/mL) (µg) (µL) Acetone Butanone 2-Pentanone 2-Hexanone 2-Heptanone 30 30 0.75 0.69 0.68 0.72 0.79 1.0 15 15 0.89 0.85 0.82 0.86 0.91 Poroshell 2.0 30 15 0.89 0.82 0.74 0.81 0.87 EC-C18 60 15 0.88 0.83 0.81 0.81 0.77

4.0 30 7.5 0.94 0.91 0.91 0.91 0.91 5 1.25 0.98 0.99 0.98 0.97 0.97 30 30 0.59 0.59 1.0 15 15 0.75 0.76 0.77 Zorbax 2.0 30 15 0.76 0.77 0.70 300Extend-C18 60 15 0.73 0.64

4.0 30 7.5 0.89 0.87 0.78 5 1.25 0.98 0.98 0.98 0.96 30 30 0.85 0.81 0.84 0.83 0.86 1.0 15 15 0.91 0.93 0.94 0.93 0.89 Zorbax 2.0 30 15 0.92 0.89 0.90 0.96 0.97 Stablebond-C18 60 15 0.91 0.88 0.85 0.85 0.83

4.0 30 7.5 0.93 0.96 0.95 0.95 0.96 5 1.25 0.97 0.96 0.96 0.96 0.97 30 30 0.56 0.55 0.59 1.0 15 15 0.86 0.81 0.76 0.74 0.73 Zorbax 2.0 30 15 0.82 0.83 0.75 0.72 0.76 Bonus-RP 60 15 0.82 0.81 0.76 0.76 0.76

4.0 30 7.5 0.92 0.89 0.89 0.88 0.87 5 1.25 0.98 0.97 0.96 0.96 0.98

50 Table 3.3. Column and solute parameters measured by the hydrophobic-subtraction model. H S* A B C(2.8) SD 1.023 -0.012 -0.139 0.021 0.202 Poroshell 120 C18 ±0.005 ±0.007 ±0.014 ±0.005 ±0.010 0.009

0.993 0.014 -0.050 0.015 0.242 Zorbax 300Extend-C18 ±0.003 ±0.004 ±0.008 ±0.003 ±0.006 0.005

1.003 -0.037 0.224 -0.001 0.214 Zorbax Stablebond-C18 ±0.004 ±0.005 ±0.010 ±0.004 ±0.008 0.007

0.881 -0.061 0.107 0.046 0.283 Zorbax 300Stablebond-C18 ±0.007 ±0.009 ±0.019 ±0.007 ±0.014 0.012 0.730 0.043 -0.435 0.261 -1.492 Zorbax Bonus-RP ±0.026 ±0.020 ±0.071 ±0.022 ±0.052 0.047 1.117 0.036 -0.010 -0.032 0.052 AMT Halo C18 ±0.005 ±0.007 ±0.014 ±0.005 ±0.010 0.009 0.947 0.014 0.046 0.018 0.194 Supelco Discovery C18 ±0.005 ±0.007 ±0.014 ±0.005 ±0.011 0.009 0.987 0.007 -0.102 -0.009 0.137 Waters XterraMS C18 ±0.001 ±0.001 ±0.003 ±0.001 ±0.002 0.002 0.988 -0.012 -0.053 0.000 -0.193 YMC PackPro C18 ±0.006 ±0.008 ±0.016 ±0.006 ±0.012 0.010 η' σ' β' α' κ' SD -1.452 -0.009 0.078 -1.205 0.065 Acetone ±0.005 ±0.215 ±0.042 ±0.255 ±0.031 0.010 -1.172 -0.372 0.038 -1.170 0.011 Butanone ±0.007 ±0.275 ±0.054 ±0.327 ±0.039 0.013 -0.927 -0.181 0.042 -0.649 0.030 2-Pentanone ±0.001 ±0.031 ±0.006 ±0.037 ±0.004 0.001 -0.695 -0.222 0.043 -0.395 0.027 2-Hexanone ±0.002 ±0.074 ±0.015 ±0.088 ±0.011 0.003 -0.463 -0.159 0.081 0.038 0.008 2-Heptanone ±0.003 ±0.106 ±0.021 ±0.126 ±0.015 0.005 -2.179 -0.053 -0.084 -2.934 1.897 Lidocaine ±0.094 ±0.803 ±0.170 ±1.875 ±0.566 0.037 -0.748 -0.335 0.024 -0.356 0.007 Acetophenone ±0.001 ±0.033 ±0.007 ±0.039 ±0.005 0.002 Acetophenone (Lit) -0.744 0.133 0.059 -0.152 -0.009

51 3.4.2 Comparison to the hydrophobic-subtraction model Column and solute parameters for the four columns and analytes studied are listed in table 3.3. To regress the solute parameters for the methyl ketones and lidocaine, five additional columns were characterized, selected randomly from on-hand supplies, in addition to the three alkyl silica columns (non-polar embedded): Agilent Zorbax 300Stablebond-C18, AMT Halo C18, Supelco Discovery C18, Waters XterraMS C18, and YMC PackPro C18. For lidocaine solute parameters, the YMC PackPro was removed from the set due to a retention factor of 0.000±0.000. Deviations from published column parameters are due to column aging, though each column gave retention time reproducibility of <0.005min and a reasonably small standard deviation for the regression, < 0.015. Compared to the literature values, each column gave a cos(θ) between 0.97 and 1.00 for the dot product, with the least being the Bonus-RP. The deviations seen in the H-S model for polar embedded phases have been described well [57], and our data follow the general trend that the Bonus-RP is significantly less acidic (lower A and C), more basic (higher B), and more polar (lower H) than type-B alkyl silica. As the carbon number is increased, the solute parameters for the methyl ketones become more hydrophobic (less negative ‘) and less hydrogen bond acidic (less negative α‘), as expected. With the remaining parameters, there are not as clear trends, and the calculated solute parameters for the acetophenone control deviate from the published literature values. The greatest standard error from the regression is found on the σ‘ and α‘ parameters, as is common for data sets with low column numbers [58]. Important to our study is the relative contribution of the hydrogen bond basicity ‘ is constant with increasing carbon number, and the hydrogen bond acidity α‘ increases with carbon number. By understanding the mechanisms of retention for these analytes, differences in the effects of the injection solvent strength regarding asymmetry can be explained. Though the solute parameters, by convention, change with separation conditions, it is not expected that the trends would be different, i.e., one would not expect the hydrophobic contribution to retention to be less for 2-heptanone than for acetone in any mobile phase. However, it can be posited that the relative contributions of each parameter to change with separation conditions, i.e., the ‘ may play less of a role than ‘ in highly aqueous phases for the retention of acetone versus the retention of 2-heptanone. As the HS model has developed into a method of column comparison, there have been no published comparisons of solute parameter

52 changes with separation conditions, to our knowledge, other than for the κ‘ value changing with pH of the mobile phase. Two trends have been noted with the plots of asymmetry vs. injection solvent: more ideal curves for low k‘ analytes with a reduction of injection volume (all columns) and a gradient of ideality with k‘ with a reduction in injection mass (few columns). We refer to ideality here in the same manner as section 4.1, where an ideal curve of the plot of asymmetry vs. injection solvent is not a function of the injection solvent strength. Any dependence on injection solvent strength is a deviation from ideality. If the change in ideality of asymmetry is a qualitative model of the change in retention parameters of the analytes within the injection solvent, then it suggests that the change with injection volume are more likely to elucidate any underlying physicochemical phenomena taking place as a function of thermodynamics, whereas the change in asymmetry with change in mass is probably a change in linearity of the isotherm caused by overloading. Reductions in the asymmetry value with weaker injection solvents than the mobile phase have been reported as caused by focusing the tail of the analytes on the head of the column, and a greater reduction has been seen for low k‘ analytes. This trend seen as a function of the ‘H contribution to retention, and is noticeably greater in figures 3.4A and 3.4B on the more hydrophobic (higher H) Poroshell and Stablebond columns, but not evident on the less hydrophobic polar embedded Bonus-RP phase. With lower injection volume, there is less time spent in the injection solvent, and therefore less focusing effect which is only evident on the least retained compounds as the time spend in the injection solvent is a function of the k‘. With injection solvents that are stronger than the mobile phase, the front of the analyte band is carried by the stronger injection solvent along some length of column. The injection solvent strength at which the asymmetry significantly deviates correlates well with the silanol activity; A term in equation 3.3. The greater silanol activity of the non-endcapped Stablebond phase allows for a greater strength of the injection solvent before significant fronting is seen on the peak when compared to the endcapped Poroshell EC or 300Extend phases. Consequently, we find a correlation of the sensitivity to the ratio of H/A for the column, as seen in figure 3.8A. Other correlations have significantly lower r2 values for the fit: H/S – 0.66, H/B - 0.49, H/C – 0.64. The Bonus-RP does not fit well with the C18 columns, which can be due to the precision of the column parameters in the method or can be indicative of a differing dependence of sensitivity on HS parameters for the amide phases. Overall, this suggests a quick

53 test to predict the sensitivity of a column to injection solvent strength or volume, and given the large number of alkyl-silica columns in the USP-PQRI database, one that may prove useful to chromatographic method development.

A

B 1.1

0.9

R2 = 0.9423 Sensitivity

2 R = 0.9824 0.7

R2 = 0.9274 30in7.5 30in15 30in30 0.5 -25 -20 -15 -10 -5 0 5 10 H/A

Figure 3.8. Comparison of sensitivity values for γ0μg of 2-heptanone in 30uL to (A) bonding density (μmol/m2), number of sorbed acetonitrile layers, and column parameters for the hydrophobic-subtraction model, and (B) to ratios of column parameters, H/A. Open shapes refer to the Zorbax Bonus-RP phase.

54

0.0015

0.0010

0.0005

0.0000

-0.0005 Poroshell EC-C18

300Extend-C18 -0.0010 Excess adsorption (umol/m^2) adsorption Excess Stablebond-C18 -0.0015 Bonus-RP

-0.0020 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 Bulk molar fraction MeCN

Figure 3.9. Measured acetonitrile excess isotherms from equation 3.4a and 3.4b.

Table 3.4. Fitting parameters for equation 3.7. Poroshell Stablebond- 300Extend- Bonus- EC-C18 C18 C18 RP t 3.04 2.37 4.09 2.54 ε 0.79 0.62 0.75 0.60 K(C18) 4.27 5.62 3.79 5.65 K(OH) 0.041 0.042 0.069 0.038 K(C18)/K(OH) 103.2 133.0 54.9 149.7

3.4.3 Comparison to the acetonitrile excess adsorption isotherm Measurement of the acetonitrile isotherm was done for the four studied columns and these are shown in figure 3.9. Positive portions of the isotherm are indicative of adsorption of acetonitrile onto the stationary phase ligand, and negative portions are due to the adsorption of water onto the residual silanols. Values from the fit of equation 3.7 are reported in table 3.4. As can be seen, endcapped phases‘ values are closer to unity and therefore are more homogenous. However the correlation between heterogeneity and sensitivity is weak, r2≈0.βη. Better correlation can be seen with number of acetonitrile monolayers (figure 3.8B, r2≈0.98) and the

55 ratio of the distribution constant between the C18 ligand and acetonitrile (KC18) to the distribution constant between the residual silanols and acetonitrile (KOH), as seen in figure 3.10. The correlation of the distribution constant ratio and sensitivity can be considered similar to the correlation seen in section 4.2 with the ratio of H/A and sensitivity. This suggests that the sensitivity to injection solvent strength is related to the ability of the column to retain both components of the hydro-organic mixture. The greater the retention of components of the injection solvent, the less the change in band shape with changes in the injection parameters. However, in all considered correlations the Bonus-RP phase is an outlier, which further supports that there is a differing dependence for polar-embedded phases. The strong negative correlations between sensitivity and bonding density and number of sorbed acetonitrile layers in figure 3.8B are both indicative of the same phenomena [53], and so offer no additional conclusions. However, the number of sorbed layers and the conclusions of reference [59] on the sensitivity to differing organic modifiers in the injection solvent in the HILIC mode can provide predictions on other modifiers not studied in this work. Ruta et al. found that the injection solvents are modeled with mobile phase solvent strength, and for HILIC can be listed as MeCN < IPA < EtOH < MeOH < H2O from weakest to strongest [59]. It is reasonable to expect the exact reverse for RPLC, and can be correlated with the thickness of the absorption layer, which can be expected to increase with carbon number. From references [52,53], roughly, MeOH forms a single monolayer, EtOH forms two monolayers, and IPA, THF, and MeCN form three monolayers. It is predicted then that injection solvent sensitivity values would trend as MeOH > EtOH > IPA ≈ THF ≈ εeCζ.

56 1.00 y = 0.0024x + 0.657 30in7.5 R2 = 0.986 0.95 y = 0.0035x + 0.5076 30in15 R2 = 0.9999 0.90 y = 0.0035x + 0.4118 30in30 2 0.85 R = 0.978

0.80

0.75

Sensitivity 0.70

0.65

0.60

0.55

0.50 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 K(C18)/K(OH) Figure γ.10. Correlation between K(C18)/K(OH) and sensitivity for values of γ0μg of β- heptanone. Closed symbols: C18 phases. Open symbols: polar embedded.

An interesting case to predict would be a ternary system where the injection solvent is not matched by the organic modifier of the mobile phase. In 2010, Coym measured the energetic contribution to retention using a ternary mobile phase of X% MeOH:(50-X%) MeCN:50% H2O [60]. It was found that the enthalpic contribution to retention remained constant, whereas the entropic contribution decreased with increasing MeOH concentration, as well as general trends in the coefficients of the Linear Solvation Energy Relationship model of retention. Contribution of cavity formation v, excess polarizability e, and hydrogen bond basicity a increase as MeOH is added to the mobile phase and favor retention. Polar interactions s and constant c decrease with MeOH concentration and favor elution, whereas the hydrogen bond acidity is roughly constant. Cavity formation and hydrogen bond basicity are the major factors in retention, so predictions can be made mostly upon the general trend with these two parameters. Then, the same prediction as above would hold within a ternary system where an injection-solvent modifier mismatches with the mobile phase: εeOH > EtOH > IPA ≈ THF ≈ εeCζ. As can be seen by molecular simulation in reference [61] by Rafferty et al., protic solvents replace sorbed water at silanols, which would serve to reduce the silanol effect on injection solvent sensitivity and relatively increase the ligand effect on injection solvent. Expectedly, then, acetonitrile as an injection solvent modifier injected into an aqueous methanol mobile phase would produce a less ideal

57 sensitivity (lower s value from equation 3.9) than the exact juxtaposition - methanol injected into an acetonitrile mobile phase.

3.4.4 Injection solvent sensitivity of lidocaine Similar to methyl ketones, lidocaine is soluble in a range of aqueous-MeCN mixtures, and with a pKa of 7.9, any effect of a change in ionization state due to the injection solvent should be apparent. It has been shown that ionizable compounds show an increased sensitivity to injection solvents compared to non-polars [59,62]. Chromatograms of lidocaine were obtained with stepped injection solvents as in section 3.4.1 with one-tenth the injection mass. As the methyl ketone data, near ideal sensitivity values (s > 0.9) are measured for the 1.βημδ injection volume on each column, and decreased ideality with increased injection volume. For the Stablebond-C18, each injection volume and strength produced an easily integrated peak that followed the general trends of the methyl ketones in that less injection strength and less injection volume produced a narrower concentration band, and thus a higher plate count. However, there was significant band splitting on the endcapped, non-polar embedded Poroshell EC-C18 and Zorbax 300Extend-C18 columns above some threshold MeCN concentration of the injection solvent. With a high organic injection solvent, the lidocaine injection band was dragged down the column producing a shoulder or a split peak. As injection volume is increased, the threshold organic concentration for the shoulder to appear was decreased, but was independent of the injection mass. Though, it can be expected that injection masses that overload the column would reduce the threshold organic concentration. It could be considered that the peak‘s shoulder could be mitigated by buffering the injection solvent, the addition of an ion pairing reagent, or by changing the column temperature. To the 10%, 20%, 30%, 40%, and 80% MeCN injection solvents, citric acid and monosodium citrate was added, isotonic with respect to the mobile phase. There was no measurable change in the peak shape. 1mM and 10mM Tetrabutyl ammonium hexafluorophosphate (TBA+, HFP-) was added to the buffered 80% MeCN injection solvents to produce a significantly fronting peak, which was found to be coelution of the fronting TBA+ with the front of the lidocaine-HFP pair. No additional change in peak shape was measured. The column temperature was increased from 25°C to 50°C to produce less of a shoulder, but not significantly. Figure 3.11 compares the results from these runs.

58

Figure 3.11. δidocaine peak shape with changes in the 1ημδ of 80% εeCζ injection solvent onto the Poroshell column. X and Y axes have been scaled to ease comparison. From top to bottom: A – 1.00mL/min flow rate, 25°C. B – 0.10mL/min flow rate. C – 1.00mL/min, 10mM citrate. D – 10mM citrate, 50°C. E – 10mg/mL TBA-HFP added. F – 1.0mg/mL TBABr only. G – 10mM citrate, 50°C, 10mg/mL TBA-HFP. H – Blank at 190nm.

59

Figure 3.12. Lidocaine peak shape with changes in the injection volume of 10% MeCN injection solvent onto the Bonus-RP column. Y axis has been scaled to ease comparison. From top to bottom: A – 1.βημδ of 0.4mg/mδ lidocaine. B – 7.ημδ of 0.4mg/mδ lidocaine. C – 1ημδ of 0.4mg/mL lidocaine. D – 1ημδ of 0.βmg/mδ lidocaine. E – 1ημδ of 0.1mg/mδ lidocaine. F – γ0μδ of 0.1mg/mδ lidocaine, day 1. G – γ0μδ of 0.1mg/mδ lidocaine, day 2.

For the Bonus-RP column, lidocaine eluted at the void in all injection solvent strengths above 10% MeCN invariant of the injection volume. Unlike the C18 columns, as the injection volume and strength were increased, the lidocaine peak produced a shoulder on the tail, invariant with injection mass. This result does not agree with an accepted reversed-phase retention mechanism, as the retention of an analyte should decrease with increasing solvent strength. Additionally, for the 10% MeCN injection solvent, the retention is a function of the injection volume, as shown in figure 3.1β. For the 1.βη and 7.ημδ injections, the lidocaine peak is predominantly at the void. For the 1ημδ injections, there is a split peak at the void and at a k‘=0.41, invariant with injection mass. For the γ0μδ injection, the lidocaine mass is predominantly at k‘=0.41, occurs in each of the triplicate injections, and is reproducible from day to day. We do not have an explanation for these anomalous results, as a change in column selectivity with injection volume is not present in theory or the experimental literature.

60 3.5 Conclusions By stepping the injection solvent strength, injection mass, and injection volume, we have measured the sensitivity of the eluting band shape to the injection solvent on four differing columns. Peak distortion occurs with increasing injection volume and solvent strength. However, with a sufficiently small injection volume (1.βημδ for 4.θmm i.d. columns) the resulting peak is mostly invariant with injection solvent strength. For methods that require larger injection volumes, the only recourse to sharpen peak shape is to weaken the injection solvent. In the case of the Poroshell EC-C18 and Zorbax Stablebond-C18 columns, the injection mass affects the change in asymmetry with respect to injection solvent strength as a function of retention factor, with the longer retained 2-heptanone being affected to a greater extent than acetone. Sensitivity of the C18 ligands to injection solvent effects was found to correlate with calculated bonding density, hydrophobic-subtraction model coefficients, thickness of the MeCN adsorbed layer, and the ratio of the distribution constants for the binary mobile phase sorbing to the stationary phase ligand or silanols. These correlations support a conclusion that decreased ligand activity (H or KC18) and increased silanol activity (A or KOH) provide a consistent peak shape with changes in injection volume or solvent strength. The sensitivity of the polar embedded phase does not correlate with any measured value, and showed anomalous behavior with the retention of lidocaine as a function of injection volume and solvent strength.

61

CHAPTER FOUR

SIMULTANEOUS TWO-DIMENSIONAL PLANAR CHROMATOGRAPHY

4.1. Introduction Davis and Giddings [8], using a statistical model, showed that the probability of a single chromatographic peak consisting of a single compound is a function of the technique‘s peak capacity. Adding a second orthogonal chromatographic dimension gives the best return for significantly increasing the peak capacity of a technique, though at the price of increased complexity. Both the method development and data analysis for a 2D technique are non-trivial when compared to that of a 1D experiment. However, for sufficiently complex samples, especially within "-omics" research, 2D techniques are measurably better suited than 1D techniques [63] The largest drawback of 2D chromatography is the lengthy time of analysis. The Murphy, Schure, and Foley [11] (M-S-F) criterion implies that the second dimension must be significantly faster than the first. The general methods to adhere to the M-S-F criterion are to slow down the first dimension, speed up the second dimension, or a combination thereof. Simultaneous separations are not subject to the M-S-F criterion and provide a solution to this problem. The remaining problem intrinsic to serial 2D techniques is the subsequent dilution of an analyte‘s concentration from one dimension to the next. Depending on the method there can be a two fold to a six fold decrease in the peak concentration of an analyte from the injector to the detector due solely to the addition of a second dimension [64]. Current chromatographic band- broadening theory applies for the dilution factor in each dimension individually. Additionally, the interface between the two dimensions is another cause for dilution within a 2D technique.

62 Planar mediums, including planar chromatography and gel electrophoresis, do not include the injector interface dilution factor and therefore have lower overall dilution factors caused by instrumentation. Dilution is a problem with analyte peak concentrations that are near the limit of detection of the detector. Parameters that combat this dilution in 2DLC include selection of column diameters and utilizing sample focusing on the second dimension. The former is well characterized and the latter is only available with specific column choices of the second dimension [12]. All previous discussions of simultaneous planar chromatography have used capillary action and electrophoresis as the separation dimensions [65-67]. None have addressed the requirement of forced flow methods to enhance the utility of the planar chromatography method in relation to modern HPLC. There are three available forced flow methods, pump driven flow, electro-driven flow, and centrifugal flow. Though, only two of these lend themselves to the requirements of a simultaneous instrument: pump driven flow and electro-driven flow. Planar Electrochromatography has been recently studied to maturity by Nurok's group [68-70], Dzido's group [71-75], and the Dorsey group [76,77]. We are reporting an innovation in 2D-TLC. By spotting the sample on one corner of a TLC plate, applying a pressure driven solvent flow in one dimension, and an electrical potential perpendicular to the pressure driven solvent flow, the sample will be separated in two dimensions simultaneously. This will give the peak capacity of a true 2D system, but with the time of analysis, the dilution factor and limits of detection of a 1D system.

63 A)

B) Neg Pos C) Waste

(-) (+)

Inject

Pump

Figure 4.1. Instrumentation of the 2D Planar Chromatography setup. A) 2D block positioned within throat of hydraulic press without TLC plate. B) Digital mockup of 2D block. C) 2D block with sealed TLC plate and glass cover. This image has been mirrored to correlate with the mockup and resulting chromatograms. All results will show an injection marker on the bottom left. However, the TLC is placed sorbent down on the block, and injected on the right hand side.

64 4.2 Instrumentation The two-dimensional instrument includes four elements: laboratory press, high voltage power supply, syringe pump, and two-dimensional block. The laboratory press is a Wabash Hydraulic Press (Wabash Metal Products Company, Wabash, IN, USA) with a Norco/KYB model 76412G 12 ton hydraulic jack with gauge port (Norco Industries, Compton, CA, USA) and is equipped with an Enerpac G2517L 0–6000 psi (0–400 bar) hydraulic pressure gauge (Enerpac, Milwaukee, WI, USA). This press was used in a previous pressurized planar electrochromatography study [77], and is depicted in figure 4.1A. Voltages were set and produced by a Glassman EH10P10.0 high voltage power supply (Glassman High Voltage, Whitehouse Station, NJ, USA). This power supply can deliver a 10 kV maximum potential and up to 10mA maximum current. Electrodes are connected to the separation block as shown in figures 4.1B and 4.1C. Current was measured by the power supply's 1/10th monitor electrodes connected to a BK Precision 390A Digital Multimeter (DMM) (Yorba Linda, CA, USA). The mobile phase was delivered to the pump inlet by an ISCO model 100DM syringe pump controlled by ISCO series D pump controller. Three blocks have been used throughout this study. The current block is the latest generation of a series of prototypes designed and built by our in house machine shop. It is a single piece of polypropylene measuring 11.43cm (4.5 in.) on each side, and 1.25cm (0.5 in.) thick. Another block of polypropylene or glass of the same dimensions as the TLC plate is used as a top cover for the 2D block. To seal the TLC plate to the block, the silicone is applied first by 10mL disposable syringe, then molded by a metal spatula. A slate of PVC slightly larger than the block, of arbitrary dimensions, aligns the 2D block within the throat of the press to ensure reproducible positioning of the block in relation to the hydraulic press head, as shown in figure 4.1C. On each of the four edges of the block is drilled a trough of nominal depth and width. Two opposite troughs are tapped from the edge for a male 10-32 finger tight HPLC fitting to allow mobile phase inlet and outlet to the plate. The other two troughs on opposite edges are outfitted with electrodes for the electrochromatography dimension with 0.51mm (0.0020 in.) platinum wire that is pushed through a small hole and pulled to span the entire length, ensuring that the potential of the entire trough is homogeneous. The wire entrance is sealed on both sides with silicone sealant to prevent mobile phase from leaking around the platinum wire. Inside the

65 trough, at the end of the platinum wire there is bead of silicone sealant to prevent any lateral or vertical movement of the fragile wire. The wire is spot welded outside the block to a female banana clip plug of arbitrary color, since it is necessary to reverse the polarity of the electrode depending on the charge of the analytes. The female banana clip plug is reinforced to the block to prevent breaking the soft platinum wire during routine maintenance. The electrode troughs are filled entirely with sea sand (Fisher Scientific, Fair Lawn, NJ, USA) to increase the pressure drop, ensuring that mobile phase entering from the pump trough does not preferentially exit through either electrode trough, but rather through the correct pump flow exit trough. The sand also decreases the extra plate volume of solvent required initially to prepare the system for an injection. To retain the sand within the troughs, a 5% acrylamide solution is added to the trough and polymerized to a hydrogel. The pump inlet trough is outfitted with a cellulose filter paper rolled to fit within the trough to act as a flow disperser such that the pressure, and therefore the flow velocity, of the mobile phase from the pump is homogeneous across the width of the plate. The trough system used here is a design derivative of the 1-D pressurized planar electrochromatography (PPEC) device devised by Dzido's group [72]. The lengths of tubing connected to each trough are connected to specific components to facilitate initial preparation of the enclosed TLC plate. The plumbing from the pump includes a high pressure switch valve, with one inlet, one outlet to waste, and a second outlet to the inlet of the 2D block. This switch valve acts as a purge valve for the pump and is later necessary to quickly remove residual fluid pressure from the block at the end of the chromatographic run. For the outlet trough, the tubing goes to a shut-off valve. This shut-off valve of the outlet trough is fully open during the initial equilibration of the system, but is partially closed during the chromatographic run to slightly increase the backpressure of the trough, giving a reasonably homogenous flow velocity across the width and breadth of the plate. Drilled into a corner of the block, near the pump inlet trough, at an upward angle is the injection port. This injection port allows spotting of the plate without removing the TLC plate from under the tangential, hydraulic pressure, which is different from all PPEC designs to date. The injection port hole is drilled 2mm in diameter and of necessary length to go from the edge of the block (entrance) to a reasonable but equal distance (exit) from the two closest troughs. Within the injection port hole is inserted a glass capillary tube, the top of which has been filed to an angle necessary to be flat against the top of the contact plate. Excess volume of the top of the

66 injection port exit where the capillary tube meets the contact plate is carefully filled in with silicone sealant, including a thin film over the filed opening of the capillary tube, to prevent mobile phase from leaking extensively from a worn injection port septum. The injection port entrance is expanded and tapped for a male 10-32 HPLC fitting. The attached HPLC fitting is capped with a modified standard PTFE/silicon headspace cap. The headspace cap is cut to proper size with a razor, and has the PTFE covering shaved off to form an 'injection port septum'. The septum has been determined empirically to be the weakest point in the sealing of the entire block, limiting the mobile phase pressure tolerance and therefore defining the maximum volume flow rate of the system. At volume flow rates greater than 3.5mL/min the silicone sealant around the septum generally fails and begins to leak significantly. Also, the septum is limited to an average of ten injections before becoming too worn to effectively seal, and will begin to leak. To qualify the designed instrumentation, three criteria must be met: 1) tangential pressure from the hydraulic press must be homogeneous across the 2D block, 2) the electric gradient must be homogeneous across the breadth of the plate but with a linear voltage drop across the width from electrode to electrode, and 3) the pumped flow must be homogeneous across the width. To test the tangential pressure, a pressure sensitive film from Sensor Products (Madison, NJ) was compressed between the 2D block and the polypropylene cover. Figure 4.2A shows the result from a 250 bar tangential pressure. The shade of pink indicates the pressure applied, with darker shades measuring a higher pressure. Across the TLC plate, the tangential pressure is roughly homogeneous, though some points see a much higher pressure. This is due to sand from the electrode trough making pits in the soft polypropylene block. These points of heterogeneity are not expected to affect the chromatography significantly. The homogeneity of the force vectors for the two dimensions will affect the chromatography significantly. Linear voltage profiles in PPEC were measured and discussed by Tate and Dorsey [76,77] and are a function of the amount of time the electric potential has been applied. After 20 minutes the voltage drop across the plate is linear (r2 > 0.99). It is reasonable to expect with the similar block used in this work will require an equivalent time, and will retain the linear voltage drop over the analysis time. The homogeneity of the pumped flow was measured by spotting FD&C blue 1 intermittently across the width of the TLC plate offline, then placed onto the block and fluid

67 forced through it. Initial spots were marked on the plastic backing of the TLC plate, and are ~1cm apart, as marked on the bottom of figure 4.1B. Measurements from the initial spot to the center-of-mass for each spot are compared in Table 4.1. The resulting migrations are noticeably inequivalent, but specifically non-parabolic. In prior prototypes, the flow velocities were of equivalent magnitude, but directed toward the outlet port, resulting in a parabolic shape. Increasing the backpressure of the outlet trough by partially closing a shut-off valve tend to homogenize the flow directions, though not ideally. Alternatives, such as multiple outlet ports were not attempted.

A)

B)

1.0 ml/min

Figure 4.2. Qualification of Simultaneous 2D Planar Chromatography instrumentation. A) Tangential pressure from hydraulic press as measured by pressure sensitive film. B) Migration of FD&C blue 1 spotted offline intermittently across the width. Only the pump flow was used.

68

Table 4.1. Migration of FD&C blue 1 from initial offline spot to center-of-mass.

Spot detection was done using a Canon Canoscan 4400F with a range of resolutions up to 4800 x 9600 (width x height) samples per inch (spi). Migration distances are measured in pixel counts from an initial spot. With a set spi for a scanned image, migration distances, dR in micrometers are calculated by the equation: dR=(dP x βη400μm)/spi, where dP is the distance measured in pixel counts. Images from the scanner are inputted into an in-house developed computer script written in the Python language. The Python script uses the Python Image Library to load pixel data into a NumPy array. The NumPy array is then labeled with SciPy's 'label' function to isolate individual spots and determine the center of mass (COM) of each. Pixel migration distances, and therefore micrometer distances, are calculated from COM of the initial spot location to the resulting final COM. The Python script will also do a background correction on the image so that only the chromatographically relevant spots are shown, which can be exported to MATLAB for 3D visualization.

4.3 Reagents Serial vs. Simultaneous study Methanol was purchased from Sigma (St. Louis, MO, USA). Sodium acetate and methylene blue were purchased from Fisher Scientific (Fair Lawn, NJ, USA). Glacial acetic acid was purchased from EMD Chemicals (Gibbstown, NJ). Water used in the buffer was purified by a Barnstead (Dubuque, IA, USA) NANOpure II water purification system to a resistance of ~18εΩ. Acetate buffer was prepared by mixing equal volumes of 10mM sodium acetate and 10mM acetic acid and adjusted to a pH of 4.7 as measured by an Orion SA520 pH meter (Orion Research Inc., Beverly, MA, USA). Buffer was then mixed with methanol to produce a 75% MeOH:25% acetate buffer mobile phase. Analtech HETLC-RPS plates (Catalog number 54377, Analtech, Newark, DE, USA) were purchased for this study. These are glass backed plates with a 1η0μm thick reversed phase

69 silica-based sorbent. The plates were cracked along the manufacturer's scoring to 5cm x 5cm dimensions. The phase was washed overnight in a methanol bath, then baked at 120 °C for 20 minutes, and left in a desiccator until used. Plates were sealed by a previously published method [77] with RTV silicone sealant ―Sensor Safe Blue RTV‖ (Pζ: θB, Permatex Inc., Solon, OH, USA) a day prior to the experiment and left to cure overnight.

Simultaneous separation study FD&C Blue #1 and Red #40 were purchased from a local grocer (McCormick brand). Acetonitrile and methanol were purchased from Sigma (St. Louis, MO, USA) and mixed, by volume, to produce a 50% MeCN:50% MeOH mobile phase. TLC Silica Gel 60 F254 plates were a gift from EMD Chemicals (Catalog number 5735-7). These 20cm x 20cm plastic backed plates were cut to four 9.5cm x 9.5cm squares using a hot soldering iron fitted with a knife tip, available at a local hobby store. Cutting the plates via hot knife, rather than scissors, prevented the phase from flaking off the plastic backing and allowed for very accurate dimensions to fit within a 11cm x 11cm x 3mm thick silicone gasket cut to fit the instrument block. The plates were stored in MeOH until used.

Amino acid separation Alanine, lysine, histidine, and arginine were purchased from Sigma (St. Louis, MO,

USA) and dissolved in 65% MeOH: 35% H2O with a concentration of 200μg/mL. The plastic backed normal phase plates were used for this study.

4.4 Results and Discussion 4.4.1 Serial vs. Simultaneous study A preliminary characterization of the technique is to run a sample in each dimension solely, measure the migration distance, then compare the migration distance when using the simultaneous 2D method. If the migration distance is reasonably similar then it can be concluded that the simultaneous 2D instrumental design has promise. Differences in migration distance of the compound in the two dimensions can give a qualitative comparison on flow rate, but without an internal standard, a retention mechanism comparison cannot be inferred.

70 The blocks used in this study were earlier 1D prototypes of the block described herein for each component dimension, or an earlier 2D prototype formed from an existing PPEC instrument published earlier [77], modified to include a pump flow dimension. Pre-sealed glass backed reversed phase TLC plates were placed sorbent down onto the block, then pressurized by the hydraulic press to a tangential pressure of ~69bar (1000psi) measured at the gauge. The flow is then turned on, either by pump, power supply, or both, for approximately 20 minutes for equilibration of the phase. Once equilibrated, the voltage and flow are halted, tangential pressure removed, and plate spotted offline with 1.0μδ of 0.1mg/mδ aqueous methylene blue solution via a 10μδ Hamilton syringe. The plate is then quickly replaced, tangential pressure reapplied, and the flow resumed for 5 minutes. After the run time, the tangential pressure is released, the plate removed, dried, and scanned. Table 4.2 lists the center-of-mass migration distance of the spots from the initial spot placement as marked by a graphite pencil. Resulting spots are visually compared in figure 4.3. The electro-driven flow migration distances vary somewhat. The ~12% difference in the PEC migration is an acceptable error considering the infancy of the technique. Systematic errors were found in the offline spotting procedure. The TLC plate must remain wet during spotting and the plate must be reproducibly aligned on the block between runs. These errors were corrected in the next prototype by online spotting through the injection port, however the proximity of the injection port to the electrode trough prevented a 1D measurement for pump driven flow on the newer blocks.

71 (-) (-)

1.0 ml/min

(+) 1.0 kV 1.0 kV (+) 1.0 ml/min

Figure 4.3. Single-dimensional migrations of methylene blue in 75% MeOH and 25% 10mM acetate buffer on reversed phase plates are comparable to the component migrations in simultaneous mode. 1D experiments were performed on prototype blocks. 5.0 min run time each.

Table 4.2. Comparison of migration distance of methylene blue between single dimension runs and simultaneous two-dimensional run. 1D 2D % difference Electro-driven flow 14.31mm 12.62 mm 11.8% Pump driven flow 5.84 mm 5.67 mm 2.9%

4.4.2 Simultaneous separation study With a sufficient proof-of-concept of the force vector addition in the previous section, the next aim was to show a separation of similarly charged dyes with the simultaneous 2D mechanism. The second block in the prototype series was produced specifically for this study. It differs from the description in section 4.2 only in that 0.95cm (0.375 in.) from the edge of the block is recessed to allow for a silicone gasket to seal the mobile phase within the apparatus. This gasket is made from 1.5mm thick, shore A 20 hardness silicone rubber sheet cut to dimensions to fit the outer edge of the block. On the ~9.5cm (3.75 in.) elevated square, and surrounded by the silicone gasket, is where the TLC plate is placed. Using a silicone rubber

72 gasket allowed TLC plates to be developed more frequently, because the sealant did not have to be given time to cure. The plastic backed normal phase plates are placed sorbent down on the 2D block positioned such that the edges of the plate cover all four troughs. The silicone gasket is placed around the plate in the recessed portion of the 2D block, and the top cover on top of the gasket. Tangential pressure is then applied and the flow and voltage switched on with the outlet shut off valve open and the switching valve set to the 2D block. Once flow is consistent from the pump flow exit, the valve is partially shut by turning the stop cock to 45°, then switching the flow and voltage off. Now that the plate is equilibrated, and sufficiently wetted, the tangential pressure is not removed, due to the new injection port. A Hamilton η.0μδ syringe with a specially ordered 76mm (3.0 in.) cemented needle is loaded with 0.5 μδ of an aqueous mixture of FD&C Blue #1 and FD&C Red #40. The syringe is pushed into the septum through until it is felt scratching the surface of the sorbent. Care must be taken such that the needle tip is facing up toward the plate, then the plunger is depressed slowly until the mixture is loaded on the plate. 0.75ml/min flow and 1.0kV potential are then reapplied for 5.0 minutes. After the run time, flow from the pump is redirected to waste via the switching valve, and the voltage is shut off. The shut-off valve is closed, and the tangential pressure is released. The top cover and gasket are removed from the block, and the TLC plate removed by carefully sliding a razor blade underneath a corner, then lifting it free. The TLC plate is dried by use of a heat gun, and then scanned. Figure 4.4 shows a five minute separation of two negatively charged species, both a total -2 charge. Red 40 has two sulfonate groups for its charge, whereas Blue 1 has three sulfonates and one iminium cation. These two compounds, under the solvent system used (50:50 ACN:MeOH), do not resolve via 1D electrochromatography (figure 4.5A) or classic TLC (figure 4.5B) with capillary flow. With the simultaneous 2D system described in this manuscript, electochromatography coupled with the pump driven dimension, the ampholyte of blue 1 is resolved from the red 40.

73

Figure 4.4. MATLAB graph of pixel data using the described simultaneous 2DPC technique. Pump flow occurred in the vertical, and electro driven in the horizontal. Initial spotting occurred at the lower left spot (160,2816), Blue #1 migrated to (1793,1067), Red #40 to (1845,831).

The classic TLC was spotted at the first mark with the same blue 1:Red 40 mixture used in the simultaneous 2D separation. The bands were focused to ~1mm thin from initial circular spots of ~2mm in diameter to a second mark, and then advanced until the solvent front migrated nearly to the end of the plate. Migrations are as follows, measured by the Canon scanner:

5.808cm solvent front distance, 5.063 cm migration distance for Red #40 (Rf = 0.872), 4.860 cm migration for Blue #1 (Rf = 0.837). Note that the variance of each spot is sufficient to prevent baseline resolution. No attempts to modify the separation parameters were attempted to achieve the separation.

74 A) B)

Figure 4.5. Spots remain unresolved using A) electrochromatography and B) conventional TLC.

The electrochromatography was run on the block described within this section, with only the voltage turned on. The dyes were migrated for 5 minutes with 1kV. The spots begin to resolve electrochromatographically as in figure 4.4, however they remain unresolved until the run time is extended to 25 minutes. Higher voltages are possible with our current instrument, and could resolve the two with a lessened run time, though the additional joule heating to the plate requires temperature control of the plate. Efforts to reproduce the separation shown in figure 4.4 showed frequent systematic errors of unknown cause. Instead of the spots migrating, a single smear of dye would progress from the injection port to the outlet along the edge of the electrode trough. However, a series of single component injections of diluted FD&C blue 1 was made and the migration due to the force vectors was measured. The migrations of three repetitive injections that did not show the systematic error are listed in table 4.3. When the system error is discarded, the 2D spot migration is reasonably reproducible. Figure 4.6 plots the vector component migrations of blue 1 as a function of run time. For each data point, a new TLC plate was equilibrated and spotted through the injection port and developed by the 2D method. After a preset run time, the flow and voltage

75 were halted and the migration measured. When the system error is discarded, the migration velocities of the components are reproducible and constant.

Table 4.3. Reproducibility of FD&C blue 1 in three runs by simultaneous method.

40.0 y = 7.5492x - 0.1724 35.0 R2 = 0.998 Pump 30.0 PEC 25.0

20.0 y = 5.715x - 0.0635 R2 = 0.9996 15.0 Migration (mm) 10.0

5.0

0.0 0.00 1.00 2.00 3.00 4.00 5.00 6.00 Run time (min)

Figure 4.6. Migration distance of blue 1 as a function of run time.

The experiment of figure 4.4 was performed again with an adjusted flow rate and run time. The flow rate was adjusted from 0.75mL/min to 1.0mL/min and the run time was adjusted from 5.0min to 3.75min. Multiplication of the two parameters suggests that within the pump dimension the spots will migrate to the same point on the plate. The 1.0kV was kept constant between the two experiments, but with the differing run times analytes will migrate less distance. The resulting spots in the reduced run time measured 74.6% of the electrochromatographic

76 migration and 97.8% of the pumped migration, as compared to the 5.0min run. When correcting for the systematic error, the force vectors can be controlled separately.

4.4.3 Amino acid separation A third block was designed and produced as outlined in section 4.2. It was noted that the placement of the silicone rubber gasket affected the reproducibility, so the current block requires a cured sealant. The elevated square in the center of the block was removed as is not necessary. To increase the number of runs that could be developed, three of these blocks were produced.

1 4 2

3

Figure 4.7. 4.0 minute separation of the four amino acids.1 – histidine, 2 – arginine, 3 – lysine, and 4 – alanine. Spot assignments were confirmed by ESI-MS.

To show a separation of a multicomponent mixture, a γ.0μδ injection of β00μg/mδ solution of each of the basic amino acids and alanine were run at 1.25mL/min with a mobile phase of 65% MeOH:35% H2O with 1% acetic acid. The electric potential was set to 0.5kV. The

77 run time was set to 4.0min. Detection was done by spraying the developed plate with 0.γη%(wt/v) ninhydrin solution, then held under a heat gun to produce a spot of Ruhemann‘s purple for detection. Peak assignments were done by cutting the spots from the plate with a razor and diluting with 100μδ of 50% MeOH:50% (H2O with 1% acetic acid). Samples were sonicated for 10 minutes to break up the silica aggregates, spun in a centrifuge, then the supernatant was filtered through a 0.βμm nylon filter. These samples were injected into a Thermo/Finnigan δCQ duo (Waltham, εA) mass spectrometer via direct infusion at ημδ/min, spray voltage at η.ηkV,

N2 sheath gas at 40 units, and capillary temperature at 200°C. One molecular ion of the protonated, singly-charged amino acid was evident per spot, along with characteristic derivatives resulting from the ninhydrin reaction.

4.5 Conclusions Simultaneous two-dimensional planar chromatography is a novel approach to separation science. Historically, all two-dimensional techniques operate in serial mode, where a first dimension‘s eluate is injected onto a second dimension. When the two dimensions are performed simultaneously the total analysis time can be shortened, there are no second dimension time limitations, and the dilution factor is reduced to a value on the order of a single dimension. Initial results on an in house design are promising for a new tool that could provide better analytical measurements. Here, we report on the efficacy of an instrument design for a simultaneous two- dimensional separation using pump driven flow perpendicular to an electro-driven flow within one stationary phase.

78

CHAPTER FIVE

SUMMARY, SIGNIFICANCE, AND BEYOND

Complex problems generally require complex . For the separation, identification, and quantitation of components within complex mixtures, two-dimensional chromatography will be the method of choice, despite the difficulty of method development. This investigation has targeted points within the two-dimensional liquid chromatography system in an attempt to better understand the fundamentals and find practical solutions. The three issues of two-dimensional liquid chromatography are: 1) To preserve the resolution produced in the first dimension, an eluting analyte peak from the first dimension must be sampled at least four times across the band width. 2) The mobile phase of the first dimension becomes the injection solvent for the second dimension. 3) Each additional dimension adds further dilution of the analyte band. The following is a discussion of the state of the chromatographic community on these issues and the perspective of the research presented.

Reduction of Reequilibration Time in Gradient Elution Reversed Phase Liquid Chromatography Regarding the speed of the second dimension, the linear flow rate and gradient selectivity are maximized by the analyst. Between one run and the next injection, the column must be reequilibrated to the initial gradient conditions. Empirical rules-of-thumb have been offered in the past that suggest a length of time required for reequilibration, or a volume of fluid that must pass through the column. Cole and Dorsey [18] conjectured that adding a small amount of 1- propanol to the mobile phase would robustly solvate the phase, requiring less volume of solvent and/or time for the reequilibration. The time and/or volume conjecture has lasted because there

79 was no need within the community to find the fundamentals of reequilibration. It was considered a solved problem. Analysts will typically do a series of chromatographic runs overnight where the length of time saved by reducing reequilibration time is small and is generally not the rate limiting step. However, in two-dimensional liquid chromatography, the number of injections made on the second dimension will be very large. Reducing the amount of time required between injections was the goal of this study, and is significant to the community. The novel conclusion drawn from the data presented in chapter 2 is that the time and volume are not the determining factor on reequilibration, nor is the diffusion of the organic modifier from the stationary phase. Because of the ability to make well timed injections at the interface of the two fluids, it was found that the limiting factor to reequilibration is the pressure applied to the aqueous phase to wet the pores of the silica. This fundamental property has been understood as the Buckley-Leverette theory in fluid dynamics to model the two-phase fluid flow in forcing oil from oil sands using water. The analogy to chromatographic reequilibration is apt, though modifications are required to apply the theory to the special case of chromatography. Future endevours would be useful to solidify the correlation between pressure and the expulsion of organic modifer. The reequilbration dependence on pressure was not tested by GC-FID. It is suspected that the initial expulsion of organic modifier occurs more quickly and with a more negative slope to the measured %MeCN in the eluent. It is suspected that the removal of acetonitrile from the Gibbs surface is pressure dependent, though one previous study by Buszewski et al. suggested differently by measuring the excess absorption isotherm under differing flow rates [54]. An alternative experiment to understand the application of the Buckley- Leverette theory to chromatography would involve measuring the variables and dependencies within the equation and solving for a maximized shock front velocity given the limitation of current instrumentation. This may prove useful to produce a simple, empirical equation to replace the multitude of rules-of-thumb that surround column reequilibration. However, despite this work, the exact chemistry of pressure effects on chromatography remains unknown.

Quantifying Injection Solvent Effects in Reversed-Phase Liquid Chromatography Every analyst that has made an injection onto a chromatographic column has given notice to the injection solvent used, however there has has been no published study to quantify the effect of the injection solvent on the chromatographic performance. Through empirical work, the

80 chromatographic community has decided that the use of the mobile phase as a diluent is best, and any deviation from this solvent choice is made on the basis of solubility alone. It was suggested at the beginning of the reported study that the new, commercially available superficially porous columns have an increased sensitivity to injection solvents due to the very sharp concentration bands being outside the linear region of the isotherm. Though reasonable, initial results proved that this posit was not the case. The methodical measurements of this work support a conclusion that the distortion of the band shape of an analye due to the injection solvent is dependent on the chemistry of the stationary phase. Decreased ligand density and increased silanol activity provide a consistent peak shape with changes in injection volume or solvent strength. This result is surprising and is in direct contrast to modern stationary phase advancements. Silanol activity has been considered the bane of liquid chromatography and is purported to be the cause of poor chromatographic performance and asymmetric peaks from surface heterogeneity. To reduce the silanol activity, there are three methods that are in common use in commercial chromatographic columns: 1) endcapping derivatized silica with a trimethylsilane reagent to neutralize residual silanols, 2) use of bulky side groups on the silane ligand to sterically inhibit residual silanols, and 3) use of highly pure silica substrate or using organic hybrid silica substrate. Along with particle diameter decreases, these methods are considered to be the main advancements of modern liquid chromatography. This work is of particular significance for two-dimensional liquid chromatography. In two-dimensional liquid chromatography, the injection volume, injection mass, and solvent strength are not set by the analyst directly. The injection volume is set by the product of the time of analysis for the second dimension and the volume flow rate of the first dimension. The time of analysis of the second dimension is optimized within the limits of the instrumentation and the required selectivity to separate the components of the sample. The volume flow rate of the first dimension is dependent on the time of analysis of the second dimension as well. Issue 1 listed above is commonly called the Murphy-Schure-Foley criterion for two-dimensional chromatography which relates the required speed of the second dimension to the thinnest concentration band of the first dimension. Given the van Deemter relationship of band spreading to flow rate, the required speed of the first dimension is a function of the attainable speed of the second dimension. A general rule-of-thumb is that the first dimension must be at least 100 times slower than the second. This leads to injection volumes onto the second column that are larger

81 than is common for the ubiquitous one dimensional liquid chromatography. The holistic effect of the injection volume and solvent strength on the resulting peak capacity of the system has not been studied, and the results of this reported study raise an interesting, experimentally-testable question: ―Is it better to use an endcapped or non-endcapped stationary phase as the second dimension column?‖ Previous work on simplifying column selection for two-dimensional liquid chromatography has focused on the orthogonality of the retention mechanisms [78,79] or selection of the first dimension column with a constant second dimension column [80], a carbon clad zirconia stationary phase. It is then suggested that future work consider the holistic, multivariate approach to maximizing peak capacity in the two-dimensional system, as well as collecting the necessary data to test the predictions of section 3.4.4.

Simultaneous Two-dimensional Planar Chromatography Simultaneous two-dimensional planar chromatography is a novel approach to separation science. Historically, all two-dimensional techniques operate in serial mode, where a first dimension‘s eluate is injected onto a second dimension. When the two dimensions are performed simultaneously the total analysis time can be shortened, there are no second dimension time limitations, and the dilution factor is reduced to a value on the order of a single dimension. Initial results on an in house design are promising for a new tool that could provide better analytical measurements. This seemingly simple approach is poised to be minimally an asterisk in the declarative statements within the two-dimensional chromatographic community. Maximally, it is a solution to all three proposed issues within multidimensional chromatography, though not without limitations. The speed of the component dimensions is only a function of set parameters, there is no solvent compatibility issue between dimensions, and the dilution factor is on the order of a one dimensional technique. It is considered that the remaining hurdles to maturity are engineering difficulties, and not the attainability of concept. This study, however, was devised from fundamental considerations and was not driven by a necessary application. Once a suitable application is married to the development process, it is suspected that the engineering hurdles shall be made trivial.

82

REFERENCES

[1] J.C. Giddings, Dynamics of Chromatography, Marcel Dekker, Inc., New York, 1965.

[2] R.K. Iler, The chemistry of silica: solubility, polymerization, colloid and surface properties, and biochemistry, Wiley, 1979.

[3] K.K. Unger, Porous silica, its properties and use as support in column liquid chromatography, Elsevier Scientific Pub. Co., 1979.

[4] U.D. Neue, HPLC Columns: Theory, Technology, and Practice, Wiley-VCH, New York, 1997.

[5] B. Fried, J. Sherma, Thin-layer Chromatography, CRC Press, 1999.

[6] B.A. Bidlingmeyer, F.V. Warren, "Column Efficiency Measurement," 56 (1984) A583.

[7] E. Grushka, "Chromatographic Peak Capacity and Factors Influencing It," Analytical Chemistry 42 (1970) 1142.

[8] J.M. Davis, J.C. Giddings, "Statistical-Theory of Component Overlap in Multicomponent Chromatograms," Analytical Chemistry 55 (1983) 418.

[9] I. François, K. Sandra, P. Sandra, "Comprehensive liquid chromatography: Fundamental aspects and practical considerations—A review," Analytica Chimica Acta 641 (2009) 14.

[10] S.A. Cohen, M.R. Schure, Multidimensional liquid chromatography: theory and applications in industrial chemistry and the life sciences, Wiley-Interscience, 2008.

[11] R.E. Murphy, M.R. Schure, J.P. Foley, "Effect of sampling rate on resolution in comprehensive two-dimensional liquid chromatography," Analytical Chemistry 70 (1998) 1585.

[12] D.R. Stoll, X.P. Li, X.O. Wang, P.W. Carr, S.E.G. Porter, S.C. Rutan, "Fast, comprehensive two-dimensional liquid chromatography," Journal of Chromatography A 1168 (2007) 3.

83 [13] L.R. Snyder, J.J. Kirkland, J.L. Glajch, Practical HPLC Method Development, John Wiley & Sons, Inc. (US), New York 1997.

[14] R.K. Gilpin, J.A. Squires, "Effect of Temperature on the Orientation of Bonded Hydrocarbon Phases in Totally Aqueous Liquid-Chromatographic Systems," Journal of Chromatographic Science 19 (1981) 195.

[15] R.K. Gilpin, M.E. Gangoda, A.E. Krishen, "Effect of Conditioning Solvent on the Orientation of Bonded Hydrocarbon Moieties in Totally Aqueous Mobile Phases," Journal of Chromatographic Science 20 (1982) 345.

[16] A.P. Schellinger, D.R. Stoll, P.W. Carr, "High speed gradient elution reversed-phase liquid chromatography," Journal of Chromatography A 1064 (2005) 143.

[17] D.F. Thekkudan, S.C. Rutan, P.W. Carr, "A study of the precision and accuracy of peak quantification in comprehensive two-dimensional liquid chromatography in time," Journal of Chromatography A 1217 (2010) 4313.

[18] L.A. Cole, J.G. Dorsey, "Reduction of Reequilibration Time Following Gradient Elution Reversed-Phase Liquid-Chromatography," Analytical Chemistry 62 (1990) 16.

[19] J.W. Coym, B.W. Roe, "Effect of temperature on gradient reequilibration in reversed- phase liquid chromatography," Journal of Chromatography A 1154 (2007) 182.

[20] A.P. Schellinger, D.R. Stoll, P.W. Carr, "High speed gradient elution reversed phase liquid chromatography of bases in buffered eluents - Part II. Full equilibrium," Journal of Chromatography A 1192 (2008) 54.

[21] F. Gritti, A. Cavazzini, N. Marchetti, G. Guiochon, "Comparison between the efficiencies of columns packed with fully and partially porous C18-bonded silica materials," Journal of Chromatography A 1157 (2007) 289.

[22] F. Gritti, I. Leonardis, D. Shock, P. Stevenson, A. Shalliker, G. Guiochon, "Performance of columns packed with the new shell particles, Kinetex-C18," Journal of Chromatography A 1217 (2010) 1589.

[23] E. Oláh, S. Fekete, J. Fekete, K. Ganzler, "Comparative study of new shell-type, sub-2 um fully porous and monolith stationary phases, focusing on mass-transfer resistance," Journal of Chromatography A 1217 (2010) 3642.

[24] G. Guiochon, F. Gritti, "Shell particles, trials, tribulations and triumphs," Journal of Chromatography A 1218 (2011) 1915.

[25] F. Gritti, G. Guiochon, "Comparative study of the performance of columns packed with several new fine silica particles: Would the external roughness of the particles affect column properties?," Journal of Chromatography A 1166 (2007) 30.

84

[26] A. Fanigliulo, D. Cabooter, G. Bellazzi, D. Tramarin, B. Allieri, A. Rottigni, G. Desmet, "Comparison of performance of high-performance liquid chromatography columns packed with superficially and fully porous 2.5 um particles using kinetic plots," Journal of Separation Science 33 (2010) 3655.

[27] C.A. Rimmer, C.R. Simmons, J.G. Dorsey, "The measurement and meaning of void volumes in reversed-phase liquid chromatography," Journal of Chromatography A 965 (2002) 219.

[28] T.H. Walter, P. Iraneta, M. Capparella, "Mechanism of retention loss when C-8 and C-18 HPLC columns are used with highly aqueous mobile phases," Journal of Chromatography A 1075 (2005) 177.

[29] E. Aker, in Department of Physics, University of Oslo, Oslo, Norway, 1996.

[30] G. Guiochon, G. Korosi, A. Siouffi, "Study of the Performances of Thin-Layer Chromatography .5. Flow-Rate in Reversed Phase Plates," Journal of Chromatographic Science 18 (1980) 324.

[31] W.J. Cheong, P.W. Carr, "The Surface-Tension of Mixtures of Methanol, Acetonitrile, Tetrahydrofuran, Isopropanol, Tertiary Butanol and Dimethylsulfoxide with Water at 25- Degrees-C," Journal of Liquid Chromatography 10 (1987) 561.

[32] S.E. Buckley, M.C. Leverett, "Mechanism of fluid displacements in sands," Transactions of the American Institute of Mining and Metallurgical Engineers 146 (1942) 107

[33] S.W. Pettersson, B.S. Persson, M. Nystrom, "General method allowing the use of 100% aqueous loading conditions in reversed-phase liquid chromatography," Journal of Chromatography B-Analytical Technologies in the Biomedical and Life Sciences 803 (2004) 159.

[34] U.D. Neue, D.H. Marchand, L.R. Snyder, "Peak compression in reversed-phase gradient elution," Journal of Chromatography A 1111 (2006) 32.

[35] N.E. Hoffman, S.L. Pan, A.M. Rustum, "Injection of Eluites in Solvents Stronger Than the Mobile Phase in Reversed-Phase Liquid-Chromatography," Journal of Chromatography 465 (1989) 189.

[36] N.E. Hoffman, A. Rahman, "Computation of Band Shape for Strong Injection Solvent and Weak Mobile Phase Combinations in Liquid-Chromatography," Journal of Chromatography 473 (1989) 260.

[37] M. Tsimidou, R. Macrae, "Influence of Injection Solvent on the Reversed-Phase Chromatography of Triglycerides," Journal of Chromatography 285 (1984) 178.

85 [38] P.W. Wrezel, I. Chion, R. Pakula, "Key factors in sample diluent selection for HPLC assays of active pharmaceutical ingredients," LC-GC North America 23 (2005) 682.

[39] E. Loeser, S. Babiak, P. Drumm, "Water-immiscible solvents as diluents in reversed- phase liquid chromatography," Journal of Chromatography A 1216 (2009) 3409.

[40] E. Loeser, P. Drumm, "Using strong injection solvents with 100% aqueous mobile phase in RP-LC," Journal of Separation Science 29 (2006) 2847.

[41] Y.V. Kazakevich, R. LoBrutto, F. Chan, T. Patel, "Interpretation of the excess adsorption isotherms of organic eluent components on the surface of reversed-phase adsorbents: Effect on the analyte retention," Journal of Chromatography A 913 (2001) 75.

[42] D.V. McCalley, "The challenges of the analysis of basic compounds by high performance liquid chromatography: Some possible approaches for improved separations," Journal of Chromatography A 1217 (2010) 858.

[43] P. Agrafiotou, C. Ràfols, C. Castells, E. Bosch, M. Rosés, "Simultaneous effect of pH, temperature and mobile phase composition in the chromatographic retention of ionizable compounds," Journal of Chromatography A 1218 (2011) 4995.

[44] J.P. Foley, J.G. Dorsey, "Equations for Calculation of Chromatographic Figures of Merit for Ideal and Skewed Peaks," Analytical Chemistry 55 (1983) 730.

[45] N.S. Wilson, M.D. Nelson, J.W. Dolan, L.R. Snyder, R.G. Wolcott, P.W. Carr, "Column selectivity in reversed-phase liquid chromatography I. A general quantitative relationship," Journal of Chromatography A 961 (2002) 171.

[46] L.R. Snyder, J.W. Dolan, P.W. Carr, "The hydrophobic-subtraction model of reversed- phase column selectivity," Journal of Chromatography A 1060 (2004) 77.

[47] N.S. Wilson, J. Gilroy, J.W. Dolan, L.R. Snyder, "Column selectivity in reversed-phase liquid chromatography - VI. Columns with embedded or end-capping polar groups," Journal of Chromatography A 1026 (2004) 91.

[48] N.S. Wilson, M.D. Nelson, J.W. Dolan, L.R. Snyder, P.W. Carr, "Column selectivity in reversed-phase liquid chromatography II. Effect of a change in conditions," Journal of Chromatography A 961 (2002) 195.

[49] N.S. Wilson, J.W. Dolan, L.R. Snyder, P.W. Carr, L.C. Sander, "Column selectivity in reversed-phase liquid chromatography III. The physico-chemical basis of selectivity," Journal of Chromatography A 961 (2002) 217.

[50] Y.V. Kazakevich, H.M. McNair, "Study of the Excess Adsorption of the Eluent Components on Different Reversed-Phase Adsorbents," Journal of Chromatographic Science 33 (1995) 321.

86

[51] Y.V. Kazakevich, H.M. McNair, "Thermodynamic Definition of HPLC Dead Volume," Journal of Chromatographic Science 31 (1993) 317.

[52] F. Gritti, G. Guiochon, "Thermodynamics of adsorption of binary aqueous organic liquid mixtures on a RPLC adsorbent," Journal of Chromatography A 1155 (2007) 85.

[53] F. Gritti, Y.V. Kazakevich, G. Guiochon, "Effect of the surface coverage of endcapped C-18-silica on the excess adsorption isotherms of commonly used organic solvents from water in reversed phase liquid chromatography," Journal of Chromatography A 1169 (2007) 111.

[54] B. Buszewski, S. Bocian, R. Zera, "Influence of temperature and pressure on the preferential adsorption of component of hydroorganic mobile phase in liquid chromatography," Adsorption-Journal of the International Adsorption Society 16 (2010) 437.

[55] K.B. Sentell, J.G. Dorsey, "On the Calculation of the Stationary Phase Volume in Reversed Phase Chromatography," Journal of Liquid Chromatography 11 (1988) 1875.

[56] 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, "A fast, convenient and rugged procedure for characterizing the selectivity of alkyl-silica columns," Journal of Chromatography A 1057 (2004) 49.

[57] N.S. Wilson, J. Gilroy, J.W. Dolan, L.R. Snyder, "Column selectivity in reversed-phase liquid chromatography: VI. Columns with embedded or end-capping polar groups," Journal of Chromatography A 1026 (2004) 91.

[58] Personal communication with Dr. Lloyd Snyder.

[59] J. Ruta, S. Rudaz, D.V. McCalley, J.L. Veuthey, D. Guillarme, "A systematic investigation of the effect of sample diluent on peak shape in hydrophilic interaction liquid chromatography," Journal of Chromatography A 1217 (2010) 8230.

[60] J.W. Coym, "Evaluation of ternary mobile phases for reversed-phase liquid chromatography: Effect of composition on retention mechanism," Journal of Chromatography A 1217 (2010) 5957.

[61] J.L. Rafferty, J.I. Siepmann, M.R. Schure, "Mobile phase effects in reversed-phase liquid chromatography: A comparison of acetonitrile/water and methanol/water solvents as studied by molecular simulation," Journal of Chromatography A 1218 (2011) 2203.

[62] J. Layne, T. Farcas, I. Rustamov, F. Ahmed, "Volume-load capacity in fast-gradient liquid chromatography: Effect of sample solvent composition and injection volume on chromatographic performance," Journal of Chromatography A 913 (2001) 233.

87

[63] D.R. Stoll, X.L. Wang, P.W. Carr, "Comparison of the practical resolving power of one- and two-dimensional high-performance liquid chromatography analysis of metabolomic samples," Analytical Chemistry 80 (2008) 268.

[64] M.R. Schure, "Limit of detection, dilution factors, and technique compatibility in multidimensional separations utilizing chromatography, capillary electrophoresis, and field-flow fractionation," Analytical Chemistry 71 (1999) 1645.

[65] G. Haugaard, T.D. Kroner, "Partition Chromatography of Amino Acids with Applied Voltage," Journal of the American Chemical Society 70 (1948) 2135.

[66] H.H. Strain, J.C. Sullivan, "Analysis by Electromigration Plus Chromatography," Analytical Chemistry 23 (1951) 816.

[67] W.J. Vanooij, "Simple Apparatus for Simultaneous Chromatography and Electrophoresis on Thin-Layers," Journal of Chromatography 81 (1973) 190.

[68] D. Nurok, "Planar electrochromatography," Journal of Chromatography A 1044 (2004) 83.

[69] D. Nurok, J.M. Koers, A.L. Novotny, M.A. Carmichael, J.J. Kosiba, R.E. Santini, G.L. Hawkins, R.W. Replogle, "Apparatus and initial results for pressurized planar electrochromatography," Analytical Chemistry 76 (2004) 1690.

[70] A.L. Novotny, D. Nurok, R.W. Replogle, G.L. Hawkins, R.E. Santini, "Results with an apparatus for pressurized planar ectrochromatography," Analytical Chemistry 78 (2006) 2823.

[71] T.H. Dzido, P.W. Plocharz, P. Slazak, A. Halka, "Progress in planar electrochromatography," Analytical and Bioanalytical Chemistry (2008) 2111.

[72] T.H. Dzido, P.W. Plocharz, P. Slazak, "Apparatus for pressurized planar electrochromatography in a completely closed system," Analytical Chemistry 78 (2006) 4713.

[73] T.H. Dzido, P.W. Plocharz, "Planar electrochromatography in a closed system under pressure - Pressurized planar electrochromatography," Journal of Liquid Chromatography & Related Technologies 30 (2007) 2651.

[74] T.H. Dzido, J. Mroz, G.W. Jozwiak, "Adaptation of a horizontal DS chamber to planar electrochromatography in a closed system," JPC-Journal of Planar Chromatography- Modern TLC 17 (2004) 404.

88 [75] T.H. Dzido, R. Majewski, B. Polak, W. Golkiewicz, E. Soczewinski, "Application of a horizontal DS chamber to planar electrochromatography," JPC-Journal of Planar Chromatography-Modern TLC 16 (2003) 176.

[76] P.A. Tate, J.G. Dorsey, "Characterization of flow and voltage profiles in planar electrochromatography," Journal of Chromatography A 1079 (2005) 317.

[77] P.A. Tate, J.G. Dorsey, "Linear voltage profiles and flow homogeneity in pressurized. planar electrochromatography," Journal of Chromatography A 1103 (2006) 150.

[78] M. Gilar, P. Olivova, A.E. Daly, J.C. Gebler, "Orthogonality of separation in two- dimensional liquid chromatography," Analytical Chemistry 77 (2005) 6426.

[79] P.J. Slonecker, X.D. Li, T.H. Ridgway, J.G. Dorsey, "Informational orthogonality of two dimensional chromatographic separations," Analytical Chemistry 68 (1996) 682.

[80] H. Gu, Y. Huang, M. Filgueira, P.W. Carr, "Effect of first dimension phase selectivity in online comprehensive two dimensional liquid chromatography (LC x LC)," Journal of Chromatography A 1218 (2011) 6675.

89

BIOGRAPHICAL SKETCH

Bradley J. VanMiddlesworth

Brad VanMiddlesworth was born in Jacksonville, Florida on July 16th, 1983. He graduated from Stanton College Preparatory School in 2001, then went onto undergraduate studies at Florida State University where he received a Bachelor of Science in biochemistry in May of 2004. After graduation, he was deployed in support of Operation Enduring Freedom in the United States Marine Corps Reserve. In May of 2005, he enrolled at Florida State Univeristy to pursue graduate studies in analytical chemistry, and joined the Dorsey Research Group in January of 2006. Brad will receive his Ph.D. in Analytical Chemistry in the fall of 2011.

90