t the end of the 1970s, HPLC reached a firstlevel of maturity as a method for instrumen-tal analysis. Breakthroughs at the base of

this progress were the availability of columns withstable microparticulate packings of the reversed-phase type, and the development second- and third-generation pumps, autosamplers, and spectrophoto-metric detectors. These modules were much morerobust and rugged than the first-generation instru-mentation that appeared in the early part of thatdecade. At that time, the optimal internal diameterfor an HPLC column was regarded to be ≥4.0 mm(normal-bore columns). A number of papers in 19801

gave comprehensive reviews of the state of the art inHPLC.

In a landmark paper,1 Knox outlined the perfor-mance characteristics of packed and open-tubularcolumns with much smaller internal diameters.2 Healso described instrumental requirements with regardto band spreading, sample injection, and detectionvolume. The main conclusion of his work was that,

provided the dispersion requirements in the systemare met, HPLC columns narrower than 1 mm i.d. willbe as efficient as columns with ≥4.0 mm i.d. In addi-tion, Knox concluded that the permeability of verynarrow columns is higher when the ratio of tube-to-particle diameter decreases. In more recent work,Knox has expanded his findings.3

In addition to these theoretical considerations, nar-row- and therefore low-volume columns have advan-tages because of minimized solvent usage, more sensi-tive detection in sample limited cases, and lessdemanding coupling requirements and higher sensi-tivity with mass spectrometric detectors. However, forpractitioners of HPLC in the 1980s and ’90s, it becameclear that commercial instrumentation was compro-mised by constraints in physics (especially optics) andmanufacturing technology. Therefore, in general, sep-aration efficiency and concentration sensitivity wereless when narrow-bore columns (3 mm i.d.) were used

than with larger-diameter columns. Cost, labor, andanalysis time benefits were marginal or not relevant(e.g., solvent usage). Most importantly, narrowcolumns established an impression for lack of robust-ness and ruggedness, were more prone to fouling, andin general were more difficult to operate. Therefore,the use of narrow-bore columns was not established asa routine methodology and has found applicationonly in some important niche areas.

In this article, the authors describe a capillaryHPLC system that offsets the aforementioned short-comings of older systems. Properties of capillaryHPLC columns for this system are discussed, and ap-plications in particular in combination with MS aregiven.

Nomenclature for HPLC columns

The ambiguous description of HPLC columns withregard to their diameter confuses practitioners ofHPLC. Terms such as “microbore” and “capillarycolumns” are used without reference to a generalnorm. Therefore, the designation of an HPLC col-umn according to its diameter and consequentlytype of chromatography as used in this article isgiven in Table 1. It would be beneficial for communi-cation among chromatographers if this ambiguitycould be removed by adhering to this (or similar)proposal as a de facto standard.

In Table 1, the internal diameter and typical flowrate range for operation of a particular column type

A

A system and columns for capillary HPLC

Gerard Rozing, Maria Serwe, Hans-Georg Weissgerber, and Bernd Glatz

Commercial instrumentation wascompromised by constraints inphysics (especially optics) andmanufacturing technology.

Table 1Nomenclature used for HPLC columns

Description Diameter Typical flow rateOpen-tubular liquid ≤20 µm i.d. <50 nL/min

chromatographyPacked capillary column liquid ≥ 50 µm < 1.0 mm i.d. 0.2–100 µL/min

chromatographyMicrobore column liquid ≥1.0 mm ≤ 2.1 mm i.d. 100–500 µL/min

chromatographySmall (narrow)-bore column >2.1 mm < 3.9 mm i.d. 500–1500 µL/min

liquid chromatographyNormal-bore column liquid ≥3.9 ≤ 5 mm i.d. 1.5–5 mL/min

chromatographySemipreparative column liquid >5 mm i.d. ≥5 mL/min

chromatography

Reprinted from American LaboratoryMay 2001

are given. The system described in this paper falls un-der the category of “Capillary column liquid chro-matography,” and consequently will be designated ascapillary HPLC system.

Requirements for a capillary HPLC system

The first step in defining the instrumental require-ments for an HPLC system is the selection of the majorfield of application of the system and the correspond-ing column. Addressing recent trends in the HPLCmarket and/or customer needs is also paramount.

There is no doubt that the current pressure on thepharmaceutical industry is to reduce time and costsfor the development of new drugs and therapies. Thishas spurred progress in the development of high-throughput instrumentation for (parallel) synthesis,screening, and analysis. The improvements can beachieved only with a simultaneous significant reduc-tion in scale of operation. The value chain in the(bio)pharmaceutical industry—unraveling themolecular base of disease, new drug discovery, drugdevelopment, clinical trials, and manufacturing—ischaracterized by the necessity to handle increasingnumbers of smaller samples for screening and analy-sis (e.g., drug metabolism pharmacokinetic [DMPK]studies).

Furthermore, the postgenomic era has begun. Pro-teomics involves the comprehensive, quantitativemeasurement of the expressed protein complementin a cell or tissue under different conditions (e.g.,normal versus tumor cell). Automated handling andanalysis of minute amounts of proteins obtainedfrom a 2-D polyacrylamide gel electrophoresis(PAGE) plate will be required.

These developments have such common require-ments as handling of very small samples, high sensi-tivity, high throughput (96- or 384-well plate com-patibility), and seamless coupling with massspectrometric detection.

With these requirements and the general consid-erations in mind, it was obvious to opt for an HPLCsystem able to operate packed capillary columns(0.3–1.0 mm i.d.). From that selection, dimensions ofsuch a system with regard to abilities of pumps, injec-tion device, column thermostat, and detector can bederived. The real challenge for development, how-ever, was to take a major step forward in ruggednessand robustness of such a system without compromis-ing overall performance to change the negative per-ception of capillary HPLC systems indicated above.

Therefore, the objective was set to develop anHPLC system on a modular basis that is able to runHPLC columns of 0.3–1.0 mm i.d. with similar func-tionality, performance characteristics, robustness,and ruggedness, i.e., one that a practitioner in thefield expects from a system that runs normal-boreHPLC columns. The system should be easy to set up,use, validate, maintain, diagnose, and repair, withhigh day-to-day reproducibility, high reliability, up-time, and high sample throughput running packed

capillary columns that have uncompromised perfor-mance and longevity.

The resulting system is shown in Figure 1. Individ-ual requirements for the modules and their realiza-tion are detailed in the following paragraphs.

Capillary HPLC pumping system

With the diameter of the columns for the capillaryHPLC system fixed, one can calculate the volumetricflow rates required to operate these columns at themobile phase velocities typical for HPLC columns(1–6 mm/sec). In practice, the maximum velocity isconstrained by the maximum pressure at which thepump can deliver the solvent viz, 400 bar. The resultof this calculation is given in Figure 2.

Figure 1 1100 series capillary HPLC system (Agilent Tech-nologies, Waldbronn, Germany).

Figure 2 Calculation of the volumetric flow rate as a func-tion of column diameter. Calculation is based on a mobilephase velocity of 1–6 mm/sec, a column porosity of 0.7,viscosity of 1 cP, and assuming a column resistance factorof 700. The column pressure will be approx. 400 bar atthe highest velocity. Sensor flow rate ranges are for the twoseparate flow rate measurement sensors that can be usedin the system.

The calculations were done with basic chromato-graphic equations for pressure drop as a function ofvelocity. In Figure 2, the horizontal arrows show thevolumetric flow rate range for a particular column di-ameter with the assumptions given in the figure cap-tion. Thus, for example, for a column with 0.3 mmi.d., a typical flow rate range of 3–18 µL/min isneeded. The overall flow rate range for packed capil-lary columns according to Table 1 will typically be1–100 µL/min, with the capability to extend thisrange below 1 µL/min and above 100 µL/min, also al-lowing the occasional usage of wider i.d. columns.For the practitioner, it is desirable that this flow raterange (1–100 µL) be well controlled and independentof the column backpressure (unlike in passive split-flow arrangements). Delay volume/time in a capillarypump must be low and allow effective mixing so thatsolvent gradients are delivered to the columnquickly, accurately, and precisely. These goals areachieved with an 1100 series high-pressure binarygradient pump equipped with electronic flow control(Agilent Technologies), as illustrated in Figure 3.

An electromagnetic purge valve (EMPV) splits thesolvent delivered by the high-pressure pump in acontrolled manner. This valve has an outlet (waste)with an adjustable flow restrictor and an outlet thatcomprises the microflow path, containing a flow sen-sor that measures the flow rate. In practice, two sepa-rate sensors are used for two different capillary flowrate ranges—one for 1–20 µL/min and one for10–100 µL/min. A feedback control loop regulatesthe flow resistance in the waste outlet in order tomaintain a constant microflow rate. For example,when the pressure decreases on the column, flow rateincreases, which is counteracted by a reduction inthe flow resistance in the waste path. In this way, thecapillary pump will be insensitive to pressure varia-tions induced by solvent change such as in gradientelution or by a gradually plugging column.

Since the flow measurement sensor response de-pends on the solvent composition, highly sophisti-

Figure 3 Schematic representation of the electronic flowcontrol in the 1100 series capillary HPLC system.

cated instrument control is needed. Time delays fromthe mixing point of the high-pressure gradient pumpto the sensor need to be known exactly to accountfor the response change when the solvent composi-tion is varied in this pump.

The flow rate before the EMPV is at normal valuesof 0.2–1.0 mL/min (primary flow rate). Therefore, thedelay time for a new solvent composition to the flowsensor and the capillary column is low, despite therelatively high delay volume (0.5–0.8 mL) in this partof the pump. The total dead volume from the sensorto the head of the column is 5 µL for the 1–20 µL/min flow rate sensor. With the 10–100 µL/min flowsensor, the volume is 14 µL (because slightly widerconnection capillaries are used in this case).

The pumping performance of the pump is givenin Figure 4. Typical values for the precision of reten-tion times at a flow rate of 4 µL/min is <0.2–0.3%. Af-ter run 24, the backpressure was increased by 80 barby placing a restriction capillary after the detector. Ascan be seen, there is barely a significant change inthe absolute retention time and precision under thesimulated higher backpressure conditions.

With this system, depending on the actual flowsensor installed, one can operate between 1 and 20µL/min or 10 and 100 µL/min with a controlled rate.It is this flow rate control that warrants the requiredrobustness and ruggedness of the pump, and givesconfidence that the microflow rate specified is actu-ally delivered independent of the backpressure. Ifone wants to operate at higher flow rates, which areactually within the specified range of the 1100 seriesbinary high-pressure gradient pump, the flow sensorcan be bypassed and normal mode selected to allow aflow rate up to 2.5 mL/min.

The pump is equipped with a two-way selectionvalve for each pump channel and a seal wash option.Moreover, the system is supplied with an on-line de-gasser that has a low volume (1 mL). Through the useof new, more effective gas exchange membranes, thedegassing efficiency is the same as that found withlarger-volume solvent degassers.

Figure 4 Flow rate precision during repetitive injection atdifferent column backpressures. Column: Hypersil ODS(Thermo Hypersil Ltd., Runcorn, Cheshire, U.K.), 5 µm,150 × 0.3 mm; flow rate: 4 µL/min; solvent: water/ace-tonitrile; gradient: 5–85% acetonitrile column; sample: iso-cratic checkout sample; injection volume: 0.1 µL; tempera-ture: 25 °C; detection wavelength: 250 nm. Pressureincreased from 80 to 160 bar after run 24.

The requirements and features of the 1100 seriescapillary HPLC pump are summarized in Table 2.

Capillary HPLC system aspects

In order to maintain the column’s separation effi-ciency, a reduction in column diameter mandates a re-duction in the sample injection volume and amountof solute applied to the column. The reduction factorequals the ratio of the squares of column diameters.For example, a 5-µL injection on a 4.6-mm-i.d. columnis equal to a 60-nL injection volume on a 0.5-mm-i.d.column. Therefore, an autosampler for a capillaryHPLC system will need to be able to inject samples

down to the nanoliter scale with good precision.Conversely, the gain in detection sensitivity in

capillary HPLC is achieved when the few available mi-croliters of sample is injected on a very low-volumecolumn, sacrificing efficiency for gain in detectionsensitivity (Figure 5). In the figure, 0.1 µL of a samplecontaining 200 ppm biphenyl is injected on columnswith decreasing i.d., illustrating a limited sample vol-ume case. It is obvious that the highest sensitivity isobtained with the narrowest i.d. column.

This is of particular importance in many life sci-ence and biopharmaceutical applications, since onlymicroliter-sized samples are available. Therefore, it isalso a requirement for a capillary HPLC autosamplerto inject a relatively large sample volume (up to sev-eral microliters) from small-volume containers withhigh recovery. In combining these requirements, the1100 series capillary HPLC autosampler allows injec-tion of sample volumes of a range of 3 orders of mag-nitude, i.e., from 30 nL to 40 µL.

Moreover, reduced volumes and optimized flow ge-ometries of all connections and capillaries are required.This will be discussed in the next section. The sampleinjection valve for the capillary injector has been re-worked to minimize dead volume and optimize flowpaths. Wear resistance of the rotor/stator assembly hasbeen significantly improved for high sample through-put. Even more importantly, though, is the ability tohandle multiple 96-(384-) well plates, which can bethermostated. Since such plates provide very high sam-ple numbers, the ability to overlap the sampling cyclewith the separation cycle is indispensable.

A typical example that reflects the precisionachievable with sample injection at a 0.1 µL sampleinjection volume level is given in Figure 6.

On the detection side, one is again confrontedwith the demand to minimize detection flow cell vol-ume and maximize pathlength for highest soluteband integrity and response. Moreover, spectral in-tegrity must be maintained with diode array spec-trophotometric detection. This demands the use offlat detection cell windows. With these considera-tions in mind, a square, Z-shaped detector flow cell(0.7 × 0.7 × 10 mm) was developed similar to thehigh-sensitivity detection flow cell for the AgilentTechnologies capillary electrophoresis system. De-spite the cell volume (0.5 µL), it provides favorableflow characteristics, conserving the peak integrity ofpacked capillary columns.

Connection capillaries

The connection capillaries are the main contribu-tors to extra-column band spreading in an HPLC sys-tem. Especially in a modularly built instrument, gov-erned by spatial and physical requirements of theindividual modules, a relatively long length of connec-tion tubing is required. This is contrary to the necessityto minimize the dead volume in a system that operateswith packed capillary HPLC columns. Fortunately, hy-drodynamics helps to cope with this contradiction.

Table 2Overview of capillary HPLC pump system

Requirement SolutionPrecise gradient delivery of 1–99% from Solvent delivery at higher primary 1 µL/min and higher independent of flows; split with flow measurementbackpressure control of column flow through

feedback control loopAccurate microflow rate delivery indepen- Flow measurement and feedbackdent of solvent mixing contraction and controlbackpressureSmall delay volume/time for fast Solvent delivery at higher primarygradients and uncompromised gradient flows, microflow channel with mini-profile mal dead volume; programmable

bypass of sample injection loopFull pressure range (0–400 bar) available Flow measurement and feedbackfor microflow delivery control. Polyetherether ketone

(PEEK)-coated fused-silica (FS)capillaries, 0.05 mm i.d. and 0.8 mm o.d. as connecting tubes

Robustness and ruggedness comparable Flow measurement and feedbackto conventional HPLC systems; user-friendly control; PEEK-coated and FS capil-handling of connections, minimal plugging, laries, 0.5 mm i.d. and 0.8 mm and diagnosis of microleaks o.d. as connecting tubes

Figure 5 Mass sensitivity benefit. Injection of the same sam-ple amount on HPLC columns with decreasing internal di-ameter. Stationary phase: ZORBAX® SB-C18; length: 150mm; solvent: water/acetonitrile, 40/60; flow rate: see dia-gram; sample: isocratic checkout sample; injection volume:0.1 µL; third peak: biphenyl, 200 ng; temperature: 25 °C;detection wavelength: 230 nm.

The influence of extra-column band spreading onthe overall dispersion in the system is given by Eq. (1):

σ2v,tot = σ2

v,col + σ2v,ext (1)

where σ2v,tot is the total volume variance of the solute band,

σ2v,col is the variance of the column itself, and σ2

v,ext is thevariance of zone broadening in the extra-column volume.

Zone dispersion in an HPLC column is describedby the well-known van Deempter equation, given inits simplified form in Eq. (2). H is the height equiva-lent to a theoretical plate; u is the solvent velocity;and A, B, and C are regression coefficients.

H (u) = A + B/u + Cu (2)

The coefficients in the van Deempter equation areindependent of the column diameter (assuming thatthe bed structure is independent of the column diame-ter). If the column diameter is reduced and the solventvelocity is kept constant by reduction of the volumet-ric flow rate by the square of the column diameter ra-tio, the same plate height is obtained in the narrowercolumn. The volume variance is then given by:

σ2v,col = = =

constant1 · d2c · H (u) (3)

where VR is the retention volume of the solute, N isthe plate number, εt is the total porosity of the col-umn, Vcol is the geometric volume of the column, Lcol

is the column length, and k′ is the capacity ratio. Thedispersion in the connection capillary is given by thewell-known Aris-Taylor equation. Neglecting thestatic diffusion contribution in this equation and

εT Vcol (1 + k′)H (u)���

Lcol

V2R

�N

supposing that only the connection capillaries con-tribute to external band spreading, the volume vari-ance due to extra-column band spreading can be de-scribed by:

σ2v,ext = = constant2 · d4

cap · Lcap · ucap

(4)

where dcap is the diameter of the connection capillary,F is the volumetric flow rate, Lcap is the length of theconnection capillary, Dm is the diffusion coefficientof the solute, and ucap is the solvent velocity throughthe capillary.

Suppose one is comparing two cases: a columnwith 4.6 mm i.d. with a total length of connectioncapillaries of Lcap and a column with 0.5 mm i.d. withthe same length of connection capillary. To maintainthe same linear velocity through the column, theflow rate will be reduced by the square ratio of thecolumn diameter, i.e., by a factor of 84. In that case,the volume variance of the solute band decreases bythe same factor (Eq. [3]). Suppose the diameter of theconnection capillary is proportionally reduced by theratio of the column diameters, i.e., from 0.25 to0.027 mm, which leads to the same velocity, ucap atthe 84× lower flow rate in the capillary than the vari-ance of the band spreading (Eq. [4]) reduces with afactor of almost 7200. Thus, the effect of externalband spreading on overall band spreading in Eq. [1]is negligible in this case.

With these conditions in mind, it becomes clearthat the internal diameter of the connection capillar-ies can be reduced less than the column diameter isreduced i.e., by half the ratio of the column diameterreduction. In this practical case, the connection cap-illary i.d. is 0.05 mm. Now, the band spreading con-tribution by the capillaries reduces by a factor of 625compared to a peak variance reduction of 84.

There are additional advantages. First, the solventvelocity in the capillary is lower, leading to lowervariance of extra-column band spreading (Eq. [4]).Second, a 2× lower pressure drop over the samelength of connection capillary under comparableconditions is obtained.

However, these considerations are not very practi-cal when stainless steel connection (SST) capillariesare used. Due to fabrication limitations, very narrow-i.d. SST capillaries have high wall roughness and areprone to plugging. Plastic capillaries have limitationswith concentricity of the tubing and will not with-stand high pressure with organic solvent, especially athigh temperatures. Fused silica would be the materialof choice, but does not have good robustness. In theauthors’ approach, the best properties of fused silicaand plastic tubing were combined by heat shrinking aPEEK outside tube onto fused-silica capillaries withthe required internal diameter. The heat shrinkingprocess generates an intimate connection to the poly-

π · d4cap · F · Lcap

���96Dm

Figure 6 Area precision of capillary HPLC autosampler.Column: ZORBAX (Agilent Technologies, Wilmington, DE)300SB-C18, 5 µm; 150 × 0.3 mm; flow rate: 4 µL/min;solvent: 0.1% trifluoroacetic acid in water/0.1% trifluo-roacetic acid in acetonitrile; gradient: 2–40% in 76 min;sample: peptide mixture; injection volume: 0.1 µL; detec-tion wavelength: 214 nm; temperature: 25 °C.

imide outside coating of fused-silica tubing, avoidingformation of small annulus space between the tubesor movement of the inside tube. In this way, the cap-illaries (0.8 mm o.d.) combine the robustness andruggedness of plastic capillaries with the favorableproperties of fused silica (very smooth inner wall andtherefore very low tendency to plugging). The avail-ability of these capillaries is essential to the ability ofthis system to run packed capillary HPLC columnsthat are uncompromised by external band spreading,while providing good robustness and ruggedness.

Columns for capillary HPLC systems

Stationary phases

At the beginning of the 1970s, microparticulate re-versed-phase-type stationary phases became availableand were a major factor in establishing HPLC as a rou-tine analysis method. Nevertheless, these initial sta-tionary phases were far from ideal in handling impor-tant compound classes such as basic molecules andbiologically important high-molecular-weight sub-stances like peptides, proteins, oligonucleotides, andnucleic acids. Workarounds were conceived, butthrough the years, especially in the late 1980s andearly ’90s, a more fundamental approach in stationaryphase development began. This has led to largely im-proved properties of column packings such as highly

uniform and homogeneous surface silanol structure,low metal ion content, wider pore structure, and morestable (wider pH range) stationary phase bonding. Inthis respect, the ZORBAX range of stationary phaseshas been expanded to comprise a series of stationaryphases for optimal usage at a particular solvent pH tomatch the chromatographic properties of the solutes.The basis for the enhanced stability of these newer sta-tionary phases is based in the type of bonding that hasbeen applied. An overview of the bonding chemistryof the phases is given in Figure 7. The optimal workingpH range for the phases is shown in Table 3.

The StableBond series has good stability at low pHsince two bulky isopropyl or isobutyl groups on thelateral silicon atom sterically protect the siloxanebond against hydrolysis and degradation. This ap-proach is also used in the Bonus RP series where, inaddition, a polar group (PG) is embedded in the alkylside chain. This provides good wetting of the station-ary phase at very low or no organic solvent contentin the mobile phase. In addition, the embeddedgroup provides for good peak asymmetry with highlypolar and ionizable solutes by minimizing unwantedinteraction with unreacted acidic silanol groups. TheEclipse XDB series has an exhaustively endcapped sil-ica surface that gives solid protection of the silicabackbone against basic hydrolysis. Therefore, the op-erating pH range is shifted up higher than the Stable-Bond series. The best stability at high pH is providedby the bidentate silane modification used by theZORBAX Extend series.4 The ring-like structure of thereacted bidentate silane, with a hydrophobic carbonchain between the silanizing atoms, effectivelyshields the silica surface from hydrolytic attack. Withthis approach, the operating pH range can be ex-tended from 2.0 to 11.5. Working at high pH has par-ticular advantages for basic compound and peptideseparations, as will be shown later.

Recently, Kirkland et al. reported results, with acomposite particle shown in Figure 8.5,6 This superfi-cially porous particle, POROSHELL, has an impervi-ous core. On the outside of the particle, a 0.25-µm

Table 3Recommended operating pH range for ZORBAX stationary phases

Figure 7 Chemical structure of StableBond, Eclipse XDB,Bonus RP, and Extend binding of alkylsilanes to silica surface.

porous layer is constructed by deposition of colloidalsilica particles on the surface. Bonding chemistries, asdescribed above, can be applied to this porous layer.In this way, particles are obtained that have an inter-particle space of a 5-µm particle but a very short in-traparticle diffusion path. Such particles are prefer-ably operated at very high mobile phase velocitywith macromolecules. Columns with POROSHELLhave the same backpressure as columns packed with5-µm particles, but separation efficiency like columnspacked with very small particles. An example of aseparation with a packed capillary column withPOROSHELL will be given later.

Column hardware

The surface finish of the inner wall of normal- andsmall-bore column tubing has to be very smooth. Invery narrow tubes such a surface finish with stainlesssteel tubing is difficult to obtain. Therefore, glass-lined, steel tubing (0.3 and 0.5 mm i.d.) was selectedfor the capillary HPLC columns. Frits are trouble-some in very narrow columns. The columns used bythe authors are terminated with metal sieves (2 µm).In order to meet dead volume requirements in the fit-tings, standard zero dead volume unions are used.The overall O.D. of the column is 1⁄8 in. to provide ro-bustness to the tube.

Many combinations of stationary phase type(StableBond, Eclipse XDB, Bonus RP, Extend, and Hy-persil), type of bonding (C18, C8, phenyl, and cyano),particle size, diameter, and length can be conceived.7

Efficiency of packed capillary HPLC columns

To illustrate the efficiency that can be achievedwith capillary HPLC columns, a height equivalent toa theoretical plate (HETP) versus mobile phase veloc-ity curve was recorded. The result is given in Figure 9and Table 4. The particular column used in the testwas a ZORBAX Eclipse XDB C18, 3.5 µm with 150mm length and 0.5 mm i.d. The regression coeffi-cients were calculated with Origin™ (Originlab Cor-poration, Northampton, MA). A few striking conclu-sions can be drawn from these results. This columnhad equal, or better, efficiency than a corresponding4.6 mm i.d. with the same phase. The calculated

value of minimum in the H-u curve was about 7 µmor a reduced plate height of 2. This result was also ob-tained for the last retained peak, showing that in the

Figure 8 Schematic representation of ZORBAX Poroshellparticle.

Figure 9 Plot of an HETP versus mobile phase velocity plot.Column: ZORBAX Eclipse XDB C18, 150 × 0.5 mm, 3.5µm; solvent: acetonitrile/water, 3/1; flow rate: 4–20µL/min; sample constitute: see diagram; injection volume:0.5 µL; all solutes: 20–60 µg/mL; temperature: 25 °C; de-tection: 250 nm, detection cell prototype; 80 nL.

Table 4Regression coefficients—van Deempter plot (for experimental details, see Figure 10)*

Hmin Velocity SlopeSolute k (µm) (mm/sec) (ms)Acetophenone 0.28 6.2 1.04 2.24Methyl benzoate 0.38 6.2 1.11 2.15Hexyl paraben 0.55 6.6 0.85 2.77Heptyl paraben 0.65 6.9 0.91 2.73Anthracene 0.84 7.1 1.56 1.84*From the plot of velocity versus pressure, a column resistance factorof approx. 350 was calculated.

Figure 10 Fast separation of alkyl parabens. Column: Hy-persil ODS, 3 µm, 35 × 0.3 mm; solvent A: water, B: ace-tonitrile; gradient: 50–90% B in 2 min; flow rate: 20µL/min; main flow: >800 µL/min; temperature: 25 °C; in-jection volume: 0.1 µL, overlap mode; sample: alkyl-parabens (1–7) + thiourea; detection wavelength: 250 nm.

min/sec. Twelve repetitive injections are shown of0.1 µL of a sample containing seven homologousparahydroxybenzoic acid alkyl esters (n = 1–7) andthiourea as a dead time marker. The gradient was50–90% in 2 min with a 0.5-min hold and 1.5-mincolumn reequilibration time. Sample injection was inthe overlap mode (i.e., the next sample is taken inthe injection needle while the previous sample is stilleluting). An overall cycle time of 4 min is possible inthis case. Precision of retention time of all peaks un-der these conditions is 0.1–0.4% RSD and of raw peakareas is 2–5% RSD. This is an illustrative example anddoes not represent the ultimate limit in speed ofanalysis with this system.

In the second example (Figure 12), repetitive injec-tions of a tryptic digest of myoglobin are shown. Thecolumn used in this case was 250 × 0.3 mm andpacked with ZORBAX 300SB-C18, 5 µm. This station-ary phase is specifically designed for the separation ofpeptides with TFA containing solvent gradients. Themain trace shows the overlay of 10 repetitive injec-tions. The RSD of the retention time over these 10 in-jections is designated on top of the peaks. Since thetime variations are so minor, the 40–41 min regionhas been expanded to illustrate the actual span of re-tention time for these peaks (0.16 min).

In Figure 13a, an example of the benefit of running apeptide separation at high pH is compared with thesame separation attempted under acidic conditions(Figure 13b). The ZORBAX Extend bonding chemistryprovides good resistance of the column at high solventpH. Therefore, ammonia-containing mobile phases canbe used. At the high pH, the charge state of a peptidechanges radically, leading to different chromatographicselectivity. A column with a stationary phase that toler-ates both low and high pH allows orthogonal separa-tions to be carried out without changing the column.Moreover, ammonia is advantageous in coupling with

capillary HPLC system, external band broadening isnegligible. In particular, this finding underlinesKnox’s prediction from 1980 that column efficiencyis not dependent on column i.d.2 Overall, the resultsillustrate that packed capillary HPLC columns with0.5 mm i.d. can be used without compromise in thisinstrument.

A second interesting observation can be made.The column resistance factor, which is the chro-matographic permeability of the column, normal-ized for the particle size and calculated from thepressure/velocity plot, is approximately a factor oftwo lower than for normal-bore columns. Again, thisis exactly in line with predictions by Knox. (This isactually a great benefit because one needs a factor oftwo lower pressure to drive the solvent at a particu-lar velocity in a packed capillary column comparedwith a normal-bore column.) With this finding inmind, one can prepare a plot of total pressure (col-umn + connecting capillaries) versus flow rate (Figure11). This plot reveals that even at the highest veloc-ity that would make sense chromatographically, apressure of less than 400 bar is required to drive thesolvent through both the column and the connec-tion capillaries. The combination of the new pump-ing concept, the relatively wider i.d. of the connec-tion capillaries, and the higher permeability ofpacked capillary columns together allow the fullpressure range of the pump to be exploited for sol-vent delivery.

Application examples

As a first example (Figure 10), a very short packedcapillary column was used to demonstrate speed andreproducibility at a very low flow rate. The columnwas a 35 × 0.3 mm column, packed with 3 µm Hyper-sil ODS. The flow rate is 20 µL/min, which equals 6.7

Figure 11 Calculation of total backpressure over packed capillary HPLC column and connection capillaries. Column length:250 mm; particle size: 5 µm; column i.d.: see graph; column resistance factor: 350; viscosity: 1 cP; total length of 0.05 mmi.d. connection capillary: 200 cm; pressure drop over wider-i.d. connection capillaries in the system neglected.

mass spectrometry. As a volatile additive, it does notpose any problem to be removed in the MS interface.

The data in Figure 13 were obtained on a 2.1-mm-i.d. column packed with ZORBAX 300 Extend C18.

The sample was a mixture of angiotensin I, II, and III.In Figure 13a, the separation is shown with an acidicTFA containing mobile phase. In Figure 13b, thesame separation is shown at high solvent pH. Thedifference in selectivity is remarkable and plausible.Angiotensin II and III differ in just one amino acid—aspartic acid (see Table 5). Thus, at low pH, the over-all charge in angiotensin II and III will be about thesame and (accidentally) the retention is the same. Athigh pH, however, the additional, aspartic acid moi-ety produces a higher charge on the molecule thanangiotensin I and III and thus is less retained.

The use of ammonia containing mobile phasessignificantly increases the signal response and re-duces the background signal and noise partly becauseof the low molecular weight of the additive ammo-nia.8 However, the complete basis for the advantageof ammonia-based mobile phases in LC-MS studieshas not yet been determined.

In Figure 14, the separation of a peptide mixture isshown on a column packed with POROSHELL 300SB-C18, 150 × 0.5 mm. The composite, layer-type struc-ture of the POROSHELL leads to rapid mass transferof macromolecules because of the much shorter dif-fusion distance compared to porous particles. The

Figure 12 Separation of a tryptic digest of myoglobin. Col-umn: 0.3 × 250 mm ZORBAX 300SB-C18, 5 µm; flowrate: 5.5 µL/min; solvent A: 0.05% trifluoroacetic acid(TFA) in water, solvent B: 0.045% TFA in acetonitrile; gradi-ent: 1–61% 0.5% B/min; sample myoglobin tryptic digest:7.5 pmol/µL; injection volume: 1.3 µL; temperature: 25 °C;detection wavelength: 214 nm.

Figure 13 Comparison of peptide separation at high andlow pH. Column: ZORBAX Extend C18, 150 × 2.1 mm;sample: angiotensin I, II, and III, 50 pmol each; injectionvolume: 2.5 µL; flow rate: 0.2 mL/min; gradient: 15–50%B in 15 min; column temperature: 35 °C; MS detection:positive ion electrospray ionization (ESI); fragmentor volt-age: 70 V; capillary voltage: 4.5 kV; nebulizing gas: N2,35 psi, 12 L/min; tip temperature: 325 °C; top acidic elu-tion: 0.1% TFA in water/0.085% TFA in 80% acetonitrile;bottom basic elution: 10 mM NH4OH in water/10 mMNH4OH in 80% acetonitrile.

a

b

Table 5Structures of angiotensin I, II, and III

Angiotensin I H- Asp- Arg- Val- Tyr- Ile- His- Pro- Phe- His- Leu- OHAngiotensin II H- Asp- Arg- Val- Tyr- Ile- His- Pro- Phe- OHAngiotensin III H- Arg- Val- Tyr- Ile- His- Pro- Phe- OH

Figure 14 Fast peptide separation by packed capillaryHPLC. Column: ZORBAX Poroshell 300SB-C18, 150 × 0.5mm; solvent: 0.1% TFA in water/0.1% TFA in acetonitrile;gradient: 5–65% in 15′; flow rate: 20 µL/min; sample:peptide mixture, about 20 ppm; injection volume: 0.1 µL;detection wavelength: 214 nm; temperature: 25 °C. Peaksidentified in figure.

porous layer particles have a higher surface area thannonporous particles; therefore, higher sample load-ing is permitted. By virtue of their structure, suchparticles are very strong and can be made in principlein a variety of particle sizes, porosities, and thick-nesses of the layer and bonding chemistry. With thisparticular stationary phase, “300” refers to the aver-age pore diameter of the silica layer; “SB-C18” refersto the StableBond type of binding of the C18-silaneto the surface.

In this case, the flow rate used was 20 µL/min,which is regarded as low flow velocity (2.4 mm/sec)for this column. The separation of the five peptideshas a much improved efficiency compared with theseparation of the mixture on a normal ZORBAX300SB column. The potential of the POROSHELL sta-tionary phases has yet to be defined.

In a final example given in Figure 15, the separa-tion of a HAEIII restriction enzyme digest of pBR322plasmid DNA is given. In this case, a prototype col-umn, 75 × 0.5 mm, was prepared with the ZORBAXEclipse dsDNA stationary phase. Typical conditionsfor this separation are the use of a triethylammo-nium acetate water/acetonitrile gradient. The separa-tion obtained is again comparable in all aspects withthis separation when executed with a 2.1- or 4.6-mm-i.d. column.

Conclusions

This paper demonstrates that capillary HPLC at aflow rate from 1 to 100 µL/min with columns of 0.3and 0.5 mm i.d. meets the requirements for a routineanalytical separations method. An innovative con-

cept allows the solvent in this flow rate range to bedelivered in a controlled manner, making the solventdelivery system rugged and robust. A significant re-duction in the diameter of the connection capillarieswas achieved with PEEK-covered fused-silica capillar-ies that are far less prone to plugging. The autosam-pler has been devised to deliver samples down to avolume of 30 nL. An optimized flow path and low-volume detection cell provided the best compromisebetween sensitivity and band spreading.

Column technology of ZORBAX phases has ad-vanced in a comparable fashion. Consequent consid-erations of the bonding chemistry have expandedthe working pH range of silica-based stationaryphases to 1–11.5. The advantage of working at thehigh pH range with ammonia as a mobile phase addi-tive is significant with respect to the orthogonal se-lectivity obtained when working at low pH. Porouslayer-type silica modified by the same new bondingchemistries, POROSHELL is very attractive for thehigh-efficiency, high-speed separation of peptidesand proteins.

Finally, a number of examples have demonstratedthat noncompromised capillary HPLC systems arefeasible, addressing specifically those analyses inwhich very low sample amounts/volumes are inher-ent to the problem.

References

1. J Chromatogr Sci 1980; 18:394–582.2. Knox JH. J Chromatogr Sci 1980; 18:453.3. Knox JH. J Chromatogr A 1999; 831:3–15.4. Kirkland JJ, Adams JB Jr, van Straaten MA, Claessens

HA. Anal Chem 2000; 70:4344–52.5. Kirkland JJ, Truszkowski FA, Dilks CH Jr, Engel GS. J

Chromatogr A 2000; 38:3–13.6. Kirkland JJ. J Chromatogr Sci 2000; 38:5359.7. www.chem.agilent.com/Scripts/Phome.asp.8. Boyes B, Apffel A, Chakel J, Hancock W. Poster presenta-

tion, HPLC 2000, Seattle, WA, June 2000.

Figure 15 Separation of DNA restriction enzyme frag-ments. Column: ZORBAX Eclipse dsDNA, 75 × 0.5 mm;solvent: 0.1 M triethylammonium acetate, 0.1 mM ethyl-enediaminetetraacetic acid (EDTA) in water/0.1 M triethyl-ammonium acetate, 0.1 mM EDTA in water/acetonitrile,3/1; gradient: 40–80% B in 30 min; flow rate: 20 µL/min;sample: HAEIII digest of pBR322 plasmid DNA; injectionvolume: 0.5 µL; temperature: 50 °C; detection wavelength:260 nm.

The authors are with Agilent Technologies, Life ScienceBusiness Unit, P.O. Box 1280, D-76337, Waldbronn, Ger-many; tel.: (49) 7243 602309; fax: (49) 7031 4302308;e-mail: Gerard-rozing@agilent.com. The authors wish toexpress their appreciation to Barry Boyes of Agilent Tech-nologies, Little Falls Analytical Division, Wilmington, DE forproviding the results of peptide separation at high pH; Kon-stantin Choikhet of Agilent Technologies, Waldbronn Divi-sion, for the separation of the HAEIII digest of pBR322plasmid DNA; and Jack Kirkland, for meticulously proof-reading the draft manuscripts. Special acknowledgement isgiven to the R&D and marketing teams at Agilent Technolo-gies for their diligence during development and support ofthe applications work.

Agilent TechnologiesPublication Number5988-3282EN