Improving HPLC Column Selection and System Performance Richard A. Henry Penn State University T410170 sigma-aldrich.com
LC Particle Innovation Leads the Way* Particle Size and Architecture (porosity) More speed, efficiency 5 µm 3 µm <2 µm Solid-Core monoliths Ascentis and Ascentis Express Particle Composition (and stationary phases) expanded ph range and selectivity silica polymers Hybrids Kromasil Eternity Other oxides Discovery Zr * From Frank Michel R. A. Henry 2010 3
Evolution of HPLC Column Particle Shapes Irregular particles Spherical particles Monoliths 1970 1980 2000 Remain the workhorse R. A. Henry 2010 4
Particle Innovation has Reduced Column Dispersion *...more efficiency, narrower peaks, higher peak capacity, more sensitivity. * T. L. Chester, American Lab, Vol. 41 No. 4, pp 11-15, March 2009. R. A. Henry 2010 5
Measured Void Volumes: Solid-Core vs Porous Void (ml) 1.00 0.90 0.80 0.70 0.60 0.50 0.40 0.30 Ascentis 3 µm Ascentis Express Data represent retention volume of uracil in 50% aqueous ACN corrected for system volume (no column). 0.20 0.10 0.00 4.6 3 2.1 mm ID 10cm x 4.6mm C18 Columns R. A. Henry 2010 6
Resolution: Peaks Separate Faster Than They Spread Maximize: Differential migration (Δt R ) - selectivity Minimize: Band dispersion (w) - efficiency R s = Δt R 0.5 (w A + w B ) Thermodynamic (chemical) Δt R L Kinetic (physical) w L 1/2 Improvements have been made in both areas R. A. Henry 2010 7
Measuring HPLC System Suitability
HPLC System Organization and Optimization Injector and detector tubing for the Agilent 1100 were 0.007 inches ID. A 10 µl flow cell was employed. Experiments are underway to examine how tubing, flow cells and other HPLC system variables affect performance. R. A. Henry 2010 15
Peak Band Spreading Equation σ 2 sys = σ2 col + σ2 instr sys = ( 2 col + 2 instr )1/2 Band spreading (dispersion) can occur in both HPLC columns and instruments leading to a system equation; instrument dispersion is also referred to as instrument bandwidth (IBW) New developments in particles have greatly reduced column dispersion; instrument brand, model and configuration now matters; system performance can be dramatically different for column and instrument combinations; can t use a column without an instrument! New instruments with higher pressure ratings and smaller volume tubing and components have been designed for modern, smaller particle columns; however, traditional instruments may also be suitable for modern columns, especially when optimized by the user; simple tests can qualify instruments for minimum dispersion. R. A. Henry 2010 16
Illustration of Instrument Dispersion A peak from the same column in three different instruments. 1 2 3 Peak 1 shows a low dispersion instrument where most of the spreading occurs inside the column. Peak 2 shows moderate instrument dispersion in blue. Peak 3 shows high instrument dispersion; peak width has doubled and column peak capacity is halved. Time spent outside the column destroys system efficiency R. A. Henry 2010 17
Origins of Column Band Spreading* (van Deemter) H = Ad p + BD m /u + Cd p2 u/d m A term- Multipath; bed uniformity; eddy diffusion B term- Longitudinal diffusion (axial in nature) C term- Rate of mass transfer from moving phase though stagnant mobile phase into stationary phase (radial in nature) Instrument impact on H and N is often neglected R. A. Henry 2010 18
Band Spreading Inside the Column Bed 2 2 2 σ col = V (1 + k ) /N V o = mobile phase column volume (µl) (unretained peak retention volume; void volume) k = peak capacity factor N = number of column theoretical plates o Small bed geometry, short retention and high efficiency favor low dispersion (dilution) within a packed column. Instrument bandwidth becomes more harmful to efficiency and resolution for short, small ID columns with low k values and high N. R. A. Henry 2010 19
Column vs System Band-Spreading The effect of system band width can be calculated from the additive relationship of variances, where the total variance of the peak is equal to the sum of the true oncolumn peak variance plus the instrument variance. σ 2 system = σ2 column + σ2 instrument σ 2 instrument = σ2 injector + σ2 detector + σ2 connector tubing σ system = Total peak dispersion in volume units (µl) σ system = W b /4 W b = 4σ is an easy way to estimate dispersion R. A. Henry 2010 20
Test Mix Chromatogram on Ascentis Express Fused-Core C18* Express peak widths (bands) at base are less than 50µL Instrument peak width should be smaller, but 50 is about typical for a traditional instrument.! N = 23,781 Sample: Uracil, benzene, toluene and anthracene Column: Ascentis Express C18, 100x4.6mm Mobile phase: 30/70 water/acn; Flow: 1.25 ml/min; T = 35 C Instrument: Waters Acquity, 220 bar (3200 psi) * T. L. Chester, American Lab, Vol. 41 No. 4, pp 11-15, March 2009. R. A. Henry 2010 21
Fused-Core Measured Plate Height for Unmodified and Modified Instrument 3 % of Total Dispersion Unmodified Modified Inj <1 1 Col 30 72 Unmodified; unsuitable N max = 8500 100,000 N/m Conn Tubes 66 8 Det* 4 19 * detector cell had a welded heat exchanger that could not be conveniently replaced (Waters Alliance). Modified; suitable N max = 20,100 200,000 N/m R. A. Henry 2010 22
What Causes HPLC Instrument Dispersion?... Time sample spends outside the column bed.
HPLC System Components that Affect Measured Peak Bandwidth* Injection Volume Injector or other Valves Pre-Col Tubing Fitting and Frit Flow Direction Column Bed Fitting and Frit * Adapted from R.P.W. Scott 8, Liquid Chromatography, Chrom-Ed Chromatography Series, Open Access Library, www.library4science.com/ Should also include column selection or bypass valves. Dispersion in injector or other valves and inlet tubing can be reduced by using weak solvent injection or gradients. Post-Col Tubing Can t remove dispersion in column fittings and frits so it must be treated as an integral part of the column bed. Detector Volume Dispersion is always a problem after the column. Detector Response and A/D Electronic and often neglected R. A. Henry 2010 24
Dispersion in Open Cylindrical Flow Path Time outside column bed is much worse than time inside 2 σ tube = 1.36 x 10-3 d 4 t L t F/D d = cylinder diameter L = cylinder length F = flow rate D = diffusion coefficient Dispersion from volume elements is constant for any given flow rate and analyte, but note that dispersion (bandwidth) increases with flow. Velocity at the wall is essentially zero under laminar flow conditions. Small inside diameter, short length, low flow and fast solute diffusion favor low dispersion in connection tubes and accessories. Larger molecules show greater dispersion (as 1/D) in connectors. R. A. Henry 2010 25
Optimizing Instrument Configuration for Elevated Temperature Operation* R. A. Henry 2010 *Agilent 1100 photo compliments of Dan Nowlan at ZirChrom 26
Performance of a Factory Agilent 1100 Instrument with Sub-2µm Zirconia A family of curves with H min ranging from <5µm to >12µm indicates strong instrument impact on H system Plate Height, H (um) 20 18 16 14 12 10 8 6 4 2 Plate Height Vs. Linear Velocity for a PBD Column Benzene Toluene Ethylbenzene Propylbenzene Butylbenzene 2 ml/min test conditions Excess instrument dispersion resembles the effect of larger particles (C-term). 0 0.03 0.13 0.23 0.33 0.43 0.53 u (cm/s) Plate height vs linear velocity, Temperature 30 ºC, Mobile phase: 50/50 ACN/water, Column: 50 x 4.6mm, Agilent 1100/UV with Standard Cell and 0.007 i.d. tubing. R. A. Henry 2010 27
Optimizing Factory Agilent 1100 by Adding Micro Flow Cell New family of curves with H min ranging from <5µm to 8µm indicates lower instrument impact on H system Plate Height Vs. Linear Velocity for a PBD Column Plate Height, H (um) 20 18 16 14 12 10 8 6 4 2 Benzene Toluene Ethylbenzene Propylbenzene Butylbenzene 2 ml/min test conditions When instrument dispersion is reduced, curves begin to overlap. 0 0.03 0.13 0.23 0.33 0.43 0.53 u (cm/s) R. A. Henry 2010 28
Optimized Factory Instrument with Micro Cell + 0.005 ID Tubing New family of curves with H min ranging from 4µm to 6µm indicates very low instrument impact on H system Plate Height Vs. Linear Velocity for a PBD Column Plate Height, H (um) 20 18 16 14 12 10 8 6 4 2 Benzene Toluene Ethylbenzene Propylbenzene Butylbenzene 1.6 ml/min test conditions Instrument dispersion is almost negligible; curves nearly overlap and become flat. 0 0.03 0.13 0.23 0.33 0.43 0.53 u (cm/s) R. A. Henry 2010 29
Factory Instrument with Micro Cell + 0.005 ID Tubing + Heat Exchanger New family of curves with H min ranging from <5µm to >7µm indicates moderate instrument impact on H system Plate Height Vs. Linear Velocity for a PBD Column Plate Height, H (um) 20 18 16 14 12 10 8 6 4 2 Benzene Toluene Ethylbenzene Propylbenzene Butylbenzene 1.8 ml/min test conditions Instrument dispersion increases again due to addition of 0.007 heat exchanger*; range of H increases and curves are no longer flat. 0 0.03 0.13 0.23 0.33 0.43 0.53 u (cm/s) * An Agilent Model 1200 heat exchanger may be surface mounted to the heater block to maintain low dispersion for heated applications. R. A. Henry 2010 30
Efficiency Plot for Agilent 1100 with Micro-Cell and Factory Tubing Plates (N) vs Linear Velocity for a sub-2um PBD Column 14000 12000 Benzene Toluene Ethylbenzene Propylbenzene Butylbenzene 10000 Number of plates (N) 8000 6000 4000 2000 0 Instrument dispersion is evident at lower k values and smaller peak volumes. 0.00 0.10 0.20 0.30 0.40 0.50 0.60 u (cm/s) Plate height vs linear velocity for retained solutes: Alkylbenzenes, Temperature 30 ºC, Mobile phase: 50/50 ACN/water; Column: 50 x 4.6mm, Agilent 1100/UV with Micro Cell and 0.007 i.d. tubing. R. A. Henry 2010 31
Beta-Blockers on Zr-PBD Sub-2µm at 75 o C Norm. 120 100 80 60 Analytes 1=Labetalol 2=Atenolol 3=Acebutolol 4=Metoprolol 5=Oxprenolol 6=Lidocaine 7=Quinidine 8=Alprenolol 9=Propranolol R = 135 Zr-PBD 50mm x 4.6mm, sub-2μm 22/78 ACN/20mM K 3 PO 4 at ph=12 F=2.5 ml/min UV=254nm T=75 o C 246 bar (3660 psi) Background pressure: ca. 25 bar Detector Response Time (0.5 sec) 40 20 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 min R. A. Henry 2010 32
Beta-Blockers Optimized with Faster Detector Response Norm. 175 150 125 100 75 Analytes 1=Labetalol 2=Atenolol 3=Acebutolol 4=Metoprolol 5=Oxprenolol 6=Lidocaine 7=Quinidine 8=Alprenolol 9=Propranolol R = 200 Zr-PBD 50mm x 4.6mm, sub-2μm 21/79 ACN/20mM K 3 PO 4 at ph=12 F=2.5 ml/min UV=254nm T=75 o C, faster sampling rate 246 bar (3660 psi) Background pressure: ca. 25 bar Detector Response Time (<0.12 sec) 50 25 0 0 0.2 0.4 0.6 0.8 1 min R. A. Henry 2010 33
Importance of Injection Volume and Gradient Focusing
Performance of Some Factory HPLC Instruments Column Plates Determined from Toluene (k = 2) Half-Height Width Isocratic Elution with 60% ACN 10 µl Injected in Mobile Phase (All instruments in standard configuration with analytical flowcells) Column Agilent 1100 Agilent 1200 Waters 2695 Column Supplier Test Data Ascentis C18 150x4.6mm 5µm 16911 (6.7% loss) 18034 (0% loss) 16874 (6.9% loss) 18119 Ascentis Express (Fused-Core) C18 100x4.6mm 2.7µm 18649 (34% loss) 22001 (22% loss) 18666 (34% loss) 28164 R. A. Henry 2010 36
Effect of Variables Using Agilent 1200 * Summary of Toluene Plates from 2.7-µm Supelco Ascentis Express FCP Column Isocratic Elution with 60% ACN on Agilent 1200 10 µl Injected in Mobile Phase* Non-Optimized (10 mm / 13 µl flowcell) 22001 Change to 6 mm / 5 µl flowcell 21912 Minimize tubing lengths (don t bother) 22104 *10x sample dilution with 10x volume increase 25738 S. Bannister studies related to ref. 7 (Xcelience Labs, Tampa, FL). R. A. Henry 2010 37
Dispersion Caused by Injections in Mobile Phase 30000 20000 10000 0 Effect of Injection Volume and Solvent on Efficiency 2.7-µm Supelco Ascentis Express C18, 100x4.6-mm Plates/Column Uracil Acetophenone Chloronitrobenzene Toluene Injected in 10% MP Injected in MP 0 20 40 60 80 100 Injection Volume, µl Sample was diluted 1:10 with water for weak solvent injection. Note that efficiency performance was greater with a 100µL injection that has been diluted with water than the original 1µL injection in mobile phase. R. A. Henry 2010 38
Estimating Instrument Bandwidth
Simple System Suitability (Bandwidth) Test...one instrument with three columns (three systems) Traditional instrument with 5 cm Ascentis Express columns 4.6 mm ID Choose test columns that can show >200,000 N/m at k = 2-3. 3.0 mm ID 2.1 mm ID 4.6mm ID columns usually look pretty good due to large void volume and peak volumes. 1.0 2.0 Time (min) R. A. Henry 2010 40
Ranking System Performance with Three Columns Measured System Plate Count N 12000 10000 8000 6000 4000 2000 Ideally, columns should show identical system efficiency. 4.6 mm ID 3.0 mm ID 2.1 mm ID This instrument is suitable for use with 4.6mm ID columns. 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 k R. A. Henry 2010 41
Another System Suitability (Bandwidth) Test...one column with three instruments (three systems) Column: Ascentis Express C18, 2.1 x 150 mm Mobile phase: 3:7, Water:Acetonitrile Column temp: 35 C Injection vol: 1 µl Sample mix: uracil, benzene, toluene, naphthalene, acenaphthene (in 1:1, water:acetonitrile) Detection: 250 nm Flow: 0.3 ml/min Pressure: Ultra System: 3500 psi (extra-column P* = 1200 psi) Traditional System: 2300 psi 2.1mm ID columns were used in the study to demonstrate differences in HPLC system performance. Most conventional instrument systems show acceptable performance with 4.6mm ID or even 3.0mm ID columns. 1.0 2.0 3.0 4.0 Time (min) R. A. Henry 2010 42
Ranking System Suitability Performance (Bandwidth) 250000 200000 The blue instrument is suitable (maybe the red one too) System 1 System 2 System 3 200993 UHPLC 226480 224184 216724 Optimized HPLC 194764 95% 90% Nobs (N/m) 150000 100000 153678 103149 170200 125002 180538 Traditional HPLC 140000 N col is the same for all plots. 172324 80% 50000 N sys shows effect of instrument bandwidth. Columns with high N expose instrument bandwidth differences. 0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 R. A. Henry 2010 43 k
Direct Method for Measuring IBW Connect injector to detector ZDV union shunt Inject small volume (µl or less) of chromophore Record peak (or retention time and N) Calculate IBW (flow in µl) or measure directly from peak retention σ = (t r x flow) / N IBW = 4σ Common mistakes data sampling rate too slow detector response time too slow flow rate too fast (or variable) calculation of N R. A. Henry 2010 44
Comparison of IBW Before and After Instrument Optimization After Before Shimadzu LC-10A (400 bar) @ flow of 100 µl/min Before- suitable for 4.6mm columns (30-50 µl) 4σ = 45 µl 0.4 0.6 0.8 1.0 1.2 Time (min) 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Time (min) Remove column and measure bandwidth at base for small volume injection Clear contrast between peak profiles before and after optimization 4σ = 15 µl 0.10 0.20 0.30 Time (min) After- suitable for most columns (15-30 µl) R. A. Henry 2010 46
Effect of Data Sampling Rate on Measured IBW 28 27 95% CI for the Mean 1 ml/min; 0.2 sec detector response time 4σ (µl) 26 25 24 23 22 21 True value? 20 5 10 20 40 Hz R. A. Henry 2010 47
Effect of Detector Response Time on Measured IBW 4σ (µl) 45 40 35 30 25 95% CI for the Mean 1 ml/min; 20 Hz sampling rate True value? 1 ml/min; 20 Hz sampling rate 0.50 sec 0.20 sec 0.10 sec 0.05 sec 20 15 10 0.05 0.10 0.20 0.50 sec 0.02 0.04 0.06 0.08 0.10 Time (min) Slow filter response time adversely effects accurate capture of peak dimensions by data system R. A. Henry 2010 48
Effect of Flow rate on Measured IBW 45 95% CI for the Mean 13.0 95% CI for the Mean 4σ (µl) 40 35 30 25 1 ml/min; 20 Hz sampling rate True value? 4σ (µl) 12.8 12.6 12.4 0.1 ml/min, 20 Hz sampling rate True value? 20 15 12.2 10 0.05 0.10 sec 0.20 0.50 12.0 0.05 0.10 0.20 sec 0.50 Flow rate of 0.1 ml/min will inherently yield less dispersion, but also permits accurate capture of peak dimensions well within capabilities of data system R. A. Henry 2010 49
2.1 mm ID x 5 cm, 0.4 ml/min N (acet) = 360 N (tol) = 1790 IBW: 42 µl Tubing ID: 0.010 Flow cell: 16 µl, 10 mm path Pressure: 1560 psi 1.0 2.0 Time (min) True N = >10,000 N (tol) = 5820 N (acet) = 1740 IBW: 17 µl Tubing ID: 0.005 Flow cell: 16 µl, 10 mm path Pressure: 1680 psi 1.0 2.0 Time (min) N (acet) = 2800 N (tol) = 7050 IBW: 12 µl Tubing ID: 0.005 Flow cell: 2.5 µl, 5 mm path Pressure: 1680 psi 1.0 2.0 Time (min) R. A. Henry 2010 51
3 mm ID x 5 cm, 0.8 ml/min N (tol) = 4400 N (acet) = 980 IBW: 42 µl Tubing ID: 0.010 Flow cell: 16 µl, 10 mm path Pressure: 1600 psi 1.0 2.0 Time (min) True N = >10,000 N (acet) = 5540 N (tol) = 8940 IBW: 12 µl Tubing ID: 0.005 Flow cell: 2.5 µl, 5 mm path Pressure: 1640 psi 1.0 2.0 Time (min) R. A. Henry 2010 52
Fekete Direct Measurements of IBW 4 0.04 min x 0.5 ml/min = 0.02 ml IBW 20 µl 4σ R. A. Henry 2010 55
Summary of Variables that Affect Instrument Bandwidth and System Suitability Instrument volume should be small with respect to column internal volume which determines peak volume at base. Reduce tubing ID and volume Reduce tubing length and volume Reduce detector flow cell volume Improve detector response time Match data collection rate to peak width in time At least 20-30 points across the peak R. A. Henry 2010 56
Dispersion References 1. R. Majors, Are you Getting the Most Out of Your HPLC Column?, LCGC NA, Vol. 21, No. 12, 1124-1133 (December 2003). 2. R. A. Henry and D. S. Bell, Important Guidelines for Optimizing Speed and Sensitivity in Small Molecule LC-UV and LC-MS, LCGC NA, Vol. 23, No. 5, 2-7 (May 2005). 3. T. Chester, Sub-2µm Performance with a Conventional Instrument Using 2.7µm Fused-CoreTM Particles, American Lab, Volume 41, No. 4, 11-15 (2009). 4. S. Fekete, et. al., Shell and Small Particles: Evaluation of New Technology, J. of Pharmaceutical and Biomedical Analysis, 49, 64-71 (2009). 5. F. Gritti, G. Guiochon, et. al., Achieving Full Performance of Columns by Optimizing HPLC Instruments, J. of Chromatogr. A, 1217, 3000-3012 (2010) 6. H. Brandes, unpublished data, Sigma-Supelco Applications Lab Reports and Notebooks (2008-2009). 7. R. A. Henry and D. Nowlan, Use of Sub-2µm Zirconia-PBD at Elevated ph and Temperature, Oral Paper, EAS 2009, Somerset, NJ. R. A. Henry 2010 57
Thanks for your attention! R. A. Henry 2010 58