Understanding the Use of Temperature Regulation to Optimize Mass Transfer in Fast Gradient Reversed Phase Liquid Chromatography

Transcription

1 Understanding the Use of Temperature Regulation to Optimize Mass Transfer in Fast Gradient Reversed Phase Liquid Chromatography Mark J. Hayward Lundbeck Research USA Chemistry : Automation, Profiling, Purification, & Structural Analysis 215 College Rd. - Paramus, NJ MHay@Lundbeck.com Summary: The speed of ACN gradient liquid chromatography is governed primarily by reaction driven mass transfer, and as a result, active temperature control can be used to increase speed at least ten fold while maintaining resolution.

2 Background Fast Gradient LC LC speed is proportional to eluent velocity (υ) through column (k per unit time). Column length (L) cannot be shortened to less than 50 mm without seriously sacrificing the separation efficiency. Unfortunately, smaller diameter (d p ) particles don t improve separation efficiency (via improved mass transfer) enough to allow further reduction in column length (as hoped).* Some folks are finding that sub 2 μm d p particles reduce velocity due to elevated pressure and thus provide slower separations. ** Fortunately, fast speeds (high υ) are readily achievable with ordinary LC systems by using a different approach (that doesn t increase pressure) for improving mass transfer efficiency. This approach uses temperature to systematically regulate diffusion (D) and is the most general way to control or otherwise increase mass transfer efficiency and thus υ. Furthermore, since it lowers pressure, it can be readily utilized with most any LC. *J.H. Knox, M. Saleem, J. Chromatogr. Sci. 1969, 7, 614. *J. Kofman, Y. Zhao, T. Maloney, T. Baumgartner, R. Bujalski, Am. Drug Discovery 2006, 1, 12. **T.L. Chester, S.O. Terami, J. Chromatogr. A, 2005, 1096, 16.

3 Background Fast LC & T The use of temperature (T) to increase speed is known.* Higher T increases B term and decreases C term in van Deemter equation without affecting A term.* The van Deemter curves simply continue to further flatten and shift the optimal flow higher as T increases (see figure).* To be answered: How does this adapt to gradient operation and how do we better understand and use this in a systematic (quantitative) approach toward increasing speed? van Deemter Curves at Two Different Diffusivities (Temperatures) Seminal Works: *F.D. Antia, Cs. Horvath, J. Chromatogr. 1988, 435, 1. *B. Yan, J. Zhao, J.S. Brown, J. Blackwell, P.W. Carr, Anal. Chem. 2000, 72, *F. Gritti, A. Felinger, G. Guiochon, J. Chromatogr. A, 2006, 1136, 57. Variance (peak width/2) 2 Room Temp (20C) Elevated Temp (30C) Velocity (flow in ml/min)

4 Background Variance, AKA peak width σ 2 observed = σ 2 injection process + σ 2 column + σ 2 extra-column σ 2 extra-column can readily be made negligible. Guiochon et.al. says: The contribution of the sampling device is particularly deleterious since, for a 2 μl injection, the maximum solute concentration in the peak that enters into the column is nearly ten-fold lower than that of the sample. [Measured by UV at injection valve.] Injection process dilutes sample fold. [Measured by UV at injection valve.] σ 2 injection process is as much as 80% of σ 2 observed in an otherwise optimized LC system (assumes good column).* i.e. column always contributes 20% or more. Also observed for gradient. [Measured by UV at injection valve.] The ultimate speed and separation efficiency in LC is not limited by mass transfer efficiency in the column (i.e. not limited by d p ).* IMPORTANT: primarily limit is σ 2 injection process which ultimately defines minimum σ 2 observed. This must be considered when interpreting trends in σ 2 observed. Nevertheless, mass transfer efficiency still plays an important role in achieving peak symmetry and the narrowest peak widths (σ 2 column cannot be made completely negligible, 20% minimum observed repeatedly). This presentation describes a reaction model for understanding conditions for minimizing σ 2 column during fast (high velocity) operation (i.e. getting σ 2 column down to near only 20% of σ 2 observed). *F. Gritti, A. Felinger, G. Guiochon, J. Chromatogr. A, 2006, 1136, 57. *Henry, R.A., in Modern Practice of Liquid Chromatography, J.J. Kirkland ed., Wiley-Interscience: New York, 1971.

5 Background: Variance AKA peak width Direct on-column injection (see below) Loop style 6-port 2-position (ordinary autosampler) The limitation imposed by σ 2 injection process on σ 2 observed in an otherwise optimized LC system has long been known! Directly from: L.R. Snyder & J.J. Kirkland, Introduction to Modern Liquid Chromatography, 1974, Wiley, New York. Original work: *Henry, R.A., in Modern Practice of Liquid Chromatography, J.J. Kirkland ed., Wiley- Interscience: New York, Same approach works with split flow with low flow side going through a 6 port injection valve. *J.J. Kirkland, W.W. Yau, H.J. Stoklosa, C.H. Dilks, J. Chrom. Sci., 1977, 15, 303. *B. Coq, G. Cretier, J.L. Rocca, Chromatographia, 1978, 11(8), 461.

6 Background van Deemter & its application to gradient operation σ 2 column = A + B/υ + Cυ van Deemter equation* where A d p B D and C d p2 /D The plate height (H) is defined as H = σ 2 column /L. Thus H simply normalizes peak width to column length without any other effect on van Deemter equation.** The plate model cannot be applied to gradient operation (due to definition of N) and is not required to make effective use of the van Deemter equation.** The ordinary definition of N factors out k dependence of B term for isocratic operation (which is α k 1 in the σ 2 column form above). Poppe*** has described the k dependence for gradient. While it can t predict gradient ramp rate effects (as written), the above general form (in terms of σ 2 ) of the van Deemter equation otherwise describes the known significant sources of dispersion in gradient reverse phase LC*** and is a better model (than plate model) for the results oriented environment where resolution per unit time and hence peak width is the parameter being optimized. ** Reaction nature of separation process is the key missing detail above for understanding the combination of T and gradient effects. *J.J. van Deemter, F.J. Zuiderweg, A. Klinkenberg, Chem. Eng. Sci. 1956, 5, 271. **J.C. Giddings, Unified Separation Science, Wiley: New York, **E.D. Katz, R.P.W. Scott, J. Chromatogr. 1983, 270, 29. ***H. Poppe, J. Paanakker, M. Bronckhorst, J. Chromatogr., 1981, 204, 77. ***F. Gritti, G. Guiochon, J. Chromatogr. A, 2007, 1145, 67. ***L.R. Synder, J.W. Dolan, J.R. Gant, J. Chromatogr., 1979, 165, 3.

7 Hypothesis Mass Transfer 2 totally different diffusion limited reactions LC Equilibrium (i.e. slowest reaction regulates mass transfer) Velocity due to bulk flow (υ) Desorption reaction (endothermic) SA S + A (desorption time also may limit rate) Diffusion of A to the flowing eluent (D m ) Unstirred eluent layer at particle surface Surface diffusion delays desorption Pore diffusion delays return to flowing eluent Particle surface (S) Velocity due to bulk flow (υ) Adsorption reaction (exothermic) A + S SA Diffusion of A to S (D m + D a ) Unstirred eluent layer at particle surface (activation barrier) Collision required for adsorption (functionally Instantaneous upon collision) Particle surface (S) Desorption limited kinetics leads to the classic, ubiquitous tailing peaks. Adsorption limited kinetics yields fronting peaks. In order to control the mass transfer rate via analyte diffusivity, T must remain balanced with υ in order to achieve minimum σ 2. The balance involves achieving sufficient mass transfer of analyte and thus peak symmetry without causing excess longitudinal diffusion and thus excess peak width. van Deemter assumed equilibrium & thus his equation may need adaptation.

8 Hypothesis: : Diffusion limited reactions, Brownian Motion (D m ) or Reaction Driven (D a )? Reaction energy coordinate for chromatographic adsorption / desorption Not Adsorbed (in flowing mobile phase) Non-equilibrium steady state Diffusion limited Isocratic Elution Kinetics are desorption (D m ) limited Desorption time (t des ), surface (D s ), and pore (D p ) difffusion can slow desorption E a Activation Energy (may be converted to translational energy toward particle surface or D a ) ΔH Binding Energy (retention indices) Adsorbed Energy well made more shallow by gradient or lowering k ΔΔH for Gradient Elution may be converted to translational energy away from particle surface such that D ΔΔH > D a > D m Desorption is caused by gradient Kinetics may not be desorption limited Kinetics may become adsorption limited and thus T activated or potentiated!

9 Hypothesis: : modified van Deemter σ 2 column = A + B/υ + C ads υ + C des υ Reaction based van Deemter equation where C ads υ = C 1 υ d p2 /(D m + D a ) and C des υ = C 2 υ d p2 /(D m + D ΔΔH ) + C 3 υ R(1 - R)t des Most models use D = D m in both B and C terms. Adaptation is simply to add in additional reaction diffusion into the split C term components (ads & des) of the ordinary van Deemter equation. D = D m + D a in the C ads and D = D m + D ΔΔH in the C des term maybe better model. Surface component of C des definition is borrowed directly from Giddings (random walk) and provides a simple way to address all desorption delays.*** Herein, D is general analyte diffusivity from all sources, D m is analyte diffusivity in the bulk eluent from Brownian motion* and D a is additional adsorption reaction and D ΔΔH is additional desorption reaction driven analyte diffusivity at the non-flowing layer of eluent at the particle surface (at the activation barrier).** For the surface component (C 3 term), f(k ) = R(1 R) for isocratic where R is retention ratio and for a gradient R or k equivalent see H. Poppe, J. Paanakker, M. Bronckhorst, J. Chromatogr., 1981, 205, 77. Convenient simplification for the hypothesis presented here is that D ΔΔH is large and t des is small and thus C = C ads for adsorption limited kinetics conditions (ACN gradient conditions). *A. Einstein, Ann. Phys. 1905, 17, 549. **S.A. Arrhenius, Z. Phys. Chem. 1889, 4, 226. ***J.C. Giddings, Unified Separation Science, Wiley: New York, ***E.D. Katz, R.P.W. Scott, J. Chromatogr. 1983,, 270, 29.

10 Hypothesis: relating separation speed to mass transfer in the modified van Deemter Since a central goal of chromatographic practice is to maximize separation efficiency and hence minimize σ 2, it can be recognized that the following relationship* for optimal eluent velocity at minimum σ 2 can be derived (assuming C = C ads to test hypotheses made here): υ optimum = (B/C) 1/2 D/d p. (D m (D m +D a )) 1/2 /d p The impact of this pivotal relationship is that increased υ optimum clearly requires decreased C term (perhaps while accepting increased B term) and thus demands increased mass transfer of analyte to/from the stationary phase. Thus, it should be possible to choose elevated υ for speed and still achieve minimum σ 2 through optimization of either or both D and d p. The relative rate of change in υ optimum (constant d p ) as a function of temperature should provide an estimate of the net rate of change in D (D net ) as a function of temperature. Thus we can test hypothesis by comparing rate of change in observed D net as a function of T with that expected from Brownian motion and estimate the potential impact of D a on mass transport. *J.C. Giddings, Unified Separation Science, Wiley: New York, *E.D. Katz, R.P.W. Scott, J. Chromatogr. 1983,, 270, 29.

11 Hypothesis Diffusion van Deemter meets Arrhenius Modeling reaction rates and even Brownian motion are frequently approximated (in a closed system) through use of the simplest form of the Arrhenius equation: D D o e -E a/rt where D o is the reaction frequency factor (only applicable over a reasonable known range of T), E a is the activation energy of the reaction, and R is the universal gas constant.* Since we can t separate the different components of D, then perhaps we should sum their vectors into a composite form of D: ΔD net D o e -ΔE net_mt/rt where ΔD net is the net mass transfer and ΔE net_mt is the composite rate of change in D net To test this hypothesis and for practical use of these findings, it is desirable to express ΔD net in terms of an experimental parameter. Since υ optimum ΔD net, then ΔD net /D o = υ optimum /υ o and thus: υ optimum υ o e ΔE net_mt/rt where υ o is the frequency factor normalized to eluent υ over a known range in the dynamic gradient chromatographic system (not a closed system). This can be used to fit data with vant Hoff plot. *S.A. Arrhenius, Z. Phys. Chem. 1889, 4, 226.

12 Practical Stuff: : How to control T? *Ordinary air ovens are totally insufficient! ΔT 12 C C for the entire column throughout the gradient is high pressure.** Even water baths are not enough!** If you have an Agilent 1050/1100/1200, Metalox,, or Selerity column oven, or other setup with preheating then you are ready! Just use the supplied preheater. Otherwise, all you need is a mobile phase preheater (inexpensive). Examples include: Selerity CaloraTherm ( ml/min) active on-demand heating. AgileSLEEVE ( 1 ml/min L=40 cm tubing req.) a compact tube oven. J-KEM Sci.. Prep ( ml/min) active but high mass dampened. Once you have active eluent preheating, simple air based column heating can keep the column steel at the desired T.*** ΔT 4 C C for the entire column throughout the fast gradient is achievable on all scales (prep too). Smaller ΔT correlates well with narrower peaks! [Also, lower mass often correlates with smaller ΔT.] *R.J. Perchalski, B.J. Wilder, Anal. Chem. 1979, 51, 774. *R.G. Wolcott et al., J. Chromatogr. A 2000, 869, 211. **A. de Villiers, H. Lauer, R. Szucs, S. Goodall; P. Sandra, J. Chromatogr. A 2006, 1113, 84. ***B.A. Jones, J. Liq. Chromatogr. Related Technol. 2004, 27, Selerity CaloraTherm Pre-Heater makes temperature control easy!

13 Results Optimum Temperatures a different way to measure optimum velocity conditions Peak Width (2 sigma in seconds) Peak Width vs. Separation Temperature at 7 mm/s Oven and Mobile Phase Temperature ( C) ACN gradient Met-Enkephalin (Inertsil - 3 um) Reserpine (Inertsil - 3 um) Reserpine (Acquity / X-bridge BEH 1.7 um) Poly. (Met-Enkephalin (Inertsil - 3 um)) Poly. (Reserpine (Inertsil - 3 um)) *H. Poppe, J.C. Kraak, J. Chromatogr. 1983, 282, 399. T control allows observation of optimization curves.* Experimentally easier than van Deemter curves to determine υ optimum as a function of T. Note T optimum seems to be independent of d p or analyte. This suggests constant T desirable for ACN gradient reverse phase LC operation. Differences in peak widths between analytes originate in injection process (seen w/uv connected to inj.). The fractional contribution of the column toward σ 2 observed appears to be independent of analyte (20%). This suggests primary mass transfer process is the same for all analytes undergoing the expected hydrocarbon to alkyl adsorptions.

14 Results Optimum Temperatures It s all about mass transfer; when you need more, T helps! Temperature (C) Optimal temperature as a function of particle diameter at 7 mm/s Particle diameter (um) Inertsil ODS3 X-Bridge BEH SunFire Luna (2) Fit line Optimum T is independent of d p when the separation is not mass transfer limited (d p = μm). As d p increases (d p = 5-8 μm) and separation becomes mass transfer limited, optimum T increases with d p. 2σ observed still <1 s for d p = 5 & 8 μm. T is an alternative to smaller d p as T is a route to at least partially achieve added mass transfer ordinarily gained from smaller d p. Observed d p 3.5 μm consistent with injection process limited. Observed d p 5 μm consistent with: υ optimum = (B/C) 1/2 D/d p.

15 Results Temperature effect on peak shape Adsorption limited (anti-langmuir) kinetics demonstrated Precise T control allows direct observation of changes in D a on chromatographic peak shape. Small reduction in RT is observed that results from the increased solvation power with less eluent modifier at higher T. Gradient speeds desorption shifting overall rate of mass transfer to adsorption regulated (fronting vs. tailing at low conc). At T < optimum (25C), peak fronting occurs because some of the analyte has insufficient time to diffuse to the stationary phase for adsorption. Fronting is corrected by raising T to optimum (45C). At higher T than optimum (60C), the peak is very symmetrical* with the width at 60.7% height increased substantially due to excessive longitudinal D. ACN gradient In ten sity Temperature Optimization at 7mm/sec 25C 45C 60C Retention Time for Met-Enke (min.) *O. Dapremont, G.B. Cox, M. Martin, P. Hilaireau,; H. Colin, J. Chromatogr. A 1998, 796, 81.

16 Results Temperature effect on peak shape Separation Temperature (C) Minimum T for Peak Symmetry vs. Velocity Peaks are symmetric Peaks are fronting Eluent Velocity (mm/s) T optimization curves at multiple velocities allows estimation of υ optimum as a function of T (also experimentally easier than van Deemter curves). Data for ACN gradient Peak symmetry known to be dependent on T and is known to occur at minimum peak width.* It is recommended that the typical T setting should have bias (< 5C) toward a slightly higher T than that of the minimum peak width. Further bias can be used to ensure symmetrical peaks and is especially beneficial at prep scale where anti-langmuir kinetics are frequently observed due to high mass loading. *O. Dapremont, G.B. Cox, M. Martin, P. Hilaireau,; H. Colin, J. Chromatogr. A 1998, 796, 81.

17 Results Example use of temperature bias It s about mass transfer; when you need more (particularly adsorption), T helps! Effect of temperature during high mass loading conditions Intensity 55C 45C 35C 25C Retention time (min) 100 mg imipiramine injected on 30 x 50 mm C 18 column (d p = 3 μm Inertsil ODS3, ACN gradient)

18 Separation Temperature (C) Arrhenius effect demonstrated Results Arrhenius effect demonstrated Activated mass transfer in LC is real! Temperature vs. Optimal Velocity Eluent Velocity (mm/s) van`t Hoff Form of Plot /Temperature (1/K) Data points = ACN gradient operation (D m + D a ) (phases: X-bridge all points, Inertsil 1-14 mm/s) **Dashed red line = ACN isocratic operation (D m ) where ΔE net_mt = 18 kj/mol (SunFire) ln (Velocity) *A. Einstein, Ann. Phys. 1905, 17, 549. **S.A. Arrhenius, Z. Phys. Chem. 1889, 4, 226. ***F. Gritti, G. Guiochon, Anal. Chem., 2006, 78, Fitting υ optimum as a function of T using the vant Hoff approach allows highly reproducible estimation of proper T at desired υ. 30 fold increase in υ optimum resulting from a 20% T increase (υ optimum (B/C) 1/2 (D m D a ) 1/2 when D a > D m ). Brownian motion / viscosity (η) reduction alone,* where η 1 /η 2 = P 1 /P 2 = D 2 /D 1, accounts for little change in υ optimum (3 fold increase in D m ) based on observed 65-70% P drop over the T range shown (red line, isocratic operation).*** Thus, the factor of 30 change in υ optimum for gradient operation can only be explained by significant additional D a (nearly 300 fold increase) that is reaction driven or Arrhenius** in nature. υ optimum υ o e -ΔE net_mt/rt ACN gradient: ΔE net_mt = 45 kj/mol & υ o = 1.6 x 10 5 m/s

19 Results Scalable speed (velocity) increases while maintaining resolution (totally routine) Intensity (AU at 275 nm) Intensity (AU at 275 nm) Peptide mixture at 7 mm/s, 45C and <250 bar Time (min) Peptide mixture at 14 mm/s, 60C and <500 bar Time (min) Waters Alliance HPLC 7 mm/s (t o = 0.12 min) 4.6 x 50 mm, d p = 3 μm Flow 5 ml/min Average 2σ 2 = 0.62 s Peptide mix (test mix for evaluating 2 hr proteomics gradient separations) Waters Acquity UPLC 14 mm/s (t o = 0.06 min) 2.1 x 50 mm, d p = 3 μm Flow 2 ml/min Average 2σ 2 = 0.32 s ACN 1-30% gradient, Inertsil ODS3

20 Examples: Open Access LC/MS 3: UV Detector: 240_400 AU AU 1.0e : UV Detector: 240_ from a 14 year old system (except computer) A random synthetic chemistry sample DAD chromatograms Combine (64:69-(49:51+82:84)) % % Combine (65:69-(49:52+82:84)) (1) 1.13 (3) /- ion mass +/- ion spectra mass spectra m/z (3) 1.40 (5) e+1 Range: 1.368e+1 Time Range: Same sample 2 weeks later (re-diluted) 2:MS ES- 1.6e+003 m/z 1:MS ES+ 1.1e+004 Time Works routinely on most any HPLC system.

21 Examples: Open Access LC/MS under ultra fast conditions (14 mm/s) A random synthetic chemistry sample DAD chromatogram 14 mm/s HLPC reaches k = 17 in 1 min. (also readily achieved with ordinary low cost LC/MS components via adaptation of HPLC pressure limits for 400 bar operating pressure, using a 7 year old system)

22 Example: Challenging Separations High Quality and Fast An isopropyl sulfonamide with >400 MW n-propyl analog (due to impurity in starting material) Fast separation data from Open Access LC/MS from a 14 year old system (except computer)

23 Example: Quantification by Fast LC/MS/MS A functional cell membrane assay (fatty acid like substrate product). 45 s (0.75 min) analysis time, nm sensitivity. Good quantitative precision (Z = 0.84 or >30x σ between controls). 372k compounds screened 10x, 60k analyses. 119 hits found. 100 % % Internal Standard Substrate Product e5 2.61e5 Drug discovery project launched

24 Conclusions So called ultra fast LC can be successfully practiced using existing main stream HPLC systems (proven for > 10 6 analyses) for velocities up to 7 mm/s at ΔP 200 bar (typical UPLC use is 5 mm/s). Higher pressure limits allow velocities up to 15 mm/s to be routinely utilized for HPLC or UPLC (still at 400 bar or less operating pressure). Mass transfer in isocratic LC is desorption limited, but in gradient LC, it can be adsorption limited, thereby allowing favorable use of activated mass transport in a reaction based van Deemter model. Recognition of Arrhenius type* mass transfer (relative to Brownian motion)** & utilization of the empirical equation makes finding and maintaining the right conditions rational and straight-forward. υ optimum υ o e ΔE net_mt/rt ACN gradient: ΔE net_mt =45 kj/mol & υ o =1.6 x 10 5 m/s New products that afford better mobile phase temperature control greatly simplify achieving these performance levels on most existing LC systems (i.e. make it push button follow recipe to set up). These also help with UPLC. Actively preheating mobile phase is crucial to achieving the best possible performance. Control of mass transfer rates via diffusion using active temperature regulation appears to be the most general & predictable approach for optimizing high υ gradient reverse phase separations (ACN 3-30 mm/s). *S.A. Arrhenius, Z. Phys. Chem. 1889, 4, 226. **A. Einstein, Ann. Phys. 1905, 17, 549.

25 Back up slides: Practical consequences of being injection process limited Expected improvements with particle size reduction level off (can even slow separation). *J. Kofman, Y. Zhao, T. Maloney, T. Baumgartner, R. Bujalski, Am. Drug Discovery 2006, 1, 12. **T.L. Chester, S.O. Terami, J. Chromatogr. A, 2005, 1096, 16. Peak Width as a Function of Particle Size for Reserpine Peak Width as a Function of Particle Size for Met-Enkephalin Peak Width (s) HPLC M easured (Luna) UPLC M easured (SunFire) Theory (C term) Peak Width (s) HPLC M easured (Luna) UPLC M easured (SunFire) Theory (C term) Particle Diameter (um) Particle Diameter (um) ACN soluble compounds: peak widths level out at 3 μm 3 μm looks like way to go. H 2 O soluble compounds: peak widths level out at 3-5 μm Minimum σ 2 exiting column slightly larger than σ 2 entering column (HPLC or UPLC, by connecting UV to inj valve). Best half dozen columns all yield about the same performance (C 18 Luna and Sunfire shown). Velocity = 7 mm/s, T = 45 C, L = 50 mm, column diameter = 4.6 mm HPLC (Luna) & 2.1 mm UPLC (SunFire).

26 Back up slides: Practical consequences of being injection process limited Ideal column diameter depends on performance of injector. Well known in literature, see: L.R. Synder & J.J. Kirkland, Introduction to Modern Liquid Chromatography, 1979, 2nd Ed., John Wiley & Sons: New York Infinite Diameter Effect or dispersion at column wall 3 decades ago Peak Width vs. Column Diameter for Met-Enkephalin at Constant Velocity and Retention Time 4.6 mm ID looks like way to go (HPLC). Variance (~peak width) Multi-path dispersion can become a primary contributor to σ 2 when HPLC column diameter is reduced (1 μl injection). Now, using Waters Alliance Column Diameter (mm) These curves can be flattened well below 1 mm diameter by using direct on-column syringe injection. *Henry, R.A., in Modern Practice of Liquid Chromatography, J.J. Kirkland ed., Wiley-Interscience: New York, 1971.

27 Back up slides: Practical consequences of being injection process limited Operation under infinite diameter conditions gives best separation efficiency. Reducing diameter below that significantly sacrifices separation efficiency. Automation required: : direct syringe on-column injection provides narrow peaks but is not sufficiently automated. Column diameter must be scaled to delivered injection volume to get best separation efficiency and speed. Delivered injection volume (2σ) ) can be measured by connecting UV detector directly to injection valve. Instrument choice is one way to reduce column diameter for improved sensitivity without sacrificing separation efficiency. Key volumes / column diameters to maintain efficiency: 2σ 50 μl col. dia mm (ordinary HPLC) 2σ 10 μl col. dia mm (example: UPLC) 2σ 0.2 μl col. dia mm (example: Eksigent Express)

28 Intensity Instrument choice as a solution to being injection process limited: Equal performance possible at column ID 0.3 mm Peptide mix separated using different column inside diameters Retention Time (min) 4.6 mm ID - Waters Alliance 2.1 mm ID - Waters UPLC 0.3 mm ID - Eksigent Express Velocity = 7 mm/s (i.e. flows = 5000, 1000, & 20 μl/min respectively). Sample: HPLC peptide mix - Sigma H-2016 (different lots). Stationary phase: Inertsil ODS3, 3 μm, 50 mm length (different lots). Injection volumes: 1000, 500, & 150 nl respectively. Mobile phase: buffer (0.2 % HOAc) & ACN ramped from 1 to 30% in 1 min. Instruments and scientists were all different in different labs. Smaller ID yields higher sensitivity α (column ID) 2 Sensitivity makes extra effort worthwhile.

29 Results - Concerns: Frictional Heating Viscous resistance to flow generates heat! Heat = Flow x ΔPressure Has been suggested to limit performance of sub 2 μm particles where ΔT between column inlet and outlet can readily be greater than 12 C.* Eluent / Column Oven T differentials reported to correct separation.** Many T differentials tested (positive & negative). ΔT 5 C C has no statistically significant impact on peak width! Larger ΔT has same negative effects as non-optimal optimal T. Thus goal is no T mismatch. In ten sity Temperature Optimization at 7mm/sec 25C 45C 60C Retention Time for Met-Enke (min.) *A. de Villiers, H. Lauer, R. Szucs, S. Goodall; P. Sandra, J. Chromatogr. A 2006, 1113, 84. **J.D. Thompson and P.W. Carr, Anal. Chem. 2002, 74, **O. Dapremont, G.B. Cox, M. Martin, P. Hilaireau,; H. Colin, J. Chromatogr. A 1998, 796, 81.

30 Results - Concerns: Frictional Heating Frictional heating was measured with low mass RTDs along the column length and the difference between column inlet and outlet was < 1 C when column oven and eluent were set at same T that is at or above the optimum. At υ 3 mm/s and when the eluent and column are not heated, frictional heating usually results in an outlet T that is significantly greater than the inlet and heats the eluent and column to > 80% of the outlet T within 1 to 2 cm from the column inlet. This suggests that once the eluent is sufficiently heated, the viscosity and P drop is lowered* to the point where relatively little more frictional heating is occurring. Lack of effect also makes sense in terms of the total energy balance. A 4.6 x 50 mm 3 μm with 5 ml/min ACN / water has a pressure drop of 200 bar and would be expected to produce less than 0.5 W of heat.** In contrast, the eluent gradient requires a measured 7-10 W to preheat. Once the eluent is preheated, ΔT between column inlet and outlet can be readily held at less than 1 C. Conclusion: Eluent preheating minimizes frictional effects to the point where they appear to be negligible.*** *A. Einstein, Ann. Phys. 1905, 17, 549. **A.D. Jerkovich, J.S. Mellors, J.W. Jorgenson, LCGC 2003, 21, 600. ***A. de Villiers, H. Lauer, R. Szucs, S. Goodall; P. Sandra, J. Chromatogr. A 2006, 1113, 84.

31 Results - Concerns: Analyte and Stationary Phase Stability Analyte stability is not an issue due to the short exposure time. * Silica based stationary phases are generally known to be stable and insoluble in the eluent up to 60 C C over a ph range from 3 to 8. Both silica and analyte stability within this range of conditions have been thoroughly confirmed with >10 6 analyses in our laboratory using the high υ approach described here. While it has been suggested that some silica bonded phases might facilitate temperature stability increases to 100 C, the stability standards often seem low (measured in k not injections) and specific stabilities in commonly used reverse phase eluents are not yet well established beyond review of a few trends. ** There are stationary phase material alternatives, including graphite, polymers, silica hybrids, and zirconia,, all of which can be more stable at higher temperatures. However, none of these materials, except perhaps silica hybrids are yet capable of delivering the high separation efficiencies that are possible with high quality silica based phases. *J.D. Thompson and P.W. Carr, Anal. Chem. 2002, 74, **Y. Yang, D.R. Lynch Jr., LCGC 2004, 22, 34. **H.A Claessens, M.A. van Straten, J. Chromatrogr. A, 2004, 1060, 23.

32 Results - Concerns: Other solvents & stationary phase stability Small changes in T have large effects on D and high speeds up to 15 mm/s are readily optimized at temperatures at or below the 60 C maximum for stability of silica when using ACN / water gradient. However, it is be noted that this eluent combination is the most favorable choice available based on low pressure as well as high D ΔΔH and organic strength for high υ gradient reverse phase LC. When other eluents are chosen for their unique selectivity properties, lower D and hence υ are the result. For example, methanol / water eluent often is observed to have desorption limited kinetics even under gradient conditions and thus υ optimum D m. (ΔE net_mt = 20 kj/mol). The use of polymer phases can further slow desorption due to surface diffusion (ΔE net_mt = 11 kj/mol polymer coated Zirconia).* When one considers the potentially valuable use of eluents resulting slow desorption like methyl or isopropyl alcohol in fast gradient LC separations, there may be real interest in operating with T increased substantially beyond those shown here.* Thus, high T silica solubility could become a limiting factor on the speed of gradient reverse phase LC separations if current trends continue toward the use of increased υ. *B. Yan, J. Zhao, J.S. Brown, J. Blackwell, P.W. Carr, Anal. Chem. 2000, 72, 1253.