Novel concepts for fast capillary gas chromatography · The most important breakthrough in gas...

Novel concepts for fast capillary gas chromatography

Citation for published version (APA):Deursen, van, M. M. (2002). Novel concepts for fast capillary gas chromatography. Technische UniversiteitEindhoven. https://doi.org/10.6100/IR554373

DOI:10.6100/IR554373

Document status and date:Published: 01/01/2002

Document Version:Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can beimportant differences between the submitted version and the official published version of record. Peopleinterested in the research are advised to contact the author for the final version of the publication, or visit theDOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and pagenumbers.Link to publication

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, pleasefollow below link for the End User Agreement:www.tue.nl/taverne

Take down policyIf you believe that this document breaches copyright please contact us at:[email protected] details and we will investigate your claim.

Download date: 15. Feb. 2021

https://doi.org/10.6100/IR554373

https://doi.org/10.6100/IR554373

https://research.tue.nl/en/publications/novel-concepts-for-fast-capillary-gas-chromatography(a1734650-71d5-471f-bc7e-f57ac8084385).html


PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de

Technische Universiteit Eindhoven, op gezag van

de Rector Magnificus, prof.dr. R.A. van Santen, voor

een commissie aangewezen door het College

voor Promoties in het openbaar te verdedigen op

woensdag 8 mei 2002 om 16.00 uur

door

MARIA MATHEA VAN DEURSEN

geboren te Weert

Dit proefschrift is goedgekeurd door de promotoren: prof.dr.ir. C.A.M.G. Cramers en prof.dr. P.J.F. Sandra Co-promotor: dr.ir. J.G.M. Janssen Paranimfen: Saskia van Deursen en Peter Lipman

Aan mijn ouders Aan Arjan

CIP-DATA LIBRARY TECHNISCHE UNIVERSITEIT EINDHOVEN Deursen, Marieke van Novel concepts for fast capillary gas chromatography/ by Marieke van Deursen. - Eindhoven : Technische Universiteit Eindhoven, 2002. Proefschrift. – ISBN 90-386-2873-0 NUGI 813 Trefwoorden: analytische chemie / gaschromatografie / "time-of-flight" massa spectrometrie; TOF-MS / voedingsmiddelenanalyse; triglyceriden Subject headings: analytical chemistry / gas chromotography / time-of-flight mass spectrometry; TOF-MS / food analysis; triglycerides Omslag en ontwerp: Jan-Willem Luiten Voorkant: Kristalglazuur pottenbakkerij dhr. H.M.B. van Deursen; foto’s meetopstelling Druk: Universiteitsdrukkerij, TUE

TABLE OF CONTENTS

1. GENERAL INTRODUCTION AND SCOPE 1References 5

2. THEORETICAL ASPECTS OF HIGH SPEED GASCHROMATOGRAPHY 7

2.1. Introduction 82.2. Theory 82.2.1. Packed versus capillary columns (influence stationary phase neglected) 112.3. Discussion 142.3.1. Low pressure drop (slow analysis, low plate number columns); P = 1 142.3.2. High pressure drop (fast analysis, higher plate number columns); P » 1 152.3.3. Speed of analysis by packed and capillary column systems 162.3.4. Alternative approaches to fast GC 182.3.4.1. Turbulent flow conditions 182.3.4.2. Vacuum outlet conditions 192.3.5. Temperature-programmed conditions (constant pressure mode) 202.3.6. Carrier gas velocities higher than the optimal velocity 212.3.7 Sample capacity 222.3.8 Band broadening 222.3.9 Multi-capillary columns 252.4. Conclusions 262.5. Trends and future perspectives 282.6. References 30

3. TEMPERATURE PROGRAMMING IN FAST CAPILLARY GASCHROMATOGRAPHY 33

3.1. Introduction 343.2. Instrumentation 373.2.1. Resistive heating analyses 373.2.2. Optimisation procedure TAG-analysis 373.2.3. Chemicals 383.3. Discussion and results 393.3.1. Retention time stability 393.3.2. Analysis of mineral oil 413.3.3. Analysis of glycols 413.3.4. Analysis of phenol and p-cresol in diesel 423.3.5. Analysis of diesel 443.3.6. Comparison of fast temperature programming to the use of short columns 463.4. Optimisation strategy of fast TAG analyses 483.4.1. Discussion and results 483.5. Conclusions 55

ii Table of contents

3.6. References 56

4. THEORETICAL DESIGN CONSIDERATIONS FOR MULTI-CAPILLARY COLUMNS IN FAST GASCHROMATOGRAPHY 59

4.1. Introduction 604.2. Theory 604.3. Experimental 624.4. Discussion and results 634.4.1. Influence of variations in diameter on the peak shape 634.4.2. Influence of variations in film thickness on the peak shape 664.4.3. Influence of variations in length on the peak shape 674.4.4. Gaussian distribution 684.4.5. Fast analyses using the multi-capillary column 684.5. Conclusions 704.6. References 70

5. FAST GAS CHROMATOGRAPHY USING VACUUMOUTLET CONDITIONS 73

5.1. Introduction 745.2. Theory 755.2.1. Sample loadability 765.3. Experimental 805.4. Discussion and results 825.4.1. Film-thickness 875.5. Conclusions 885.6. References 89

6. EVALUATION OF TIME-OF-FLIGHT MASS SPECTROMETRICDETECTION FOR FAST GAS CHROMATOGRAPHY 91

6.1. Introduction 926.2. Theory 946.3. Experimental 956.3.1. Reagents 966.4. Discussion and results 976.4.1. Cryogenic focusing inlet system 976.5. Conclusions 1026.6. References 102

7. COMPREHENSIVE TWO-DIMENSIONAL GASCHROMATOGRAPHY COUPLED TO TIME-OF-FLIGHT MASSSPECTROMETRIC DETECTION FOR CHARACTERISATIONOF OIL SAMPLES 105

7.1. Introduction 1067.1.1. Selectivity in GC×GC-TOFMS 1067.2. Experimental 1087.2.1. Instrumentation 1087.2.2. Chromatographic conditions 1097.2.3. Chemicals and samples 109

Table of contents iii

7.2.4. Software 1107.3. Discussion and results 1107.4. Quantification in oil analyses 1237.4.1. Quantification with FID 1237.4.2. Quantification with MS 1257.5. Discussion and results 1277.6. Conclusions 1367.7. References 136

8. THE USE OF A SPLIT/SPLITLESS INJECTOR AS A VAPORISATIONINTERFACE FOR COMPREHENSIVE LC××××GC SEPARATIONS OFTRIGLYCERIDES IN FOOD SAMPLES 139

8.1. Introduction 1408.2. Experimental 1428.2.1. Off-line large volume experiments 1428.2.2. On-line comprehensive LC×GC experiments 1438.3. Off-line large volume injections 1468.3.1. Discussion and results 1468.4. On-line comprehensive LC×GC analyses 1508.4.1. Discussion and results 1508.5. Conclusions 1578.6. References 158

SUMMARY 159

SAMENVATTING 163

DANKWOORD 167

CURRICULUM VITAE 169

BIBLIOGRAPHY 171

iv Table of contents

CHAPTER 1GENERAL INTRODUCTION AND SCOPE

Within the family of chromatography-based methods gas chromatography (GC) is oneof the most widely used techniques. It was first described by James and Martin in1952 [1] and has become one of the most important tools for the separation of volatilecompounds. Gas chromatography has gained widespread acceptance in numerousapplication areas, such as process control in chemical plants, quality control in thefood industry, monitoring sample composition in the oil-industry, environmental and(bio-) medical sciences. These are just a few examples in which gas chromatographyhas been applied. The combination of speed, sensitivity and a high resolving power ingas chromatography provides a very adequate technique for the separation of complexsamples. Moreover, the coupling to spectrometric methods such as mass spectrometry(MS) for direct identification of unknown compounds is easy to establish.

The most important breakthrough in gas chromatography was the introduction ofopen tubular columns by Golay in 1958 [2]. Compared to packed columns thepermeability of capillary columns provides a higher obtainable plate number. As anexample, Schutjes et al. [3] obtained a plate number of 106 for a 70 m × 50 µm i.d.capillary column. In 1962, Desty et al. [4] demonstrated the potential of capillarycolumns with a reduced inner diameter for high speed separations. The introduction offused silica columns by Dandeneau and Zerenner in 1979 [5] has greatly enhanced theacceptance of capillary columns in gas chromatography. In contrast to glass capillarycolumns fused silica columns are more flexible and easy to mount in a gaschromatograph. The development of deactivation methods provided a better inertnessof the column wall. Since then it is possible to analyse a wide range of compounds,from non-polar to polar substances. Fused silica columns are now routinely used inevery analytical laboratory.

With the introduction of capillary columns with a small inner diameter morestringent instrumental requirements have to be met. Especially sample introductionhas always been a critical factor. Split and splitless injections have been applied since

2 Chapter 1

the beginning of capillary gas chromatography and they are still the most widely usedsample introduction systems. A disadvantage of these systems is that discrimination ofsamples containing compounds with a wide boiling point range is hard to avoid.Quantification of this type of samples was improved by on column injectiontechniques developed by Grob [6] and Schomburg [7]. For this purpose alsoprogrammed temperature vaporisers (PTV) are used, where the solvent is evaporatedat a low temperature and the sample components are completely introduced onto thecolumn. These techniques allow the introduction of large sample volumes (up to 1mL) and are therefore suitable for trace analyses e.g. in environmental applications.

The most popular detector in gas chromatography is the flame ionisation detector(FID). Its high sensitivity, fast response, wide dynamic range and simplicity ofconstruction make it the most widely used detector. Another detector with a largepopularity in gas chromatography is the mass spectrometer. GC-MS is very adequatefor the identification of unknown compounds, but has a more limited dynamic rangethan the FID. Several other detectors that are available which offer high selectivity arethe electron capture detector (ECD) for halogen containing compounds, thethermionic detector (TID) for nitrogen and phosphorous containing compounds or thechemiluminescence detector (SCD) for sulphur containing compounds.

Due to the enormous growth of the number of samples required to be analysed,nowadays a lot of interest is shown in speeding up a gas chromatographic analysis.The development of methods for fast GC to obtain a high sample throughput isemerging. The theory of fast GC has already been developed in the 1960s [4,8,9]. Thepractical implementation with dedicated instrumentation however, took place only 10-15 years ago. The development of better GC ovens and resistive heating techniquesprovided high temperature programming possibilities in GC. Multi-capillary columns(900 narrow bore capillaries with an inner diameter of 40 µm in parallel) offer thehigh analysis speed of narrow bore columns combined with an increased samplecapacity. Injection techniques have been developed for the use of fast narrow borecolumns. Time-of-flight mass spectrometry with high acquisition rates has beendeveloped for coupling with fast GC.

The scope of this thesis is to give an overview of the possibilities and limitationsof fast GC methods. In the present work new techniques for fast GC have beendeveloped and dedicated instrumentation is successfully used to increase the speed ofanalysis for several industrial samples. The gain in speed that can be obtained withfast GC methods as compared to the use of conventional GC is discussed.

In chapter 2 the theoretical aspects of fast GC will be described. The speed ofanalysis using capillary columns compared to packed columns is discussed.

General introduction and scope 3

Additionally, the effect of using columns with a reduced inner diameter, vacuumoutlet pressures, fast temperature programming, multi-capillary columns and highcarrier gas velocities on the speed of analysis is described theoretically. Also, theimportance of narrow input band widths in fast GC and the consequences of fast GCfor the sample capacity will be discussed.

In chapter 3 the performance of fast temperature programming used to increasethe analysis speed for several industrial samples is discussed. For this purposededicated GC ovens and resistive heating columns capable of high temperatureprogramming rates, have been used. The temperature at which a solute is eluting froma temperature programmed GC column is amongst other things determined by thetemperature programming rate. A rule of thumb is that the optimum temperatureprogramming rate is 10°C/void time [10]. At higher rates, not only the resolution willdecrease, but also the elution temperature will increase. Therefore, increasing theprogramming rate is not an option for faster GC for high boiling compounds that willhave elution temperatures higher than the maximum allowable temperature of thestationary phase. A second drawback of rapid programming is the fact that theincreased elution temperatures may affect samples that contain thermally labileanalytes.

In chapter 4 the use of a multi-capillary column, which consists of a bundle of 900narrow bore capillaries, for fast separations is described. The advantages of the multi-capillary column in comparison to a single narrow bore column are the highervolumetric flow rate, the higher sample capacity and, consequently, the lowerminimum detectable concentration of the solute. To obtain the maximum performancefrom a multi-capillary column, the column has to meet very stringent requirements.For example, each of the capillaries should have exactly the same diameter. If the 900capillaries significantly differ in diameter, this causes serious band broadening. Otherimportant parameters that have to be equal are the length and the film thickness of theindividual capillaries. Some examples are shown of high speed separations with themulti-capillary column of samples with a wide boiling point range.

In chapter 5 the use of vacuum outlet GC for short wide bore columns coupled tomass spectrometric detection as a method towards fast GC will be described. Animportant trend in GC is the ever-increasing need for positive identification and theneed for more flexible systems that allow the analysis of a wide variety of samples onone system. These last two trends clearly result in a strong requirement for massspectrometric detection. Combination of fast GC with mass spectrometric detection isby no means trivial. Several types of mass spectrometers are available for coupling toGC systems. These systems differ in the way that the ion-fragments, formed by

4 Chapter 1

electron-bombardment of the molecules eluting from the GC column, are separatedaccording to their mass. Important mass analysers are the ion trap, the magnetic sectorinstrument, the quadrupole and the time-of-flight mass spectrometer. The performanceof each of the resulting mass spectrometers show differences in terms of acquisitionrates, detection limits, mass spectrometric resolution and quality of the mass spectraobtained [11]. The choice of the most suitable MS is very much dependent on thecomposition of the sample, the detection limits, and the speed of separation. In thiswork a time-of-flight mass spectrometer was used, which is capable of highacquisition rates. By operating a capillary column at low pressure conditions, thediffusion coefficient of the solutes in the mobile phase and thus the analysis speed isincreased. The gain in speed of columns operated at vacuum conditions as comparedto columns operated at atmospheric outlet conditions will be discussed. The influenceof the stationary phase thickness on the obtainable gain in speed in vacuum GC isdescribed in this chapter. Also the sample loadability, which for wide bore columns ismuch higher than for narrow bore columns, is discussed in detail.

Analyses performed with a cryogenic injection system, producing very narrowinjection band widths, are described in chapter 6. To minimise the contribution of theinput band width to the total band broadening, the injected sample plug has to benarrow in comparison to the total chromatographic band broadening. In case of fasteranalyses the residence time of the components in the column is reduced which resultsin extreme small peak widths. Injection hence becomes more critical. This isespecially true for isothermal analyses. In temperature programmed separations zonefocusing will occur at the column inlet due to the low initial oven temperature.Typical injection band widths (σi) required for very fast analyses on short narrow borecolumns are approximately 1-3 ms [12]. Especially for the narrow bore option for fastGC injection band width is critical. In this chapter the performance of a time-of-flightmass spectrometer for ultra fast separations (milliseconds range) on short narrow borecolumns will be evaluated.

One of the most interesting applications of very fast GC is in comprehensive two-dimensional GC (GC×GC). In chapter 7 a description is given of comprehensive two-dimensional GC for the analysis of complex oil samples. In GC×GC two independentGC separations are applied to an entire sample. The sample is first separated on anormal-bore high-resolution capillary GC column in the programmed temperaturemode. All of the effluent of this first column is then focused in very many, verynarrow fractions at regular, short intervals and subsequently re-injected onto a secondcapillary column, which is short and narrow to allow for very rapid separations in theseconds range. The second column is temperature programmed independent of the

General introduction and scope 5

first column. The resulting chromatogram has two time axes (retention on each of thetwo columns) and a signal intensity corresponding to the peak height. This techniqueprovides a very high peak capacity and is therefore very adequate for separatingcomplex samples. It is shown that by coupling a GC×GC-system to a time-of-flightmass spectrometer, the selectivity for group-type analyses in oil is increased.

Liquid chromatography (LC) coupled to GC also provides a very high separationpower for complex samples. In chapter 8 the use of comprehensive LC×GC for thecharacterisation of fats (triacylglycerides: TAG) in food samples is described.Fractions of the LC system were injected onto the GC column using a new interfacetype: the hot vaporisation chamber (hot split injector). During GC analysis the LCflow was stopped (stop-flow principle). The coupling of LC and GC is necessary to beable to obtain an adequate separation of the very complex TAG sample, which is ofimportance in the food industry. The first separation on a silver (Ag) phase LCcolumn is based on the number of double bonds present in the TAG-molecule. Thesubsequent GC separation is based on the boiling point of the solutes. Because thetotal analysis time largely depends on the separation speed of the second dimension,the speed of the GC-analysis was increased [13]. To enable a fast GC analysis a fastinjection is required. The speed of injection is a major advantage of using a splitinjection in the second dimension analysis. Several examples are shown of LC×GCanalyses of TAG in butter and margarine. A comparison is made with theconventional off-line LC×GC analyses.

REFERENCES

1. A.T. James and A.J.P. Martin, Biochem. J., 50 (1952) 679.2. M.J.E. Golay, in “Gas Chromatography”, V.J. Coates et al. (Eds.), Academic

Press, New York (1958) 1.3. C.P.M. Schutjes, E.A. Vermeer, J.A. Rijks and C.A. Cramers, J. Chromatogr., 253

(1982) 1.4. D.H. Desty, A. Goldup and W.T. Swanton, in “Gas Chromatography”, N. Brenner

et al. (Eds.), Academic Press, New York, 1962.5. R.D. Dandeneau and E.H. Zerenner, J. High Resolut. Chromatogr. Chromatogr.

Comm., 2 (1979) 351.6. K. Grob and K. Grob Jr., J. Chromatogr., 151 (1978) 311.7. G. Schomburg, H. Behlau, R. Dielmann, F. Weeke and H. Husmann, J.

Chromatogr., 142 (1977) 87.

6 Chapter 1

8. J.H. Knox and M. Saleem, J. Chromatogr. Sci., 7 (1969) 614.9. J.C. Giddings, Anal. Chem., 34 (1962) 314.10. L.M. Blumberg and M.S. Klee, J. Microcolumn Separations, 12(9) (2000) 508.11. P.G. van Ysacker, High-speed narrow-bore capillary gas chromatography, thesis,

Eindhoven University of Technology, The Netherlands, 1996.12. A. van Es, J. Janssen, C. Cramers and J. Rijks, J. High Resolut. Chromatogr.

Chromatogr. Comm., 11 (1988) 852.13. M.M. van Deursen, H.-G. Janssen, J. Beens, G.A.F.M. Rutten and C.A. Cramers,

J. Microcolumn Sep., 13(8) (2001) 337.

CHAPTER 2THEORETICAL ASPECTS OF

HIGH SPEED GAS CHROMATOGRAPHY1

SUMMARY

An overview is given of existing methods to minimise the analysis time in gaschromatography being the subject of many publications in the scientific literature.Packed and (multi-) capillary columns are compared with respect to their deploymentin fast gas chromatography (GC). It is assumed that the contribution of the stationaryphase to peak broadening can be neglected (low liquid phase loading and thin filmcolumns, respectively). The treatment is based on the minimisation of the analysis timerequired on both column types for the resolution of a critical pair of solutes(resolution normalised conditions). Theoretical relationships are given, describinganalysis time and the related pressure drop. The equations are expressed in reducedparameters, making a comparison of column types considerably simpler than with theconventional equations.

Reduction of the characteristic diameter, being the inside column diameter foropen tubular columns and the particle size for packed columns, is the best approachto increase the separation speed in gas chromatography. Extremely fast analysis isonly possible when the required number of plates to separate a critical pair of solutesis relatively low. Reducing the analysis time by reduction of the characteristicdiameter is accompanied by a proportionally higher required inlet pressure. Due tothe high resistance of flow of packed columns this seriously limits the use of packedcolumns for fast GC. For fast GC hydrogen has to be used as the carrier gas and insome situations vacuum-outlet operation of capillary columns allows a furtherminimisation of the analysis time. For fast GC the columns should be operated nearthe conditions for minimum plate height (i.e. at appropriate carrier flow velocity).Linear temperature programmed fast GC requires high column temperatureprogramming rates. Reduction of the characteristic diameter affects the samplecapacity of the 'fast columns'. This effect is very pronounced for narrow bore columnsand in principle non-existing in packed columns. Multi-capillary columns (a parallelconfiguration of some 900 narrow bore capillaries) take an intermediate position.

1 Partially published as “High-speed gas chromatography: an overview of various concepts” by C.A. Cramers,H.-G. Janssen, M.M. van Deursen and P.A. Leclercq in J. Chromatogr. A, 856 (1999) 315-329.

8 Chapter 2

2.1. INTRODUCTION

The theory of fast gas chromatography (GC) has already been developed in the 1960s[1-3]. The practical implementation in daily routine analyses however, took place onlyin the last 10 years. Currently GC methods are applied routinely using fasttemperature programming (resistive heating), short columns (multi-capillarycolumns), vacuum outlet GC or narrow bore columns. These methods can provide a 5-20 times shorter analysis-time compared to conventional analyses. In this chapter anoverview is given which will allow a thorough evaluation of the advantages anddisadvantages of several approaches to fast gas chromatography. It includes acomparison of capillary, multi-capillary and packed columns, based on reduced(dimensionless) parameters, which characterise the chromatographic and flowphenomena.

2.2. THEORY

In gas chromatography the primary goal is to achieve sufficient separation betweenthe compounds of interest. This means that the required number of theoretical plates(Nreq) of a chromatographic column should be sufficient to achieve completeseparation. It is obvious, that a separation requiring only a low number of plates (≈1000), is faster than the separation of a complex sample, requiring millions of plates.In order to enable comparison of the different methods for fast GC it is necessary touse the concept of resolution normalised conditions. This means that Nreq for a givenseparation is assumed to be constant. Thus, the optimisation criterion for fast gaschromatography which is used in this chapter is achieving the maximum speed ofanalysis while retaining a given Nreq.

To achieve a separation of two compounds the mixture is injected onto achromatographic column. Here each compound has its characteristic interaction withthe stationary phase of this column and therefore travels with its own specific speedthrough the column. Compounds travelling at different velocities elute at differenttimes and thus a separation is established. The retention time tR of a compound can bewritten as:

)1( ktt MR += (2.1)

Theoretical aspects of high speed gas chromatography 9

Where tM is the time required for the elution of an unretained component and k is theretention factor, which is partly related to the thermodynamic parameters governingthe distribution process of the compound into the mobile and stationary phase. This tM

can be written as follows:

uLtM = (2.2)

Where L is the column length and ū is the average carrier gas velocity. By definitionthe degree of separation (Rs) between a critical pair of components can be expressedas:

σ4)( 1,2, RR

s

ttR

−= (2.3)

Where tR,2 and tR,1 are the retention times of the second and first peak of the criticalpair in the chromatogram and σ is the average standard deviation (in units of time, s)the two peaks. The chromatographic separation process is counteracted by bandbroadening processes during migration of the solute through the column. Thechromatographic band broadening σ is related to the theoretical plate height H and tothe number of theoretical plates N, which can be described by:

2

=σ

RtN (2.4)

HLN = (2.5)

Substitution of equations 2.1 and 2.4 in 2.3 leads to the following well-knownexpression:

22

2

22

1116

−

+=αα

kkRN sreq (2.6)

where α is the relative retention α = k2 / k1. For peaks of equal size, baseline separationis achieved at Rs > 1.5, but separation is usually considered to be sufficient at Rs = 1.

10 Chapter 2

To minimise the required plate number excessive resolution should be avoided as wellas capacity factors close to zero and a relative retention close to1.

When the primary demand for a specific resolution between a critical pair ofcomponents is fulfilled, a second aim is to obtain the separation in a minimum of time.The following expression for tM can be used:

20 fuL

uLtM == (2.7)

where uo is the linear carrier gas velocity at the column outlet and f2 is the gascompressibility factor, defined in table 2.1. L can be replaced by N·H (equation 2.5)leading to:

[ ]

+=

2

)1(fu

HkNto

R (2.8)

Table 2.1: Gas compressibility correction factors

Pa = 1 P » 1

23

24

1 111

89

)(P)(P)(Pf

−−−= 1

89

)(P)(Pf

11

23

3

2

2 −−= 1

P23

)(Pf 121

3 += 1 P21

32 ff 143

a P = pi / po (ratio inlet to outlet pressure)

Combining equations 2.6 and 2.8 the analysis time tR can be expressed as:


−

+=2

2

2

32

1)1(16

fuH

kkRt

oSR α

α (2.9)

For baseline separation Rs = 1.5, equation 2.9 yields:

−

+=2

2

2

3

1136

fuH

kk)(t

oR α

α (2.10)

Equations 2.8, 2.9 and 2.10 contain two parts. The first term of the right hand side isequal to the required number of plates multiplied by 1+k. This part reveals theimportance of optimising k (between 1.7 and 3 [4,5]) and maximising the relativeretention α by proper selection of the stationary phase. The retention factor k can betuned by selecting appropriate liquid phases, varying phase ratios, and columntemperatures. This thermodynamic treatment will not be discussed in the presentchapter.

The second part H/ū = H/uof2 is the Purnell criterion [6] or the plate duration [7],i.e. the time spent by the carrier gas to traverse one theoretical plate. Minimising theanalysis time for a given required plate number is equivalent to minimising the peakduration. This gives the highest speed for Nreq (Nreq, not L is kept constant). Evaluationof the peak duration is difficult, because it is a complex function of retention factors,gas velocities, pressure drops, particle/column dimensions and solute diffusioncoefficients.

2.2.1. Packed versus capillary columns (influence stationary phase neglected)

Two types of basic equations are needed to the calculation of the numerical value ofH/ū to compare packed and capillary columns. (a) Plate height equations whichdescribe the dependence of the experimental plate height upon linear carrier gasvelocity and column parameters, e.g. the Horvath-Lin equation [8] for packed columnsand the Golay-Giddings equation [9,10] for capillary columns. (b) Flow equations,e.g. the Kozeny-Carman [11] and Hagen-Poiseuille [12] equations for packed andcapillary columns, respectively.

These theoretical relationships can be considerably simplified, if they areexpressed in terms of reduced parameters (dimensionless numbers) as proposed byGiddings [13]. In this way the dependence of H on ū can be approximated by a single

12 Chapter 2

curve in reduced parameters for packed columns and another one for capillarycolumns. Thus, irrespective of particle size or column inside diameter, nature of themobile phase or nature of the components, respectively. Identical curves can beexpected only [2], if the following conditions are fulfilled: (1) Contribution to plateheight of the stationary phase is negligible (small liquid phase loading in packedcolumns and thin film open tubular columns). A more concise treatment includingarbitrary film thickness in capillary columns can be found in ref. [8]. (2) Structure ofthe packing in a packed column is independent of particle size.

The reduced, dimensionless parameters are defined in table 2.2, where d is thecharacteristic diameter and equals the particle size dp in packed columns and theinside column diameter dc for open tubular columns; uo and Dm,o represent the lineargas velocity and the binary solute/carrier gas diffusion coefficient at column outletpressure po; pi is the column inlet pressure, and ϕ is the dimensionless flow resistanceparameter:

LΔp.

ηd.

v

21><

=ϕ (2.11)

with <v> the mean cross sectional fluid flow velocity [14].

Table 2.2: Definition of reduced parameters

Reduced plate heightdHh =

Reduced carrier gas velocityomD

duν,

0=

Column resistance factor ϕ (see text, equation 2.11)

Retention factorM

MR

tttk −=

Reduced pressureo

i

ppP=


The retention time equation 2.8 can now be written in terms of reduced parameters:

)k(Dfd.

νhNt

omreqR += 1

,2

2

(2.12)

The flow equations in reduced form read:

23

,

dfDhN

ppp omreqoiΔ

ηϕν=−= (2.13)

where η is the mobile gas phase dynamic viscosity.Combining equations 2.12 and 2.13 it follows:

32

22

ffhNtΔp req

Mηϕ= (2.14)

where h2ϕ = E is defined as the separation impedance, which represents the elutiontime per plate (plate duration) for an unretained solute times the pressure drop perplate, corrected for viscosity.

Table 2.3: Approximated values of the reduced parameters under optimal conditions

Packed

column

Packed

capillary

column

Capillary

column

Reduced plate height, h 2 2 0.8

Reduced velocity, ν 3 3 5

Column resistance factor, ϕ ≈ 1000 ≈ 500 32

Separation impedance, E ≈ 4000 ≈ 2000 20

14 Chapter 2

For ideal gases the relevant compressibility correction factors f2 and f3 are given intable 2.1. Using equations 2.12 and 2.13 the appropriate dimensionless numbers h, νand ϕ have to be included. In table 2.3 the approximate values under optimumseparation conditions are given. The consequences of working at higher than optimumvalues of the reduced velocity will be discussed later.

2.3. DISCUSSION

Some interesting conclusions can be drawn from equations 2.12–2.14. Theseequations show a complex pressure dependence. The optimum conditions can only becalculated numerically. Only under boundary conditions, such as a very low or veryhigh inlet to outlet pressure ratio and a negligible influence of the stationary phase,can explicit relationships be obtained.

2.3.1. Low pressure drop (slow analysis, low plate number columns); P = 1.

The compressibility correction factors f2 and f3 approach a value of 1. Including thedimensionless numbers relevant for packed and open tubular columns, it follows fromequation 2.12 that the analysis time can be reduced proportional to dp

2 for packedcolumns or to dc

2 for open tubular columns. As shown by equation 2.13 the price is aproportionally increasing pressure drop.

Comparing the separation impedance E for packed (≈ 4000) and capillary (20)columns, it can be concluded that if both column types are operated with the samepressure drop ∆p a capillary column will allow a 200× faster analysis. The influenceof the selection of the carrier gas on speed of analysis (equation 2.12) is given in table2.4 for P = 1, showing the advantage of using hydrogen. As demonstrated above,chromatographic performance in terms of speed of analysis for a fixed plate numbercan be substantially improved by miniaturisation: decreasing the characteristicdiameter and increasing the pressure drop proportionally. Therefore, in fast gaschromatography, especially for larger required plate numbers, Nreq, the assumptionP = 1, f2 = f3 = 1 becomes invalid.


Table 2.4: Analysis time for different carrier gases relative to hydrogen.

Carrier gas P = 1∝ 1/Dm,o (equation 2.12)

P » 1∝ (η/Dm,o)1/2 (equation 2.15)

H2 1 1

He 1.2 1.6

N2 4.0 2.8

CO2 5.6 2.8

2.3.2. High pressure drop (fast analysis, higher plate number columns); P » 1.

As can be seen from table 2.1, f2 can now be replaced by 3/(2P) and f2.f3 equals 3/4.Substitution of these values now leads to the following expressions:

k)(pD

ηdν

hNtoom

reqR +

= 1

21

,21

21

23

23 ϕ (2.15)

d)pD(η)ν(hN

Δp oomreq2

1

,2

12

1

34 ϕ

= (2.16)

ηENhNtΔp reqreqM222

34 == ηϕ (2.17)

= constant at a given temperature

Under high pressure drop conditions (P » 1) the analysis time tR now becomes linearlyproportional to the characteristic diameter (dc or dp), instead of d2 as in the lowpressure situation. Again separation speed tM (tR) is traded off against pressure drop∆p. A similar relationship for the analysis time (equation 2.15) was reported byTijssen [15] and Guiochon [16]. Knox and Saleem [2] published equations 2.15-2.17as early as 1969. The influence of the carrier gas type at P » 1 is reflected now in themobile phase diffusion coefficient and the carrier gas viscosity (equation 2.15). Using

16 Chapter 2

diffusion coefficients according to Fuller et al. [17] and dynamic viscosities [18] therelative analysis times are given in table 2.4.

In figure 2.1 an example is shown of a fast GC analysis of natural gas on a narrowbore capillary column.

min0.4 0.6 0.8 1 1.2 1.4

pA

11.5

12.0

12.5

Figure 2.1: Fast analysis of natural gas on a narrow bore column (8.6 m × 50 µm i.d.× 0.1 µm non-polar stationary phase). Vinj: 300 µL (gas), splitflow: 650 mL/min, inlet-pressure: 10 bar, ū = 37 cm/s, carrier-gas: helium, plate number (benzene, k = 2):120,000, detector: FID, Toven = 35°C (isothermally).

2.3.3. Speed of analysis by packed and capillary column systems

Some interesting conclusions can be drawn from equations 2.15 and 2.16 about thespeed of packed and capillary column systems. Substituting the optimum values of hand ν from table 2.3, it follows that the ratio dc/dp at which the analysis times areequal is 29. Inserting the proper values of h and ν and dc = 29 dp in equation 2.16shows that under equal analysis time conditions the pressure drop of a packed columnsystem is 200 times larger. The same conclusion also directly follows from the ratio ofthe separation impedances Epacked/Ecapillary (equation 2.17).

n-C5

n-C6 Benzene

C1-C4


Under optimum conditions the column length required to separate a critical pair isgiven by the product of Nreq and Hmin or:

L = Nreq h d (2.18)

If Nreq is assumed to have the same value for the packed and the open tubularsystem, it follows that under the equal analysis time conditions (dc/dp = 29), usingvalues for hmin and νopt from table 2.3:

611.LL

packed

capillary = (2.19)

Fast separations of simple mixtures with packed columns were already achievedin the 1960s. An example of extreme speed has been published by Jonker et al. [19].They showed a separation of 4 components in only 0.15 s with a 3.2 cm long columnpacked with 10 µm particles. The plate number was 650 with a pressure drop of 63bar. The column was operated under the conditions h = 4.9 at ν = 6.4 with helium asthe carrier gas.

Under optimum conditions (h = 2 and ν = 3) this column would have yielded 1600plates with an analysis time of 0.23 s (k = 2) and a pressure drop of 43 bar (accordingto equations 2.15 and 2.16). These equations also show that a capillary column,operated under vacuum outlet conditions (P » 1), can generate the same required platenumber of 1600 using a 290 µm inside diameter column with a length of 37 cm, and apressure drop of only 0.21 bar. For this open tubular column, operated at atmosphericoutlet pressure, however, the requirement P » 1 is not valid anymore (caused by therelatively low plate number in this case) and for the open tubular column equation2.12 with f2 = 1 becomes valid.

The column diameter dc giving 1600 theoretical plates, a retention time of 0.23 sfor a solute with retention factor 2 can be calculated, if Dm,o is included. It appears thatan open tubular column with an inside diameter of 70 µm and a length of 9 cm givesexactly the same analysis time and plate number (see table 2.5). The pressure drop(equation 2.13, f3 = 1) is now only ≈ 0.1 bar. This example clearly illustrates the greatsuperiority of capillary columns as compared to packed columns from the viewpointof speed of analysis related to pressure drop.

It should be emphasised again that in the above treatment the stationary phaseterm in the plate height equations is neglected. For capillary columns this means that

18 Chapter 2

thin film columns are considered. For fast analysis this is obvious, since for thick filmcolumns the (reduced) plate height is increased and the corresponding optimal(reduced) velocity is decreased considerably.

Table 2.5: Comparison of packed and capillary columns

Packed column Open tubular column

Outlet pressure Atmospheric Atmospheric Vacuum

L (cm) 3.2 9 37

d (µm) dp 10 dc 70 dc 290

∆p (bar) 43 0.1 0.1

T (K) 373 215* 200*

tR = 0.23 s for k = 2; N = 1600; carrier gas He.* In the data presented in this table k = 2 is kept constant. For thin film open tubular columns thistheoretical exercise leads to sub-ambient column temperatures. In the calculation of pressure drop byusing equation 2.17 the accompanying change in carrier gas viscosity with temperature is accountedfor.

2.3.4. Alternative approaches to fast GC

2.3.4.1. Turbulent flow conditions

Another way to obtain very short analysis times in capillary GC is to create turbulentflow. Only a few experimental results on turbulent flow in GC are known, some datingfrom about 30 years ago. The results were not as promising as expected, possibly due toinstrumental contributions or a not negligible influence of the stationary phase [20,21].With turbulent flow the velocity profile is largely flattened. Furthermore, the effectiveradial dispersion is considerably increased by convective contributions.

Experiments show that low plate heights can be obtained under very high speedconditions (pi = 50 bar, ū ≈15 m/s; Re ≈104) [22]. Unfortunately, the dependence ofthe plate height on the retention factor is significantly higher than under laminar flowconditions, limiting the use of turbulence to solutes with a low retention factor. Alsothe instrumental requirements with respect to time constants of sample introduction


and detection are very severe in turbulent gas chromatography. Taking into accountthe high pressure drop required for turbulent flow, a reduction of the column diameteris a better approach to increase the analysis speed in capillary GC.

2.3.4.2. Vacuum outlet conditions

As already demonstrated by Giddings in the sixties [3,23] true time optimisation willrequire vacuum column outlet conditions. This can be concluded directly fromequation 2.8. Under vacuum outlet conditions P = pi/po » 1, f2 can be replaced by3/(2P) yielding:

m,ireqR D

d.νh.N.t

2

32= (2.20)

where Dm,i is the binary solute carrier gas diffusion coefficient at inlet pressure.The largest value of Dm,i will be obtained, if a given column is operated at vacuum

outlet conditions generating Nreq theoretical plates, where the absolute value of thecolumn inlet pressure will be minimal. The gain in speed of analysis (G) by vacuumoutlet pressure conditions compared to atmospheric outlet pressure conditions of agiven column can be expressed by [24]:

232

3

1

1

)(p

pG

i,opt,atm

i,opt,atm

−

−= (2.21)

Where pi,opt,atm is the absolute inlet pressure (in bar) under optimal conditions atatmospheric outlet pressure (po = 1 bar).

The optimum absolute inlet pressure under vacuum outlet conditions pi,opt,vac

expressed in bar can be found from:

122 −= i,opt,atmi,opt,vac pp (2.22)

G is seen to increase with decreasing values of pi,opt,atm (table 2.6). The largestgains are obtained with sub-atmospheric inlet pressures. Thus, vacuum outlet will be

20 Chapter 2

of particular interest for high permeability (open tubular) columns, with a large innerdiameter and/or a short length.

For narrow bore capillary columns and packed columns requiring high inletpressures the gain in speed of analysis becomes negligible. The subject of vacuumoutlet gas chromatography has been extensively treated in reference [24]. As pointedout above under vacuum outlet conditions always the condition P = pi/po is met andtherefore equations 2.15-2.17 have to be used. According to these equations it appearsthat the same analysis time as observed by Jonker on a packed column withmicroparticles (Nreq = 1600, tR = 0.23 s, ∆p = 43 bar) can be produced by an opentubular column under vacuum outlet conditions, with the following dimensions: L =37 cm; i.d. 290 µm, inlet pressure 0.1 bar absolute (see table 2.5).

Table 2.6: Gain in speed of analysis by vacuum outlet operation as a function of theoptimum inlet pressure (pi,opt,atm) at atmospheric pressure outlet.

pi,opt,atm [bar, absolute] pi,opt,vac [bar, absolute] Gain G1.01 0.14 10.61.05 0.32 4.81.1 0.46 3.41.5 1.12 1.72.0 1.73 1.353.0 2.83 1.154.0 3.87 1.08

Using the computer program as described in references [25,26] for timeoptimisation of capillary columns, these conditions could be verified. This program isbased on the underlying theory; no simplifications are made, except from theassumption of ideal gas behaviour.

2.3.5. Temperature-programmed conditions (constant-pressure mode)

In the foregoing treatment, analysis time optimisation was discussed for isothermalanalysis. From the viewpoint of practical gas chromatography this assumption is toorestrictive. It has been estimated that 80% of all GC-analyses include temperature-


programming. In an extensive treatment Schutjes et al. [27] proved that thedependence of the analysis time on the column inside diameter for a capillary columnor particle size in a packed column is equal both for isothermal and lineartemperature-programmed analysis.

At P = 1 the analysis time, tR , in the programmed mode will be proportional to d2

(dc or dp), whereas at values of P » 1 a linear dependency on d will be found. For theprogramming rate r = (∂T/∂t) it follows:

isoMtr

,

1∝ (2.23)

Or for columns that are operated with the same phase ratios, stationary phase andcarrier gas: r ∝ 1/d2 for P = 1 and r ∝ 1/d for P » 1

Consequently, in linear temperature-programmed analysis the programming ratehas to be significantly increased when using narrow bore (or small particle) columnsas compared to standard ones, requiring instrumentation offering these highprogramming rates (e.g. by resistive heating, described in chapter 3 of this thesis). Thevalidity of this theory was confirmed by several experiments [27]. It should beemphasised here again, that extreme high speed of analysis both in isothermal andprogrammed-temperature analysis can only be obtained if Nreq is relatively small(equations 2.12 and 2.15).

2.3.6. Carrier gas velocities higher than the optimal velocity

The theoretical relationships presented before proceed from the assumption of gaschromatography performed at the optimum carrier gas velocity. At higher velocitiesthe analysis time is reduced, but simultaneously the column plate number willdecrease due to a larger plate height. In order to restore the ensuing loss in peakresolution a longer column has to be selected, thus re-establishing the original platenumber. However, this opposes the decreased analysis time. It can be concluded thatthe speed of analysis will be improved as long as the increased carrier gas velocityoverrules the required column length increment. This subject was already recognisedby Scott and Hazeldean in 1960, introducing the concept of "optimum practical gasvelocity" [28]. It can be deduced (equations not described in this chapter) that whenthe carrier gas velocity of a fast GC separation is 40% above the optimal value only an8% faster analysis can be obtained [29]. Blumberg in a series of publications [7] on

22 Chapter 2

the theory of fast capillary gas chromatography came to the same conclusions. Also inchapter 3 of this thesis it is shown that increasing the carrier-gas velocity mostly is notthe best option to increase the speed of analysis.

2.3.7. Sample capacity

Another important factor is the effect of characteristic diameter reduction on thesample capacity of fast columns. In principle this is not a problem for packedcolumns; the speed of analysis is related to dp, but the sample capacity can be varied atwill by increasing the column diameter. A large advantage in this respect for packedcolumns. In open tubular columns the speed as well as the sample capacity are relatedto the inside column diameter dc.

In an extensive theoretical and practical study on the sample capacity of opentubular columns Ghijsen et al. [30] came to the following conclusion: the maximumsample capacity, Cmax for columns with an equal phase ratio leading to maximally10% peak broadening is given by:

3max cd.β C ∝ (2.24)

Where β is a proportionality factor: 0.05 < β < 1.8. β ≈ 1.8 for solutes and stationaryphases with similar chemical functional groups and β ≈ 0.05 for solutes and stationaryphases with very different structure. The sample capacity is thus drastically reduced[∝ dc

3] for narrow bore columns. This is a serious limitation for narrow bore opentubular GC.

2.3.8. Band broadening

Band broadening in fast GC separations, especially when using narrow bore columns(i.d. < 100 µm), is very low. Instrumental requirements therefore are very stringent infast GC. Sample introduction often is the most critical step in fast separations.Moreover, the high inlet pressures required for optimal operation of narrow borecolumns is an additional problem.

When a negligible contribution of the detector volume, interfaces and electronicscan be assumed, the total band broadening σt can be expressed as a function of theinput band width (σi) and the chromatographic band width (σc). Because thesecontributions are independent, the variances can be summed:


222cit σσσ += (2.25)

From the Golay-Giddings equation (equation 4.3 in chapter 4 and from equations 2.4and 2.5) σc can be calculated and σt can be determined experimentally from thechromatogram. For example, for a column with an inner diameter of 50 µm and aplate number of 105 an input band width of approximately 10 ms is required in orderto keep the loss in column efficiency lower than 1%.

Several introduction systems have been developed in the past, which arecompatible with narrow bore columns. The most widely used injector is thesplit/splitless injector. Input band widths are determined by the injection speed of thesample and band broadening in the injector [31]. The latter one can be reducedsignificantly by increasing the split flow. Applying a high split flow reduces theresidence time of the compounds in the injector and therefore minimises bandbroadening. An example of an analysis on a narrow bore column of a mixture ofalkanes using a high split flow is depicted in figure 2.2. The injection was performedusing a gas sample valve and therefore a plug injection takes place. The input band-width can then be calculated using the following relation:

12/22 wi =σ (2.26)

where w is the width of the injected plug. The influence of the split flow on the inputband width is shown in figure 2.3. From this figure it can be seen that the measuredinput band width is somewhat higher than is expected from plug injection.

Another possible injection system is the cryogenic inlet system. This injector typecan be used for separations in the (milli-)seconds range, and is described in chapter 6of this thesis.

24 Chapter 2

min0.2 0.4 0.6 0.8 1 1.2

pA

11.4

11.8

12.2

12.6

C1

C2

C3

C4 C5

C6

Figure 2.2: Analysis of a C1-C6 mixture (100 ppm) on a narrow bore column (8.6 m ×50 µm i.d. × 0.1 µm non polar phase) using a gas sampling valve. Vloop = 235 µL, splitflow = 750 mL/min, Toven = 35 °C (isothermally), inlet-pressure = 10 bar, N =120,000, detection: FID, GC: HP 6890 (Agilent, Wilmington, DE, USA). Somediscrimination was observed for the very volatile compounds C1-C3.

0

40

80

120

160

200

0 200 400 600 800 1000Splitflow [ml/min]

Sigm

a [m

s]

Figure 2.3: Influence of the split flow on the input band width (σi) of methane on ashort narrow bore column (1 m × 50 µm i.d. × 0.1 µm). T = 35°C, Vloop = 125 µL.××××: σi calculated with equation 2.24. A chromatographic band width σc of 5 ms wasdetermined and the σt of CH4 was measured from the chromatogram.∆: σi calculated from equation 2.25 assuming plug injection.

× : σi measured∆ : σi calculated from plug injection


2.3.9. Multi-capillary columns

A very interesting development in column technology is the introduction of the multi-capillary column ([32], chapter 4 of this thesis). A parallel configuration consisting of919 coated capillaries with an i.d. of 40 µm and a length of 1 m. Speed and pressuredrop are dictated by the 40 µm inside diameter. Flow rate and sample capacity arehowever theoretically 919 times larger because of the parallel operation. Theseparation of a BTEX-mixture within 0.8 minutes is shown in figure 2.4.

From our work and work done by others [33-35] it follows that due to very slightinequalities in the parallel columns hmin ≈ 1.8-2.0 instead of hmin ≈ 0.8 for an ideal thinfilm open tubular column. It should be emphasised here, however, that also for asingle narrow bore column with an i.d. of 50 µm or less, it is very difficult to obtain aminimum value of h of 0.8. Peak broadening in multi-capillary columns is furtherdescribed in chapter 4 of this thesis.

A real impact of the concept of multi-capillary columns can only be anticipated, ifmore flexibility can be offered. This applies to available column lengths, insidediameters and number of columns in the parallel configuration.

min0 0.2 0.4 0.6 0.8

pA

0

1000

2000

3000Benzene

Toluene

Ethylbenzene

m- + p-Xylene

o-Xylene

Figure 2.4: Separation of a BTEX mixture on a multi-capillary column; L = 1 m, 919parallel capillaries of 43 µm i.d.; df = 0.2 µm; SE-30. Inlet pressure: 375 kPa(helium); column flow = 200 ml/min; temperature program: 40°C (0.5 min) to 200°Cat 25°C/min. Detector: FID. Injection: headspace 30 µL; splitflow = 800 mL/min.

26 Chapter 2

2.4. CONCLUSIONS

The basic principle of decreasing analysis times in gas chromatography isminiaturisation: a decrease of the particle size in packed columns or the columndiameter in capillary columns. This implies that for a given required plate number,Nreq, the column diameter can be decreased proportional to the decrease incharacteristic diameter. Reduction of the characteristic diameter heavily increases thedemand on the pressure capability of the instrumentation. For packed columns thisimplies that higher plate numbers at high speed are out of reach.

It should be emphasised that it is assumed that the contribution of the stationaryphase term to the plate height can be neglected (thin films). This assumption iscorrect, when discussing fast gas chromatography.

The effect of characteristic diameter reduction on sample capacity is verypronounced for capillary columns and in principle non-existing in packed columns.Multi-capillary columns take in this respect an intermediate position.

Table 2.8 summarises the dependence of analysis time, pressure drop, samplecapacity and volumetric flow rate at column outlet conditions and linear temperature-programming rate on the characteristic diameter.

Table 2.8: Dependence on characteristic diameter in high speed GC (P » 1)

Capillary

column

Multi-capillary

column

Packed column

Analysis time dc dc dp

Pressure dropcd

1

cd1

pd1

Sample capacity * dc3 n.dc

3 column diameter

Volumetric flow rate * dc n.dc column diameter

Temperature programming

rate (r) cd1

cd1

pd1

* n = the number of parallel capillaries


High speed GC involves high pressure drops if Nreq is not too small, therefore, forthis table the condition P » 1 is used (equations 2.15 and 2.16). Turning back toequation 2.15 it appears that the analysis time is very dependent on the required platenumber (Nreq

3/2). Extremely fast analysis can only be obtained, if Nreq is small.Minimisation of Nreq is therefore always a primary task, this is done by properstationary phase and column temperature selection (kopt ≈ 2); this optimisation is notthe subject of this chapter.

For fast analysis hydrogen is the carrier gas of choice, with helium as second best(about 60% slower). Vacuum outlet operation also enhances the diffusivity in thecarrier gas, since the column inlet pressure is lowest for po → 0. This implies a largeraverage diffusion coefficient in the mobile phase and thus a shorter analysis time. Thiseffect is most pronounced for wide bore and/or short columns. For longer narrow borecolumns the inlet (and average pressure) is hardly affected by a change in outletpressure from atmospheric to vacuum and no gain in analysis time is obtained.

For a well optimised column (L = Lreq), just able to separate the critical pair,increasing of the carrier gas velocity to higher values than uopt is counterproductive:the resolution will decrease. Increasing the column length to counteract the loss inresolution will yield a marginal reduction of 8% faster analysis time. The generalconclusion is that high speed open tubular columns should be operated very close tothe optimum conditions.

If a given column offers a resolution for the critical pair that is greater than needed(e.g. Rs = 3), of course increase of the flow rate above the optimum can drasticallyreduce the analysis time. Situations as this are often described in literature. However,a reduction of the length of that column to a value where the resolution is decreased to1.5 would yield under optimum conditions the same analysis time as the (too long)column described above.

If linear temperature programming is used the programming rate of the columnhas to be increased proportionally to the reduction in the characteristic diameter,requiring ovens that allow for very high programming rates.

On the instrumental side also the reduction of analysis times by miniaturisationmust be accompanied by a proportional decrease of the time constants of theinstrument (i.e. of detection and sample introduction systems). For very fast analysisthis requires adapted instrumentation.

28 Chapter 2

2.5. TRENDS AND FUTURE PERSPECTIVES

As increasing the speed of a gas chromatographic separation is economicallyadvantageous, it will also be introduced in routine analysis. In most applications opentubular columns of 100 to 150 µm inside diameter will be used. These columns offer acompromise with respect to analysis time and instrument compatibility. Commercialinstrumentation that can operate with such columns is now available from differentmanufacturers. A significant amount of work has been done on methods for fasterchromatography during the last decades. Even more work, however, remains to bedone.

Up till now in the analytical chain: sampling - sample pre-treatment - sampleintroduction - separation - detection most attention has been paid to understanding andoptimising the kinetic aspects of separation as summarised in this overview. Theeffects of miniaturisation are well understood for both packed and open tubularsystems. As has been shown open tubular columns are superior to packed systemswith respect to analysis times. The only exception being the application to simplemixtures, where both systems can be used.

In recent history the emphasis in open tubular column technology has been oncolumns with a high resolving power. Analytical problems were tackled by a mereovershoot of theoretical plates. The need for very specific stationary phases wasdramatically reduced, compared to the packed column period during the start of gaschromatography.

For fast GC there will be a revival of interest in tailoring stationary phaseselectivity for target separations (see discussion below equation 2.10). The logicalchoice of packed systems thereby is of limited interest due to the required high inletpressures and activity of the support materials. Improved column technology (e.g.sol/gel technology) will allow the production of a wider choice of selective opentubular columns. Fine tuning of selectivities can be obtained by (electronically)adjusting the mid point pressure between two serially connected columns with widelydifferent selectivities [36]. In this way flexibility can be built in the one-dimensionalcolumn system: and selectivity tuning can be performed automatically instead oftedious column replacement.

High speed gas chromatography using open tubular columns in the futurecertainly will be important in field-portable GC instruments [37] for on-site,environmental and industrial hygiene applications. Micro-machining techniques allowthe production of sampling valves and TCD devices compatible with the special needsof microbore columns. The even more stringent environmental regulations will


necessitate the development of sample preparation and more sensitive detectiontechniques for field portable instruments.

Considering the growing importance of quality issues in analysis, the use of GC-MS for positive identification will grow. In addition to improving the integrity of theanalytical results, this adds additional selectivity to the system: GC-MS can rapidlyand automatically detect and resolve overlapping peaks, thereby further reducinganalysis times compared with analytical systems that require or demand adequatecomponent separation.

Furthermore, it can be anticipated that the use of (ultra) high speed GC isaccompanied by a decrease in reproducibility of retention data, certainly if high speedtemperature or pressure programming is used. This makes the combination with massspectrometry even more important. Due to the small peak widths in high speed GC,conventional scanning mass spectrometers are not compatible. Time-of-flight MS orspatial array detection are the methods of choice [38]. An example of a high speedGC-TOF MS analysis is shown in figure 2.5.

0 1 2 3 4 5 6 7Time [s]

Tota

l ion

cur

rent

1 6

5

43

2

Figure 2.5: Reconstructed chromatogram of a hydrocarbon mixture of 6 alkanes(recorded at 400 spectra/second). Column: L = 2 m; 50 µm i.d.; df = 0.1 µm OV-1; ū= 180 cm/s. Inlet pressure = 500 kPa (helium); splitflow = 400 ml/min. Isothermallyat 80°C. Detector: time-of-flight mass-spectrometer (LECO, MI, USA); scanned massrange: 30-400 m/z (mass to charge ratio); Tion source: 200ºC. Compounds: 1. 2,3-dimethylbutane, 2. hexane, 3. heptane, 4. methylcyclohexane, 5. 2,3,4-trimethylpentane, 6. octane.

30 Chapter 2

2.6. REFERENCES

1. D.H. Desty, A. Goldup and W.T. Swanton, in N. Brenner et al. (Eds.), GasChromatography, Academic Press, New York, (1962) 105.

2. J.H. Knox and M. Saleem, J. Chromatogr. Sci., 7 (1969) 614.3. J.C. Giddings, Anal. Chem., 34 (1962) 314.4. G. Guiochon, Anal. Chem., 38 (1966) 1020.5. E. Grushka and G. Guiochon, J. Chromatogr. Sci., 10 (1972) 649.6. J.H. Purnell, "Gas Chromatography", J. Wiley & Sons Inc., New York, (1962).7. L.M. Blumberg, J. High Resolut. Chromatogr., 20 (1997) 679.8. C. Horvath and H.J. Lin, J. Chromatogr., 126 (1976) 401.9. M.J.E. Golay, in "Gas Chromatography" D.H. Desty (Ed.), Butterworths, London

(1958) 36.10. J.C. Giddings, S.L. Seager, L.R. Stucky, G.H. Stewart, Anal. Chem., 32 (1960)

867.11. P.C. Carman, "Flow of gases through porous media", Butterworths, London,

(1956).12. A.E. Scheidegger, "The physics of flow through porous media", University of

Toronto Press, Toronto (1960).13. J.C. Giddings, J. Chromatogr., 13 (1964) 301.14. J.C. Giddings, Dynamics of Chromatography, Part 1. Principles and Theory,

Marcel Dekker, New York (1965).15. R. Tijssen, Anal. Chem., 59 (1987) 1007.16. G. Guiochon, Anal. Chem., 50 (1978) 1812.17. E.N. Fuller, P.D. Schittler and J.C. Giddings, Ind. Eng. Chem., 58 (1966) 19.18. L.S. Ettre, Chromatographia, 18 (1984) 243.19. R.J. Jonker, H. Poppe and J.F.K. Huber, Anal. Chem., 54 (1982) 2447.20. J.C. Giddings, W.A. Manwaring and M.N. Myers, Science, 154 (1966) 146.21. F. Doue and G. Guiochon, Sep. Sci., 5 (1970) 197.22. A.J.J. van Es, J.A. Rijks and C.A. Cramers, J. Chromatogr., 477 (1989) 39.23. J.C. Giddings, Anal. Chem., 36 (1964) 741.24. C.A. Cramers and P.A. Leclercq, CRC Critical Reviews in Anal. Chem., 20 (1988)

117.25. P.A. Leclercq and C.A. Cramers, J. High Resol. Chromatogr. & Critical Comm. 8

(1985) 764.26. P.A. Leclercq, J. High Resolut. Chromatogr., 15 (1992) 531.


27. C.P.M. Schutjes, E.A. Vermeer, J.A. Rijks and C.A. Cramers, J. Chromatogr., 253(1982) 1.

28. R.P.W. Scott and G.S.F. Hazeldean, in R.P.W. Scott (Ed.) "Gas Chromatography1960", Butterworths, London, U.K. (1960) 144.

29. C.A. Cramers and P.A. Leclercq, J. Chromatogr., 842 (1999) 3.30. R.T. Ghijsen, H. Poppe, J.C. Kraak and P.P.E. Duysters, Chromatographia, 27

(1989) 60.31. M. van Lieshout, M. van Deursen, R. Derks, H.-G. Janssen and C.A. Cramers, J.

High Resol. Chromatogr., 22(2) (1999) 116.32. W.S. Cooke, Today's Chemist at Work, ACS, January (1996) 16.33. P. Sandra, B. Denoulet and F. David, in Proc. 18th Int. Symp. on Cap.

Chromatography, Vol. I, Riva del Garda, Italy, Hüthig, Heidelberg, (1996) 479.34. I.R. Pereiro, V.O. Schmitt, R. Lobinski, Anal. Chem., 69 (1997) 4799.35. M. van Lieshout, M. van Deursen, R. Derks, J.G.M. Janssen and C.A. Cramers, J.

Microcol. Sep., 11 (1999) 155.36. H.L. Smith and R.D. Sacks, Anal. Chem., 69 (1997) 5159.37. H. Pham Tuan, H-G. Janssen, C.A. Cramers, Ph. Mussche, J. Lips, A. Handley,

and N. Wilson, J. Chromatogr. A., 791 (1997) 197.38. P.A. Leclercq and C.A. Cramers, Mass Spectrom. Rev., 17 (1998) 37.

32 Chapter 2

CHAPTER 3TEMPERATURE PROGRAMMING

IN FAST CAPILLARY GAS CHROMATOGRAPHY1

SUMMARY

In the first part of this chapter it is described that the application of fast temperatureprogramming by using resistive heating techniques is a very adequate way to speed upa gas chromatographic analysis. With this heating technique programming rates up to20°C per second can be reached. A relative standard deviation of retention timesbetter than 0.2% is obtained. Using fast temperature programming the analysis timesof a mineral oil sample, an industrial oligomer sample and toxic compounds in dieselhave been reduced 5 to 20 times, compared to a standard temperature programmedanalysis. However, to reduce the analysis time of a complex sample, resistive heatingin most cases cannot be applied. The use of fast temperature programming isfavourable to the use of short columns and columns operated at above-optimumcarrier gas velocities.

In the second part of this chapter numerical calculation procedures are used topredict the optimum experimental conditions for rapid-programming fast capillaryGC. Retention times and peak widths are calculated from experimentally determinedthermodynamic compound properties. Errors in the predicted values are below 2 and15% for the retention times and peak widths, respectively. It is shown that rapidtemperature programming is a powerful tool for speeding up the separation ofsamples containing homologous series of compounds such as e.g. triglycerides. Theoptimal programming rate and the minimum required column length to obtain theshortest possible analysis time with sufficient resolution (Rs = 1.5) are predicted.Using a wide-bore column with a length of 10 m, a programming rate of 4°C/s and acarrier gas velocity of approximately 250 cm/s, the shortest possible analysis time fora triglycerides sample can be obtained while still maintaining sufficient resolution.With these instrumental settings the analysis time of a triglycerides separation can bereduced from 15 to approximately 5 minutes.

1 The first part of this chapter has been published as “Possibilities and Limitations of Fast TemperatureProgramming as a Route towards Fast GC”, by: M.M. van Deursen, J. Beens, H.-G. Janssen and C.A. Cramers,J. High Resol. Chromatogr., 22(9) (1999) 509-513.The second part of this chapter has been published as “Design considerations for rapid-heating columns appliedin fast capillary gas chromatography”, by: M.M. van Deursen, H.-G. Janssen, J. Beens, G. Rutten and C.A.Cramers, J. Microcol. Sep., 2001.

34 Chapter 3

3.1. INTRODUCTION

Significant savings in time and money can be obtained through the reduction of theoverall analysis time in gas chromatography. Especially for laboratories where manyroutine samples are analysed on a daily base, it is very attractive to speed up theseanalyses. Also in situations where a short “time-to-result” is needed, it can be highlybeneficial to use fast gas chromatographic separation systems. In chapter 2 variousroutes towards a faster separation method have already been described theoretically.These methods (most of them successfully applied in analytical laboratories) comprisethe use of: narrow bore columns, fast temperature programming (described in thischapter), short columns, vacuum outlet operation, above optimal carrier gas velocities,the use of hydrogen as a carrier gas, reduced film thicknesses, selective stationaryphases and turbulent flow conditions. Fast GC using turbulent flow conditions however,has been generally applied in a research environment but is not yet used for routineanalyses.

From the extensive list of possible options to speed up GC separations, theconclusion could be drawn that the various options are equally suited for everyanalytical problem. In fact the opposite is true. The question of which approach shouldbe selected for reducing the analysis time for a given separation is certainly not trivial.Turbulent flow for example, can only be applied in case of extremely simple samplescontaining compounds with low retention factors. Another example is that of amixture containing a limited number of compounds widely differing in boiling points.Such a sample can be successfully separated on a short column with rapid temperatureprogramming. On the other hand, in order to separate a complex mixture containing alarge number of components, the use of: I) a long narrow bore column, II) slowtemperature programming, III) applying vacuum-outlet operation, or IV) by usinghydrogen as the carrier gas, is necessary. When considering methods for the reductionof the overall GC analysis time one should always bear in mind that the best way tospeed up a separation is to optimise the resolution. Separating compounds with ahigher resolution than strictly needed is a waste of time.

In this chapter, investigations on the possibilities and limitations of obtainingfaster GC separations through the use of higher heating rates, higher carrier gas flowsand shorter columns are described. Since the first temperature programmable GCbecame commercially available in the late 1950s, the heating capacity of the ovens hasbeen significantly improved. In the current generation of GC instruments themaximum temperature programming rate is limited to approximately 1 to 2ºC/s. Withthe standard oven and heater designs cooling times are long and higher rates aredifficult to obtain due to temperature stability problems. To enable the use of higher

Temperature programming in fast capillary gas chromatography 35

heating rates, while still maintaining sufficient temperature stability and acceptablecooling times, a number of systems based on resistive heating were recentlydeveloped. To adopt such a technique, a capillary column can be coated with aconductive material, a coil can be wrapped around or run in parallel to the column, orthe capillary column can be inserted into a metal tubing.

Jain and Phillips [1] successfully applied a conductive coating onto the capillarycolumn, and obtained a fast heating program with heating rates up to 20ºC/s. By usingaluminum-clad columns, Hail and Yost [2] achieved the separation of C7-C10 in 2seconds. Ehrmann and coworkers [3] tested two concepts of resistive heating. Onewas based on the use of a metal tube as a heater (coaxial at-column heater) and thesecond one used a metal wire that runs in parallel to the column as the heater (colinearat-column heater). It was found that a colinear at-column heater resulted in a morereliable system than the coaxial heater and also enabled rates of 20ºC/s to be achieved.The latter suffered from differences in expansion coefficients of the metal and fusedsilica, causing damage of the fused-silica column.

The use of higher heating rates results in a loss of resolution, which implies thatthis method is not suitable for the analysis of very complex samples that need highseparation efficiencies. This is shown by the analysis of a diesel mixture. In thischapter several examples of analyses are mentioned that can be performed faster byusing fast temperature programming. The first shows the separation of an industrialoligomer sample containing four components having widely differing boiling points.In the second, the analysis of the mineral oil content of environmental samples isperformed, and it is investigated whether the time of an environmental mineral oilanalysis can be reduced by using fast temperature programming while still meeting therequirements described in the Dutch regulatory methods [4]. In the third, the analysisof the toxic compounds phenol and p-cresol (para-cresol: 4-methylphenol) in diesel isinvestigated.

A further method for the reduction of the analysis times studied here is the use ofshort columns at high carrier gas velocities. Analysing samples using short columns islimited to simple samples because of the low plate number of such columns. Theanalysis time obtained on short columns in comparison to the use of rapid temperatureprogramming is described in the first part of this chapter.

As mentioned previously the use of high temperature programming rates is themethod of choice for samples containing only a limited number of peaks covering awide range of elution temperatures, such as e.g. triglycerides (tri-acylglycerides:TAG1) samples or hydrocarbon mixtures [5]. These samples are generally over-

1 The molecular structure of a TAG molecule is shown in chapter 8 of this thesis.

36 Chapter 3

resolved using conventional GC conditions and some resolution can be traded inagainst time. TAG samples, or other homologous series of e.g. n-alkanes, are more orless over-resolved in a normal analysis using the maximum possible heating rate ofcurrent standard GC instruments. For these types of samples rapid temperatureprogramming can increase the speed of analysis substantially, while still providingsufficient resolution.

Resolution in a (rapid) temperature programmed GC separation is a verycomplex function of i) column parameters such as length, diameter and film thickness,ii) operational settings such as linear carrier gas velocity and temperatureprogramming rate, and finally iii) solute properties such as the temperaturedependency of the vapour pressure. Because of this complexity, the optimisation of atemperature programmed gas chromatographic analysis is a tedious and time-consuming task. A large number of experiments are generally required to locate theoptimum that provides just sufficient resolution in the shortest possible analysis time.To aid in this process, various calculation methods have been described in literaturethat predict retention times in linear temperature-programmed GC [6,7]. Some ofthese methods allow the additional prediction of peak widths, which means that theresolution of a given separation can be calculated [7]. A parameter that requiresspecial attention in this process is the elution temperature of the last eluting analyte.Rapid-programming of the column will result in an increasing elution temperature ofthe heavy analytes, especially if the increase of the programming rate is notaccompanied by either an increase of the linear velocity or a decrease of the columnlength. Maximum speed, optimum resolution and acceptable elution temperatures areonly obtained at mutually compatible operational settings (velocity and programmingrates) and column parameters (mainly length).

In the second part of this chapter the optimisation strategy for optimisingtemperature programmed GC analyses developed by Snijders et al. [8] is used topredict the influence of column parameters and operational settings on the resolutionand analysis time in fast temperature programmed GC. In this way the optimumsystem and instrument settings for rapid-programming fast GC can be selected. Thetemperature programmed GC-separation of a homologous series of TAG (CN32-CN501) serves as an illustration. The GC optimisation method described by Snijders isused to calculate both retention times and peak widths in a temperature programmedanalysis of triglycerides. The optimal programming rate and the minimum requiredcolumn length to obtain the shortest possible analysis time with sufficient resolution(Rs = 1.5) are predicted. The influence of carrier gas velocity is discussed. Special 1 The carbon number (CN) is the total number of carbon atoms in the three fatty acid chains of a TAG molecule(described in chapter 8 of this thesis).


attention is devoted to the elution temperatures of the analytes, a parameter that isrelevant since the poly-unsaturated triglycerides have a limited thermal stability.Experimental data are used to validate the theoretically predicted data. It will beshown that a substantial gain in speed can be obtained in the GC-analysis of a TAGsample by using fast temperature programming.

3.2. INSTRUMENTATION

3.2.1. Resistive heating analyses

For the fast temperature programming experiments the TDX EZ-Flash fromThermedics Detection Inc. (Chelmsford, MA, USA) was used. This resistive heatingsystem consists of a stainless steel tube that contains the fused silica column. Thefused-silica column is directly heated by the metal tube. Furthermore, the EZ-Flashsystem consists of a computer-controlled power module (software version 1.1.0) and aGC-specific interface. The column was installed in an HP 6890 gas chromatograph(Agilent Technologies, Wilmington, DE, USA) equipped with an OPTIC 1 injector(ATAS Veldhoven, The Netherlands) and an FID detector. The data system used wasthe HP Chemstation. The GC column had a length of 5 m, an inner diameter of 320µm and was coated with an Rtx-5 film of 0.25 µm thickness. The column was linkedto the injector and detector with two pieces of deactivated fused silica. These 250 µmfused-silica capillaries had a length of about 8 cm at the injector side, and 10 cm at thedetector side. In the oven, on the bottom of the injector and detector, interface blockswere installed which can be heated separately. An injector liner with a small innerdiameter of 1.2 mm, was used to minimise band broadening in the injector [9].

For the experiments on short columns a GC 8000 series type 8380 of FisonsInstruments (Milan, Italy) was used. This GC was equipped with a split/splitlessinjector and an FID detector. The injector liner that was used had an inner diameter of0.7 mm. The columns (5, 4, and 3 m) had an inner diameter of 320 µm and a filmthickness (OV-1) of 0.25 µm (Varian/Chrompack, Middelburg, The Netherlands).

3.2.2. Optimisation procedure TAG-analysis

Three GC instruments were used. The first one, used for on-column injections, was aShimadzu GC type GC-17A (Tokyo, Japan) equipped with a flame ionisation detector

38 Chapter 3

and an on-column injector. The maximum programming rate of the oven was ≈20°C/min, in the temperature range between 40 and 425°C. The analytical columnwas a 5 m, 530 µm i.d. high temperature Ultimetal SIMDIS column, coated with 0.17µm of a non-polar phase (Varian/Chrompack). Helium was used as the carrier gas.Injections were performed manually, with a 10 µL syringe suitable for on-columninjections. The injected volume was 1 µL. Data-processing was done using Class VPver. 4.3 (Shimadzu, Tokyo, Japan). The oven temperature was programmed from60°C (1 minute hold) to 425°C (1 minute hold) at a programming rate of 0.33°C/s(20°C/min). The FID was operated at 430°C, with a hydrogen- and air-flow of 40 and400 mL/min, respectively. The sample was a solution of triglycerides ranging fromCN22 to CN52 at approximately 50 ppm each in hexane.

The second instrument, used for increasing the analysis-speed of a TAG sample,was a HP 6890 GC (Agilent Technologies) equipped with a flame ionisation detectorand an on-column injector. An extra oven insert (Agilent) to decrease the oven volumeand to provide extra insulation for high temperature programming rates was used. Thecolumn was the same as used in the Shimadzu GC. Hydrogen was used as the carriergas. Injections were performed using an autosampler (HP 7683 Series). The oven wasprogrammed from 60°C to 200°C at 1.67°C/s (100°C/min) and from 200°C to 355°C(30 seconds hold) at 1.1°C/s (65°C/min). The on-column injector was operated in theoven-track mode: the temperature was approximately 3°C ahead of the oventemperature. The FID was operated at 440°C with a hydrogen- and air-flow of 40 and450 mL/min, respectively.

Experiments to determine the thermodynamic values of the solutes wereperformed on a Carlo Erba/Fisons GC 8000 Series, type 8380 (Milan, Italy), equippedwith a flame ionisation detector and a split/splitless injector. Analyses were doneisothermally at 300, 350 and 390°C (duplicate). The column (10 m long) that wasused was similar to the column described above. An injector- and detector-temperature of 430°C was used and a split-flow of approximately 50 mL/min wasapplied. The injected volume was 1 µL.

3.2.3. Chemicals

The solvents that were used (hexane and methanol) had a purity of 99% or higher(Merck, Darmstadt, Germany). A diesel sample was obtained from a gas stationduring normal service. The mineral oil sample was obtained from the Chemiewinkelat the Eindhoven University of Technology (Eindhoven, The Netherlands). The n-


alkane standards (C10-C42), triethylene glycol (teg), tetra-ethylene glycol (tteg) (Fluka,Buchs, Switzerland), phenol, p-cresol, mono-ethylene glycol (meg) and diethyleneglycol (deg) (Sigma-Aldrich, Steinheim, Germany) all had a purity of 98% or higher(GC-grade). The TAG sample was a standard sample obtained from UnileverResearch (Vlaardingen, The Netherlands).

3.3. DISCUSSION AND RESULTS

3.3.1. Retention time stability

In the resistive heating technique large amounts of heat are produced in systems thathave very low thermal masses. Small changes in heat production hence can result inmajor temperature variations, which in turn can result in poor retention time stability.In a first series of experiments the retention time stability in the EZ-Flash system wasstudied. Ehrmann and coworkers [3] reported a relative standard deviation (RSD) ofretention times of better than 1%. MacDonald [10] reported an even better retentiontime RSD of lower than 0.2%. The latter value is comparable to that found using aconventional GC-oven operated at high heating rates. In this work the RSD value ofthe retention times was lower than 0.3% (table 3.1). The typical average peak width athalf height was about 0.4 seconds at a heating rate of 4ºC/s.

Table 3.1: Repeatability of retention time, peak area and peak width using atemperature programming rate of 5ºC/s. Compound: n-C22H46.

Retention time[minutes]

Area [pA⋅s]Peak width at50% height [s]

Temperature program:5ºC/s

0.6420.6420.6430.6440.6430.6410.6450.643

89.7*

90.598.192.391.995.9

121.8*

90.2

0.1940.1940.1940.1930.1930.200*

0.1940.194

AverageSTDEVRSD %

0.6430.001250.194

93.23.163.40

0.1940.000488

0.252

* These values are not included in the calculations

40 Chapter 3

min0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

pA

0

200

400

600

800

1000

1200

1400C10

C40

C12

C24

Figure 3.1: Fast analysis of a normal alkane mixture (C10-C42) within 1.5 minutes.Temperature program EZ-Flash: 80ºC –375ºC at 4ºC /s, inlet-pressure = 300 kPa,split-flow = 500 ml/min, ū = 200 cm/s, temperature-program GC-oven: 80ºC –140ºCat 30ºC/min, injector temperature program: 350ºC -550ºC at 16ºC /s.

An example of a fast analysis of C10-C42 at a temperature programming rate of4ºC/s, is shown in figure 3.1. The elution time of C42 here is less than 1.3 minutes.Elution of the same compound in a conventional GC-analysis using a maximumheating rate of 40oC/min roughly takes 10 minutes.

The detection limit and the sample capacity of the EZ-Flash system wereestimated at approximately 10 pg and 200 ng (per component), respectively. Theresulting working range, being the region between detection limit and sample amountat which overloading of the column starts to occur, is extremely large. Normally infast GC poor working ranges are found [11]. The good results for the working rangeobtained here are the result of the wide diameter column that is used.

From figure 3.1 it can be seen that some band broadening occurs for the highboiling compounds C40 and C42. This is most likely caused by the fact that these twocompounds elute isothermally at 375°C. Another possible explanation is the presenceof cold spots between the injector and the EZ-Flash column or between the columnand the detector. Similar effects were also observed by Ehrmann [3]. As discussed inthe introduction, the typical application area for fast heating techniques is that ofsamples consisting of a low number of compounds with widely differing boilingpoints. Below, the use of resistive heating for such samples as well as for samples notmeeting these requirements, is discussed.


3.3.2. Analysis of mineral oil

As already mentioned in the introduction, a possible application for fast analysis usinghigh temperature programming rates, is the determination of mineral oilcontamination in water samples. In the official Dutch methods [4] for thequantification of mineral oil content of water samples, the water is extracted withpetroleum ether or hexane. The organic phase is then injected onto a non-polar GCcolumn. The total peak area between the elution positions of the normal alkanes C10

and C40 is then used as a measure for the oil content, which means that little resolutionis needed. An example of a fast mineral oil analysis is shown in figure 3.2. Thisanalysis, including the cool-down time, takes less than 2 minutes. This means thatwith fast temperature programming a reduction of the analysis time of approximately10 times could be obtained compared to the conventional mineral oil analysis ofalmost 25 minutes. In comparison to figure 3.1, the retention time of C40 in figure 3.2is longer because of the lower initial temperature.

min0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

pA

0

200

400

600

800

C10C40

Figure 3.2: Fast analysis of mineral oil dissolved in hexane within 1.5 minutes.Temperature program EZ-Flash: 60ºC –375ºC at 4ºC/s, inlet-pressure = 300 kPa,splitless injection, temperature-program GC-oven: 60ºC –140ºC at 30ºC/min, injectortemperature program: 350oC-550oC at 16oC/sec.

3.3.3. Analysis of glycols

The second application that was investigated was the separation of an industrial glycolmixture. This mixture consists of four oligomers with large differences in boiling

42 Chapter 3

range: (meg) monoethylene glycol (C2H6O2), (deg) diethylene glycol (C4H10O3), (teg)triethylene glycol (C6H14O4) and (tteg) tetra-ethylene glycol (C8H18O5) dissolved inmethanol. The boiling points of these compounds are 198, 244, 285 and 328oC,respectively. An analysis of this sample with a standard GC on a standard columntemperature programmed from 40 to 320oC at 15oC/min, took about nine minutes.Using a fast temperature program with the EZ-Flash, the analysis time was reduced toless than 0.5 minutes. An example of a chromatogram is shown in figure 3.3.

min0 0.1 0.2 0.3 0.4 0.5 0.6

pA

0

20

40

60

80

100

megtteg

deg

teg

Figure 3.3: Fast analysis of (meg) monoethylene glycol (C2H6O2), (deg) diethyleneglycol (C4H10O3), (teg) triethylene glycol (C6H14O4) and (tteg) tetra-ethylene glycol(C8H18O5). Inlet pressure: 45 kPa, splitflow: 300 ml/min, TEZ-Flash: 40ºC-300ºC at15ºC/s, Toven: 40ºC -100ºC at 120ºC/min.

3.3.4. Analysis of phenol and p-cresol in diesel

The third application is the analysis of phenol and p-cresol in diesel-oil. The presenceof these poisonous compounds in oil is determined by a gas chromatographic analysis.Compared to a standard temperature programmed analysis, the analysis time can bereduced significantly by using high temperature programming rates. Examples ofthese analyses are presented in figures 3.4 and 3.5. In figure 3.4 an overall temperatureprogramming rate of 1°C/s was applied. For the analysis shown in figure 3.5 however,first a slow temperature program (4°C/s) is applied during elution of phenol and p-cresol to obtain a good separation and detection of these compounds. Then a very high


temperature program (12°C/s) is applied to elute the remaining compounds, which arenot of interest. In this way, the analysis time could be reduced 5 times.

min0 1.0 2.0 3.0 4.0

pA

10

20

30

40

Phenol p-Cresol

Figure 3.4: Analysis of phenol and p-cresol in diesel (dissolved in hexane): TEZ Flash =60ºC – 350ºC (1ºC/s), inlet pressure: 100 kPa, splitflow: 100 ml/min. Detection: FID.For direct identification of phenol and p-cresol in diesel a mass-spectrometer or anO-FID (oxygen selective) is required.

min0 0.2 0.4 0.6 0.8

pA

0

100

200

Phenol p-Cresol

Figure 3.5: Analysis of phenol and p-cresol in diesel (dissolved in hexane): TEZ Flash =60ºC – 156ºC (4ºC/s), 156ºC – 350ºC (12ºC/s). Inlet pressure: 100 kPa, splitflow: 100ml/min. Detection: FID.

44 Chapter 3

3.3.5. Analysis of diesel

In the previous paragraphs a number of GC applications have been described that canbenefit from the use of fast temperature programming. Now an analysis is shown forwhich fast temperature programming is not a suitable method to reduce the analysistime. In figure 3.6 a conventional temperature programmed analysis of diesel isshown. The temperature programming rate used here is 0.5ºC/s. A fast analysis at8ºC/s is shown in figure 3.7. For both analyses the same columns (resistively heated)are used, in order to be able to compare the resolution at the two applied temperatureprogramming rates. A detail comparison is shown in figure 3.8A and 3.8B. As alreadymentioned before, the drawback of the use of high temperature programming rates isthe loss of resolution. When comparing the analysis of diesel at a temperature programof 8ºC/sec with that obtained at 0.5ºC/s, this loss of resolution is clearly visible fromthe loss of separation for the critical pair C18 and phytane in diesel. Fast temperatureprogramming hence is not an option for reducing the analysis time of this sample.

min0 1 2 3 4 5 6 7

pA

10

15

20

25

Figure 3.6: Analysis of diesel in hexane within 8 minutes. TEZ-Flash = 60°C – 350°C at0.5°C/s. Inlet pressure: 55kPa, splitflow = 200 mL/min. Toven= 60-140°C at 8°C /min.Critical pairs C17 with pristane and C18 with phytane are shown.

C17 Pristane

C18 Phytane


min0 0.2 0.4 0.6 0.8

pA

20

40

60

80

100

Figure 3.7: Fast analysis of diesel in hexane within 1 minute. Temperature program:TEZ-Flash = 60°C – 350°C (8°C/s), Toven = 60°C – 140°C (30°C/min), inlet pressure:55kPa, splitflow = 300 ml/min.

Figure 3.8: Separation of the critical pair 1) C18 and 2) phytane in diesel using hightemperature programming rates with a 320 µm i.d. column: (A) 0.5°C/s and (B) 8°C/s(from chromatogram in figure 3.6 and 3.7) compared to the separation on: (C) 50 µmi.d. column at 2°C/s.

For speeding up this separation clearly a method that preserves resolution isrequired, e.g. hydrogen as the carrier gas, vacuum outlet operation or the use of anarrow bore column. A baseline separation of the critical pair C18 and phytane indiesel is shown in figure 3.8C. Here a 10 m × 50 µm i.d. narrow bore column wasused at a temperature programming rate of 2ºC/s (fast GC-oven heating). The plate

A: 0.5ºC/si.d.: 320µm

B: 8ºC/si.d.: 320µm

C: 2ºC/si.d.: 50µm1

2

11

22

C17

C18

46 Chapter 3

number for the short (5 m × 320 µm i.d.) resistive heating column was estimated atapproximately 15,000, whereas the narrow bore column provides a much higherseparation efficiency with up to 150,000 plates.

3.3.6. Comparison of fast temperature programming to the use of short columns

A realistic question at this point is whether the use of fast temperature programming ispreferable over the use of short columns for example. This was investigated bystudying the results from columns with lengths of 5, 4 and 3 meter (320 µm i.d.).When using column lengths shorter than 3 meter, the mono-ethylene glycol (meg)peak is not separated from the solvent peak, so the 3 meter column is the shortestpossible column for a sufficient separation of the oligomer sample. Operating theshort columns at carrier gas velocities above the optimum also leads to a badseparation and ill-shaped, tailing peaks. Also, at higher inlet pressures it becomes verydifficult to inject the sample properly. Therefore only the analysis times at theoptimum inlet pressure are compared to the analysis times at fast temperatureprogramming.

By calculating the Trennzahl number (TZ), which indicates the number of peaksthat can be placed between two consecutive members of a homologous hydrocarbonseries, the influence of temperature programming and length of the column on theseparation efficiency was estimated. The Trennzahl number was calculated accordingto equation 3.1, where tR1, tR2, wh1 and wh2 are retention times and peak widths(baseline level) of two consecutive oligomers (for example between the meg and degpeak).

1)()( 21

12 −

+−

=hh

RR

wwtt

TZ (3.1)

From table 3.2 it can be concluded that the best solution to speed up the analysisof the present sample is the use of fast temperature programming. Even at a heatingrate of 15ºC/s, the separation of the oligomers on the resistive heating system is betterand the analysis time is shorter compared to the column with a length of only 3 meter.The high TZ value of the meg-deg separation on a 3 m column in comparison to theTZ values of the 4 and 5 m columns, is caused by the fact that the chromatographicconditions for the 4 and 5 meter columns probably were not adequately optimised forthis separation.


Table 3.2: Comparison of the minimum analysis time and Trennzahl number (TZ) ofthe analysis of four oligomers: (meg) monoethylene glycol (C2H6O2), (deg) diethyleneglycol (C4H10O3), (teg) triethylene glycol (C6H14O4) and (tteg) tetra-ethylene glycol(C8H18O5) using fast temperature programming and short columns.

Columnlength

[m]

Platenumber

Optimuminlet

pressure[kPa]

Analysis timeat popt and

49.9oC/min[min]

TZ meg-degat 20oC/min

and popt

TZ deg-tegat 20oC/min

and popt

TZ teg-ttegat 20oC/min

and popt

5*

4*

3*

30,00016,00012,000

403020

2.562.742.64

6.444.396.32

22.510.48.22

27.215.311.3

Analysis timeat 15oC/sec

[min]

TZ meg-degat 15oC/sec

and popt

TZ deg-tegat 15oC/sec

and popt

TZ teg-ttegat 15oC/sec

and popt

5# 30,000 40 0.491 13.0 6.48 6.12

*: Analyses performed with Fisons GC 8000 Series.#: Analyses performed with resistive heating system.

48 Chapter 3

3.4. OPTIMISATION STATEGY OF FAST TAG ANALYSES

3.4.1. Discussion and results

An optimisation strategy as developed by Snijders and co-workers for the optimisationof temperature-programmed GC separations is used to predict the influence ofoperational and instrumental parameters on the separation obtained in fast temperatureprogrammed GC. The principles of the method are described in detail in a paper ofSnijders et al. [8]. In short, the chromatographic separation process is modelled as asequence of very short time segments. In a time segment the column temperature isassumed to be constant. The calculation procedure starts with the calculation of thetemperature in the time segment at the given temperature program. Next, the retentionfactor at that temperature is calculated and flow equations are used to calculate thelinear velocity at the position of the compound in the column. From these parametersthe migration distance of the solute in the time segment is calculated. This procedureis repeated until the compound reaches the column exit. The model takes into accountpressure drop across the column, pressure dependency of velocity and diffusivity,band broadening caused by a non-uniformly coated stationary phase and temperaturedependency of the viscosity. The heart of the calculation model is the equation thatdescribes retention as a function of temperature (equation 3.2):

RTHak ∆+=

βlnln (3.2)

Where a = exp(∆S/R), β is the column phase ratio, R is the universal gas constant, Tthe absolute temperature, ∆H the molar enthalpy of the solution and ∆S the molarentropy of solution. By plotting ln(k) versus the reciprocal of the absolute temperature,the enthalpy- and entropy-term can be determined from the intercept and the slope,respectively. This was done by performing retention measurements at three isothermaltemperatures: 300, 350 and 390ºC. The thermodynamic data obtained from thesemeasurements are presented in table 3.3. In addition to the thermodynamic properties,other parameters are required for the calculations, such as the dimensions of thecolumn, the dead-time at the initial temperature, the temperature program, themolecular structures of the solutes and the coating efficiency of the column.


Table 3.3: Entropy and enthalpy-terms of the triglycerides (TAG) determined on a 10m × 530 µm × 0.17 µm column. Injections were performed on a split/splitless injector,with a splitflow of approximately 50 mL/min.

Number of C-atomsof triglyceride (CN)

Entropy terma/β (× 10-11)

Enthalpy term∆H/R (K)

32343638404244464850

142.24101.7094.2099.7226.1316.7732.0620.9311.819.27

12125125281274712896139451440014140145881512515439

In figure 3.9 and 3.10 a comparison is shown of an experimental and a calculatedchromatogram of a TAG separation. The sample consists of 16 components of ahomologous series of triglycerides, ranging from CN22-CN52. For the calculations onlyCN32-CN50 were selected to reduce the amount of data. The calculated chromatogram(figure 3.10) was produced from the calculated retention times and peak widths. Toreduce the total analysis time even further the temperature programming rate wasincreased from 20 to 65°C/min while the carrier gas velocity was increasedsimultaneously from 60 to 280 cm/s. This resulted in a reduction of the analysis timefrom 16 to 4 minutes. The resulting chromatogram is shown in figure 3.11. In table 3.4the calculated and experimental values for the retention times and peak widths arecompared. From the figures 3.9 and 3.10 and the values presented in table 3.4, it can beconcluded that the calculation method used here is suitable for predictingchromatograms of TAG analyses on wide bore columns. Calculated and experimentalvalues show a good agreement, especially with regard to the retention times. There issome discrepancy between the calculated and experimentally determined peak widths.The errors in predicting the retention times and peak widths in the present workgenerally are below 2 and 15% respectively (table 3.4). Values reported by Snijders andco-workers are 4 and 15% respectively [8]. The peak widths for the CN36, CN42,CN44 and CN50 triglycerides are not given in table 3.3 since they could not bedetermined accurately. These TAG consist of two or more isomers, which results in

50 Chapter 3

double peaks. If the calculated and experimental results of the fast chromatogram(figure 3.11) are compared, errors for the peak widths are somewhat higher (table 3.4:20-45%). Most likely this is caused by injection band broadening.

11.5 12.0 12.5 13.0 13.5 14.0 14.5 15.0 15.5Time [min]

Res

pons

e FI

D

C32

C34

C36

C38

C40

C42

C44C46

C48

C50

8.5 11.0 13.5 16.0Time [min]

Res

pons

e FI

D

Figure 3.9: Experimental chromatogram of TAG (CN32 - CN52), at a temperatureprogramming rate of 0.33°C/s (20°C/min). Toven: 60°C (1 min) - 20°C/min - 425°C(1 min). Inlet pressure: 10kPa, ū: 60 cm/s, injection: on-column, injection volume:1 µL, column: 5 m × 530 µm × 0.17 µm. Insert: complete chromatogram.

11.5 12.0 12.5 13.0 13.5 14.0 14.5 15.0 15.5Time [min]

Cal

cula

ted

resp

onse

C32

C34

C36

C38

C40

C42

C44

C46

C48

C50

Figure 3.10: Calculated chromatogram according to the method described bySnijders et al. [8]. Input-data: Column: 5 m × 530 µm × 0.17 µm, temperatureprogram: 60°C (1 min) - 20°C/min - 425°C (1 min), dead-time: 0.14 min (measuredfor hexane).


2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8Time [min]

Res

pons

e FI

D

C32

C50

C48

C46

C44

C42C40

C38

C36

C34

Figure 3.11: Fast chromatogram of TAG (25 ppm in hexane) showing CN32 to CN50.Toven: 60°C - 100°C/min - 200°C - 65°C/min - 355°C. Inlet pressure: 20 kPa(hydrogen), ū: 280 cm/s, injection: on-column, injection volume: 2 µL.

From figure 3.11 it can be seen that the analysis time of the TAG separation can bereduced by a factor of 4 by using high temperature programming rates. The total cycletime (time between two injections) was now approximately 13 minutes1. This is causedby the long cooling times for both the oven and on-column injector. The retention timeand the peak area repeatability of this analysis was 0.1 and 5 % (RSD), respectively (8injections). The obtained resolution (average between all neighbouring pairs in thechromatogram) is still acceptable (Rs = 2-3). The sample composition (100% method2)found using the fast method was in excellent agreement with that of the standard on-column analysis of TAG.

From the present results it is clear that to reduce the analysis time, it is moreappropriate to increase the temperature programming rate rather than increasing thecarrier gas velocity. In this work we have been able to use the high programming rate ofrecently developed GC-ovens. However, it would be even more efficient to use the veryhigh heating rates of a resistive heating system as described earlier in this chapter. Themaximum temperature of the resistive heating system (375°C) used in this work was nothigh enough for a fast elution of the high boiling TAG compounds. For this purposeresistive heating systems capable of heating to at least 425°C should be developed.

1 In chapter 8 of this thesis it is shown that by using a hot split injection of TAG and high temperatureprogramming rates the overall GC analysis time could be reduced to approximately 5 minutes.2 In the 100% method the amount of TAG is expressed as the peak area of one compound relative to the totalpeak area of the chromatogram. Because an FID was used as the detection device, the response factors of theTAG are assumed to be all equal (unity).

52 Chapter 3

Using the calculation procedures described above it is possible to select the columnlength, diameter, programming rate etc. that provide the required resolution in theshortest possible analysis time. Additionally, the elution temperature can be predicted.

Table 3.4: Comparison of calculated and experimental values for the retention timeand peak width (σ). The values used in this table correspond to the chromatogramsshown in figure 3.9 (slow), 3.10 (calculated) and 3.11 (fast).

Retention-time [min] Peak-width (σσσσ) [min]

Experimental Calculated ∆∆∆∆ (%) Experimental Calculated ∆∆∆∆ (%)

Carbonnumber

TAG(CNX)

X= a* b* a b a b a b a b a b

32

34

36

38

40

42

44

46

48

50

11.73

12.22

12.68

13.15

13.60

14.00

14.35

14.76

15.14

15.47

2.47

2.62

2.76

2.90

3.04

3.16

3.28

3.41

3.54

3.65

11.82

12.32

12.71

13.10

13.75

14.19

14.40

14.83

15.25

15.54

2.52

2.67

2.79

2.91

3.11

3.25

3.31

3.44

3.57

3.66

-0.72

-0.80

-0.18

0.41

-1.09

-1.35

-0.38

-0.45

-0.72

-0.47

-1.90

-2.05

-1.12

-0.30

-2.38

-2.64

-0.89

-0.85

-0.99

-0.31

0.0178

0.0187

d.p.#

0.0221

0.0191

d.p.

d.p.

0.0225

0.0195

d.p.

0.012

0.012

0.013

0.015

0.012

0.013

0.014

0.014

0.012

0.014

0.0179

0.0179

0.0181

0.0184

0.0177

0.0176

0.0182

0.0182

0.0180

0.0180

0.0098

0.0098

0.0099

0.010

0.0096

0.0096

0.0099

0.0099

0.0097

0.0097

-0.36

4.24

d.p.

11.9

8.0

d.p.

d.p.

11.3

8.8

d.p.

20

25

32

46

25

36

43

40

25

47

a* = Slow analysis. Conditions: Pi = 10 kPa, ū = 60 cm/s, Toven = 60°C (1 min) - 20°C/min - 425°C.b* = Fast analysis. Conditions: Pi = 20 kPa, ū = 280 cm/s, Toven = 60°C - 100°C/min - 200°C -65°C/min -355°C (30 s).#d.p. = double peaks: co-eluting isomers. Peak widths could not be calculated accurately.

Calculations were performed for a 530 µm wide-bore column at differentprogramming rates (1 – 4°C/s), different column lengths (5 and 10 m) and differentcarrier gas velocities (50-350 cm/s). The influence of these parameters on resolution,elution temperature and analysis time was investigated with the aim to select the set ofconditions that provides the shortest analysis times and acceptable elution temperatures.The results of the calculations are shown in figure 3.12 and 3.13. Figures 3.12A-3.12Cand figures 3.13A-3.13C represent the results of the calculations for the 10 and 5 mcolumn, respectively. From figures 3.12A and 3.13A it can be seen that at carrier gasvelocities exceeding 200 cm/s, resolution approaches the minimum required value of1.5 (baseline separation). At higher carrier-gas flows, separation of the TAG is nolonger acceptable. From the results shown in figures 3.12B and 3.13B it can beconcluded that an increase of the carrier gas velocity from 25 to 250 cm/s reduces the


0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

25 50 75 100 125 150 175 200 225 250 275 300 325 350

Carrier gas velocity [cm/s]

Res

olut

ion

0.5 degrC/s1 degrC/s2 degrC/s4 degrC/s8 degrC/s

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

11.0

12.0

13.0

25 50 75 100 125 150 175 200 225 250 275 300 325 350


Elut

ion-

time

[min

]


300

325

350

375

400

425

450

475

25 50 75 100 125 150 175 200 225 250 275 300 325 350


Elut

ion

tem

pera

ture


0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

25 50 75 100 125 150 175 200 225 250 275 300 325 350


Res

olut

ion


1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

11.0

12.0

13.0

25 50 75 100 125 150 175 200 225 250 275 300 325 350


Elut

ion

time

[min

] 0.5 degrC/s1 degrC/s2 degrC/s4 degrC/s8 degrC/s

300

325

350

375

400

425

450

475

25 50 75 100 125 150 175 200 225 250 275 300 325 350


Elut

ion

tem

pera

ture


Figure 3.12 (left) and 3.13 (right): Calculated influence of the temperatureprogramming rate and linear carrier-gas velocity on: A) the resolution (CN46 andCN48), B) analysis time (CN50) and C) elution temperature (CN50) of a TAGanalysis on a 10 (left, figure 12) and 5 m (right, figure 13) wide bore column. Data atcarrier gas velocities lower than 200, 100 and 50 cm/s (at programming rates of 8, 4and 2°C/s, respectively) are not mentioned in this figure because for these settings thecalculated elution times would correspond to elution temperatures >450°C, which ishigher than the maximum operation temperature of the ultimetal column used in thiswork.

3.12A 3.13A

3.12B 3.13B

3.12C 3.13C

54 Chapter 3

analysis time by approximately 25% only. Figures 3.12 and 3.13 present the results ofthe calculations for linear velocities up to 3.5 m/s. It should be emphasised here thatsuch velocities are of limited practical relevance. Carrier gas velocities of 350 cm/s andhigher require high inlet pressures, even for 530 µm i.d. columns. Moreover, at thesehigh column outlet flows the flame ionisation detector is readily extinguished. Finally,leaks at the injection port are more likely to occur at high gas pressures.

Faster temperature programming clearly is a more attractive way to reduce theanalysis time. Upon doubling the temperature programming rate, the analysis time isreduced by around 50% for both the 5 m and the 10 m column. In figures 3.12C and3.13C the influence of the carrier-gas velocity and temperature programming rate on theelution temperature of the last eluting compound (CN50) is shown. From the figures itis apparent that the elution temperatures increase significantly when higherprogramming rates are used together with low linear velocities. The general conclusionhence is that an increase of the programming rate should always be accompanied by anincrease of the carrier gas velocity. Failure to do so results in significantly higherelution temperatures, which can cause problems, especially for thermally labileanalytes. Finally, it can be concluded that programming rates higher than 4°C/s cannotbe applied for the TAG sample described in this chapter. As can be seen from Figures.3A and 4A, programming rates in excess of 4°C/s result in insufficient resolution (Rs <1.5). In work of Blumberg et al. [12] it is described that the temperature increase in atime interval equal to the column void time should not exceed about 10-15°C. Whenthis criterion is applied to the work presented in this paper, the optimal temperatureprogramming rate should not exceed approximately 3-4°C/s. This is in accordance withour conclusions.

Table 3.5 summarises the conclusions that can be drawn from figures 3.12 and 3.13.The table lists optimum carrier gas velocities as well as elution temperatures and elutiontimes of the last eluting compound at a constant resolution of 1.5 for column lengths of5 and 10 m. Elution temperatures in excess of 425ºC should be avoided to reduce therisk of thermal degradation of the triglycerides. When comparing the elution times of a5 and 10 m column, it can be concluded that a shorter column does not yieldsubstantially shorter analysis times as compared to longer columns. For on-columninjection the 10 m column is more suitable because of the improved componentrefocusing in the flooded zone. Because on-column injection has to be used for the highmolecular weight TAG, the choice of the column dimensions was restricted to largerbore columns [13]. Narrow-bore columns might give faster separations, but only at theexpense of a poorer compatibility with on-column injection. From the results shown inTable 3.5 it can be concluded that to obtain baseline separation (Rs = 1.5) for a TAG


sample a programming rate of 4ºC/s can be applied at a carrier gas velocity of 265 cm/s.The analysis time will then be reduced from 16 minutes obtained with a conventionalanalysis (figure 3.9) to 2.5 minutes when using rapid temperature programming. Thetotal cycle time, i.e. analysis time and cooling of the GC-oven, will be around 5minutes.

Table 3.5: Effect of programming rate and linear carrier gas velocity on the elutiontemperature and elution time at a constant resolution of 1.5 for a 10m and 5m widebore (530 µm) column, calculated according to Snijders et al. [8].

Temperatureprogramming rate

[ºC/s]

Linear carrier gasvelocity [cm/s]

Elutiontemperature a [ºC]

Elution time a

[min]

L = 10 m, Rs = 1.5 b

0.5 395 355 10.01 380 350 5.92 340 370 3.64 265 410 2.5

8 c 175 470 1.9L = 5 m, Rs = 1.5 b

0.5 295 325 9.71 275 340 5.72 245 360 3.54 190 390 2.4

8 c 150 430 1.8

a Elution-time and -temperature of CN50 triglyceride;b Resolution (Rs) of CN46 and CN48;c Required resolution of 1.5 was not obtained at this programming rate. The values for velocity,elution temperature and elution time are valid for a resolution of 1.25.

3.5. CONCLUSIONS

It is possible to increase the speed of analysis substantially by using fast temperatureprogramming. The speed of analysis of mineral oil, the oligomer sample and of

56 Chapter 3

p-cresol and phenol in diesel are increased substantially. The performance of theresistive heating system used in this work is high, but has to be improved forcompounds higher than C36. The maximum operation temperature of 375°C is nothigh enough for very high boiling compounds like the n-alkanes C36-C42 or for thehigh boiling TAG compounds. A gain in speed of approximately 5 to 10 times can beobtained using fast temperature progamming.

For samples similar to the oligomer sample shown here, the use of rapidtemperature programming might be a better option than the use of a short column. Foreach separation it should be carefully considered which method to reduce the analysistime is most appropriate. A method that might give the best performance for oneapplication, might very well be unsuitable for another.

A numerical calculation method previously developed to optimise temperatureprogrammed GC separations was successfully used to select the optimum column andoperational settings for the gas chromatographic analysis of TAG. Calculated andexperimental values of retention times and peak widths are in good agreement. From theobtained results it can be concluded that fast temperature programming is the method ofchoice for reducing the analysis time of a TAG sample. There is however a maximumheating rate above which sufficient resolution can no longer be obtained. The optimumheating rate depends on both the column length as well as the linear velocity. By usingthe simulation procedure, the number of experiments required to find the optimumconditions could be reduced substantially.

3.6. REFERENCES

1. V. Jain and J.B. Phillips, J. Chromatogr. Sci., 33 (1995) 55.2. M.E. Hail and R.A. Yost, Anal. Chem., 61 (1989) 2410.3. E.U. Ehrmann, H.P. Dharmasena, K. Carney and E.B. Overton, J. Chromatogr.

Sci., 34 (1996) 533.4. NVN 6678 (Dutch standards), Determination of mineral oil content by gas

chromatography, 1 (1997).5. M. v. Deursen, J. Beens, H.-G. Janssen and C.A. Cramers, J. High Resol.

Chromatogr., 22 (9) (1999) 509.6. G. Castello, P. Moretti and S. Vezzani, J. Chromatogr. A, 635 (1993) 103.7. D.E. Bautz, J.W. Dolan, W.D. Raddatz and L.R. Snyder, Anal. Chem., 62 (1990)

1560.8. H. Snijders, H.-G. Janssen, and C.A. Cramers, J. Chromatogr. A, 718 (1995) 339.


9. M. v. Lieshout, M. v. Deursen, R. Derks, J.G.M. Janssen and C.A. Cramers, J.High Resol. Chromatogr., 22 (2), (1999) 116.

10. S.J. MacDonald and D. Wheeler, International Laboratory, September, (1998)20c.

11. P.G. Van Ysacker, H.-G. Janssen, H.M.J. Snijders, P.A. Leclercq and C.A.Cramers, J. Microcol. Sep.

12. L.M. Blumberg and M.S. Klee, J. Microcolumn Separations, 12 (9) (2000) 508.13. H.-G. Janssen, H. Steenbergen, J. Oomen, and J. Beens, J. Microcolumn

Separations, 12 (10) (2000) 523.

58 Chapter 3

CHAPTER 4THEORETICAL DESIGN CONSIDERATIONS

FOR MULTI-CAPILLARY COLUMNS IN FAST

GAS CHROMATOGRAPHY1

SUMMARY

In this chapter the design criteria for a new type of column for fast gaschromatography are evaluated. The column under investigation is the so-called multi-capillary column. This column consists of some 900 capillaries with small innerdiameters combined in a bundle. Simulations are performed to investigate theinfluence of variations in film thickness, diameter and length on the performance ofthe system. These calculations are performed using the theoretical plate heightequation of Golay and Giddings for capillary columns. The simulations convincinglyshow that non-uniform film thicknesses, diameters and to a lesser extent lengths,result in significant band broadening. The performance for the multi-capillary columnfor fast separations is discussed and some examples of analyses are shown.

1 Partially published as “Theoretical design considerations of multi-capillary columns in fast gaschromatography”, by: M.M. van Deursen, M. van Lieshout, R. Derks, H.-G. Janssen and C.A. Cramers in J.High Resol. Chromatogr., 22(2), 119-122, 1999.

60 Chapter 4

4.1. INTRODUCTION

Reduced analysis times can most conveniently be achieved by decreasing the innerdiameter of the capillary GC column [1-3] (chapter 2 of this thesis). Capillary columnswith reduced inner diameters, typically in the range of 50-150 µm, have indeed beensuccessfully used to obtain extremely fast separations. As already demonstrated inchapter 2, the use of such columns is, unfortunately, far from easy. Injection anddetection have to be extremely fast, dead volumes have to be strictly minimised andthe detector should be highly sensitive.

In 1997 a new type of column that is very suitable for fast GC has beenintroduced: the multi-capillary column. This column is made by combining some 900capillaries with an inner diameter of 40 µm into a bundle. The advantages of themulti-capillary column in comparison to a single narrow-bore column are the highertotal flow rate, the higher sample capacity and, consequently, the lower minimumdetectable concentration of the solutes. To obtain the maximum performance of amulti-capillary column, the column has to meet very stringent requirements. Forexample, each of the capillaries should have exactly the same diameter. If the 900capillaries would significantly differ in diameter, this would cause serious bandbroadening. Other important parameters that should remain constant are the length andthe film thickness of the capillaries.

In this chapter a simulation of extra column band broadening caused by variationsin film thickness, diameter and length of the individual capillaries in the ‘bundlecolumn’ is made with the help of theoretical plate height equations and relatedequations describing retention in gas chromatography. The aim of these simulations isto obtain insight in the requirements that have to be imposed on the manufacturingprocess of the multi-capillary column. Finally, some examples of analyses performedwith a multi-capillary column are shown and the performance of the column isdiscussed.

4.2. THEORY

A chromatogram can be represented as a series of peaks, each with its own retentiontime, peak width and height. The retention time and the peak width are fundamentalparameters that are only affected by the thermodynamics of the system (retentiontime) and its kinetics (peak width). In this chapter equations will be derived that allow

Theoretical design considerations for multi-capillary columns in fast gas chromatography 61

the prediction of chromatograms for single capillary columns and for sets of singlecapillaries, that eventually form a multi-capillary column.

First of all the retention time tR and the chromatographic peak width σc arecalculated. The retention time tR of the retained component can be calculated usingequation 2.1 and 2.2 in chapter 2. In order to be able to calculate tR, first k and ū haveto be determined. The retention factor k is defined as the number of solute moleculesin the stationary phase divided by the number of molecules in the mobile phase:

c

f

m

s

mm

ss

dd

KVV

KVCVC

k4⋅≈⋅== (4.1)

Where dc is the diameter of the capillary, df is the film thickness, K is the distributioncoefficient, Vs is the volume of stationary phase and Vm is the volume of mobile phase.The average carrier gas velocity ū is given by equation 2.7 (chapter 2). The columnoutlet velocity of the carrier gas can be calculated using equation 4.2:

)1(64

22

−= PL

pdu oc

o η(4.2)

Where η is the dynamic gas viscosity. Once uo and f2 are known, ū can be calculated.Now the retention time tR can be calculated from equations 2.1 and 2.2. Equation 4.1can be used to correct the retention factor k for differences in column diameter or infilm thickness.

Using a second set of equations the other representative of the peak shape, thechromatographic peak width σc can be calculated according to equation 2.4 (σ in thisequation is equal to σc (chapter 2)). The number of theoretical plates can be calculatedby using equation 2.5.

For a single capillary column, Golay and Giddings [4,5] described the relationbetween plate height H and gas velocity uo using the expressions:

2

2

21,

2

2

2,

)1(32

)1(9616112 f

Dud

kkf

Dud

kkk

uD

Hs

of

om

oc

o

om ⋅+

+⋅

+

+++= (4.3)

Where f1 is the pressure correction factor according to Giddings [5], f2 the pressurecorrection factor according to James and Martin [6] (see table 2.1, chapter 2), Ds thediffusion coefficient of the solutes in the stationary phase and df the film thickness.

62 Chapter 4

Throughout this chapter gas diffusion coefficients will be calculated according tothe method developed by Fuller et al. [7]. The following approximate equation is usedto calculate the diffusion coefficient of the solute in the stationary phase:

4, 105 ⋅≈s

om

DD

(4.4)

From the equations 2.5, 4.3 and 4.4 the number of plates can be calculated. Thechromatographic peak width σc can then be obtained from equation 2.4.

From the peak widths and retention times obtained from the calculationsdescribed above, the total chromatogram can now be calculated. To do so, the tR andσc values for each peak have to be substituted in the Gaussian distribution function(equation 4.5). Summation of these functions gives the predicted chromatogram. Inthe calculations performed in this research work, the amplitude of the peak, themaximum peak height hMAX, is an arbitrarily chosen value.

2

2

2)(

)( c

Rtt

MAX ehth σ⋅−

−

⋅= (4.5)

where h(t) is the height of the peak at time t, hMAX is the height of the peak in itsmaximum and t is a given point in time.

4.3. EXPERIMENTAL

The GC used for this work was an HP 6890 GC (Agilent Technologies, Wilmington,DE, USA), equipped with a split/splitless injector. The flame ionisation detector wasoperated at a temperature of 295°C with a hydrogen and air flow of 100 and 400mL/min, respectively. Because of the relatively high column outlet flow of 200mL/min, an increased hydrogen flow (100 mL/min instead of 40 mL/min) was appliedto obtain a stable flame. The data acquisition rate of the FID was set at 50 Hz.Samples of 1 µL (≈ 100 ppm) were injected in the hot split mode at 250°C using anHP 6890 Series Auto-injector. A split flow of approximately 1000 mL/min wasapplied and a septum purge flow of 1.5 mL/min. The system was operated in theconstant pressure mode. The carrier gas was helium. The multi-capillary column ismounted to the injector and detector side with two deactivated fused silica connector


capillaries with a length of 0.1 m and an inner diameter of 250 µm each. HPChemstation was used as data acquisition and processing software.

The multi-capillary column was purchased from Alltech (Deerfield, Il, USA). Thelength of the column was 1 m, and it consisted of 919 capillaries in parallel each withan inner diameter of 40 µm. The stationary phase film thickness was 0.2 µm of SE-30(non-polar phase). The maximum operation temperature of the column (MAOT) was200°C.

All solvents and chemicals (p.a., for analysis quality) were purchased from Merck(Darmstadt, Germany).


With the equations described in section 4.2 chromatograms for single capillarycolumns can be simulated. To simulate chromatograms for the multi-capillary column,chromatograms of individual capillaries are summed. The simulations described hereare performed in order to be able to answer the question what would happen to thebandwidth and retention time of the final peak if the diameter, film thickness andlength of the 900 capillaries in the multi-capillary column show small differences.This question is addressed in more detail in the paragraph below. The calculations areperformed in Microsoft Excel. The time resolution applied was 0.01 sec. In table 4.1the characteristics of the multi-capillary column are listed.

Table 4.1: Nominal characteristics of individual capillaries of the multi-capillarybundle

Ldc

df

1 m40 µm0.2 µm

4.4.1. Influence of variations in diameter on the peak shape

If each of the 900 individual capillaries in the multi-capillary column would beexactly equal, the chromatogram obtained from this column could not be distinguishedfrom that of a single capillary with identical dimensions. If however, the 900

64 Chapter 4

capillaries are not exactly identical, 900 different chromatograms will be obtained.The final chromatogram now will be the sum of the 900 individual chromatograms.

In a first series of calculations the influence of variations in diameter wassimulated. In these simulations it is assumed that the inlet pressure applied to theindividual capillaries is the same and that their lengths are exactly equal. It is furtherassumed that the capillaries have been coated statically and thus the phase ratio isconstant. Here the phase ratio β is defined as:

f

c

s

m

dd

VV

⋅≈=

4β (4.6)

Equation 4.1 can now be rewritten as:

βKk = (4.7)

Equation 4.7 implies that the retention factor k is constant irrespective of the varyingdiameter. The calculations are performed for columns which deviate 1 µm (both plusand minus) from the nominal diameter. The constants used in these calculations arelisted in table 4.2.

Table 4.2: Constants used in the calculations

Constants multi-capillary column

Dm,o

Ds

kβKpi

po

f1

f2

η

3⋅10-5 m2/s6⋅10-10 m2/s4 [-]50 [-]200 [-]514 kPa100 kPa1.09696 [-]0.28287 [-]1.95⋅10-5 kg/(m.s)


As described in the theory section, in order to be able to predict chromatograms tR andσc have to be calculated. From equation 4.2 it can be seen that uo depends on thecolumn diameter dc. Hence for each diameter the elution time tR will be different. Theretention factor k and the distribution coefficient K, on the contrary, remain constant.

In figure 4.1 the effect of a 1.0 µm variation in capillary diameter of some of thecapillaries is illustrated. In the figure the curve for 10× shows the situation that for atotal of 900 capillaries, 10 capillaries have a diameter that is 1.0 µm larger than thenominal diameter of 40 µm, 10 capillaries have a diameter that is 1.0 µm smaller thanthe nominal diameter and the remaining 880 capillaries have the actual “correct”diameter. For each point in time t, h(t) is calculated for all three diameters. Next, thefollowing relation is used to calculate the overall height at time t:

h(t)multicap = 10⋅h(t)39 µm + 880⋅h(t)40 µm + 10⋅h(t)41 µm (4.8)

4.9 5.1 5.3 5.5 5.7 5.9 6.1t [s]

h(t)

[-] 0 x10 x100 x

Figure 4.1: Effect of variations in diameter on the peak-shape. 0× (thin line): theideal situation that all capillaries have exactly the same diameter of 40 µm. 10×(dotted line): the situation that for a total of 900 capillaries, 10 capillaries havediameters which are 1.0 µm larger than the nominal diameter of 40 µm, 10 capillarieshave diameters which are 1.0 µm smaller than the nominal diameter and theremaining 880 capillaries have the actual “correct” diameter. 100× (thick line): thesituation that for a total of 900 capillaries, 100 capillaries have diameters which are1.0 µm larger than the nominal diameter of 40 µm, 100 capillaries have diameterswhich are 1.0 µm smaller than the nominal diameter and the remaining 700capillaries have the actual “correct” diameter.

66 Chapter 4

In the first (predicted) chromatogram three peaks appear. The first peak shows theelution from the 10 capillaries with a diameter of 41 µm, the large second peakcorresponds to the 880 “correct” capillaries and the third peak results from the 10capillaries with a diameter of 39 µm. From figure 4.1 it can be concluded that in caseof non-uniform diameters band broadening occurs. The simulation presented above isan ideal case in which only one variation (1.0 µm) is considered. Although theassumption of e.g. a Gaussian distribution in diameter makes the models somewhatmore complicated, the equations derived above can also be used to simulate such asituation (described in section 4.4.4 of this chapter).

4.4.2. Influence of variations in film thickness on the peak shape

In the second series of simulations the effect of small variations in film thickness inthe 900 capillaries is investigated. In these model calculations it is assumed that alimited number of the capillaries have a film thickness that is 0.01 µm thicker orthinner than the nominal value.

4.9 5.1 5.3 5.5 5.7 5.9 6.1t [s]

h(t)

[-] 0 x10 x100 x

Figure 4.2: Effect of variations in film thickness on the peak-shape. 0× (thin line): theideal situation that all capillaries have exactly the same film thickness of 0.2 µm. 10×(dotted line): the situation that for a total of 900 capillaries, 10 capillaries have filmthicknesses which are 0.01 µm thicker than the nominal film thickness of 0.2 µm, 10capillaries have film thicknesses which are 0.01 µm thinner than the nominal filmthickness and the remaining 880 capillaries have the actual “correct” film thickness.100× (thick line): the situation that for a total of 900 capillaries, 100 capillaries havefilm thicknesses which are 0.01 µm thicker than the nominal film thickness of 0.2 µm,100 capillaries have film thicknesses which are 0.01 µm thinner than the nominal filmthickness and the remaining 700 capillaries have the actual “correct” film thickness.


To a good approximation tM and ū are now independent of the film thickness. Theretention factor k, however, changes with both in- and decreasing film thickness asbecomes evident from equation 4.1. Using the constant distribution factor K, theactual retention factor can be calculated with equation 2. Using the new value of k, tR

and N, σc can now be calculated for each set of capillaries. Similar as described in theprevious paragraph the final total chromatogram can be calculated. The results ofthese calculations are schematically shown in figure 4.2. These calculations once moreshow that high demands are imposed on the repeatability of the individual capillaries.

4.4.3. Influence of variations in length on the peak shape

The last series of calculations addresses the effect of variations in the length of theindividual capillaries. The results of these calculations are shown in figure 4.3. Inthese calculations length variations of 1 mm are considered. Figure 4.3 clearly showsthat such a flaw has no significant consequences for the performance of the multi-capillary column. Very tight manufacturing-specifications have to be imposed on theprecision of the capillary diameter and the film thickness.

4.9 5.1 5.3 5.5 5.7 5.9 6.1t [s]

h(t)

[-] 0 x10 x100 x

Figure 4.3: Effect of variations in length on the peak-shape. 0× (thin line): the idealsituation that all capillaries have exactly the same length of 1.0 m. 10× (dotted line):the situation that for a total of 900 capillaries, 10 capillaries have lengths which are1.0 mm longer than the nominal length of 1.0 m, 10 capillaries have lengths which are1.0 mm shorter than the nominal length and the remaining 880 capillaries have theactual “correct” length. 100× (thick line): the situation that for a total of 900capillaries, 100 capillaries have lengths which are 1.0 mm longer than the nominallength of 1.0 m, 100 capillaries have lengths which are 1.0 mm shorter than thenominal length and the remaining 700 capillaries have the actual “correct” length.

68 Chapter 4

4.4.4. Gaussian distribution

In reality it is more likely that a Gaussian distribution occurs of the variations incolumn diameter, film-thickness and length. In figure 4.4 calculations are performedfor diameters of 40 ± 1µm (1 µm is the standard deviation of the diameter variation).

A general conclusion that can be drawn from the simulation results shown aboveis that a drastic decrease in sensitivity occurs when production of this type of columnis not optimal.

4.8 5.3 5.8 6.3Time [s]

h(t)

[-]

Gaussiandistribution ofd(c)

No variation indiameter

Figure 4.4: Influence of a Gaussian distribution of the variations in column diameter(39µm < dc < 41 µm) on the peak shape.

4.4.5. Fast analyses using the multi-capillary column

In figure 4.5 an example is shown of a fast analysis of n-alkanes obtained with amulti-capillary column. Peak shapes obtained in this chromatogram are verysatisfactory. The minimum detectable amount for dodecane was determined atapproximately 10 pg (S/N = 3) [8]. The sample capacity of this column was 1000 ng.A high dynamic range could be obtained: 105. The linearity of the calibration curveobtained with solutions ranging from 2 mg/mL to 40 µg/mL (duplicate) was 0.9994.The optimum average carrier gas velocity of the column was determined at 100 cm/s.A maximum plate-number of 12,500 could be obtained, which is lower than thetheoretical plate-number of 20,000. Most likely some extra band broadening is causedby irregularity effects of the column as described previously in this chapter, but also


min0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

pA

4000

8000

12000 C8

C10C12

C14 C16

Figure 4.5: Fast analysis of n-alkanes (C8-C16) on a multi-capillary column. Columndimensions: 1 m × 40 µm × 0.2µm SE-30. Inlet-pressure: 375 kPa, column-flow: 200mL/min, split-flow: 1000 mL/min, Toven: 40°C - 120°C/min - 200°C (1 min), Vinj :1 µL(1000 ppm).

min0.8 1 1.2 1.4 1.6 1.8 2

pA

1000

3000

5000

7000C12

C14 C15C16

C17

C18

Figure 4.6: Fast analysis of FAMES (Fatty Acid Methyl Esters, C12-C18) on a multi-capillary column. Inlet-pressure: 375 kPa, column-flow: 200 mL/min, split-flow: 1000mL/min, Toven: 75°C - 120°C/min - 200°C (1 min), Vinj :1 µL (1000 ppm).

70 Chapter 4

some extra injection band broadening can occur. The repeatability (% RSD) of thepeak area and retention time of 6 repetitive injections of dodecane was 6.5 and 0.5,respectively.

In figure 4.6 an example is shown of a fast FAME (fatty acid methyl esters)mixture. The increasing peak width for the last two compounds (C17 and C18) can beexplained by the fact that these compounds were eluting isothermally. From bothchromatograms it can be concluded that the performance of the column is satisfactoryfor mixtures containing compounds with a wide boiling point range. For speeding upthe analysis of a complex mixture however, the multi-capillary column is not suitable.In this case only a narrow bore column will provide the required high separationefficiency and a high speed analysis. However, for a narrow bore column a muchlower linear dynamic range (100 times lower) can be obtained [8].

4.5. CONCLUSIONS

From the results described in this chapter it is clear that very stringent requirementshave to be met for producing multi-capillary columns. From the examples shown inthis work it can be concluded that by using this type of column analysis times can bereduced substantially. The performance of the column in terms of sensitivity,loadability and gain in speed was very satisfactory. However, due to the short lengthof the column and the low plate number this method is found to be suitable forspeeding up analyses of relatively simple mixtures only.

4.6. REFERENCES

1. C.P.M. Schutjes, High speed, high resolution capillary gas chromatography,thesis, Eindhoven University of Technology (1983).

2. A.J.J. van Es, High speed narrow-bore capillary gas chromatography, thesis,Eindhoven University of Technology (1990).

3. P. v. Ysacker, High speed narrow-bore capillary gas chromatography, thesis,Eindhoven University of Technology (1996).

4. M.J.E. Golay, Gas Chromatography, D.H. Desty, ed., Butterworths, London,(1958) 36.

5. J. C. Giddings, S.L. Seager, L.R. Stucki, and G.H. Stewart, Analytical Chemistry,32 (7) (1960) 867.


6. A.T. James and A.J.P. Martin, Biochem. J., 50 (1952) 6797. E.N. Fuller, P.D. Schettler and J.C. Giddings, Ind. Eng. Chem., 58(5) (1966) 19.8. M. v. Lieshout, M. v. Deursen, R. Derks, H.-G. Janssen and C. Cramers, J.

Microcolumn Separations, 11(2) (1999) 155.

72 Chapter 4

CHAPTER 5FAST GAS CHROMATOGRAPHY

USING VACUUM OUTLET CONDITIONS1

SUMMARY

In this study the possibility of operating a wide-bore column with an inner diameter of530 µm at vacuum outlet conditions was investigated as a route towards fast gaschromatography. High analysis speeds can be obtained as a result of the sub-ambientpressure conditions present in the column and the resulting high diffusion coefficient ofthe solute in the mobile phase. The advantage of using a wide-bore column is the highsample capacity. To enable vacuum operation of the column it was coupled to a massspectrometer. For this purpose three methods for restricting the column flow wereinvestigated. The first method uses a narrow-bore pre-column (60 cm length × 100 µmi.d.) coupled to the wide-bore column. The second method relies on the use of a taperedon column restrictor (SFC-restriction). The third method makes use of a fast rotatingmicro-injection valve to enable direct injection of the sample onto the column at sub-atmospheric pressures. From the results of the comparative study it was concluded thatthe performance of a short narrow-bore column as the restriction is somewhat betterthan the use of a SFC-type restriction. Examples of chromatograms in the seconds- andminutes-range are shown, with peak-widths ranging from 0.1 to 2 seconds at halfheight. The influence of the film-thickness to enlarge the sample capacity was studied. Itwas concluded that with a film-thickness above 1.5 µm, the advantage of the increase inanalysis-speed due to low pressures is nullified due to retention time increase on a thickfilm column.

1 This chapter has been published as “Fast gas chromatography using vacuum outlet conditions”, by: M.M. vanDeursen, H.-G. Janssen, J. Beens, P. Lipman, R. Reinierkens, G. Rutten and C. Cramers, J. Microcol. Sep.,12(12), 613-622, 2000.

74 Chapter 5

5.1. INTRODUCTION

In gas chromatography, the application of vacuum column-outlet conditions is anattractive way to increase the speed of analysis. As already demonstrated in chapter 2of this thesis a considerable gain in speed is possible, especially for short and/or wide-bore columns [1-4]. For such columns a gain in speed of a factor of 3-5 can easily beobtained. Moreover, in contrast to other methods for fast GC, the vacuum outlet routehas a significantly increased sample loadability. This mainly arises because wide-borecolumns can be used.

Despite the attractive speed- and loadability characteristics of wide-bore columnsoperated under vacuum outlet conditions, this approach towards faster analysis so farhas received little attention. Most likely, this is a result of the experimental difficultiesassociated with the use of wide-bore columns operated at vacuum outlet conditions.Vacuum outlet conditions are most readily obtained by using a mass spectrometer as thedetection device. Direct coupling of a short and/or wide-bore column to an MShowever, will result in operational problems. Firstly the carrier-gas inlet and theinjection system have to be operated at sub-ambient pressures. Secondly, the highcolumn outlet flow might increase the pressure in the ion source of the massspectrometer to a level exceeding the tolerable limit. Typical (optimum) column outletflow-rates of a wide-bore column with an inner diameter of 530 µm, and a length of 10m are approximately 7-10 ml/min. In general, the maximum pumping capacity of anMS is already reached at a column flow-rate of 5 ml/min. If, on the other hand, thepumping system of the MS does have sufficient capacity to maintain pressure at anacceptable level, a complication will be that the inlet pressure in the injector willdecrease to sub-ambient values. Although advantageous in terms of speed, this mightcause additional practical problems.

If short wide-bore columns are to be used in conjunction with MS instruments, arestriction can be coupled to the column inlet. The flow is now restricted to anacceptable level, the injection system can operate at above-atmospheric pressures andstill low-pressure conditions prevail throughout the entire column. Such a column inletrestriction can be readily obtained by coupling a narrow-bore column (e.g. 60 cm x 100µm) at the inlet position of the analytical column, using a zero dead-volume connector[5]. A disadvantage of such a coupling is the possible occurrence of dead volumes. Analternative solution is the use of a SFC integral tapered restriction prepared from thecolumn inlet. In the latter case the restriction is an integrated part of the column whichimplies that no coupling piece is needed [6]. This restriction system can act as a bench-mark in terms of maximum achievable column efficiencies. Therefore in this work it

Fast gas chromatography using vacuum outlet conditions 75

was investigated whether this SFC-type of restriction could be an alternative in couplinga wide-bore to an MS.

When the capacity of the pumps is high enough, it can also be considered to couplea wide-bore column directly to the MS, without using a pre- or post-column flowrestriction. Sample introduction now has to take place at sub-atmospheric inletpressures. In this work the possibilities and limitations of wide-bore columns at vacuumconditions as a means towards fast GC will be demonstrated. It will be shown that a fastrotating micro-injection valves (injection times < 100 ms), allow very fast injections. Byusing such a valve, injection band broadening can be minimised.

5.2. THEORY

In gas chromatography, band broadening in capillary columns can be accuratelydescribed by the Golay-Giddings equation (equation 4.3 in chapter 4 of this thesis).The Golay-Giddings equation satisfactorily describes the relation between the columnoutlet velocity and the plate height. However, the carrier-gas velocity at column outletconditions is not exactly known because the pressure at the column outlet at vacuumconditions can not be accurately determined. Therefore it is more practical to use theaverage carrier-gas velocity which can be readily assessed experimentally, rather thanthe outlet velocity. Both the velocity u and the diffusion coefficient Dm vary inverselywith pressure:

pDpDpD mim,iom,o == (5.1)

pupupu iioo == (5.2)

Where Dm,i, ui and pi are the diffusion-coefficient, carrier-gas velocity and pressure atinlet-conditions. The average velocity ū can also be expressed as a function of uo andf2 (equation 2.7 in chapter 2 of this thesis).

By using the plate-height equation (4.3), equations 5.1, 5.2 and 2.7, and byneglecting the influence of the stationary phase on band broadening, the followingrelation can be derived:

1611138 2

2

+++=

kkk)(

dDu

c

mopt (5.3)

76 Chapter 5

From this equation it can be seen clearly that for a given column the optimum averagecarrier gas velocity ūopt is proportional to the average diffusion coefficient mD . Fromequation 2.20 (chapter 2 of this thesis) it was already concluded that a high diffusioncoefficient results in short analysis times. Short wide-bore columns can be operated atvery low pressures over the entire length of the column, and therefore provide a higheranalysis-speed than the same column operated at atmospheric outlet-pressures. As acontrast, if long and narrow-bore columns are operated at vacuum outlet conditions,only a fraction of the column length is operated at sub-ambient pressures. This meansthat the gain in speed for long narrow-bore columns with high plate-numbers,operated under vacuum conditions, can be neglected.

5.2.1. Sample loadability

An advantageous effect of wide-bore columns is the high sample loadability of suchcolumns. The maximum sample capacity Cmax (also see equation 2.24) can bedescribed by using the following relation (Ghijssen et al. [7]):

Hddρk

)k(βπC cfs'' ⋅⋅⋅

+= 2

0

20

max

12

5 (5.4)

Where β´´ is a solute-liquid phase specific factor, k0 the retention factor at infinitedilution and ρs is the density of the stationary phase. In this work a similar equationwas used to calculate the sample capacity [8,9]. For thin film capillary columns theplate-height H is proportional to dc [10]. At constant phase-ratio (β = dc/(4⋅df)) thefilm-thickness df is also proportional to the column diameter dc. From these relations itfollows that Cmax is proportional to dc

3 (volume of one theoretical plate). The samplecapacity is thus drastically increased when using wide-bore columns. It might beinteresting to increase the sample loadability even more by increasing the film-thickness. The disadvantage, however, of an increasing film-thickness is theincreasing Cs-term in the plate-height equation (4.3). This means that at a certain film-thickness, the influence of slow diffusion in the stationary phase on the plate heightcan no longer be neglected. In other words, with too high a film-thickness theseparation becomes less efficient and separation time increases. If a column isoperated at vacuum outlet conditions, the stationary phase starts to significantlycontribute to band broadening at much lower thicknesses than in case of atmosphericoutlet operation.


The influence of the film-thickness on the efficiency of a GC separation can bedescribed using the following relation [11]:

)yf(CfCuH

sm,oo,opt

221min 22 ++= (5.5)

In this equation Hmin is the minimum plate height at optimum conditions, uo,opt is theoptimum carrier-gas velocity at outlet conditions, Cm,o is the resistance to masstransfer in the mobile phase at column outlet conditions, Cs is the resistance to masstransfer in the stationary phase and y2 a function of P [11]. At vacuum outlet: P → ∞ ,f1 = 9/8, f2 = 3/(2P) and y2 = -1/2⋅f2 [11]. With f2⋅uo,opt = ūopt (2.7) this leads to:

23

818

2

min sm,o

opt

Cf

CuH += (5.6)

This equation can be rewritten as:

+=

m,o

sm,o

opt CfC

fC

uH 2

2

min

23

818 (5.7)

With the following equations for Cm,o and Cs [12]:

m,o

cm,o D

d)(k

kkC2

2

2

1961611 ⋅

+++= (5.8)

s

fs D

d)(k

kC2

2132 ⋅+

= (5.9)

and together with equations 5.10 and 5.11 [1]:

mm,o DDf =2 (5.10)

32 ipp = (5.11)

78 Chapter 5

equation 5.7 can be rearranged to:

⋅

++

⋅

+++⋅=

s

f

m,

ic

opt Dd

)(kk

Dppd

)(kkk

uH 2

211

2

2

2min

11241611

83 (5.12)

Where p1 and Dm,1 are the pressure and diffusion-coefficient at atmosphericconditions. Together with the following relation,

k)(u

HNt

opt

reqR +

⋅= 1min (5.13)

where Nreq is the required plate number, the total analysis time can be calculated. Withthis equation it can be determined at which film-thickness the negative influence ofslow mass transfer in the stationary phase is larger than the positive influence of lowerinlet pressures. This is shown in figure 5.1.

0

50

100

150

200

0.0 0.5 1.0 1.5 2.0 2.5 3.0

Film thickness [µm]

Ret

entio

n tim

e [s

]

tR

k)(1·

N·D

d·

1)(k

k

s

2f

2

+

+ k)

(1 N·Dp

pd·

1)24(k

16k

11k·

83

m,11

i2c

22

+

+

++

Figure 5.1: Influence of the film-thickness in a wide-bore column (10 m × 530 µm) onthe analysis time (calculated using equations 5.12 and 5.13). Drawn line: totalretention time; dotted line (large): influence stationary phase on analysis time; dottedline (small): influence diffusion in mobile phase on analysis time. Plate-number:20000, compound: nonane, T = 60°C, detector: mass spectrometer. The inlet pressurepi was calculated using the equation described in work of Leclercq [12].


From this figure it can be concluded that, to take full advantage of the gain inspeed when operating a wide-bore column (10 m × 530 µm) at vacuum conditions, thefilm-thickness should not exceed approximately 1.5 µm.

In table 5.1 a comparison is made between the analysis speed on a wide-borecolumn used at vacuum outlet conditions and two narrow-bore columns operated atambient outlet pressures. Results were obtained by using a computer-programdeveloped in house [12]. From these data the conclusion can be drawn that a 50 µmi.d. column with a length of only 80 cm in combination with a flame ionisationdetector, would be faster than a wide-bore column at vacuum outlet conditions, whichhas the same plate-number and phase ratio. This short 50 µm column however, is noteasy to use in practice. Sample-introduction for such a column has to be extremelyfast, only minute sample amounts can be injected and the detector should be highlysensitive and fast. Given the experimental difficulties of working with a short 50 µmcolumn, a comparison with a 50 µm column with a length of 5 meters would be moreappropriate. From a comparison of the results shown in table 5.1, it is evident that theefficiency of the wide-bore column is less than that of the narrow-bore column. Forvery complex mixtures the narrow-bore column clearly is more suitable. From thistable it can also be seen that the sample loadability of a wide-bore column isapproximately 100-1000 times higher than that of a 50 µm narrow bore column. Thecalculated very low sample capacities for the 50 µm column were experimentallyconfirmed by Van Ysacker [13] who found column overloading for 50 µm columns atinjected amounts of only 2-3 ng.

Table 5.1: Comparison between the analysis speed and sample capacity of a wide-bore column operated at vacuum conditions and two 50 µm columns at ambient outletpressures. Calculations were performed using the computer-program developed byLeclercq [12]. Compound: nonane, k = 1.6 for column 1 and 3, k = 7 for column 2,temperature: 60°C.

Columnpo

[kPa]N[-]

ū[cm/s]

tR

[s]Q

[ng]1) 9 m × 530 µm × 0.25 µm2) 5 m × 50 µm × 0.1 µm3) 0.8 m × 50 µm × 0.02 µm

0100100

2000010000020000

12040110

20902

5005

0.5

80 Chapter 5

5.3. EXPERIMENTAL

In this work two wide-bore columns were used. The first column was 9 m long, had aninner diameter of 530 µm and contained a CP Sil 8 CB stationary phase with athickness of 0.25 µm (Varian/Chrompack, Middelburg, The Netherlands). A non-coated restriction column of 60 cm with an inner diameter of 100 µm was connectedto the inlet of the column. The second column was 10 m long, had an inner diameterof 530 µm and a BP-1 stationary phase with a thickness of 3.25 µm (Perkin-ElmerCorporation, Norwalk, Connecticut, USA).

The GC used for the experiments with wide-bore columns under vacuumconditions was an HP6890 (Agilent Technologies, Wilmington, DE, USA) equippedwith a 7673 autosampler. Injections were performed using an OPTIC 2 PTV injector(Atas, Veldhoven, The Netherlands), operated in the hot split mode at an injectiontemperature of 250oC. For the experiments on the wide-bore column with the column-restriction and the SFC-restriction a split-ratio was applied of 1:100. The system wasoperated in the constant pressure mode. The GC used for experiments with wide-borecolumns at atmospheric outlet pressure was a GC 8000 Series (Fisons Instruments,Rodano, Italy). Injections were performed using a split/splitless injector, operated inthe hot split mode (ratio 1:100) at an injection temperature of 250oC. In allexperiments helium was used as the carrier gas.

Injections on the third restriction system were performed with a micro-injectionvalve (Valco Europe, Schenkon, Switzerland). In order to be able to inject the sampleat sub-atmospheric pressures, which is the optimum inlet pressure of a wide-borecolumn, a fast rotating micro-injection valve was used. The micro-injection valvecould be operated at very low pressures, and at temperatures up to 275°C. In order toenable carrier-gas regulation at sub-atmospheric pressures, a non-coated column witha length of 15 m and an inner diameter of 250 µm was installed between the pressureregulator and the carrier-gas inlet of the valve. By applying pressure to the 250 µmcolumn, the inlet pressure of the wide-bore column could be regulated very accurately.The pressure was measured by a precision pressure gauge (Wallace & Tiernan,Günzburg/Do., Germany). All analyses were performed isothermally at 60oC. Theinternal sample loop of the valve has a volume of 60 nL and is actually an engravedconnecting slot on the rotor. The rotor was switched electrically, with switching times(load→inject→load) of less than 100 ms. The valve-loop produces input band widthsof approximately 0.1 ms (equation 2.25). The analyses on the wide-bore columnrequired fast mass spectral data acquisition. Therefore a time-of-flight massspectrometer was used in this work (Pegasus II TOF-MS, LECO, St. Joseph, MI,USA). In all experiments the ion-source temperature and the transfer-line temperature


were kept at 180°C and 275°C, respectively. The pressure inside the ion-source was10-7 Torr. Information about scan-ranges and scan-speeds is given in the figurelegends.

The SFC restriction on the inlet of the wide-bore column was produced by closingthe column-inlet at high temperatures (melting) using a micro-torch flame [6]. Whileapplying a pressure of 1 bar (helium) at the open end of the column, the closed end ofthe capillary tube was abraded manually by using a piece of sandpaper. The sandpaperwas located in a metal beaker filled with water. The occurrence of bubbles indicatedthat the column was open. Then the flow was measured using a soap-film flow-meter.The polishing of the column was continued until a flow of approximately 12 mL/minwas measured. Each time a new orifice had to be produced, the same procedure wasfollowed. In this way reproducible restrictions could be prepared. In figure 5.2 theelectron microscope photograph is shown of the SFC-restriction which was used inthis work. The orifice had an inner diameter of approximately 20 µm.

In figure 5.3 the narrow-bore pre-column, the SFC- [6] and valve-restrictionsystem are schematically shown.

Figure 5.2: SEM-photograph of the in-house produced SFC-restriction of a wide-borecolumn (530 µm), reduced to approximately 20 µm.

n-Hexane and n-heptane (Merck, Darmstadt, Germany) were injected in thevapour-phase. Carrier-gas velocities were measured using methane, which wasinjected simultaneously with the hydrocarbons. Solutions of 250 ppm of n-undecane

100 µmRestrictionopening

82 Chapter 5

(Merck, Darmstadt, Germany) and hexachlorobenzene (HCB) (Janssen Chimica, Geel,Belgium) in hexane were prepared.

TOF-MSInlet

Inlet TOF-MS

A. Column-restriction

B. SFC-restriction

C. Valve restriction

Inlet

TOF-MS

P

Figure 5.3: Schematic drawing of the three restriction types: (A) narrow-bore pre-column (Varian/Chrompack), (B) SFC-restriction (produced in house) and (C) valverestriction. Restriction column at inlet: 15 m × 250 µm (uncoated).


As mentioned previously, three types of restrictions were used to control the columnflow and to enable coupling of the wide-bore columns to a mass spectrometer: anarrow bore pre-column, a SFC-restriction and a micro-injection valve. For the SFCrestriction no coupling is needed, this restriction is directly fabricated on the column.This means that leaks can not occur and dead-volume problems are absent. Thedisadvantage of this type of restrictor is that it is a rather time-consuming task toproduce. The retention time repeatability obtained with the SFC-restriction was betterthan 0.5% RSD. The area repeatability varied between 5 and 10% (measured forhexachlorobenzene and undecane). Results are shown in table 5.2.

Internal sampleloop: 60 nL


Table 5.2: Repeatability (RSD) and mean values of the retention-time (seconds) andpeak-area of analyses performed on the thin film wide-bore column using a SFC-restriction, and analyses performed on the thick film wide-bore column using themicro-injection valve (both vacuum outlet-conditions). n = Number of analyses.Compounds: hexachlorobenzene (HCB), undecane (C11), hexane and heptane.

SFC-restriction Micro-injection valve

HCB C11 Hexane Heptane

n = 10 Area tR Area tR Area tR Area tR

Mean

RSD %

1145737

8.7

285.48

0.1

5879070

10.1

81.66

0.3

30221

8.8

31.84

0.3

28223

13.8

55.72

0.2

A high efficiency can be obtained only by minimising band broadening caused bythe injection system. This means that sample introduction has to be very fast. For thetwo restriction methods using the narrow-bore pre-column and the SFC-inlet, sampleintroduction was performed in the split mode at high split flows (approximately 10 to200 mL/min). The split-ratio is around 2 to 100. In the third (valve) method, sampleintroduction took place by electrically actuating the sample valve. The performance ofthis valve for use in high-speed separations on narrow-bore columns has been studiedpreviously [14]. Due to the extremely small loop volume and the short valve rotationtime, injection pulses and input band widths (σi) are in the milliseconds range and thushave a negligible contribution to the total band width. Because the valve can beoperated at temperatures up to 275°C, it can well be used as an injection device forfast GC. However, only gaseous samples can be injected, which means that the rangeof compounds that can be analysed, unfortunately, is restricted. In figure 5.4 theperformance of these three restriction methods are shown. The average carrier-gasvelocities that were observed are approximately 200 cm/sec. In table 5.2 it is shownthat the repeatability for injections with the micro-injection valve for the compoundshexane and heptane is acceptable.

The performance of the three restriction methods in terms of speed and efficiencyare comparable. The maximum theoretical plate-number of a wide-bore column of 10meters is approximately 20,000 plates (table 5.3). As expected, with the SFC-restriction the maximum attainable plate-number was obtained. When using a narrow-bore pre-column as restriction method this efficiency is also obtained. Dead-volumeproblems hence are absent. The minimum plate-height and optimum gas velocity forthe micro-injection valve was higher than the corresponding values for the narrow-bore pre-column and SFC-restriction.

84 Chapter 5

Table 5.3: Comparison of the optimum average carrier-gas velocity (ūopt) , plate-number and inlet-pressure for a thick-film (3.25 µm) and a thin film (0.25 µm)column. Experimental conditions for the thick- and thin-film wide-bore column atvacuum conditions: injection with micro-injection valve, Toven = 60°C, injection-volume ≈ 50 µL vapour, split flow = 500 mL/min, Tinj = 250°C, compound = hexane.Experimental conditions for the thick- and thin-film wide-bore column at atmosphericconditions: no restriction used, Toven = 60°C, injection-volume = 50 µL vapour,splitflow = 500 mL/min, Tinj = 200°C, Tdet,FID = 200°C, compound = hexane.

Calculated Observed Calculated ObservedCompound: hexaneColumn: 10 m × 530 µm

Vacuumoutlet

Vacuumoutlet

Atmosphericoutlet

Atmosphericoutlet

ūopt , df = 0.25 µm [cm/s] ūopt , df = 3.25 µm [cm/s]Nmax , df = 0.25 µm [-]Nmax , df = 3.25 µm [-]pi,opt , df = 0.25 µmpi,opt , df = 3.25 µm

17080

2000070005525

200100

200005000

6030

6020

2200015000

110105

5818

2500013000

115110

0.0

0.5

1.0

1.5

2.0

2.5

3.0

0 50 100 150 200 250 300Average carrier-gas velocity [cm/s]

Plat

ehei

ght [

mm

]

valveSFCColumn

Figure 5.4: Comparison of the experimental HU-curves of a thin film wide-borecolumn coupled to a TOF-mass spectrometer using the three restriction devices at thecolumn inlet side. �: Narrow-bore pre-column-restriction; �: SFC-restriction; �:micro-injection valve. Compound: hexane, k = 0.3, temperature: 60ºC, column: 9 m ×530 µm × 0.25 µm CP Sil 8 CB. Scanned mass-range: 40-200; acquisition-rate: 50spectra/second.


Contrary to what was expected, the performance of the valve injection was notbetter than that of the other two methods. The plate number obtained with theinjection valve was approximately 30% lower than that obtained with the other tworestrictor types. Most likely this is caused by extra column band broadening effects inthe coupling of the column to the valve (dead volume or a poor column cut). On thebasis of these results it can be concluded that the best restriction method is the use of anarrow-bore pre-column connected to the analytical column using a low dead-volumecolumn connector. For compounds that might exhibit interaction with the connectormaterial, the use of a SFC integral tapered restrictor is a good alternative. The plate-number that was obtained ranged from 17000 to 20000 plates at average carrier-gasvelocities of 150 to 180 cm/s. In figure 5.5 an example is shown of analyses in theseconds range, producing peak-widths at half height of approximately 100 ms. Forthese type of chromatograms a fast TOF-MS (> 50 full scans/s) is required. Thesample analysed is a mixture of methane, hexane and heptane (vapour).

0 10 20 30 40 50 60

Tim e [s]

A

B

1

2

3

1

23

4.3 4 .4 4 .5 4 .6 4 .7 4 .8 4 .9 5 .0Tim e [s ]

1

Figure 5.5: Difference in analysis speed using a thin film (0.25 µm) wide-bore column(A) and a thick film (3.25 µm) wide-bore column (B). Injections were performed usingthe micro-injection valve. Compounds: (1) methane, (2) hexane and (3) heptane, oven-temperature = 60°C. Scanned mass-range: 40-200; acquisition-rate: 50 spectra persecond. In the enlargement (peak no. 1) the number of datapoints in one peak isshown.

86 Chapter 5

140 190 240 290 340 390 440Time [s]

Mas

s 61

1

7

3

2

6

5

4

98

ISTD

Figure 5.6: GC-TOFMS analysis of an industrial derivatised alcohol-standard,dissolved in ethylacetate, ranging from C9 to C17 (compound no. 1 to 9: acetic acid,alkyl esters + internal standard) using a wide-bore column with SFC restriction.Inlet-pressure = 3 bar, split-flow = 16 mL/min, Toven = 70°C (2 min) → 20°C/min →240°C, Vinj = 1 µL (autosampler). MS-conditions: scanned mass-range = 40-400,acquisition-rate = 25 spectra/s.

10 15 20 25 30 35Time [s]

TIC

1

7

6

5

4

3

2

Figure 5.7: GC-TOFMS analysis of gasoline (pure) using a wide-bore column withSFC restriction. Inlet-pressure = 4 bar, split-flow = 50 mL/min, Toven = 60°C →20°C/min → 150°C, Vinj < 0.1 µL (manual). MS-conditions: scanned mass-range =41-400, acquisition-rate = 50 spectra/s. Compounds: (1) 2-methyl-butane; (2) 2-methoxy-2-methylpropane (MTBE); (3) benzene; (4) toluene; (5) ethylbenzene; (6)m+p-xylene and (7) o-xylene.


In figure 5.6 and 5.7 examples are shown of slightly slower analyses in theminutes range, for which a scanning MS (quadupole or ion-trap) would be fastenough. These are analyses of an industrial alcohol-standard (figure 5.6) and of BTEXin gasoline (figure 5.7) using a wide-bore column with SFC restriction. Analysis timesare 8 minutes and 35 seconds, respectively. Analysing these samples on aconventional system with longer normal-bore columns using atmospheric outletconditions would take approximately 3-5 times longer. The use of a wide-bore columncoupled to a mass spectrometer is therefore very suitable for speeding up the analysis-time of simple mixtures. The working range of a wide-bore column coupled to TOF-MS ranges from 50 pg to 50 ng.

5.4.1. Film-thickness

The advantage of using thick film columns is the higher sample loadability comparedto thin film columns. However, as was mentioned previously, the analysis speed willbe reduced drastically when the film-thickness of the wide-bore column exceeds acertain threshold value. It was deduced from theory that at film-thicknesses above 1.5µm, retention times increase too much to take full advantage of the vacuum conditionsin the column. In figure 5.5 a comparison is shown of the analysis-speed of a thickfilm wide-bore column (df = 3.25 µm) versus a thin film column (df = 0.25 µm) withan identical diameter. Typical plate-numbers of a thick film and a thin film wide-borecolumn operated at vacuum outlet conditions are 5000 and 20000 plates respectively.More details of the comparison are shown in table 5.3. In this table calculated valuesof two wide-bore columns coated with a thick and a thin film of stationary phase arecompared with observed values. In the table also calculated values of a wide-borecolumn operated at atmospheric outlet conditions are inserted. The data given in thetable allow making a comparison between operation of a wide-bore column at vacuumconditions and the same column operated at atmospheric outlet conditions. From thistable it can be concluded that the analysis speed of a wide-bore column at vacuumconditions is approximately 3 to 5 times higher than that of a wide-bore columnoperated at atmospheric conditions.

A distinct disadvantage of wide-bore columns is that the plate-number is not veryhigh. This system therefore is not very suitable for complex separations. For mixtureswith a low number of compounds and a wide boiling point range (figure 5.6), thesewide-bore columns operated at vacuum outlet conditions can result in a significantlyfaster separation. In figure 5.8 HU-curves are compared based on experimental and

88 Chapter 5

calculated data for thick and thin film columns. The experiments shown in figure 5.8were performed with the micro-injection valve. From this figure it follows that theperformance of the column is somewhat lower than can be predicted by calculations.

0.0

1.0

2.0

3.0

4.0

5.0

6.0

0 100 200 300 400Average carrier-gas velocity [cm/s]

Plat

e-he

ight

[mm

] C6, df = 3.25µmC6, df = 0.25µmCalculated 3.25µmCalculated 0.25µm

Figure 5.8: Comparison of experimental (valve-injection) and calculated HU-curvesfor a thick-film wide-bore column and thin-film wide-bore column at vacuum outletconditions. �: film-thickness of 3.25 µm; �: film-thickness of 0.25 µm; compound:hexane. Calculations are performed using in-house developed software [12].

5.5. CONCLUSIONS

Operation of a wide-bore column at vacuum outlet conditions results in a significantlyfaster analysis. Especially the coupling of wide-bore columns directly to a massspectrometer is an attractive route. Such a coupling can only be realised byincorporating some modifications to enable sub-atmospheric carrier-gas control andthus to increase the diffusion coefficient of the solutes in the mobile phase. Once theseconditions are fulfilled, very fast analyses and high sample loadabilities can beobtained. The different ways to adapt the system to these conditions show acomparable performance, so that it is a matter of practical implications which of theadaptations to choose. The plate-number, however, of approximately 20,000 of the


thin-film wide-bore columns operated at vacuum outlet conditions used in this work,is only suitable to separate relatively simple mixtures containing a low number ofcompounds. Compared to analyses operated at atmospheric outlet pressures by usingvacuum outlet pressures a gain in speed of approximately 3 to 5 could be obtained.

5.6. REFERENCES

1. C.A. Cramers, G.J. Scherpenzeel, P.A. Leclercq, J. Chromatogr. A, 203 (1981)207.

2. M.M. van Deursen, H.-G. Janssen, J. Beens, P.A. Leclercq, C.A. Cramers, J.Chromatogr. A., 878 (2000) 205.

3. A. Amirav, N. Tzanani, S.B. Wainhaus, S. Dagan, Eur. Mass Spectrom., 4 (1998)7.

4. H. Smith, E.T. Zellers, R. Sacks, Anal. Chem., 71 (1999) 1610.5. J. de Zeeuw, J. Peene, H.-G. Janssen, X. Lou, 21st Int. Symposium Capillary

Chromatogr. and Electrophoresis, Park City, Utah, June 20-24, Proceedings, 16(1999).

6. E.J. Guthrie, H.E. Schwarz, J. Chromatogr. Sc., 24 (1986) 236.7. R.T. Ghijssen, H. Poppe, J.C. Kraak, P.P.E. Duysters, Chromatographia, 27(1/2)

(1989) 60.8. P.A. Leclercq, J. High Resolut Chromatogr., 15 (1992) 531.9. C.A. Cramers, P.A. Leclercq, CRC Critical reviews in Analytical Chemistry, 20(2)

(1988) 117.10. C.P.M. Schutjes, High speed, high resolution capillary gas chromatography,

thesis, Eindhoven University of Technology, The Netherlands (1983).11. C.A. Cramers, F.A. Wijnheymer, J.A. Rijks, J. High Resolut. Chromatogr./

Chromatogr. Comm., 2 (1979) 329.12. P.A. Leclercq, C.A. Cramers, J. High Resolut. Chromatogr./ Chromatogr. Comm.,

8 (1985) 764.13. P.G. van Ysacker, H.-G. Janssen, H.M.J. Snijders, P.A. Leclercq, C.A. Cramers, J.

Microcol. Sep., 5 (1993) 413.14. A.J.J. van Es, High speed narrow bore capillary gas chromatography, thesis,

Eindhoven University of Technology, The Netherlands (1990).

90 Chapter 5

CHAPTER 6EVALUATION OF

TIME-OF-FLIGHT MASS SPECTROMETRIC

DETECTION FOR FAST GAS CHROMATOGRAPHY1

SUMMARY

Separations below one second of a mixture of organic compounds ranging from C5 toC8 have been performed to investigate the performance of a time-of-flight massspectrometer in fast gas chromatography. The gaseous samples were focused on acold trap, and then re-injected by thermal desorption to obtain the required narrowinput band widths. Also, to obtain a very fast separation, a short narrow bore columnwas used operated at above optimum inlet pressures. With this system it was possibleto identify 10 compounds within 500 ms, showing peak-widths (2.354⋅σ) as narrow as12 ms. The spectral acquisition rate used for these analyses was 500 Hz. The qualityof the recorded spectra and their similarity with library spectra was very high.Deconvolution algorithms offer the possibility to identify overlapping peaks. It isshown that the spectral acquisition rate of the TOF-MS is high enough for very fastseparations.

1 This chapter has been published as “Evaluation of time-of-flight mass spectrometric detection for fast gaschromatography”, by: M.M. van Deursen, J. Beens, H.-G. Janssen, P.A. Leclercq, C.A. Cramers, J. Chromatogr.A., 878, 205-213, 2000.

92 Chapter 6

6.1. INTRODUCTION

The combination of a chromatographic separation technique with mass spectrometricdetection is a very powerful tool for the study and identification of organic compoundsin complex samples. Also in case the identity of a compound is known, it can beadvantageous to use MS detection. With MS-detection target peaks can readily beidentified in crowded chromatograms. By using extracted ion traces, non-separatedpeaks can even be quantified. Finally, MS detection greatly simplifies methoddevelopment as compounds of interest can easily be identified in groups of interferingpeaks. In gas chromatography mass spectrometry is nowadays widely used as adetection device. Several types of mass spectrometers are available for coupling to GCsystems. These systems differ in the way that the ion-fragments, formed from themolecules eluting from the GC column, are separated according to their mass. Importantmass analysers are the ion trap, the sector instrument, the quadrupole and the time-of-flight mass spectrometer. The resulting mass spectrometers show differences in terms ofacquisition rates, detection limits, mass spectrometric resolution and quality of the massspectra obtained. The choice of the most suitable MS is very much dependent on thecomposition of the sample, the detection limits, and the speed of separation. Currently,there is a clear trend towards faster methods for analysis. Several methods for faster GChave been described in literature [1-7] and in the previous chapters. Other importanttrends in GC are the ever increasing need for positive identification and the need formore flexible systems that allow the analysis of a wide variety of samples on onesystem. These last two trends clearly result in a strong requirement for massspectrometric detection. Combination of fast GC with mass spectrometric detection isby no means trivial.

For an accurate description of a chromatographic peak in a chromatogram, at least15-20 datapoints across the peak are required [8-9]. Typical acquisition rates ofscanning mass spectrometers like the ion trap, the quadrupole and the sectorinstrument, range from 10 to 20 spectra per second in the full scan mode. Hence, onlychromatographic peaks with a width of 0.5 second or more can be accuratelyrepresented.

The most important method to achieve fast GC is the use of columns with areduced inner diameter (50-150 µm) [1-2]. Unless very short columns are used, thepeak-widths obtained from such columns are usually slightly above 1 second.Scanning mass spectrometers are hence capable of offering the speed required for MSdetection. In table 6.1, peak-widths are calculated for analyses on a standard column, anarrow bore column and two extremely short columns. Here, the influence on band

Evaluation of time-of-flight mass spectrometric detection for fast gas chromatography 93

broadening of only the column itself is taken into account. From this table it is clearthat for very fast separations on short columns the spectral acquisition rate of scanningmass spectrometers is too low.

Table 6.1: Comparison of calculated plate number, retention-time and peak-width atbaseline level for a standard, fast, very fast and ultra fast separation. Calculations areperformed using the Golay-Giddings equation, incorporated in the computer programdeveloped by Leclercq [15]. Conditions for these calculations are: β = 62.5,compound = hexane, T = 330 K, carrier-gas = helium, peak width at half-height(2.354⋅σ).

Type of analysis Column innerdiameter[µm]

Columnlength[m]

Platenumber

Retentiontime[s]

Peak-width athalf height[s]

po* = 100 kPaStandardFastVery fastUltra fast

po = 0 kPaStandardFastVery fastUltra fast

320 50 50 50

320 50 50 50

251010.3

251010.3

90 000260 000 25 000 7000

75 000260 000 24 000 6500

160602.00.40

100602.00.30

10.20.030.01

0.70.20.030.01

* po is the pressure at column outlet conditions.

The use of an MS compared to using an atmospheric outlet like an FID willsignificantly increase the speed of analysis (chapter 5 of this thesis). The highest gainin speed is obtained when using short wide bore columns. Very short columns incombination with MS can provide separations in the seconds- or even sub-secondsrange, yielding peak-widths of 5-15 milliseconds. For these separations an extremelyfast spectral acquisition rate is required. Time-of-flight mass spectrometers canprovide up to 500 full spectra per second. In this work separations in the sub-secondsrange were performed using a very short narrow bore column. The primary goal ofthis work was to investigate the performance of the TOF mass spectrometer in high-speed separations and to explore the limits of TOF mass spectrometric detection.

The terms fast GC, very fast GC and ultra fast GC are often used in literature, butuntil now they have not been well defined with regard to analysis times and peak-

94 Chapter 6

widths. Dagan and Amirav [10] already defined a speed enhancement factor to divideanalyses in the three fast GC categories. This factor is the increase in speed that can beobtained by using a shorter column and a higher carrier gas velocity in comparison tothe same analysis on a conventional GC column under optimum carrier gas velocityconditions.

In this chapter a classification is made based on peak-widths (2.354⋅σ) and totalanalysis-times. This is shown in table 6.1.1) Fast GC: separation in the minutes range; peak-width: several seconds.2) Very fast GC: separation in the seconds range; peak-width: 30-200 ms.3) Ultra fast GC: separations in the sub-seconds range; peak-width: 5-30 ms.As already mentioned before, fast separations typically can be obtained from columnswith an inner diameter of 50-150 µm. Very fast separations can be obtained by usingshort columns of about 1 m with inner diameters ranging from 50-320 µm. Typicalanalyses are shown in work of Dagan and Amirav [10] and of Davis et al. [11].Typical ultra fast analyses are separations in the milliseconds range, with peak-widthsranging from a few to 10 ms. Jonker and Poppe [12] obtained ultra fast separations ona very short packed column with very small particle diameters. In their work ultra fastseparations were shown with peak-widths of 4 ms and a separation time of 150 ms forthe separation of four compounds. In the present chapter an example will be shown ofa separation of 10 compounds in 500 milliseconds, and the applicability of the TOF-MS for GC separations in the 0.01 to 1 seconds range is investigated.

6.2. THEORY

In 1981 Cramers et al. [13] derived an equation describing the gain in speed ofanalysis using vacuum outlet as compared to atmospheric outlet operation. Thisequation reads:

( )( ) 2/32

,,2

3,,

3

,

,

,

,

atmatmopti

atmatmopti

vacopt

atmopt

atmopt

vacopt

pppp

pp

uuG

−−=== (6.1)

where: ūopt,vac and ūopt,atm are the average optimum gas velocity through a capillarycolumn at vacuum outlet and atmospheric outlet conditions, respectively; vacoptp , and

atmoptp , are the average optimum column pressure at vacuum outlet and atmosphericoutlet conditions; pi,opt,atm is the optimum inlet pressure at atmospheric outlet


conditions and patm is the atmospheric pressure. From this equation it is clear that forshort wide bore columns the gain in speed is much higher than for long narrow borecolumns. For the short wide bore column the pressure inside the column is low overits entire length. The resulting high diffusion coefficients result in a high speed ofanalysis as is shown in figure 6.1. A complication factor when using a short wide borecolumn with MS detection, is the high resulting column outlet flow. This flow caneasily be as high as 10 ml/min, which is too high for the pumps to maintain thevacuum in the system at an acceptable level. Summarising, there are various optionsfor fast GC separations. For the present purpose, assessing the limits of TOF-MSdetection, we opted for the use of short 50 µm columns operated at above-optimuminlet pressures to achieve the highest separation speed possible.

0

2

4

6

8

10

12

14

16

0 100 200 300 400 500 600Column diameter [µm]

Gai

n in

spe

ed

L = 20 m

L = 0.3 m

L = 5 m

Figure 6.1: Influence of column diameter and column length on gain in speed using avacuum outlet compared to an atmospheric outlet.

6.3. EXPERIMENTAL

The GC used in this work was an HP6890 (Hewlett Packard, Wilmington, DE, USA)equipped with a Gerstel PTV CIS-4 injector (Gerstel, Mülheim a/d Ruhr, Germany).Injections were performed in the hot split mode, at an injection temperature of 250oC.The system was operated in the constant pressure mode with an inlet-pressure of 450kPa. The injections took place at a split-ratio of 1:100. The analyses were performedisothermally at 75oC, unless stated otherwise. The column used was a 30 cm x 50 µm

96 Chapter 6

capillary column with a 0.17 µm film of a non-polar OV-1 stationary phase (MEGA,Fisons, Milan, Italy). Helium was used as the carrier-gas.

The time-of-flight mass spectrometer used was the Pegasus II MS (LECOCorporation, St. Joseph, MI, USA). This is a reflectron-type TOF equipped with anelectron impact (EI) ionisation source. After ionisation the fragments are pushed intothe flight tube by a push pulse electrode, at a rate of 5000 pulses per second. Themaximum possible data-rate obtainable with the TOF is 500 spectra per second, whichmeans that a single acquired spectrum consists of a sum of 10 transients. The totallength of the flight path of the ions is approximately 1 meter. The upper mass rangefor detection in the Pegasus TOF is m/z = 1000. The flight time of this ion will beapproximately 170 µs. The temperature of the transfer-line between the GC and theMS was maintained at 275oC. The ion source temperature was 200oC. The pressure inthe system was 10-8–10-7 Torr, which was achieved by two turbo-molecular pumps,one located at the ion-source and the other one at the reflectron side of the flight tube.Data processing was performed using LECO Pegasus II software, version 1.10.

Headspace samples (1 µL) were injected manually with a 10 µL syringe. Theinjected amount was approximately 1 ng for each compound. After injection thecompounds were trapped and cryo-focused on the head of the column by using a cold-trap, similar to the system described by van Es et al. [14]. A schematic drawing of thetrap is shown in figure 6.2. Helium, cooled to about -70oC by passing it through liquidnitrogen, was used as a cooling medium. Re-injection of the trapped sample tookplace by flash-heating a metal capillary surrounding the fused silica column using apulsed resistive heating device made in house. By putting a voltage of about 11 Vacross the metal capillary, heating rates of 4000oC per second could be reached [14].The final temperature of the trap was approximately 165oC. The flow of cooling gaswas not interrupted during the heating step.

6.3.1. Reagents

Pentane, hexane and heptane were purchased from Merck (Darmstadt, Germany),octane, 2,3-dimethylbutane, 1,4-dimethylcyclohexane (cis- + trans-) and methyl-cyclohexane from Fluka (Buchs, Switzerland) and toluene and benzene fromPolyscience Corporation (Niles, IL, USA). All chemicals had a purity higher than99%, except for pentane, which had a purity of 95%.


To mass-spectrometer

Column: 30 cm × 50µm

��

��

��

��

Helium in,T = 20oC

Helium out,T = -70oC�� Cold trap

Pulsed resistive heating device

��

Metal capillary

LiquidNitrogen

��

��

Figure 6.2: Schematic diagram of the cryogenic focusing inlet system.


6.4.1. Cryogenic focusing inlet system

In previous work a description is given of a cryogenic focusing inlet system capable ofgenerating input band widths in the milliseconds range [14]. An example of a very fastanalysis is shown in figure 6.3. This chromatogram shows the separation of 10compounds in 500 ms. The plate number of the column was approximately 3500which is close to the theoretical value. This chromatogram was recorded from mass 40to mass 200 at 500 spectra per second, which is the maximum acquisition rate of theTOF. Peak-widths obtained in this chromatogram are approximately 12 ms. Similarresults have been obtained in work of van Es et al. [14], by using an FID as thedetection method. Peak-widths were obtained of 10 to 20 ms, for compounds rangingfrom C6 to C8, which is comparable to the values obtained with the mass spectrometer.In table 6.2 the peaks (figure 6.3) are identified and the similarity with library spectrais shown. From this table it is clear that even for such extremely fast separationsidentification of the compounds can take place with an excellent library similarity formost of the compounds. This means that the quality of the collected spectra is veryhigh. An example of the good spectral quality of the TOF is shown in figure 6.4. Thisfigure shows a comparison between a recorded spectrum and the library spectrum.

98 Chapter 6

0.1 0.2 0.3 0.4 0.5Time [s]

Abun

danc

e

1

2

3

4 5

6

7

8 9 10

Figure 6.3: Ultra fast analysis of ten compounds within 500 milliseconds on a 0.3 m ×50 µm column with a 0.17 µm thickness non polar OV-1 phase, using a cryogenicfocusing inlet system. Inlet pressure = 450 kPa, carrier gas: helium, split flow = 400ml/min, Tinjector = 250oC, Toven = 75oC. Injection and sampling: 1 µL, headspace (≈ 1ng/compound). Detection: time-of-flight mass spectrometer, acquisition-rate = 500spectra/second, from mass 40 to 200. Compounds: (1) pentane; (2) 2,3-dimethylbutane; (3) hexane; (4) benzene; (5) heptane; (6) methylcyclohexane; (7)toluene; (8) trans-1,4-dimethylcyclohexane; (9) octane; (10) cis-1,4-dimethylcyclohexane.

Table 6.2: Library search results (%) from the analysis shown in figure 6.3.

Peak number Compound name Similarity12345678910

PentaneButane, 2,3-dimethyl-HexaneBenzeneHeptaneCyclohexane, methyl-TolueneCyclohexane, 1,4-dimethyl-(trans)OctaneCyclohexane, 1,4-dimethyl-(cis)

93.991.092.795.792.192.789.089.890.779.8


Figure 6.4: Example of a measured spectrum compared to a NIST-library spectrum,taken from the ultra fast analysis shown in figure 6.3 (the data-acquisition was startedbefore the injection).

Table 6.3: Comparison of experimental data and calculated values, for a short 50 µmi.d. column and a short 320 µm i.d. column.

dc = 50 µm,l = 0.3 m

dc = 320 µm,l = 0.3 m

Experimental Optimumcalculated

Optimumcalculated

Absolute inlet pressure [kPa]Average gas velocity [cm/s]Number of platesColumn flow [mL/min]

550 2003600 0.8

270 1509000 0.3

55 6002400 10

In table 6.3 the experimental plate heights, optimal velocities etc. for the 0.3 m ×50 µm column are compared to theoretical values. The data for the 50 µm column arecompared to calculated values for a 320 µm column. Calculations were performedusing a computer program developed by Leclercq [15]. In the experimental work thecolumn was operated at an inlet pressure of 550 kPa. From table 6.3 it can be seen that

100 Chapter 6

this is significantly above the optimum pressure of 270 kPa. A pressure exceeding theoptimum value was selected for speed reasons. The low plate number (3600 comparedto 9000) is also a consequence of operating the column at above optimum velocities.

The column outlet-flow, reduced to atmospheric pressure, was calculated usingequation 6.2 and 6.3.

( )atmPPurQ /2π= (6.2)

32

2

io pfpp == (6.3)

where: Q is the column outlet flow; r is the column radius; p is the average columnpressure; f2 is the pressure correction factor (= 3/(2P) for P → ∞ at vacuum outlet),with P = pi/po as the ratio of inlet to outlet pressure.

From the flow calculations it can be concluded that a very short 320 µm columncan only be coupled to a mass spectrometer when using a narrow bore restriction.Recently it was shown that optimum performance is obtained if this restriction ismounted at the column inlet [16]. Another possibility would be to split the columnflow before entering the mass spectrometer, but this will decrease sensitivity.

The use of mass spectrometric detection in chromatography offers an additionalmeans of distinguishing different solutes. Non separated peaks can be “separated” onthe basis of mass spectral differences. Mass spectrometry hence offers an alternativeroute towards faster analysis. Spectral deconvolution of the Leco Pegasus II softwareoffers the possibility to deconvolute and identify overlapping peaks. Somechromatographic separation is however required for the deconvolution algorithm torecognise the presence of two or more compounds in one peak [17, 18]. Figure 6.5shows the identification of the co-eluting compounds octane and cis-1,4-dimethylcyclohexane (see figure 6.3). The separation here amounts to 10 ms or 5spectra (rectangle). Figure 6.6 shows the influence of the spectral collection rate onthe peak resolution and the deconvolution performance of the software. The sampleanalysed is the same as shown in figure 6.3. At 500 spectra per second, the first threepeaks are baseline separated. If now the same separation is repeated at an acquisitionrate of only 50 spectra per second, these three peaks can no longer be distinguished. Intable 6.4 the similarity index of the experimentally spectra collected at 500 spectra persecond is compared to that of spectra recorded at 50 spectra per second. At 50 spectraper second only 3-5 spectra are collected over the peak. This means that it is no longerpossible to identify or deconvolute overlapping peaks. At 50 spectra per second all the


0.40 0.42 0.44 0.46 0.48Time [s]

Abun

danc

e

mass 85mass 97

10. cis-1,4-Dimethylcyclohexane

9. Octane

Figure 6.5: Deconvolution of octane and cis-1,4-dimethylcyclohexane, peak number 9and 10 of the ultra fast separation shown in figure 6.3. In this graph the uniquemasses 85 and 97 are plotted, which are specific for the two compounds.

Table 6.4: Similarity (%) to library spectra for analyses recorded at 500 and 50spectra per second. (a) automatically; (m): manually.

Name 500 spectra/s 50 spectra/s

1

2

3

4

5

6

7

8

Pentane

2,3-Dimethylbutane

Hexane

Benzene

Heptane

Methylcyclohexane

Toluene

Octane

94.1 (a)

91.9 (a)

93.8 (a)

96.2 (a)

93.5 (a)

92.7 (a)

90.4 (a)

92.2 (a)

85.1 (m)

92.0 (a)

94.2 (m)

94.9 (a)

89.9 (m)

92.6 (a)

84.9 (m)

92.6 (a)

102 Chapter 6

500 spectra/second

0E+00

1E+05

2E+05

3E+05

4E+05

0.1 0.3 0.5 0.7

Time [s]

50 spectra/second

0E+00

1E+06

2E+06

3E+06

4E+06

5E+06

0.1 0.4 0.6 0.9 1.1

Time [s]

12

3

4 5

6

7

8

Abu

ndan

ce

1

2 3

45

6

7

8

Figure 6.6: Comparison of chromatograms recorded at 500 and 50 spectra persecond respectively. Analysis conditions: see figure 6.3, Toven = 60oC. Compounds: 1)pentane; 2) 2,3-dimethylbutane; 3) hexane; 4) benzene; 5) heptane; 6)methylcyclohexane; 7) toluene; 8) octane.

compounds could be identified, but for 50% of the compounds this had to be donemanually. Deconvolution certainly is a helpful tool in the separation of complexsamples and for the identification of co-eluting peaks. Spectral deconvolution methodshowever should be applied with some care.

From figure 6.6 it can be seen that the sensitivity increases at decreasingacquisition rate. When acquiring data at a speed of 500 spectra per second only 10transients are summed. At an acquisition rate of 50 spectra per second 100 transientsare summed and thus the signal intensity increases ten times. Signal theory states thatthis results in a gain in signal to noise ratio of √10. In conclusion, the optimumacquisition speed is a compromise between preserving resolution and maximisingsensitivity.

6.5. CONCLUSIONS

From the results presented here it can be concluded that TOF-MS is very suitable as adetection method for very fast separations. With TOF-MS, it is possible to accuratelydetect peaks with peak-widths in the milliseconds range. Even for peaks as narrow as12 ms, the quality of the spectra is very high. Deconvolution techniques can be used toadvantage, but nevertheless some caution has to be taken into account.

The limits of the system have been explored, and at even faster separations morethan 500 spectra are required. Such fast analyses are not yet performed on a dailyroutine base, but might become of more interest in the future.


6.6. REFERENCES

1. P.G. van Ysacker, J. Brown, H.-G. Janssen, P.A. Leclercq and A. Phillips, J. HighResol. Chromatogr., 18 (1995) 517.

2. P.G. van Ysacker, H.-G. Janssen, H.M.J. Snijders, P.A. Leclercq and C.A.Cramers, J. Microcol. Sep., 5 (1993) 413.

3. H. Wollnik, R. Becker, H. Götz, A. Kraft, H. Jung, C.-C. Chen, P.G. van Ysacker,H.-G. Janssen, H.M.J. Snijders, P.A. Leclercq and C.A. Cramers, Int. J. MassSpectrom. Ion Processes, 130 (1994) L7.

4. M. v. Lieshout, M. v. Deursen, R. Derks, H.-G. Janssen and C.A. Cramers, J.Microcolumn Separations, 11(2) (1999) 155.

5. M. v. Deursen, M. v. Lieshout, R. Derks, H.-G. Janssen and C.A. Cramers, J. HighResol. Chromatogr., 22, 2 (1999) 119.

6. M. v. Deursen, J. Beens, H.-G. Janssen and C.A. Cramers, J. High. Resol.Chromatogr., 22, 9 (1999) 1999.

7. C.A. Cramers, H.-G. Janssen, M.M. v. Deursen and P.A. Leclercq, J. Chromatogr.A, 856 (1999) 315.

8. J. Novák, Quantitative Analysis by Gas Chromatography (Second Edition, Revisedand Expanded), Chromatogr. science series, 41 (1988) 187.

9. N. Dyson, J. Chromatogr. A, 842 (1999) 321.10. S. Dagan and A. Amirav, J. Am. Soc. Mass Spectrom., 7 (1996) 737.11. S.C. Davis, A.A. Makarov and J.D. Hughes, Rapid Commun. Mass Spectrom., 13

(1999) 237.12. R.J. Jonker and H. Poppe, Anal. Chem., 54 (1982) 2447.13. C.A. Cramers, G.J. Scherpenzeel and P.A. Leclercq, J. Chromatogr. A, 203 (1981)

207.14. A. v. Es, J. Janssen, C. Cramers and J. Rijks, J. High Resol. Chromatogr. & CC,

11 (1988) 852.15. P.A. Leclercq and C.A. Cramers, J. High Resol. Chromatogr. & CC, 8 (1985) 764.16. J. de Zeeuw, J. Peene, H.-G. Janssen, and X. Lou, 21st Int. Symp. Capillary

Chromatogr. and Electrophoresis, Park City, Utah, proceedings (1999) 16.17. C. Leonard and R. Sacks, Anal. Chem., 71 (1999) 5177.18. Masslib Reference Manual, Version 7.1 (1992).

104 Chapter 6

CHAPTER 7COMPREHENSIVE TWO-DIMENSIONAL GAS

CHROMATOGRAPHY COUPLED TO TIME-OF-FLIGHT

MASS SPECTROMETRIC DETECTION FOR

CHARACTERISATION OF OIL SAMPLES1

SUMMARY

A comprehensive two-dimensional gas-chromatographic system (GC×GC) wascoupled to a time-of-flight mass spectrometer (TOF/MS) for the analysis of oil-samples with boiling points ranging from 150°C to 450°C. Group-types present in oillike the alkanes, saturated cyclic compounds (naphthenes) and aromatic compounds,are shown separately by selecting their unique masses. By selecting appropriate ion-fragments with this method also the determination of sulphur containing species in oilcan be performed. It is shown that co-eluting compounds like benzothiophenes (BT)and naphthalenes or paraffins and olefins compounds can be positively identified.Former results obtained by using FID detection could be confirmed with massspectrometry. After proper selection of unique ions in GC×GC-TOFMS bothselectivity and sensitivity increase.

Quantitative results have been obtained for aromatic target compounds (mono-,di- and tri-aromatic) and for (di-)benzothiophenes (DBT) in oil. The concentration ofthe target compounds in oil ranged from 2 to 8000 mg/kg oil. The minimum detectableamount obtained for these compounds was approximately 2-5 pg and the lineardynamic range of the system was 103. Obtained RSD values for concentrations oftarget compounds in oil are in the order of 2-20%. The obtained results are inagreement with results of oil analyses performed with other analytical techniques.Mass spectrometric response factors for the tested aromatic compounds differ widelywithin each group-type. For quantification of group-types therefore, it is moreappropriate to use the flame ionisation detector (FID) because of the uniformresponse factors of this detector. The amount of DBT present in “low sulphur” dieselwas near to the detection limit. 1 The first part of this chapter has been published as “Group-type identification of oil samples usingcomprehensive two-dimensional gas chromatography coupled to a time-of-flight mass spectrometer (GC×GC-TOF)”, by: M.M. van Deursen, J. Beens, J.C. Reijenga, P.J.L. Lipman and C.A. Cramers, J. High Resol.Chromatogr., 23 (7/8), 2000, 507-510.The second part of this chapter has been submitted as “Quantification of aromatic and sulphur compounds inpetroleum products using comprehensive two-dimensional gas chromatography coupled to time-of-flight massspectrometry”, by: M.M. van Deursen, J. Beens, J. Blomberg and C.A. Cramers in J. Chromatogr. A.

106 Chapter 7

7.1. INTRODUCTION

In this chapter the applicability of GC×GC-TOFMS for oil analyses is described. Inthe first part of this chapter (§7.1.1) the selectivity obtained with this system for theidentification of group-types present in oil will be shown. In the second part (§7.4)quantification of target compounds in oil will be discussed.

7.1.1. Selectivity in GC××××GC-TOFMS

In recent years many techniques became available for oil analyses. Separation anddetection can be performed by high performance liquid chromatography (HPLC)coupled to ultra-violet detection (UV). HPLC in normal phase mode can separate fuelsamples into three large chemical group-types: aliphatic hydrocarbons, aromaticcompounds of various ring–structures, and polar compounds [1]. Also supercriticalfluid chromatography (SFC), fluorescence spectroscopy and nuclear magneticresonance (NMR) are techniques which can be used for oil analyses. High-resolutioncapillary gas chromatography with flame ionisation detection (GC-FID) and withmass spectrometric (GC-MS) detection are the most widely used techniques for thispurpose now. Excellent separation and direct identification can only be obtained byusing the latter technique [2,3]. For determination of specific compounds, likenitrogen and sulphur containing components, selective detectors (FPD, SCD forsulphur-containing compounds) and (TID or NCD for nitrogen-containingcompounds) can be used [2].

The most promising techniques for group-type analysis of petroleum samples areamong the coupled techniques. A very adequate technique is the on-line coupling ofliquid chromatography with gas chromatography (LC-GC). After a pre-separationwith LC based on chemical structure, the fractions are then injected subsequently ontothe GC column, where separation within the groups based on boiling point (or carbonnumber) takes place [1, 4-7]. Furthermore, multidimensional techniques like the so-called PIONA-analyses (Paraffins, Isoparaffins, Olefins, Naphthenes and Aromatics)have been developed successfully for characterisation of oil [8,9]. In a review ofBertsch [10] the history of two-dimensional GC techniques for the analysis ofcomplex samples has been described.

Recently a new technique became available for oil characterisation:comprehensive two-dimensional gas chromatography (GC×GC). This technique,which is the subject of the present chapter, offers a high resolving power that is

Comprehensive two-dimensional gas chromatography 107

necessary for the analysis of complex samples [11]. GC×GC is a technique highlysuited for the separation of complex mixtures such as petroleum-, environmental- andflavour-samples [12-14]. In this chapter the characterisation of petroleum fractionsusing this high-resolution technique will be described.

Separation in the first non-polar dimension column is mainly based on volatilityof the analytes, whereas separation in the second dimension polar column is mainlybased on polarity or molecular structure of the analytes. Identification of compoundscan be made by using standard solutions with known compounds, together withchemical logic like boiling point, polarity and origin of the sample. As shown byBeens et al. [12] the different groups of compounds present in kerosene, like alkanes,naphthenes and aromatics can well be determined with GC×GC using a FID.

An additional advantage of GC×GC for oil analyses compared to conventionaltechniques like the multidimensional PIONA-analysis or the use of a one-dimensionalhigh-resolution column [15], is the gain in speed that can be obtained. The analysis ona PIONA or high resolution system takes approximately 180 minutes, whereas theanalysis on a GC×GC-system with a much higher efficiency generally takes 60minutes. Moreover, the PIONA system is only suited for the lower boiling fractions(final boiling point (FBP) < 200°C), whereas the GC×GC system can analyse fractionswith FBP up to 450°C and is thus able to identify much more and complex fractions.The GC×GC-system, which consists of two capillary GC-columns coupled in seriesand using a FID-detector, offers a far higher resolution than the use of a single columnGC-MS system [16]. Re-injection from the first column onto the second column cantake place using a heated [11,17,18] or a cryogenic modulation system [19-21].Chromatographic performance of both modulation systems is more or less similar, andthey both enhance sensitivity approximately 10 times compared to a one-dimensionalsystem. Since approximately five years equipment for this type of analysis iscommercially available. GC×GC coupled to a flame ionisation detector (FID) isapplied now in several laboratories as the analytical technique for group-typeseparation of oil samples [11,18].

During the last few years there is an increasing interest in coupling the GC×GC-system to a mass spectrometer for direct identification of the (separated) compounds.This already has been done by Frysinger et al. [22] who coupled the system to aquadrupole mass spectrometer. However, from their work the conclusion can bedrawn that for an accurate detection of the very narrow second dimension peaks, afaster mass spectrometer is required. Therefore in this work a GC×GC-system wascoupled to a time-of-flight mass spectrometer, which has a maximum spectralacquisition rate of 500 spectra per second.

108 Chapter 7

The aim of the present work is to show that using GC×GC-TOFMS enhancesselectivity in identification of group-types in middle distillate analyses. It is shownthat even co-eluting compounds of different classes like paraffins and olefins ornaphthalenes and benzothiophenes can positively be identified with MS. On the two-dimensional (2D) contour plots it is apparent immediately whether or not an oilsample has undergone a hydrotreatment. The identification of several group-types isshown in two-dimensional plots. With this system it is possible to confirm theidentification of compounds previously found with FID.

7.2. EXPERIMENTAL

7.2.1. Instrumentation

The GC×GC-system used in this work was an HP 6890 Gas Chromatograph (AgilentTechnologies, Wilmington, DE, USA) equipped with an OPTIC 2 PTV injector(ATAS, Veldhoven, The Netherlands) a FID and a thermal modulation assembly(Zoex Corp., Lincoln, NE, USA). This modulation assembly consisted of a separateheating oven for the temperature programming of the second dimension column,independent of the first dimension column heating. Focusing and re-injections ontothe second column were performed by rotating a heated modulator (sweeper), whichwas operated at a temperature 100°C higher than the oven temperature [17].

A time-of-flight mass spectrometer was used as the detection device (LecoCorporation, St. Joseph, MI, USA). The mass spectrometer was operated at anacquisition-rate of 50 spectra/s, with an ion-source temperature of 180°C and a transfer-line temperature of 275°C. The pressure inside the flight tube was about 10-7 Torr. Thescanned mass range was from mass 40 to 280. With this mass spectrometer a unit massresolution was obtained. The following column combination was used:

1st dimension column: 10m × 0.25mm × 0.25µm DB-1 (J&W Scientific, Folsom,CA, USA).

Modulation capillary: 7cm × 100µm × 3µm SE-30 (Quadrex, New Haven, CT,USA).

2nd dimension column: 0.6m × 100µm × 0.1 BPX-50 (SGE, Ringwood, Australia).

Columns were coupled using special glass micro press-fits (Techrom, Purmerend, TheNetherlands). To prevent too high linear gas flows through the second column caused


by the low pressures in the MS, the second dimension column was coupled to the MSthrough a deactivated column of 1 m long and an inner diameter of 100 µm. This wasalso done to obtain near atmospheric outlet conditions for the second dimensioncolumn to be able to compare this relatively new GC×GC-TOFMS separation patternwith former results obtained with GC×GC-FID. Injections were done manually with a10 µL liquid syringe. Data-acquisition of the GC×GC-TOFMS system was startedmanually.

7.2.2. Chromatographic conditions

The inlet pressure used in these experiments was 75 kPa. Helium 6.0 was used as thecarrier-gas. Injections were performed manually in the hot split-mode, at a split-ratioof approximately 50. The injected volume of all samples was 1 µL. The PTV wasprogrammed from 200°C to 350°C at a programming rate of 10°C/s. The sweepervelocity was set at 0.25 sweeps/s, the pause-time of the sweeper at the end of themodulation capillary was one second and the total modulation time was 7.5 seconds.The oven temperature-program was:

1st dimension: 40°C (5 min) → 3°C/min → 250 °C (5 min).2nd dimension: 55°C (5 min) → 3°C/min → 265 °C (5 min).

7.2.3. Chemicals and samples

Several kerosene fractions, a cycle oil and its hydrotreated equivalent were obtainedfrom Shell Global Solutions International (Shell Research and Technology Centre,Amsterdam, The Netherlands). Two diesel fuel samples have been obtained from aShell gas station, during normal service: a diesel sample containing a normal sulphurlevel and a diesel sample containing a low sulphur level. The compounds used as thecalibration standards (table 7.2) are: alkylbenzenes (no. 1-7) (Polyscience, Niles, Il,USA), naphthalene, acenaphthene, anthracene, benzo- and dibenzothiophene (Aldrich,Steinheim, Germany), biphenyl, fluorene, phenanthrene, monochlorobenzene (Fluka,Buchs, Germany), toluene, 1,2,3,4-tetrachlorobenzene (Merck, Darmstadt, Germany)and naphthalene-d8 (Supelco, Bellefonte, PA, USA). The purity of all compounds was98% or higher.

110 Chapter 7

Calibration solutions of the aromatic compounds in toluene, with concentrationsranging from 2 to 200 µg/mL, were prepared volumetrically. Chlorinated aromaticcompounds were used as internal standards (table 7.2). The concentration of theinternal standards in each calibration solution was ≈ 95 µg/mL. Calibration solutionsof benzothiophene (BT) and dibenzothiophene (DBT) in toluene were preparedgravimetrically, ranging from 5 – 300 µg/mL. For these compounds naphthalene-d8

was used as an internal standard, with a concentration of 90 µg/mL in each solution.For the determination of the aromatic compounds and sulphur compounds in theuntreated and hydrotreated cycle oils, 4 separate solutions were prepared in toluene ≈200 µg/mL. The concentration of the chlorinated and deuterated internal standards inall samples was 90 µg/mL. For the determination of BT and DBT in “normal” sulphurdiesel fuel and low-sulphur diesel, solutions were prepared of 700 mg/mL diesel and150 µg/mL naphthalene-d8 in toluene.

7.2.4. Software

The thermal modulator was controlled using 2D-GC software (Zoex corp.). Data-collection, identification of compounds and integration of the chromatograms wasperformed with Pegasus II software (version 1.33) developed by Leco Corp. Thecomputer contained a 300 MHz processor and 192 Mb memory, which are theminimum requirements to handle the large data-files (100-200 Mb) that are generated.The disk-space used in the present work was 5 Gb. However, for convenience it isadvised to use a much larger disc-space. Software developed in-house was used toconvert the one-dimensional array of data produced by the mass spectrometer, into anASCII matrix file in which each row represents a second dimension chromatogram.Contour plots and second-dimension chromatograms of these files were generatedthrough Noesys Transform (Creaso, Apeldoorn, The Netherlands).


Figure 7.1 represents a two-dimensional contour plot of the separation of a kerosenesample with GC×GC-TOFMS. In this figure the total ion current (TIC) is shown, whichis comparable with the results obtained from a similar GC×GC analysis with FID-detection [12]. The different group-types present in this sample are represented withnumbers 1 to 4: (1) alkanes; (2) mono-naphthenes; (3) di-naphthenes and (4) aromatic


compounds. In this analysis 580 second dimension chromatograms of 7.5 seconds havebeen recorded, which resulted in a total analysis time of 73 minutes. The total numberof collected spectra during analysis amounted to 210,000, which resulted in file sizes upto 150-200 Mb. It is therefore very important to have a powerful computer with a largedisk-space. After data collection, the large array of TOF/MS-data was converted to amatrix of 580 rows (first dimension) and 375 columns (second dimension) with in-house developed software. This enabled the possibility to obtain a 2D-contour plot forvisualisation of the separated sample.

1st d

imen

sion

rete

ntio

n tim

e [m

inut

es]

2nd dimension retention times [s]

Figure 7.1: Two-dimensional contour plot of the analysis of a kerosene sample usingGC×GC-TOFMS (total ion current). (1) alkanes; (2) mono-naphthenes; (3) di-naph-thenes; (4) aromatic compounds.

Figure 7.2 is a typical example of a second dimension chromatogram of a GC×GCanalysis of an oil sample, which shows the fast separation of eight compounds withinfour seconds. For each compound the unique mass is shown. For example mass 85 is acharacteristic ion fragment of alkanes (C6H13

+). In table 7.1 the identification of thecompounds and in figure 7.3 the molecular structure of the compounds is shown. As

1 2 3

4

112 Chapter 7

can be seen from the number of datapoints in each peak (figure 7.2), it can be concludedthat the acquisition-rate of 50 spectra per second is enough for an accurate detection ofthe peaks. Moreover, at higher acquisition rates the file-size will be too large foradequate data-handling and the limit of detection (LOD) will be substantially higher[23].

The carrier-gas velocity in the second column was estimated to be about 150cm/sec, which is not far from the optimum velocity of a short 100 µm column. Peak-widths at half height are approximately 150 ms, which resulted in a plate-number of3000 plates for compound no. 8 at a retention-time of 3.2 seconds (figure 7.2). This islower than the theoretical plate-number of about 10000 plates for a short 100 µmcolumn, but most likely extra band-broadening occurs during injection on the secondcolumn or due to the extra restriction column coupled between the second dimensioncolumn and MS. Peaks are not baseline separated in this chromatogram, but togetherwith automatic peak-deconvolution of the MS-software, compounds can be identified.

0 1 2 3 4

Retention time second dimension [s]

Res

pons

e M

S

8511118016691145118131

Figure 7.2: Example of a second dimension chromatogram from the analysis shownin figure 7.1, with the unique ion-traces of each compound shown separately. The firstdimension retention-time was about 32 minutes. Compound identification is shown intable 7.1.

1

2

34

56

7

8


Table 7.1: Identification of the peaks in the chromatogram shown in figure 7.2.

Compound no. Formula Group12345678

C13H28

C13H26

C13H24

C12H22

C12H18

C12H16

C12H16

C11H14

AlkanesMono-naphthenesDi-naphthenesDi-naphthenesMono-aromaticsMono-aromaticsMono-aromaticsMono-aromatics

1.

8.

7.

2.

6.

4.

3.

5.

Figure 7.3: Molecular structure of compounds (table 7.1).

The advantage that can be obtained from coupling GC×GC to TOF/MS is thepossibility to separately show the different groups present in oil-samples by selectingthe extracted ion traces, which are unique for each group of compounds. In figure 7.4Athe alkanes are shown. To this end the ions 71, 85, 99 and 113 (which are unique foralkanes) were summed. In figures 7.4B, 7.4C and 7.4D the mono-naphthenes, di-naphthenes and mono-aromatic compounds (respectively) are shown separately. Fromthese figures it can be concluded that the identification that has been made previously

114 Chapter 7


1st d

imen

sion

rete

ntio

n tim

e [m

inut

es]


1st d

imen

sion

rete

ntio

n tim

e [m

inut

es]

Figure 7.4, A (left): Contour plot of alkanes in kerosene (see figure 7.1). Selectedions: 71, 85, 99 and 113 (summed), unique for alkanes. B (right): Contour plot ofmono-naphthenes in kerosene (see figure 7.1). Selected ions: 69, 83, 97, 98, 111, 112,153 (summed), unique for mono-naphthenes.


1st d

imen

sion

rete

ntio

n tim

e [m

inut

es]


1st d

imen

sion

rete

ntio

n tim

e [m

inut

es]

Figure 7.4, C (left): Contour plot of di-naphthenes in kerosene (see figure 7.1). Selectedions: 67, 81, 95 and 109 (summed), unique for di-naphthenes. D (right): Contour plot ofmono-aromatic compounds in kerosene (see figure 7.1). Selected ions: 77, 91, 105(summed), unique for mono-aromatic compounds.

A B

C D


by GC×GC-FID on the basis of spiking a limited number of compounds and onchemical logic [12], is now confirmed by the present method.

By selecting the appropriate ion-fragments, it is now also possible to determinesulphur containing species in oil. For this purpose a middle distillate petroleum fractionwith a high concentration of sulphur-containing compounds was analysed. The contourplot of this analysis is shown in figure 7.5A. In figure 7.5B an enlargement is shownfrom the contour plot in figure 7.5A. Here, ions with mass 161 and 176 are selected thatshow the presence of sulphur-containing compounds (benzo-thiophenes, BT). The roof-tile structure of dimethyl-benzothiophenes (C10H10S), ethyl-methyl-benzothiophenes(C11H12S) and di-ethyl-benzothiophenes (C12H14S) is clearly visible. In the total ioncurrent plot of figure 7.5A this group of compounds can hardly be detected. Therefore,a substantial increase in selectivity and in sensitivity is obtained when the ions areselected properly. For the determination of oxygenates in oil the same procedure can beapplied.

The co-elution of sulphur compounds (benzothiophenes) with di-aromatic com-pounds is one of the problems that are encountered in oil analyses. In figure 7.5A di-aromatic compounds are visible (no. 1-4). In figure 7.6 the co-elution of methyl-ethyl-benzothiophene and trimethyl-naphthalene is shown in a second dimensionchromatogram. Automatic peak deconvolution (incorporated in the MS software) canpositively identify both peaks. The spectra of both peaks together with the comparinglibrary spectra are shown in figure 7.7A and 7.7B. The quality of the deconvolutedspectra and the similarity with the library spectra is high. Deconvolution therefore is avery useful tool in complex analyses where peaks are overlapping. Complete baselineseparation is not required, however results have to be considered with care.

Figure 7.8A and 7.8B are two-dimensional contour plots, of a cycle oil (CCCO)and its hydrotreated equivalent. Some important reactions that take place during thehydrotreatment are [24]:

compoundsaromaticMonocompoundsaromaticDiSHBiphenylopheneDibenzothi

SHneEthylbenzeheneBenzothiop

H

H

H

−→−

+→

+→

2

2

2

.3

.2

.1

2

2

In general, the non-saturated compounds are converted into saturated compounds andsulphur compounds are converted into aromatic compounds.

116 Chapter 7

1

2nd dimension retention times [s] 2nd dimension retention times [s]

1st d

imen

sion

rete

ntio

n tim

e [m

inut

es]

1st d

imen

sion

rete

ntio

n tim

e [m

inut

es]

S

Figure 7.5: (A, left) GC×GC-TOFMS analysis (TIC) of a petroleum fraction containinga high sulphur concentration. In the TIC mode no sulphur compounds are visible(circle). (1) naphthalene; (2) methyl-naphthalenes; (3) dimethyl-naphthalenes; (4)trimethyl-naphthalenes. (B, right) Selected ions of benzothiophenes: mass 161+176(enlarged from figure 7.5A). (3*) dimethyl-benzothiophene and (4*) methyl-ethyl-benzothiophenes, co-eluting with naphthalenes.

2471.5 2472.5 2473.5 2474.5Time [s]

Res

pons

e

S

Figure 7.6: Co-elution of (1) trimethyl-naphthalene (Mw = 170) and (2) methyl-ethyl-benzothiophene (Mw = 176) in a middle distillate petroleum fraction (figure 7.8A). Thischromatogram shows a part of a second dimension separation. The first dimensionretention-time here was 41 minutes. Peaks are deconvoluted mathematically.

Mass 170

Mass 1761

2

2

3

4

4*

3*

A B


Figure 7.7A: Deconvoluted spectrum (peak true) of 1,4,5 trimethyl-naphthalenecompared to the library spectrum (corresponding to figure 7.6).

Figure 7.7B: Deconvoluted spectrum (peak true) of 2-ethyl-7-methyl-benzothiophenecompared to the library spectrum (corresponding to figure 7.6). The presence offragment m/z = 155 from co-eluting peak no. 1 is still visible.

From these plots one of the advantages of GC×GC compared to one-dimensional GCbecomes evident immediately after comparing the two pictures. The untreated productcontains unsaturated compounds (olefins) co-eluting with the paraffins and aromaticcompounds, which are partly converted into saturated paraffinic and saturated cyclic

A

B155

118 Chapter 7

compounds after hydroprocessing. In figure 7.8A the first band (labeled no.1)represents the (saturated) normal and branched alkanes and the co-eluting unsaturates.In figure 7.8B a “second” band can be observed (labeled no. 2), which represents thesaturated cyclic alkanes. In figure 7.8A this band is hardly visible. In figure 7.9 onesecond dimension chromatogram is shown of the (partly) overlapping paraffinic andolefinic compounds. With deconvolution both peaks can be positively identified, withan obtained library similarity of 87%. Measured spectra of both compounds are shownin figure 7.10.

The original cycle oil product contains high concentrations of di-aromatics. Thisis shown in figure 7.8A in the no. 5 labeled band. From this band it can be seen thatthe intensity of the peaks is high (bright yellow colour). Hydrotreating the cycle oilpartly converts the di-aromatics into mono-aromatics with a fused cyclohexane group.The intensity of the di-aromatics in figure 7.8B therefore, is much lower. Likewise,the mono-aromatics are partly converted into saturated cyclic compounds. Theintensity of the mono-aromatics in figure 7.8B is increased accordingly (band labeled6).

Firs

t dim

ensi

on re

tent

ion-

time

[min

utes

]

8A

12

3 4

5

8B

12

3 4

5

6 7

2nd dimension retention-times [s] 2nd dimension retention-times [s]

Firs

t dim

ensi

on re

tent

ion-

time

[min

utes

]

Figure 7.8: Comparison of a cycle oil (A, left) and its hydrotreated equivalent (B,right) analysed with GC×GC-TOFMS. (1) Alkanes, (2) saturated cyclic alkanes, (3)1,2,3,4-tetrahydronaphthalene, (4) naphthalene, (5) di-aromatics, (6) mono-aroma-tics, 7) bi-phenyl.


1502.5 1502.8 1503.0 1503.3 1503.5 1503.8 1504.0 1504.3Time [s]

Res

pons

e (T

IC, m

ass

83, m

ass

85)

mass 85mass 83TIC

Figure 7.9: Second dimension chromatogram showing the co-elution of paraffinic andolefinic compounds in cycle oil: (1) 5-propyldecane and (2) 6-dodecene (Z). Firstdimension retention-time is 25 min (from figure 7.8A). The extracted masses 85 and83 are unique mass fragments of saturated and unsaturated paraffins, respectively.The peak eluting before the alkane peak (total ion current) is stationary phasematerial (-Si-O- based compound). Data-acquisition rate was 50 spectra per second,from mass 35 to 300. In the TIC-trace the datapoints are visible.

Figure 7.10: Deconvoluted spectra of (A) 5-propyldecane and (B) 6-dodecene (Z).Obtained library similarities were 87%.

In figure 7.11 and 7.12 two second dimension chromatograms of the hydrotreatedand the original product are compared. These chromatograms were recorded at firstdimension retention times of 24 and 25 minutes respectively and they correspond tothe number 3 and 4 labeled peaks indicated in figure 7.8A and 7.8B.

12

A B

120 Chapter 7

0 1 2 3 4 5 6 7Time [s]

Res

pons

e (T

IC)

Hydrotreated CCCO CCCO

Satu

rate

d cy

clic

s

Alka

ne

Alky

lben

zene

Figure 7.11: Two second dimension chromatograms of a GC×GC-TOFMS analysis ofcycle-oil and hydrotreated cycle-oil at a first dimension retention time of 24 minutes,showing the formation of 1,2,3,4-tetrahydronaphthalene (spot no. 3 in figures 7.8Aand 7.8B) from naphthalene during hydrotreatment. (Note: injected amounts for bothchromatograms may differ slightly. Relative intensity of peaks is important.)

0 1 2 3 4 5 6 7Time [s]

Res

pons

e (T

IC)

Hydrotreated CCCO CCCO

Alka

ne Satu

rate

d cy

clic

s

Alky

lben

zene

Figure 7.12: Two second dimension chromatograms of a GC×GC-TOFMS analysis ofcycle-oil and hydrotreated cycle-oil at a first dimension retention time of 25 minutes,depicting the formation of an alkylated mono-aromatic compound (spot no. 4 infigures 7.8A and 7.8B), and showing the decreasing amount of naphthalene, due toreaction during hydrotreatment. (Note: injected amounts for both chromatograms maydiffer slightly. Relative intensity of peaks is important.)

As already mentioned before, di-aromatic compounds are converted into mono-aromatic compounds after hydrotreating. Therefore the last eluting peak in figure 7.11


has an increased intensity. Likewise the intensity of the last eluting peak in figure 7.12has decreased, due to conversion of naphthalene into the mono-aromatic compound1,2,3,4-tetrahydronaphthalene (figure 7.11). When the second dimensionchromatogram of the hydrotreated product in figures 7.11 and 7.12 is compared to thatof the non-treated product, also the production of the saturated cyclic compounds fromaromatic compounds becomes visible.

Finally, as shown in the reaction scheme on page 115, sulphur containingcompounds like benzothiophenes and di-benzothiophenes are converted intoalkylbenzenes and bi-phenyls respectively. Sulphur containing compounds co-elutewith naphthalenes (band no. 5, figure 7.8A). In figure 7.6 the overlapping peaks areshown in a second dimension chromatogram. In figure 7.8B no sulphur compoundscan be detected anymore, and the intensity of biphenyl has increased (peak 7, figure7.8B).

As was already shown in figure 7.5A and B, an improved way of visualising thebenzothiophenes in a two-dimensional contour plot is by extracting unique ionsspecific for benzothiophenes. In figure 7.13 a comparison is shown of the presence ofBT in a non-treated (cycle) and hydrotreated oil. In this figure benzothiophenes areshown by extracting the molecular masses of methyl-benzothiophene, dimethyl-benzothiophene and methyl-ethyl-benzothiophene (masses 148, 162 and 176respectively). This was done for both the cycle and the hydrotreated cycle oil. Thebenzothiophenes (numbers 1 to 3) could not be detected anymore in the hydrotreatedproduct. Moreover, the reaction-products of the benzothiophenes into alkylbenzenescan be seen from the increased bands (circled part of numbers 4 to 6) in the right partof the figure. The fact that next to the benzothiophenes the alkylbenzenes are alsovisible, can be explained by the fact that the selected masses (molecular masses ofbenzothiophenes) are equal to the molecular masses of the alkylbenzenes. Forexample, the molecular mass of both benzothiophene and pentylbenzene is 148. Incontrast to benzothiophenes and naphthalenes, benzothiophenes and alkylbenzenes arewell separated and can be shown separately.

In this way, by comparison of contour plots only, the type of oil can bedistinguished very easily whereas this is more difficult in a complex one-dimensionalchromatogram. Especially for environmental analyses this can be very advantageousin determining the source of, for example, oil spills.

Unique masses that were used are not as unique as it seems. There is of course acertain overlap of ions in the spectra of the compounds in the different groups that arementioned above. Unique masses therefore have to be chosen with great care to showthe differences between two groups, for example between mono- and di-naphthenes.

122 Chapter 7

Second dimension relative retention-times [s]

Firs

t dim

ensi

on re

tent

ion-

time

[min

]

1

2

3

4

5

6

Cycle oil Hydrotreated cycle oilMass 148

Mass 162

Mass 176

Mass 148

Mass 162

Mass 176

Figure 7.13: Two-dimensional contour plots of sulphur containing compounds(benzothiophenes) from cycle-oil and hydrotreated cycle-oil. Selected masses 148, 162and 176 are molecular masses of C1, C2 and C3-benzothiophenes (BT) (no. 1, 2 and 3respectively). No. 4, 5 and 6 (circles) represent the alkylbenzenes produced from BT.

There are a number of possibilities to improve this selectivity between differentgroups. A soft ionisation method can be used to obtain a high intensity of themolecular ion and a low amount of fragment-ions. Soft ionisation methods are forexample field ionisation or chemical ionisation methods. Because in this work onlyelectron ionisation was available, the influence of a low ionisation energy (20 eV) anda low ion-source temperature (90ºC) on the fragmentation pattern has beeninvestigated (no figures shown). This method however, will decrease the intensity ofall fragments equally and has little or no effect on the intensity of the molecular ion.For future research therefore it is recommended to investigate the possibilities of thementioned soft ionisation methods for an increased sensitivity.


7.4. QUANTIFICATION IN OIL ANALYSES

In the petroleum industry the quantification of the groups of organic compoundspresent in petroleum and its (distillate) fractions has been the subject of many studies.Detailed characterisation of these fractions can be used to predict the productproperties and provides a better understanding of the reaction kinetics of conversionprocesses such as hydro-processing and -cracking. The quantitative analysis of thesehydrocarbon mixtures has been applied since the early 1950s [25]. The first massspectrometers, with a unit resolution power and a mass range up to a few hundred inmass, were not suited to measure high molecular mass compounds. The mainapplication area was the identification of group-types in lower boiling oil fractionswith final boiling points (FBP) < 150°C. Soon however, instruments with increasedpotential were developed, which enabled the analysis of higher molecular weightcompounds [26]. With high resolution mass spectrometry it was possible to identifytwo compounds with the same nominal masses (molecular weight), which areseparated based on the difference in their exact masses. Moreover, for two compoundswith the same mass like decane (C10H22) and methylnaphthalene (C11H10), massfragmentation patterns in electron impact ionisation often show differences. Thisenables the possibility to mathematically deconvolute the spectra and thus identify thecompounds. For very complex mixtures like petroleum distillates, this procedurebecomes very complicated. Teeter [27] reported the use of a high-resolution massspectrometer with a resolving power of 5000, for type analysis of complexhydrocarbon mixtures. It is reported that 22 group-types present in petroleumdistillates, ranging from C6 to C21, have been identified and quantified directly withoutthe use of a preliminary separation.

7.4.1. Quantification with FID

Quantification is an important issue in group-type analyses of oil samples. When it isassumed that the complete distillate oil sample is transferred to the detector,quantification can be performed by internal normalisation of the chromatograms. Thesum of the peak areas of the separate peaks represents the peak area of the totalsample amount. Because the response of the detector is different for all compounds,response factors have to be determined to unify all peak areas. Response-factors thatare used in FID analyses, can be calculated or can be determined based on group-types. Within those groups, for example the paraffinic compounds, response factors

124 Chapter 7

are fairly constant. Therefore it is estimated that only around 20 group-responsefactors have to be determined for quantification with a FID.

The relative response factor for a FID (fim) can be approximated with the

following relation:

ni

imi C

Mf = (7.1)

where Mi is the molecular mass and Cni is the number of carbon atoms. Thiscalculation is valid for hydrocarbon groups present in oil like aliphatic, aromatic andsulphur-containing compounds. For oil analyses performed with GC-FID the massfraction of a compound i (gi) can then be calculated according to equation 2 [28]:

∑=

⋅

⋅= k

j

mjj

mii

i

fA

fAg

1

(7.2)

Where Ai is the calculated peak-area of compound i and j denotes any compoundpresent in the sample.

Examples of quantitative oil analyses performed with an FID detector aredescribed in work of Johansen et al. [15]. In their work a detailed investigation of thequantitative group-type analysis of a light petroleum product (FBP < 250°C) usinghigh resolution capillary GC coupled to FID detection was reported. For allcompounds the same response factor was used (i.e. 1.000). Quantitative group-typeanalyses for high boiling petroleum distillates (FBP 350-500°C) using HPLC coupledto moving-wire FID were reported by Qiang and Lu [29]. Cebolla et al. [30,31]reported quantitative results of the characterisation of fossil fuel products using thinlayer chromatography coupled to FID detection.

Quantitative group-type analyses with LC-GC-FID or GC×GC-FID, using FIDresponse factors, have been reported in several papers [1,28,32-34]. Jiang et al. [1]obtained quantitative results for poly-aromatic hydrocarbon compounds ranging frombenzene to benzo[b]fluoranthene (tetra-aromatic). FID-factors were determinedrelative to an internal standard. Beens et al. [28], reported quantitative results (basedon equations 7.1 and 7.2) of GC×GC group-type separations which are in goodagreement with results from a LC-GC separation of the same samples. For thispurpose data processing software was developed to handle the vast amount of datathat is produced in a GC×GC run. Results were obtained for a heavy gas-oil,


containing compounds ranging from C8 to C22. Unfortunately, with GC×GC-FID itwas not yet possible to quantify compounds with higher boiling points than C14 (tri-aromatics) because of the temperature limits of the heated modulator that was used.For temperatures higher than approximately 250°C column bleed will increase thebaseline-level and therefore hinder accurate quantitative results. The same workersalso investigated the quantification of sulphur containing compounds in middledistillates, using LC-GC-FID-SCD (sulphur chemiluminescence detector) [32].Sulphur compounds ranging from C1-benzothiophenes (one aromatic ring) to C2-benzonaphthothiophene (four aromatic rings) were determined quantitatively.Frysinger and Gaines [33,34] investigated the quantification of oxygenates, BTEX andtotal aromatic compounds in a gasoline using GC×GC-FID. In this workquantification was based on calibration curves of standard solutions, with carefullyselected internal standards. This means that these compounds should possess a similarchemical structure as the compounds of interest, they should not be present in the oilsample and they should be well resolved from all gasoline compounds. The number ofsuitable internal standards in oil analyses is rather limited because of the complexityof the chromatograms. Quantitative results obtained in this work compare favourablywith accepted standard methods.

7.4.2. Quantification with MS

As mentioned previously, quantitative group-type analyses of oil samples can also beperformed using a mass spectrometer as the detection device. This can be donedirectly by using high resolution MS, but the most widely used technique forcharacterising and quantifying oil samples is GC coupled to MS. The advantage of theMS is the possibility of direct identification of compounds. The problems that areencountered in quantification with MS, are in the area of the determination ofresponse factors.

To enable MS-quantification it is possible to use the total ion current (TIC).Determining response factors then would be similar to the calculations performed fordetermining FID-response factors. Quantification could then be accomplished usingdeuterated compounds as internal standard, which represent a specific group-type.However, as was investigated in the present chapter, response factors of the TIC arenot as uniform as obtained with an FID.

The difficulty that is encountered in quantification in group-type analyses withMS is the innumerable number of compounds that are present. For distillate fractions

126 Chapter 7

containing compounds with carbon-numbers ranging from C8 – C20 the number ofpossible isomers can easily amount to 105 [7]. Each compound shows a characteristic,but different fragmentation pattern, with different specific ions. A specific fragment,for example the unique m/z = 85 in paraffinic compounds, has a different intensity inall different isomers. This also counts for specific ions of the other group-types. It is avirtually impossible task to develop calibration curves or response factors for eachcompound. In the following paragraph methods for quantifying oil-analyses whichhave been used up to now are evaluated.

In most cases GC-MS is primarily used to identify peaks detected by GC-FID.However, target compounds like polyaromatic hydrocarbons (PAHs), alkylated PAHsand sulphur containing compounds like dibenzothiophenes (DBT) can be quantifiedby using GC-MS in the selected ion monitoring (SIM) mode. Deuterated internalstandards (PAH) can then be used to obtain calibration curves. For each group-type(for example the alkyl homologues of naphthalene or the phenanthrene alkyl series) adifferent set of unique ions is selected [35,36].

Dzidic et al. [2] identified and quantified nitrogen and sulphur compounds inheavy gas oils using isobutane/chemical ionisation (CI) mass spectrometry. Resultsproved to be similar to those obtained with a nitrogen and sulphur selective detector.Poirier and Smiley [37] reported quantitative results of sulphur compounds in naphtha(FBP 30-200°C) and middle distillate (200-350°C) fractions by GC-MS. Theadvantage of using chemical ionisation MS is that it is a “soft” ionisation technique,where the amount of fragment ions produced is much lower than in electron impactionisation because little energy is used to ionise the molecule. This means that MH+

(molecular ion plus hydrogen atom) can be used for quantification and identification.In CI the molecular ion in the mass spectrum has a much higher intensity than in EI.Overlap with spectra of other compounds is less likely to occur, and thus selectivitybased on the molecular ion is much higher. For group-type analyses of petroleumdistillates it is therefore advantageous to use CI above EI. However in this chapteronly EI is described, because the CI option was not (yet) present on the TOFinstrument used in this work.

In another paper Dzidic et al. described the Townsend discharge nitric oxidechemical ionisation GC-MS method for hydrocarbon analysis of the middle distillates[38]. In this method alkyl substituted aromatics only yield the molecular ions (M+),whereas the paraffins and cyclic compounds yield M-1 ions together with fragmentsof the same mass as produced in EI but now in a much lower intensity. This method isnot suitable for olefin containing samples, however. All ion intensities in the mass


spectrum of a compound are corrected for the naturally occurring 13C isotopes. Thisincreases the signal and thus increases sensitivity.

7.5. RESULTS AND DISCUSSION

As described previously, quantification of group-types is often done using an FIDdetector [28]. Response factors for compounds within a specific group-type are fairlyconstant. This in contrast to the use of a mass spectrometer as the detection device:within a group-type responses of compounds are widely differing. Especially whenextracted ions would be used for quantification. The intensity of a specific ion differsfor each molecule, and therefore thousands of calibration curves would be needed forquantification of group-types. However, for quantitative determination of targetcompounds the mass spectrometer can be used very well. This was done in the presentwork.

Compounds of interest in hydrotreatment processes are aromatic compounds andsulphur-containing compounds (BT and DBT). During hydrotreatmentbenzothiophenes are mainly converted into mono-aromatic compounds. For a betterunderstanding of the hydrotreatment-step it is advantageous to know the concentrationof the mentioned compounds. Compounds that were quantified in this work arementioned in table 7.2. Five-point calibration curves were generated, eachconcentration measured in duplicate. In the first quantitative experiments chlorinatedbenzenes (mono-chlorobenzene and 1,2,4,5-tetrachlorobenzene, see table 7.2) wereused as internal standard, because originally this compound is not present in oil and asatisfactory separation from the complex sample is obtained. Moreover, forchlorinated benzenes ions can be extracted which are solely unique for thesecompounds. Another possibility of quantification with MS is the use of deuteratedcompounds as internal standard. These compounds obtain similar chemical propertiesas the oil-compounds, can be separated from the sample and provide unique ionswhich are not present in the remaining compounds. Deuterated naphthalene(naphthalene-d8) was used as internal standard for benzothiophene anddibenzothiophene.

In tables 7.2 and 7.3 data are shown of calibrations obtained with the GC×GC-TOFMS system. For this purpose, depending on the type of compound, masses wereselected (see table legend).

128 Chapter 7

Table 7.2: Calibration compounds and equations of the calibrations (based on massconcentrations). Calibration solutions of the aromatic compounds and i.s. (1-12,15,16) were prepared in toluene volumetrically; for the benzothiophenes and i.s.(13,14,17) gravimetrically. Each first dimension peak was re-injected 3 to 4 times: thetotal peak area for each compound was calculated by summation of the separate peakareas of each re-injection. Injections were performed manually (1 µL). A split-ratio of1:50 was applied. Internal standards used: mono-chlorobenzene (for mono-aromaticcompounds, no. 1-7); 1,2,3,4-tetrachlorobenzene (for di- and tri-aromatic compounds,no. 8-13); naphthalene-d8 (for benzothiophenes: 13, 14).

Calibration equation

CompoundConcentrationstock-solution[µg/mL] Selected ions

Total ion current(TIC)d

Aromatic compounds:1. 1-Methylethylbenzene2. Propylbenzene3. 1,3,5-Trimethylbenzene4. Butylbenzene5. Hexylbenzene6. Octylbenzene7. Decylbenzene8. Naphthalene9. Biphenyl10. Acenaphthene11. Fluorene12. PhenanthreneSulphur-compounds:13. Benzothiophene (BT)14. Dibenzothiophene (DBT)

333246266272332319231251224363108141

1125424

y = 0.7791x + 0.0352a

y = 1.4316x + 0.0441a

y = 1.0367x + 0.0552a

y = 0.8927x + 0.0552a

y = 0.6275x + 0.0349a

y = 0.4315x + 0.0225a

y = 0.2637x + 0.0268a

y = 5.1863x + 0.1192b

y = 4.1434x + 0.1497b

y = 3.5663x + 0.2298b

y = 1.4782x + 0.0410b

y = 1.1333x + 0.0758b

y = 1.0984x – 0.0611c

y = 0.3654x – 0.0003c

y = 0.8078x + 0.0681y = 1.3693x + 0.1039y = 1.1526x + 0.1257y = 1.1979x + 0.1343y = 0.9160x + 0.0857y = 0.7062x + 0.0644y = 0.5269x + 0.0497

----------

----

Internal standards:15. Chlorobenzene16. 1,2,3,4-

Tetrachlorobenzene17. Naphthalene-d8

954948

886

----

--

----

--

a Selected ions: 77+91+105+112 (summation). M112= Mw (chlorobenzene).b Selected ions: 28+153+154+166+178+216 (summation). M216= Mw (tetrachlorobenzene).c Selected ions: 134+136+184 (summation). M136= Mw (naphthalene-d8).d TIC: all scanned masses from 40 to 280 are summed.


Table 7.3: Regression coefficients of the calibrations.

Regression coefficient (r)

CompoundSelected ions

Total ioncurrent (TIC)

Aromatic compounds:1. 1-Methylethylbenzene2. Propylbenzene3. 1,3,5-Trimethylbenzene4. Butylbenzene5. Hexylbenzene6. Octylbenzene7. Decylbenzene8. Naphthalene9. Biphenyl10. Acenaphthene11. Fluorene12. PhenanthreneSulphur-compounds:13. Benzothiophene (BT)14. Dibenzothiophene (DBT)

0.99600.99720.99700.99870.99590.99730.99520.99240.99310.99230.98400.9966

0.99950.9974

0.99240.99440.99230.99220.99310.99230.9854

----------

----

Internal standards:15. Chlorobenzene16. 1,2,3,4-Tetrachlorobenzene17. Naphthalene-d8

------

------

For example, for alkylbenzenes the unique masses 77, 91, 105 were selected.Mass 112 is the molecular mass of the internal standard mono-chlorobenzene. Fromtable 7.2 it can be concluded that the responses for all compounds are different, evenwithin a group (i.e. propyl-, trimethyl- and methyl-ethylbenzene: C3-benzenes).Recalculation of the mass concentrations in molar concentrations did show someimprovement in similarity of the response factors, however not sufficient. It appearsthat low molecular weight compounds have high MS-responses and vice versa.Quantitative determination of group-types using GC×GC-TOFMS hence will be avery cumbersome task. Therefore, this method is only suitable for determining targetcompounds in oil samples by extracting unique ions.

130 Chapter 7

Instead of extracting unique masses also the total ion current can be used forquantification purposes. It is expected that response factors using this method aremore similar. Results of calibration curves are shown in table 7.2. The expectedsimilarity is still not high enough to use the TIC for quantification purposes of group-types. Experiments on a quadrupole mass spectrometer showed more or less similarresults as obtained with the TOF-MS. As was expected, the response factors of theFID were found to be constant (not shown in this table). This means that this detectoris suitable for quantification of group-types in oil.

The total peak-area that is taken into account for the calculation of ratios ofcompound peak area and internal standard peak area, is the summation of areas of thesuccessive re-injections of the first dimension peak onto the second column. Hence,compared to a one-dimensional GC system each compound was re-injectedapproximately 3-4 times (triplet/quartet-peaks). Integration of each peak was donemanually because no software for handling GC×GC-TOFMS data was available yet.During this procedure it is very well possible that relatively large errors have beenintroduced in the integration of the smallest peaks in each triplet and/or quartet. Thiscan also explain the relatively low correlation coefficients (r), which are not betterthan 0.998 (extracted ions) and 0.994 (TIC). In the latter case the signal to noise ratiois decreased roughly 4 times compared to the extracted ions signal. This is depicted infigure 7.14. To verify the accuracy of the calibration solutions the same experimentswere repeated on a one-dimensional GC-FID system (single injections). Better resultsfor the linearity were obtained, but the conclusion could be drawn that the highestconcentration but one showed the largest error. Because the quality of the spectra washigh for all concentrations, overloading of the ion-source as a possible cause for thelow linearity was excluded. Nevertheless, for quantification purposes in this work thefive-point calibration curve was used to obtain the highest accuracy. The repeatabilityof the retention times was high: in most cases < 0.5 % RSD.

In figure 7.15 quantitative results are shown of mono-, di- and tri-aromaticcompounds and sulphur containing compounds BT and DBT in the parent andhydrotreated cycle oil. Sulphur compounds present in car fuel can pollute the catalyticconverter present in cars. Moreover, the emission of sulphur in car exhaust also has anegative effect on the air quality and the production of SO2/SO3 causes acid rain.Therefore in the oil industry it is very important to remove these compounds fromfuel. In figure 7.15 the reduction of the di- and tri-aromatic and the sulphurcompounds is visible clearly (see reaction scheme on page 115). In the hydrotreatedoil sample only traces of sulphur compounds were visible, whereas in the untreated oilsample the concentration of BT and DBT is very high (500 and 2000 ppm


respectively: figure 7.15). The same accounts for acenaphthene. Also the increase ofbiphenyl in hydrotreated oil due to reaction of DBT is visible (reaction 2, page 115).The reduction of aromatic compounds (di- and tri-aromatic) into mono-aromaticcompounds is also visible, due to the increase of mainly propyl-benzene and methyl-ethylbenzene. The amount of decylbenzene in oil was below the minimum detectableconcentration of ≈ 1 ppm.

The relative standard deviation (RSD) of concentrations of target compounds inoil of 5 successive analyses is 2-20%. The minimum detectable amount (MDA) wasdetermined at approximately 1 and 5 pg (based on S/N = 3) determined forbutylbenzene and BT respectively. Table 7.4 shows concentrations of targetcompounds in oil obtained in this work compared to values found in literature. It canbe concluded that results obtained in this work are similar to results obtained in workusing different analytical methods.

Also the presence of BT and DBT in two types of diesel fuel was determined:normal and low sulphur diesel fuel. The concentration of BT in normal diesel fuel wasdetermined at approximately 20 ppm. For this purpose 3 µL of diesel fuel was injectedpurely at a splitratio of 1:50. Next to BT also methyl-, and dimethyl-BT weredetected, however these compounds were not quantified in the present work. Traces ofDBT were detected in normal diesel fuel, however not quantified. In low sulphurdiesel fuel BT and DBT were not detected. The presence of BT in normal diesel andthe absence of BT in low sulphur diesel is shown in the chromatograms of figures 7.16and 7.17.

1025.0 1025.5 1026.0 1026.5 1027.0

Time [s]

1025.0 1025.5 1026.0 1026.5 1027.0

Time [s]

Figure 7.14: Influence of selected ions on signal to noise ratio. (A) Summed ions:77+91+105+112, signal to noise ratio: 80. (B) Summed ions: TIC, signal to noiseratio: 20. Compound: 1,3,5-trimethylbenzene.

A: B:S/N = 80 S/N = 20

132 Chapter 7

0

1

2

3

4

5

6

7

8

9

Benzene, (1-methylethyl)-

Benzene, propyl-

Benzene, 1,3,5-trimethyl-

Benzene, butyl

Benzene, hexyl-

Benzene, octyl-

Benzene, decyl-

Naphthalene

Biphenyl

Acenaphthene

Fluorene

Phenanthrene

Benzothiophene

Dibenzothiophene

Con

cent

ratio

n [µ

g/m

g oi

l]

CCCO oil Hydrotreated CCCO oil

Figure 7.15: Comparison of concentrations (µg/mg oil) of target compounds (no. 1-14 from table 7.2) in parent and hydrotreated middle distillate cycle oil.

Table 7.4: Concentration of target compounds in middle distillate oil determined byGC×GC-TOFMS compared to some literature values.

Concentration [µg/mg oil]experimental

Concentration [µg/mg oil]literature data

CompoundCCCO-oil

n = 5

HydrotreatedCCCO-oil

n = 3

ASMBoil1

Dieseloil2

Middledistillate

oil3

1. Naphthalene2. Biphenyl3. Fluorene4. Phenanthrene5. BT6. DBT

5.910.422.048.230.52.0

0.300.960.170.310.000.00

0.660.0820.0890.197

-0.188

3.7----------

--------

4.84

1.54

1 Wang et al. [36], ASMB = Alberta Sweet Mix Blend.2 Bundt et al. [40].3 Beens et al. [7].4 Sum of all methyl substituted benzo- and dibenzothiophenes [7].


1702.0 1704.5 1707.0 1709.5 1712.0 1714.5 1717.0 1719.5 1722.0Time [s]

Res

pons

e se

lect

ed m

asse

s

Mass 134 Mass 148

SS

7.5 s

Figure 7.16: Analysis of sulphur compounds (BT) in “normal” diesel fuel.Concentration: approximately 20 ppm. Internal standard: d8-naphthalene. Selectedmasses 134 and 148 are molecular masses of BT and C5-benzene (1-methyl-4-(1-methylpropyl)-benzene). Two successive second dimension chromatograms areselected, each of 7.5 seconds duration.

1703.0 1705.5 1708.0 1710.5 1713.0 1715.5 1718.0 1720.5 1723.0Time [s]

Res

pons

e se

lect

ed m

asse

s

Mass 134 Mass 1487.5 s

Figure 7.17: Analysis of sulphur compounds (BT) in “low sulphur” diesel fuel.Selected masses 134 and 148 are molecular masses of BT and C5-benzene. BT was notdetected in low sulphur diesel.

For an optimal performance of car-engines however, the presence of some sulphuris required in diesel fuel. It is also known that in hydrotreatment processes 4,6- and2,3-dimethyl-dibenzothiophene (C2-DBT) are sulphur compounds that are verydifficult to remove from oil [39]. C2-DBT was positively identified in the normal andlow sulphur diesel samples (figures 7.18 and 7.19). It can be seen from figure 7.19

134 Chapter 7

that only low amounts around 1 ppm (near the detection limit) of the compound arestill present in low sulphur diesel. The shift in retention time of 15 seconds is high, butcan be explained by the fact that the samples were analysed at two separate days. Infigures 7.16 and 7.17 a retention time shift of 1 second was observed. The separationefficiency decreases at high retention times (> 3200 s). Most likely some overloadingeffects occur at an injected amount of 3 µL of diesel fuel.

3550 3555 3560 3565 3570Time [s]

Res

pons

e se

lect

ed m

asse

s

mass 212 TICx0.003

7.5 s

1

3

2

S

Figure 7.18: Identification of dimethyl-dibenzothiophene (C2-DBT: C14H12S) in“normal” sulphur diesel fuel. Selected masses 212 (= molecular weight of C2-DBT)and the total ion current (TIC). From C2-DBT the triplet peaks (no. 1, 2 and 3) areshown.

3565 3570 3575 3580 3585Time [s]

Res

pons

e se

lect

ed m

asse

s

mass 212 TICx0.00047.5 s

13

2S

Figure 7.19: Identification of C2-DBT in low sulphur diesel fuel. Selected masses 212(= molecular weight of C2-DBT) and the total ion current (TIC). From C2-DBT thetriplet (no. 1, 2 and 3) is shown.


For a comparison with our GC×GC-TOFMS system, the same low sulphur dieselsample was also analysed in the Shell laboratories (Amsterdam, The Netherlands)with a specific sulphur detector (SCD). With this detector 4,6-dimethyl-dibenzothiophene was also positively identified (figure 7.20). The concentration wasdetermined at approximately 2 ppm. The total sulphur concentration was determinedat 12 mg S/liter oil (≈ 12 ppm). Ion source overloading was not observed (high qualityspectra) at an injected amount of 1 ng of pure compound, from which it was estimatedthat the linear dynamic range of the system is around 103.

Figure 7.20: Analysis of “low sulphur” diesel (same sample as in figures 7.17 and7.19) on a conventional (one-dimensional) GC system coupled to an SCD (sulphur)detector (analysis performed at Shell Amsterdam, The Netherlands). Injection: 1 µl(pure), hot-split at 250°C, split-ratio: 1:20, inlet pressure: 100 kPa (helium), column:SPB-1 sulphur poly(dimethylpolysiloxane), 30 m × 0.32 mm ID × 4.0 µm df (Supelco,Bellefonte, PA, U.S.A.), detector: Antek model 704E (Antek Instruments, Houston, TX,U.S.A.), oven temp: 35°C (1 min) - 3.0°C/min - 300 °C - 10 min. Compounds: 1) 4,6-dimethyl-dibenzothiophene, 2) 2,3-dimethyl-dibenzothiophene, 3) 1,2-dimethyl-dibenzothiophene, 4, 5, 6) C3-DBT, 7, 8) C4-DBT.

136 Chapter 7

7.6. CONCLUSIONS

GC×GC-TOFMS is a highly suitable technique for characterisation of complex oilsamples. By selecting unique ion fragments it is now possible to show the group-typespresent in oil samples in separate contour plots. By using TOFMS the transitionbetween for example mono- and di-naphthenes can be indicated more clearly than byusing an FID as the detection device. The high acquisition rate of the TOFMS used inthis work was very suitable for an accurate detection of the very fast seconddimension separation.

For quantification of target compounds GC×GC-TOFMS is very useful. For thispurpose extracted ions should be selected to obtain the highest selectivity andsensitivity. However for quantification of group-types in oil the FID should be usedbecause of its uniform response factors for a wide range of compounds. Forquantitative determination of specific sulphur compounds in low sulphur diesel fuel,which are present in very low quantities, the sensitivity of the present system is nothigh enough. For this purpose a specific sulphur detector (SCD) is more appropriate.

7.7. REFERENCES

1. T. Jiang and Y. Guan, J. Chromatogr. Sci., 37 (1999) 255.2. I. Dzidic, M.D. Balicki, I.A.L. Rhodes and H.V. Hart, J. of Chromatogr. Sci., 26

(1988) 236.3. I. Skjevrak, S. Larter, G. van Graas, M. Jones and E. Berge, Advances in Org.

Geochem., 22(3-5) (1993) 873.4. F. Munari, A. Trisciani, G. Mapelli, S. Trestianu, K. Grob Jr. and J.M. Colin, J.

High Resol. Chromatogr.& Chromatogr. Comunn., 8 (1985) 601.5. A. Trisciani and F. Munari, J. High Resol. Chromatogr., 17 (1994) 452.6. I.L. Davies, K.D. Bartle, P.T. Williams and G.E. Andrews, Anal. Chem., 60 (1988)

204.7. J. Beens and R. Tijssen, J. Microcol. Sep., 7 (1995) 345.8. H. Boer, P. v. Arkel and W.J. Boersma, Chromatographia, 13(8) (1980) 500.9. P. v. Arkel, J. Beens, H. Spaans, D. Grutterink and R. Verbeek, J. of Chromatogr.

Sci., 25 (1987) 141.10. W. Bertsch, J. High Resol. Chromatogr., 22(12) (1999) 647.


11. J. Blomberg, P.J. Schoenmakers, J. Beens and R. Tijssen, J. High Resol.Chromatogr., 20 (1997) 539.

12. J. Beens, J. Blomberg and P.J. Schoenmakers, J. High Resol. Chromatogr., 23(3)(2000) 182.

13. J. Beens, J. Dallüge, M. Adahchour, R.J.J. Vreuls and U.A.Th. Brinkman, J.Microcol. Sep., 13(3) (2001) 134.

14. J.D. Dimandja, S.B. Stanfill, D.L. Ashley, J. Grainger and D.G. Patterson Jr.,218th ACS National Meeting, Book of abstracts (1999).

15. N.G. Johansen, L.S. Ettre and R.L. Miller, J. Chromatogr., 256 (1983) 393.16. Z. Liu and J.B. Phillips, J. of Chromatogr. Sci., 29 (1991) 227.17. J.B. Phillips, R.B. Gaines, J. Blomberg, F.W.M. van der Wielen, J.M. Dimandja,

V. Green, J. Granger, D. Patterson, L. Racovalis, H.J. de Geus, J. de Boer, P.Haglund, J. Lipsky, V. Sinha and E.B. Ledford, J. High Resolut. Chromatogr.,22(1) (1999) 3.

18. M.M. van Deursen, J. Beens, J. Reijenga, P. Lipman and C. Cramers, J. HighResol. Chromatogr., 23 (7/8) (2000) 507.

19. R.M. Kinghorn and P.J. Marriott, J. High Resol. Chromatogr., 21(1) (1998) 32.20. R.M. Kinghorn and P.J. Marriott, J. High Resol. Chromatogr., 22(4) (1999) 235.21. J. Beens, M. Adahchour, R.J.J. Vreuls, K. van Altena and U.A.Th. Brinkman, J.

Chromatogr., 919 (2001) 127.22. G.S. Frysinger, R.B. Gaines, J. High Resol. Chromatogr., 22(5) (1999) 251.23. R.J.J. Vreuls, J. Dallüge and U.A. Th. Brinkman, J. Microcol. Sep., 11(9) (1999)

663.24. A. Fafet and J. Magné-Drisch, Revue de l’institut Français du petrole, 50(3)

(1995) 391.25. R.A. Brown, Anal. Chem. 23 (1951) 430.26. M.J. O’Neal Jr. and T.P. Wier, Anal. Chem., 23 (1951) 820.27. R.M. Teeter, Mass Spectrometry Reviews, 4 (1985) 123.28. J. Beens, H. Boelens, R. Tijssen and J. Blomberg, J. High Resol. Chromatogr., 21

(1998) 47.29. D. Qiang and W. Lu, J. of Petroleum Sci. and Eng., 22 (1999) 31.30. V.L. Cebolla, J. Vela, L. Membrado and A.C. Ferrando, J. Chromatogr. Sci., 36

(1998) 497.31. J. Vela, L. Membrado, V.L. Cebolla and A.C. Ferrando, J. Chromatogr. Sci., 36

(1998) 487.32. J. Beens and R. Tijssen, J. High Resol. Chromatogr., 20 (1997) 131.33. G.S. Frysinger and R.B. Gaines, J. High Resol. Chromatogr., 22(4) (1999) 195.

138 Chapter 7

34. G.S. Frysinger and R.B. Gaines, J. High Resol. Chromatogr., 23(3) (2000) 197.35. Z. Wang and M. Fingas and K. Li, J. of Chromatogr. Sci., 32 (1994) 361.36. Z. Wang and M. Fingas and K. Li, J. of Chromatogr. Sci., 32 (1994) 367.37. M.A. Poirier and G.T. Smiley, J. Chromatogr. Sci., 22 (1984) 304.38. I. Dzidic, H.A. Petersen, P.A. Wadsworth and H.V. Hart, Anal. Chem., 64 (1992)

2227.39. J. Beens, Chromatographic couplings for unraveling oil fractions, thesis,

University of Amsterdam, The Netherlands (1998).40. J. Bundt, W. Herbel, H. Steinhart, S. Franke and W. Francke, J. High Resol.

Chromatogr., 14 (1991) 91.

CHAPTER 8THE USE OF A SPLIT/SPLITLESS INJECTOR AS A VAPORISATION

INTERFACE FOR COMPREHENSIVE LC××××GC SEPARATIONS OF

TRIGLYCERIDES IN FOOD SAMPLES1

SUMMARY

In this chapter the performance of a new interface type for on-line comprehensiveLC×GC is described: the hot split injector. Comprehensive LC×GC was used for thecharacterisation of fats in food samples. An important step in coupling LC with GC isthe large volume injection of the LC effluent into the GC system. The performance ofthe hot split injector was tested with off-line large volume injections of standardtriglycerides (TAG) samples using a syringe pump. Discrimination of high boilingtriglycerides was absent.

On-line comprehensive LC×GC analyses for the characterisation of TAG inbutter have been performed. Obtained results with the present system are comparableto the conventional off-line LC×GC analyses of TAG. The first dimension LCseparation was performed on a silver-based stationary phase where separation of theTAG molecules was achieved based on the number of double bonds. The seconddimension GC separation was based on the boiling point or carbon number of theTAG molecules. For the on-line comprehensive LC×GC analyses the use of two LCcolumns, one with an inner diameter of 4.6 mm and one with 2 mm, were tested. Thevolumes of the LC fractions that were injected into the GC system ranged from 100(4.6 mm LC column) to 40 µL (2 mm LC column). The second dimension GC analyseswere performed on a short wide bore column. Compared to the conventional off-lineanalysis, the total analysis time of the on-line LC×GC analyses was reduced fromapproximately 24 to approximately 11 hours.

1 Paper in preparation: “The use of a split/splitless injector as a vaporisation-interface for comprehensiveLC×GC separations of triglycerides in food samples”, by M.M. v. Deursen, H.-G. Janssen, J. Beens, G. Ruttenand C.A. Cramers.

140 Chapter 8

8.1. INTRODUCTION

The combination of liquid chromatography (LC) with capillary gas chromatography(GC) is an extremely powerful tool for the chromatographic analysis of complexsamples. In an LC-GC set-up the LC part can selectively isolate the components ofinterest. High resolution GC can then be used to separate the individual compounds inthe fraction of interest and provide selective and sensitive detection. In comprehensiveLC×GC an on-line coupling of LC with GC is established where the complete LCanalysis is transferred in small fractions to the second dimension GC-system. In thepresent work a new interface type for on-line comprehensive LC×GC, the hot splitinjector, was used.

Key to successful on-line LC-GC is the large volume injection step that forms thebridge between the LC and GC. In addition to this, for comprehensive LC×GC thelarge volume step has to be fast. A conventional LC column, generally operated at aflowrate of 1 mL/min, will provide peaks of one minute wide. In order to preserve theresolution obtained in the first dimension LC separation in comprehensive LC×GC,fractions of approximately 200 µL will have to be transferred into the GC system.

For the successful injection of large sample volumes specific requirements areneeded for the GC injection system. An adequate separation has to be obtainedbetween the solvent and the solutes, and subsequently the solutes have to betransferred onto the GC column quantitatively. In literature several techniques havebeen described which allow the introduction of large volumes into the GC [1]. Thedirect “on-column” technique with the use of a retention gap is especially suitable forthe analysis of thermally labile compounds or for very volatile analytes. On-columntechniques with partially concurrent solvent evaporation have been used for thetransfer of LC fractions up to 500 µL [2]. The loop-type interface allows the injectionof similar sample volumes. Because this method uses fully concurrent solventevaporation it is less suited for samples that contain volatile analytes [3].

Next to on-column techniques, programmed temperature vaporising systems(PTV) have been developed for transferring large volumes of up to 10 mL into a GC-system. Here the sample is injected into a cooled vaporisation chamber [4-7]. In thevaporisation chamber the sample is evaporated at a temperature somewhat below thesolvent boiling point. The analytes are then transferred to the column by rapidlyincreasing the temperature of the chamber.

In each of the systems for large volume LC-GC transfer described above, asolvent-solute pre-separation is realised either in the column inlet (loop-type or on-column) or slightly before the column inlet (PTV). After this pre-separation thesolvent vapours are discharged from the system using a vapour exit and only the

The use of a split/splitless injector as a vaporisation interface for comprehensive LC×GC 141

solutes are transferred to the analytical column. These large-volume methods areeminently suited for trace analyses, where the concentration of the analytes is verylow. In the present work however, the concentration of the analytes in the LC effluentis high. Therefore, other methods for interfacing LC and GC can be considered aswell.

The use of a hot vaporisation chamber as the interface for on-line LC and GC wasdescribed by Grob et al. [8-10]. The advantage of using a high temperature is thatstrong cooling effects, normally observed in the injector upon solvent evaporation[11], can be prevented. To benefit from this temperature effect, but also to be able tovaporise the high boiling solutes, the vaporisation interface used in this work wasoperated at high temperatures.

In the first part of this chapter (§8.3) the performance of this new interface typefor comprehensive LC×GC experiments is described. For this purpose, prior to theactual coupling of LC with GC, off-line large volume GC-experiments have beenperformed using the hot split injection technique. Because the total analysis time inLC×GC largely depends on the separation speed of the second dimension, the speed ofthe GC-analysis was maximised first. This was already described in chapter 3 [12]. Toenable a fast GC analysis and to minimise the contribution of the injection band widthto the total band broadening, a fast injection is required. The high speed of injection isa major advantage of the new interface tested in this work. High injection speeds canonly be obtained by using split injections with high split flows. Conventional largevolume injections take place using pre-columns or long solvent venting times, whichwill increase the analysis-time. In the hot split chamber however, the solvent and apart of the solutes are vented through the split line very rapidly. The high boilingcompounds are refocused at the entrance of the analytical column that is kept at a lowtemperature. Results of large volume injections up to 100 µL in the off-line modeusing a syringe pump are presented. A comparison with conventional on columntechniques will be given.

In the second part of this chapter (§8.4) the feasibility of comprehensive LC×GCfor characterisation of triglycerides (triacylglycerides: TAG) in food samples using theinvestigated split injector interface, is discussed. Comprehensive LC×LC is alreadyused in several application areas such as the separation and characterisation ofbiomolecules, polymers or environmental samples [13-16]. In a comprehensiveanalysis, fractions of the first dimension effluent are sampled by the second dimensionseparation system in such a way that the separation of the first dimension is notdestroyed. The coupling of LC and GC is necessary to be able to obtain amultidimensional separation of the very complex TAG sample, which is of great

142 Chapter 8

importance in the food industry. The first separation on a silver (Ag) phase LCcolumn is based on the number of double bonds present in the TAG-molecule. Thesubsequent GC separation is based on the boiling point of the solutes. During GCanalyses the LC flow was stopped (stop-flow principle). Results will be compared tothose of off-line LC×GC analysis of fats in food samples. It will be shown that withthe comprehensive set-up described here, the analysis time could be reduced from 24to less than approximately 11 hours.

8.2. EXPERIMENTAL

8.2.1. Off-line large volume experiments

For off-line large volume experiments an HP 6890 GC (Agilent Technologies,Wilmington, DE, USA) equipped with an OPTIC 2 PTV injector (ATAS, Veldhoven,The Netherlands) and a flame ionisation detector (FID) was used. The GC instrumentwas equipped with an extra oven insert to reduce the oven volume and thus to increasethe maximum attainable temperature programming rate of the oven. The injector wasoperated in the hot split mode at a constant temperature of 425°C. The split flow andseptum purge were set at 200 and 2 mL/min, respectively. The hydrogen, air andmake-up flow (N2) of the FID (operated at 425°C) were set at 40, 450 and 40 mL/min,respectively. Helium was used as the carrier gas. Large volume injections wereperformed with a programmable Digisampler syringe pump (Gerstel GmbH, Mülheima/d Ruhr, Germany). The injection volume was 100 µL at an injection rate of 500µL/min. The analytical column that was used was a 5 m, 530 µm i.d. high temperatureUltimetal SIMDIST column, coated with 0.17 µm of a non-polar stationary phase(Varian/Chrompack, Middelburg, The Netherlands). The GC-oven was temperature-programmed from 150°C to 400°C (1 minute hold) using ballistic heating. In thistemperature range an actual average heating rate of 1.2°C/s was measured. Theheating profiles obtained showed a good repeatability. The TAG sample used for theoff-line experiments was a standard solution of triglycerides ranging from carbonnumber 24 (CN24) to carbon number 56 (CN56) at approximately 5 ppm each inhexane. Data-processing was done using HP Chemstation software (AgilentTechnologies). The inlet pressure applied was 20 kPa, which resulted in a carrier gasflow of approximately 20 mL/min.


8.2.2. On-line comprehensive LC××××GC experiments

For the on-line comprehensive LC×GC experiments an HP 1100 LC was coupled toan HP 6890 GC (Agilent Technologies). Two silver phase LC columns were used: (1)25 cm × 4.6 mm i.d. and (2) 25 cm × 2.0 mm i.d. Both columns were packed with 5µm Ag+-ion loaded silica particles obtained from Varian/Chrompack. The first seriesof experiments (analyses of fats in butter) were performed using column 1. Theinjected volume on LC column 1 was 10 µL (10 mg/mL). The column was operated ata flow of 0.5 mL/min and fractions of ≈100 µL were transferred to the GC system. Ina second series of experiments (comparison to off-line LC×GC analyses) the 2 mmi.d. column was used to obtain lower fraction volumes (40 µL). The injected volumeon column 2 was 5 µL (1 mg/mL). The eluent that was used is a methylene chloride/acetone mixture (eluent A: 99.5/0.5 v/v). After approximately 60 GC injections (≈ 5.5hours), i.e. after elution of the first three TAG-fractions (SSS, SOS and SSO, seesection 8.4.1.), acetonitrile (eluent B, mixed with 50% A) was used to elute the last(rest) fraction. The GC column was the same wide-bore column as described above.An evaporative light scattering detector (ELSD) (Cedex, Sedère, France) was used formonitoring the LC analysis prior to the LC×GC experiments, to obtain better insightin the retention times of the TAG fractions on the silver loaded LC column. It wasdisconnected during an on-line comprehensive LC×GC run. The solvent gradient usedfor an LC-ELSD analysis is shown in table 8.1.

Table 8.1: Solvent gradient used for LC-ELSD analyses (figures 8.10, 8.14B and8.15B).

Time [minutes] % B*

01520253050

00505000

* Solvents: (A): methylene chloride / acetone (99.5 / 0.5 vol%), (B): Acetonitrile (ACN)

144 Chapter 8

LC-column

Restriction capillary

FID

GC-column

PTV injector

1

Figure 8.1: Schematic drawing of the LC×GC set-up. The LC effluent is injected via adeactivated capillary (no. 1, 30 cm × 100 µm) into the GC injector.

The LC column was coupled to the GC via a six-port valve. A schematic drawingof the system is shown in figure 8.1. A deactivated fused silica capillary (30 cm × 100µm i.d.; no.1 in figure 8.1) was used to transfer the LC effluent directly into theOPTIC 2 PTV injector (ATAS). The liner used had an inner diameter of 3.4 mm andwas packed with a plug of silanised glass wool (Perkin-Elmer, Norwalk, Conn., USA).The restriction capillary was a deactivated fused silica capillary (2 m × 100 µm i.d). Inthe flow position of the valve (drawn lines), transfer of the LC fraction into the GCtakes place. In the stop-flow position (dotted lines) the LC flow is stopped and the GCseparation is performed. A small leak flow of carrier gas is obtained through therestriction capillary. In this way the liquid remaining in the transfer capillary (no. 1) isdischarged to waste and diffusion into the injector is prevented. The volume of eluentdischarged through this capillary during each second dimension run is approximately2 µL. This amount can be neglected compared to the injected amount (40-100 µL).

The GC settings used in the on-line LC×GC experiments are the same as for theoff-line experiments described above. However, the needle valve used to control thesplitflow was positioned as close as possible to the injector. Moreover, it was heatedto approximately 90°C to prevent re-condensation of the (large amount of) solvent inthe tubing between injector and split valve. Transfer of the LC fractions of 40 µL (LCcolumn 2) to 100 µL (LC column 1), was performed at a split flow of 100 and 200mL/min, respectively.


An example of a pumping cycle corresponding to one injection (programmed inHP Chemstation software) is shown in figure 8.2. During each injection-cycle the flowis gradually increased from 0 to 0.2 mL/min (and decreased from 0.2 to 0 mL/min atthe end of an injection) at a speed of 4 mL/min/min (3 seconds). By applying thisgradient and thus preventing pressure pulses the lifetime of the LC column will beincreased. This is of importance in a comprehensive LC×GC separation where 140injections have to be performed. A schematic diagram of a fully automated LC×GCoperation is shown in figure 8.3. From the flow-diagram in figure 8.2 it can becalculated that for LC column 2, fractions of 40 µL are injected into the GC-system.The relative standard deviation (RSD) of the collected fractions was lower than 5%(determined by weighing).

The samples of real butter and diet butter were all obtained from a local grocerystore. A 13 mg/mL solution of real butter was prepared by melting the butter,dissolving it in methylene chloride (DCM) and finally filtering. From the diet butteran 11 mg/mL solution was prepared by melting the butter, separating the fat from thewater phase and dissolving the fat phase in DCM. HPLC grade chemicals were used(Merck, Darmstadt, Germany). For the comparison of the described on-line LC×GCsystem with off-line LC×GC, TAG samples from Unilever Research (Vlaardingen,The Netherlands) were analysed.

0 0.05 0.1 0.15 0.2 0.25Time [min]

Flow

[mL/

min

]

Posi

tion

valv

e

0.20

0

On

Off

Injection

Figure 8.2: Schematic diagram of a (flow-programmed) injection and position of thevalve. The column flow is programmed from 0 to 0.2 mL/min at the beginning of acycle and decreased from 0.2 to 0 mL/min at the end of a cycle in 3 seconds. Apumping cycle can be programmed in the HP Chemstation software.

Flow

Valve

146 Chapter 8

0 1 2 3 4 5 6 7 8 9 10 11Time [min]

Heating

oven

Cooling oven

Heating

oven Cooling oven

Overall GC run-time : 5.25 min

Inje

ctio

n

Inje

ctio

n

LC-flow on

TGC= 400°C

LC-flow off

TGC= 150°C

Figure 8.3: Analysis diagram of a LC×GC operation (two consecutive injections).

8.3. OFF-LINE LARGE VOLUME INJECTIONS


The first step in the construction of a fully automated comprehensive LC×GC systemis the development of a suitable large volume transfer technique. Because sensitivityis not an issue in TAG-analysis, a hot vaporisation/split interface was proposed. Theadvantages of the split injector system used here compared to conventional largevolume techniques were already mentioned in the introduction of this chapter.Approximately 90% of the sample, both solvent and solutes, is homogeneouslydischarged via the split line. The splitting of a large amount of solvent most certainlyhas a positive effect on the lifetime and performance of both GC column and detector.This is of major importance in comprehensive LC×GC analyses where hundreds ofGC injections are performed.

Solute refocusing is important to prevent band broadening. Refocusing of theanalytes in the second dimension GC separation is automatically performed since thesample only contains relatively high boiling solutes (CN22 and higher). These highboiling solutes (bp > 250°C) are re-condensed into small bands at the beginning of theanalytical column. At this point the temperature of the oven is approximately 250ºCbelow the injector temperature. It is however significantly higher than the solventboiling point. Hence, the solvent will not re-condense and leave the columnunretained.


In figure 8.4 an example is shown of an injection of 100 µL of a TAG sample(CN22-CN54: only even-numbered carbon numbers) at an injection-rate of 500µL/min. A short 5 m wide bore column was used operated at a relatively high carriergas flow rate to allow rapid solvent discharge. Injecting large sample volumes into ahot injection chamber results in extremely large solvent-vapour flows that have to bedischarged through the split exit. As an example, at an injection rate of 500 µL/min,the rate of vapour generation for methylene chloride is approximately 250 mL/min.This means that, neglecting the difference in the viscosity of the vapour and the carriergas flow and taking into account the temperature of the injector and inlet-pressure, thetotal flow (split and column flow) should be at least 250 mL/min to discharge allsolvent vapour. In this work a split flow of 200 mL/min was applied because a fairamount of vapour is transported out of the system by the large column flow (≈ 20mL/min). From figure 8.4 it can be seen that the separation efficiency of this fastTAG-analysis is acceptable. Also for the solvent peak no peak tailing was observed(figures 8.5A and 8.5B). This indicates that the solvent vapour could be dischargedfrom the system sufficiently fast and that the required refocusing of the TAG at thecolumn inlet took place satisfactory.

min1 1.25 1.5 1.75 2 2.25 2.5 2.75 3

pA

100

300

500

CN24 CN52

Figure 8.4: Fast GC analysis of a standard TAG sample (10 ppm in DCM) using hotsplit conditions. Injection volume: 100 µL at 500 µL/min, splitflow: 200 mL/min.

In figure 8.5A it can be seen that the width of the solvent peak is only 30 seconds,and that the signal very sharply returns to base-line level after evaporation hasfinished. The dip in the middle of the peak is most likely caused by a pressure increase

148 Chapter 8

due to formation of solvent vapour during injection, subsequently followed by apressure decrease to correct the pressure back to the original setting (constant pressuremode). It was observed that after an injection of 100 µL or more the pressure shortlyincreased from 20 to 30 kPa, then decreased to 0 kPa and went up again to 20 kPa.This was also observed in work of Pocurull et al. [17]. This phenomenon will notinfluence the analysis of the analytes (TAG-compounds), which elute one minutelater. In figure 8.5B it is shown that for injections of 40 µL this pressure decrease wasabsent.

min0.5 1.0 1.5 2.0 2.5 3.0

pA

5*104

15*104

25*104

min0.2 0.4 0.6 0.8 1 1.2

pA

0

200000

400000

Figure 8.5: Comparison of solvent peak profiles at (off-line) GC injections of(8.5A): 100 µL and (8.5B): 40 µL.

Results of large volume hot split injections of TAG in DCM were compared to thestandard on-column injection technique and a splitless injection. This is shown infigure 8.6. The three techniques showed very similar results. The relative standarddeviation of the sample composition was below 5%. From these results it wasconcluded that discrimination of high and low boiling compounds was absent, and thatthis type of interface could be used for quantitative transfer of compounds from theLC to the GC system. Other solvents (hexane or hexane/toluene 50%/50% v/v)showed similar results. The relative standard deviation of absolute peak areas (RSD)is 7% (6 injections). The RSD of the retention time is around 1% (6 injections). Thelower repeatability of the retention times is most likely caused by the fact that theoven was heated ballistically.

In the current set-up the column flow was approximately 20 mL/min. At a splitflow of 200 mL/min this results in a split ratio of around 1 to 10 during sampleintroduction. Hence, approximately 10% of the solvent and the compounds areintroduced onto the column. This was also determined experimentally by comparing

A B


peak areas of a split injection and a normal volume splitless injection. In ‘normal’ LC-GC, i.e. LC-GC performed to isolate and quantify trace compounds in a complex andinterfering matrix, this sensitivity loss would clearly be unacceptable. In triglycerideLC×GC, as mentioned previously, the aim of the LC step is not to isolate thetriglycerides from the rest of the sample. The sample solely consists of triglycerides.The purpose of the LC step is to separate the TAG according to a certain propertyfollowed by a further GC separation according to a second, preferably orthogonal,property.

0

5

10

15

20

25

24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54Carbon number

% T

AG

(1) On-column (2) hot split 100 µL (3) splitless 5 µL

Figure 8.6: Comparison of three GC injection techniques of a standard TAG samplein DCM: (1) On-column injection, injection volume: 50 µL, column: 10 m × 530 µm ×0.17 µm (results obtained from reference [18]); (2) hot split injection, injectionvolume: 100 µL, column: 5 m × 530 µm × 0.17 µm (figure 8.4); and (3) (hot) splitlessinjection, injection volume: 5 µL, column: 5 m × 530 µm × 0.17 µm.

150 Chapter 8

8.4. ON-LINE COMPREHENSIVE LC××××GC ANALYSES


Because of the complexity of fat mixtures, the characterisation of TAG in foodsamples is a complicated task that requires an analytical method with a highseparation efficiency. The structure of a TAG molecule is shown in figure 8.7. The so-called carbon number (CN) is the total number of carbon atoms in the three fatty acidchains. A TAG molecule containing three saturated chains is named SSS, in which the“S” refers to saturated. An SOS- or SSO-TAG has one double bond present in themiddle and the outer hydrocarbon chain, respectively. The “O” here refers to oleic butalso to other fatty acids with one double bond. TAG molecules containing two doublebonds in one of the side chains and two saturated fatty acids at the other positions arenamed after “linoleic” acids: SLS or SSL. The number of different molecules in aTAG mixture is determined by the number of double bonds, the position of the doublebond in the molecule and the CN-number, and can thus easily amount to severalhundreds of compounds. The elution order in the first dimension LC separation isbased on the (complex forming) interaction of the double bonds in the TAG moleculewith the silver phase.

O

O

O

O

O

O

Figure 8.7: Molecular structure of a saturated TAG molecule (SSS, C18:0, C18:0,C14:0).

The more double bonds or the better the accessibility of the double bonds, themore retained the molecules will be. This results in the following elution order: (1)SSS; (2) SOS; (3) SSO; (4) Rest: SOO, SLS, etc. The second dimension GCseparation is based on the boiling point and thus carbon number of the molecules. Theelution order ranges from CN24 to CN54 or even higher, depending on the type offood sample. In figure 8.8 this is presented in a (schematic) two-dimensional plot.


SSSSOS

SSO

SLS

First dimension (LC): double bonds

Seco

nd d

imen

sion

(GC

): C

N

Figure 8.8: Two-dimensional separation plot of an LC×GC analysis of edible fats.Area of each GC peak is represented by the area in each circle (dotted).

min1.4 1.8 2.2 2.6 3

pA

100

200

300

400

Figure 8.9: Two consecutive GC chromatograms of an LC×GC analysis of TAG inreal butter. LC column: 4.6 mm i.d., transferred LC fractions: 100 µL, LC flow:0.5 mL/min, TAG-fraction: SSS.

In figure 8.9 two second-dimension GC chromatograms (overlay) are shown ofGC analyses of two consecutive LC fractions (4.6 mm LC column) during an LC×GCanalysis of real butter. The shift from high boiling compounds to low boilingcompounds can clearly be observed. Most likely the retention of the smaller moleculeson the LC column is higher compared to larger compounds as a result of the lower

CN56

CN28

First fractionSecond fraction

152 Chapter 8

steric hindrance. It can be seen that the repeatability of the retention times of these twoconsecutive injections is acceptable. Peaks in the second dimension separation areidentified by injecting standard solutions of triglycerides after each LC×GC run hasfinished. By comparison of retention-times, for each peak the carbon number (CN)could be assigned.

One of the disadvantages that was observed was the strong corrosion of the FIDafter only two or three completed LC×GC analyses with the 4.6 mm i.d. LC column.Also deposition of soot in both the injector and detector occurred during analysis.Methylene chloride, in combination with hydrogen from the detector flame, mostlikely forms corrosive HCl. It was observed that after 200 to 300 large volume GCinjections, the FID background level increased drastically caused by thin layers ofsoot creating a short-circuit. To prevent this, the air-flow and make-up gas (N2) wereincreased from 400 to 450 and from 20 to 40 mL/min, respectively. However, this didnot completely solve our problem. After approximately two or three LC×GC runs thedetector had to be cleaned.

The above mentioned problem only occurred when using the LC column with thelarger inner diameter (4.6 mm). In the LC×GC analyses using this column, fractions of100 µL were sampled. When using the 2 mm i.d. LC column, fractions of only 40 µLwere sampled. Here no baseline problems were seen, not even after 10 completedLC×GC analyses, or approximately 1500 injections of 40 µL.

As was already mentioned in the experimental section, an ELSD was used formonitoring the LC separation. In this way, by studying the peak widths of the LCchromatogram, the volume of the fractions to be transferred to the GC system wasestimated. In figure 8.10 an example is shown of an LC-ELSD analysis of a TAGsample containing the four groups, SSS, SSO, SOS and SLS (rest). The solventgradient that was used is shown in table 8.1. From figure 8.10 it can be seen that thepeak width at baseline level is approximately 1 minute (peak volume of 200 µL). In acomprehensive two-dimensional separation the number of samples transferred to theGC system should be high enough to keep the separation in the first dimension intact.For a first dimension peak of the LC analysis shown in figure 8.10 this means that atleast 4 to 5 re-injections into the GC system are required. Therefore the volume of thetransferred LC fractions was 40 µL.

The total analysis time of the first dimension LC separation is ≈ 27 minutes(figure 8.10). When a sampling volume of 40 µL is chosen, a total of 135 fractionshave to be transferred to the GC system in a LC×GC analysis of the four triglyceridesgroups. Every first 10 fractions of an on-line analysis however are discharged, sincethey do not contain compounds of interest. When a second dimension separation time


of 5.5 minutes is assumed the total analysis time amounts to 11 hours. Compared tothe conventional off-line analysis (24 hours) this means a decrease in analysis-time ofapproximately 13 hours. For future research however, the LC separation should beoptimised (by changing the solvent composition) to further reduce the elution-time ofthe rest fraction to less than 15 minutes. Then a total analysis-time of approximately 7hours is within reach.

min10 20 30 40

mV

8

12

16

Figure 8.10: LC-ELSD analysis of a standard TAG sample containing the fourgroups: SSS, SOS, SSO and SLS (rest) in methylene chloride (≈ 100 ppm). LC column:25 cm × 2 mm, LC-flow: 0.2 mL/min, nebuliser temperature ELSD: 70°C, pressure(N2): 2.5 bar, Vinj,LC = 5 µL . Solvent gradient: table 8.1.

In figure 8.11 an example is shown of a two dimensional contour plot of an on-line LC×GC analysis (using the novel interface) of the same TAG sample as shown inthe LC-ELSD analysis of figure 8.10. The separation of the triglycerides in the fourgroups can clearly be seen in both figures. Also the differences in intensity betweenthe groups (for example the low concentration of the SSS-group) is very similar inboth figures.

A comparison was made between the automated comprehensive LC×GC set-upused in this work and manual off-line LC×GC. In the off-line mode the firstdimension LC separation was performed on a column with an inner diameter of 4.6mm and fractions ranging from 0.3 to 4 mL were collected during analysis (seecaption figure 8.12). After completion of the LC analysis, second dimension GCanalyses were performed by injecting approximately 3 µL from each of the collectedfractions (on-column). Figure 8.12 shows a two-dimensional contour plot of such ananalysis. In this figure the same TAG sample was analysed as is shown in figures 8.10and 8.11. The elution patterns in both figures are very similar.

SSS

SOS SSO

Rest

154 Chapter 8

38

40

42

44

46

48

50

52

54

56

58

60

0 10 20 30 40 50 60 70 80 90 100 110

Fraction no.

Car

bon

num

ber (

CN

)

SSS

RestSOS SSO

Figure 8.11: Two-dimensional contour plot of an on-line comprehensive LC×GCanalysis of the standard TAG sample shown in figure 8.10. LC column: 2 mm i.d., LCflow: 0.2 mL/min, transferred fraction: 40 µL, Vinj,LC = 5 µL. Total number oftransferred fractions: 100. The first 400 µL of the LC-effluent after starting the LCanalysis was discharged, then the coupling was made and the actual LC×GC analysiswas started (rest = SLS). Insert: enlargement of SOS and SSO groups.

38

40

42

44

46

48

50

52

54

56

58

60

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28

Fraction Number

Car

bon

Num

ber

Figure 8.12: Two-dimensional contour plot of an off-line comprehensive LC×GCanalysis of a standard TAG sample of figure 8.10. The SSS fraction was collectedfrom 3-6 minutes: total fraction volume = 3.0 mL. The critical SOS/SSO elution areawas divided in 300 µL fractions. The rest fraction was collected from 41-45 minutes ina fraction volume of 4 mL (from: Unilever Research).

SOS SSO Rest

SSS


0

1000

2000

3000

4000

5000

6000

0 5 10 15 20 25Time [minutes]

Area

sum

med

SSS

Rest

SOS SSO

Figure 8.13: Reconstructed first dimension LC separation from the comprehensiveLC×GC analysis shown in figure 8.11 (rest = SLS).

In figure 8.13 the reconstructed first dimension chromatogram of the on-linecomprehensive LC×GC analysis from figure 8.11 is shown. This chromatogram isobtained by summation of the peak areas of the fractions. The elution pattern of thisfigure corresponds very well to the LC-ELSD analysis of figure 8.10. However, theelution times of the different groups are not exactly the same. Most likely this iscaused by the fact that in the comprehensive LC×GC analysis the stop flow principleis used, whereas in the ELSD analysis a continuous flow was applied. Anotherexplanation could be that the true injected volume is somewhat smaller than wascalculated from the flow program. For the reconstruction of the first dimensionseparation of figure 8.12 a volume of 40 µL was used. This can explain the higher(calculated) retention times of the SOS and SSO groups. The earlier elution of the rest(SLS)-group is caused by a change in the solvent composition for eluent B (100%ACN instead of 50%).

In figure 8.14A and 8.15A examples are shown of on-line LC×GC analyses of realbutter and diet butter, with the corresponding LC-ELSD chromatograms in 8.14B and8.15B. Again the elution pattern of the TAG groups in the contour plot could beconfirmed with ELSD detection. The differences between the real and diet butter canbe seen clearly from the two-dimensional contour plots in figures 8.14A and 8.15A.For example the concentration of SSS TAG in real butter is much larger than in dietbutter. The SOS and SSO TAG groups are present in real butter but are absent in dietbutter. The LC-ELSD chromatograms in figures 8.14B and 8.15B both confirm theseobservations.

156 Chapter 8

24

28

32

36

40

44

48

52

56

60

Car

bon

num

ber [

CN

]

Figure 8.14A: Oanalysis of fracti

mV

6

8

10

ADC1 A

Figure 8.14B: Lfigure 8.10.

SSS SOS SSO Rest

0 10 20 30 40 50 60 70 80 90 100

Fraction no.

n-line comprehensive LC×GC analysis of TAG in real butter. Afteron no. 46 the LC eluent was changed from A to 100% B (ACN).

, ADC1 CHANNEL A (MARIEKE1\ELSD2.D)

SSS Rest

SOS SSO

min5 10 15 20 25 30 35 40 45

C-ELSD analysis of TAG in real butter (≈ 100 ppm). Conditions: see


6101418222630343842465054586266

0 10 20 30 40 50 60 70 80 90 100

Fraction no.

Car

bon

num

ber (

CN

)

Figure 8.15A: On-line comprehensive LC×GC analysis of TAG in diet butter. Afteranalysis of fraction no. 11 the LC eluent was changed from A to 100% B (ACN).

min 5 10 15 20 25 30 35 40 45

mV

6

10

14

18

ADC1 A, ADC1 CHANNEL A (MARIEKE1\ELSD1.D)

Figure 8.15B: LC-ELSD analysis of TAG in diet butter. Conditions: see figure 8.10.

8.5. CONCLUSIONS

The on-line coupling of an LC and GC system to obtain a comprehensive LC×GC set-up by using a hot split injector as the interface was successfully applied for theanalysis of triglycerides. The performance of the novel interface was satisfactory.Important in the coupling of LC with GC is miniaturisation of the LC separation. Thisenables the possibility to transfer and inject smaller fractions (40 µL) into the GC

Rest

SSS

Rest

SSS

158 Chapter 8

system, which is advantageous to an optimal use of the detector. Results obtained inthis work are very similar to the results obtained using an off-line LC×GC system.The technique provides a high separation efficiency for the analysis of the verycomplex TAG samples. Moreover, the total analysis time compared to theconventional analysis could be decreased from 24 to 11 hours.

8.6. REFERENCES

1. K. Grob, M. Biedermann, J. Chromatogr. A, 750 (1996) 11.2. F. Munari, P.A. Colombo, P. Magni, G. Zilioli, S. Trestianu, K. Grob, J.

Microcolumn Separtions, 7 (1995) 403.3. K. Grob, On-line coupled LC-GC, Hüthig Buch Verlag Heidelberg (Germany),

(1991) 34.4. L. Mondello, P. Dugo, G. Dugo, A.C. Lewis, K.D. Bartle, J. Chromatogr. A, 842

(1999) 373.5. W. Engewald, J. Teske, J. Efer, J. Chromatogr. A, 842 (1999) 143.6. J. Teske, J. Efer, W. Engewald, Chromatographia, 47, 1/2 (1998) 35.7. J. Villén, F.J. Señoráns, M. Herraiz, J. Microcolumn Separations, 11(2) (1999) 89.8. K. Grob, M. Biedermann, J. Chromatogr. A, 897 (2000) 237.9. K. Grob, M. Biedermann, J. Chromatogr. A, 897 (2000) 247.10. U. Boderius, K. Grob, M. Biedermann, J. High Resol. Chromatogr., 18 (1995)

573.11. H.G.J. Mol, H.-G. Janssen, C.A. Cramers, U.A.Th. Brinkman, J. High Resol.

Chromatogr., 18 (1995) 19.12. M.M. van Deursen, H.-G. Janssen, J. Beens, G. Rutten, C. Cramers, J.

Microcolumn Sep., 13(8) (2001) 337.13. G.J. Opiteck, J.W. Jorgenson, M.A. Mosely III, R.J. Anderegg, J. Microcolumn

Separations, 10(4) (1998) 365.14. R.E. Majors, J. Chromatogr. Sci., 18 (1980) 571.15. F. Erni, R.W. Frei, J. Chromatogr. A, 149 (1978) 561.16. W.W.C. Quigley, G.F. Carlos, R.E. Synovec, J. Microcolumn Separations, 12(3)

(2000) 160.17. E. Pocurull, M. Biedermann, K. Grob, J. Chromatogr. A, 876 (2000) 135.18. H.-G. Janssen, H. Steenbergen, J. Oomen, J. Beens, J. Microcolumn Separations,

12(10) (2000) 523.

SUMMARY

Due to the growing demand for high throughput analyses, the past ten years a lot ofemphasis has been put on the development of fast gas chromatographic (GC)instrumentation. In this thesis a description is given of the development of newtechniques for fast GC analyses. The use of dedicated instrumentation to reduce theGC analysis time for several samples of industrial origin is described.

In chapter 2 the theoretical aspects of fast GC are described. From the theory ofchromatography it can be deduced that many methods are possible to increase thespeed of GC analyses. These methods comprise the use of: narrow bore columns,short columns, high carrier gas velocities, vacuum conditions, hydrogen as carrier gas,multi-capillary columns, fast temperature programming or turbulent flow conditions.The method of choice however, has to be selected with great care. To speed up theanalysis of a complex mixture the high separation efficiency of a narrow bore columnand the use of hydrogen as the carrier gas is needed. For increasing the speed ofanalysis of a mixture with a low number of compounds with widely differing boilingpoints requiring only a low plate number, shorter columns, higher carrier gasvelocities or fast temperature programming should be the method of choice. In generalover-separation of a sample (resolution > 1.5 for a critical pair) is a waste of time andshould clearly be avoided in fast GC. An important issue in fast GC is to minimise thecontribution of the injection band broadening to the total band width of a gaschromatographic peak. For this purpose adequate injection systems were developed inthis work. A drawback of fast GC using narrow bore columns is the low samplecapacity and hence the low dynamic range that can be obtained.

In chapter 3 fast temperature programming by using a resistively heated column isdescribed. This method is applied to speed up the analysis of several industrialapplications. With this instrument heating rates of up to 20°C per second could beobtained, compared to a maximum of 0.5°C per second obtained for a conventionalGC oven. As fast temperature programming will decrease the separation efficiency thetechnique is applied for several samples requiring a low plate number. By using thistechnique, in general a reduction of the analysis time of approximately 5 to 10 timescould be obtained. The disadvantage of the resistive heating system used in this workis the relatively low maximum operation temperature (275°C). Therefore this systemis less suitable for mixtures containing compounds with high boiling points

160 Summary

(> 500°C). With in-house developed software, based on the plate-height equation, theoptimal conditions were predicted for a fast separation of a mixture of triglycerides.For this purpose a short wide bore column and a GC oven capable of fast temperatureprogramming (1°C per second) and capable of heating up to 425°C was used. Theanalysis time could be reduced from 17 to less than 3 minutes.

In chapter 4 the use of a so-called multi-capillary column for fast separations isdescribed. High analysis speeds could be obtained using this column, which consistsof a bundle of 919 short narrow bore columns (1 m × 43 µm). For producing this typeof column, very stringent manufacturing requirements have to be met in order toobtain a uniform film thickness and inner diameter in each of the individualcapillaries. It is shown that minor deviations in these parameters cause significantextra column band broadening. Examples of fast separations within two minutes of n-alkanes and fatty acid methyl esters (FAMES), performed with the multi-capillarycolumn, are shown. The advantage of the multi-capillary column compared to a singlenarrow bore column is the much higher sample capacity (1000 ng compared to 1 ng).With this column a plate number of 12,500 could be obtained. The column is thereforeonly suitable for speeding up analyses of relatively simple mixtures.

The use of vacuum outlet conditions as a route towards fast GC is discussed inchapter 5. Low pressure conditions in the analytical column were used to increase thediffusion process of solutes in the mobile phase. For this purpose a wide bore column(10 m × 530 µm) was coupled to a mass spectrometer. To reduce the column flow andto be able to operate the column at vacuum conditions two types of flow reductionwere applied: a narrow bore pre-column and an SFC-restriction. In addition to this,injections at sub atmospheric pressures were performed using a fast rotating injectionvalve. With this method the analysis time of several industrial samples could bereduced 3 to 4 times compared to a conventional analysis using atmospheric outletpressures. With this method a plate number of 20,000 could be obtained. The methodtherefore is only suitable for relatively simple mixtures.

In GC the use of mass spectrometric detection for direct identification of unknowncompounds is very important. In fast GC the detection device has to be fast in order toaccurately describe a chromatographic peak. In chapter 6 the performance of a time-of-flight mass spectrometer (TOF-MS) was tested for very fast separations on a 1 m ×50 µm i.d. column. A cryogenic injection system was used which was capable ofproducing the required narrow input band widths (σi) of several milliseconds. Withthis system a separation of 10 compounds in 500 ms was obtained. It is shown thatTOF-MS is required for an adequate identification of compounds in separations in the

Summary 161

milliseconds range. The quality of the obtained spectra is high. Consequentlyautomatic deconvolution can be used for identification of co-eluting compounds.

An important recent application of fast GC is in comprehensive two-dimensionalGC (GC×GC). With this method (described in chapter 7) a very high separation poweris obtained which is required for separating complex samples. The method is thereforevery suitable for characterisation of oil samples. The first dimension separation isbased on the boiling point of the compounds and is performed on a non-polar 10 m ×250 µm i.d. capillary column. Each peak eluting from the first dimension column isre-injected 2 to 3 times onto a polar second dimension column (70 cm × 100 µm), anda separation based on polarity or chemical structure takes place. In the seconddimension a very fast separation (seconds range) is obtained and therefore a fastdetection method is required. In this work GC×GC-TOFMS was used for thecharacterisation of oil samples. Using this method a high selectivity was obtained forthe analysis of group-types, which could be presented in separate contour plots byselecting unique ion fragments for each group. This was also achieved for sulphurcontaining compounds in oil. By using deconvolution co-eluting compounds likenaphthalenes and benzothiophenes or alkanes and olefins could be identifiedpositively. From the elution pattern of a two-dimensional contour plot the type of oilcan be distinguished much easier than from a complex one-dimensional GC-MSchromatogram of an oil analysis. The concentration of aromatic (mono-, di- and tri-aromatic) and sulphur compounds in oil, which were determined using GC×GC-TOFMS, ranged from 2 to 8000 mg/kg oil.

In chapter 8 the development of an on-line comprehensive LC×GC system forcharacterisation of complex triglycerides (TAG) mixtures from the food industry isdescribed. The crucial step in coupling liquid chromatography (LC) with GC is thelarge volume sampling step for transferring LC effluent into the GC system. In thiswork a new interface type was developed: the hot split vaporisation chamber.Compared to conventional large volume on-column or PTV large volume injectiontechniques, the hot split injection technique used in this work will allow a fastinjection into the GC system. A high injection speed is required to minimise injectionband broadening in the fast second dimension GC separation. Due to the highconcentration of the TAG in the LC effluent it was not necessary to transfer all solutesinto the GC system. Thus the use of a split injection was justified. In a comprehensiveLC×GC system the first dimension separation was performed on a silver basedstationary phase LC column where triglycerides molecules are separated according tothe number of double bonds present in the molecule. Subsequently a GC separation isperformed on a short wide bore column based on the boiling point of the molecules.

162 Summary

With this method TAG analyses of butter samples have been performed. Results arecompared to the conventional off-line LC×GC analyses. Using the system describedhere the analysis time could be reduced from approximately 24 to 11 hours comparedto the conventional analysis.

SAMENVATTING

Door de steeds stijgende vraag naar het analyseren van grote aantallen monsters in eenzo kort mogelijke tijd, is de ontwikkeling van instrumentatie voor snelle gaschromatografische technieken de afgelopen tien jaar enorm gegroeid. In dit onderzoekzijn nieuwe technieken voor snelle GC ontwikkeld en is speciaal voor snelle GContworpen instrumentatie met succes toegepast om de analyses van verschillendemonsters van industriële origine aanzienlijk te versnellen.

In hoofdstuk 2 worden de theoretische aspecten van snelle GC beschreven. Aan dehand van de theorie kan worden afgeleid dat er vele methoden en technieken mogelijkzijn om een GC analyse te versnellen. Hiertoe behoren: capillaire kolommen met eenkleine diameter, korte kolommen, hoge draaggas snelheden, vacuüm condities,waterstof als draaggas, multi-capillaire kolommen, snelle temperatuur programmeringof turbulente stroming. Gezien de specifieke voor- en nadelen van elk van dezetechnieken moet de uiteindelijke keuze weloverwogen worden gemaakt. Om descheiding van een complex mengsel te versnellen zijn de hoge scheidingsefficiëntievan een nauwe kolom en het gebruik van waterstof als draaggas noodzakelijk. Eeneenvoudige analyse van een monster met weinig componenten die zeer verschillendekookpunten hebben en waarvoor het vereiste schotelgetal laag is, kan eenvoudigversneld worden door het gebruik van korte kolommen, hoge draaggas snelheden ofsnelle temperatuur programmering. Een algemeen geldende regel in snelle GC is dateen analyse met een te hoge scheiding (d.w.z. een resolutie > 1.5 voor een kritischpiekpaar) onnodig veel tijd kost en dus moet worden voorkomen. Een ander belangrijkpunt in snelle GC is het feit dat de bijdrage van de injectie bandbreedte aan de totalechromatografische piekbreedte zo klein mogelijk gehouden moet worden. Daarom zijnin dit onderzoek injectie methoden ontwikkeld die passend zijn voor snelle GC. Eennadeel van snelle GC bij het gebruik van kolommen met een kleine diameter is de lagemonster capaciteit en het daardoor beperkte dynamische bereik van de methode.

In hoofdstuk 3 is het gebruik van snelle temperatuur programmering met behulpvan directe verwarming van de kolom d.m.v. een weerstandsdraad beschreven. Dezemethode is gebruikt om de analyse van diverse industriële monsters te versnellen. Metdit systeem kan een opwarmsnelheid tot 20ºC per seconde worden bereikt, wat veelhoger is dan de opwarmsnelheid van 0.5ºC per seconde van een conventionele GCoven. Omdat bij snelle temperatuur programmering de scheidingsefficiëntie omlaaggaat wordt de techniek alleen toegepast voor analyses die een laag schotelgetalvereisen. Gebruik makend van deze techniek kunnen GC analyses met een factor 5 tot

164 Samenvatting

10 worden versneld. Het nadeel van het snelle temperatuur systeem is het feit dat demaximaal toegestane temperatuur slechts 275ºC is, waardoor dit systeem mindergeschikt is voor componenten met een zeer hoog kookpunt (> 500ºC). Tevens wordtin dit hoofdstuk een optimalisatie uitgevoerd voor een snelle scheiding vantriglyceriden (vetten). Berekeningen zijn uitgevoerd met software gebaseerd op deschotelhoogte vergelijking. Voor de GC analyses werd gebruik gemaakt van een kortekolom met een wijde diameter en een snelle GC oven die met een snelheid van 1ºCper seconde werd opgewarmd tot 425ºC. De analyse tijd van triglyceriden kon wordenteruggebracht van 17 naar minder dan 3 minuten.

In hoofdstuk 4 is het gebruik van een zogenaamde multi-capillaire kolom voorsnelle scheidingen beschreven. Met deze kolom, die bestaat uit een bundel van 919parallel geschakelde kolommen met een lengte van 1 meter en een diameter van 43µm (1 m × 43 µm) kunnen hoge analyse snelheden worden bereikt. Bij de productievan een dergelijke kolom moet echter goed gelet worden op een uniforme diameter,filmdikte en lengte van elk individueel capillairtje. In dit hoofdstuk wordt beschrevendat afwijkingen in een van deze parameters extra bandverbreding kan veroorzaken.Het voordeel van de multi-capillaire kolom in vergelijking met een enkelvoudigenauwe kolom is de veel hogere monster capaciteit die wordt verkregen (circa 1000 ngin vergelijking met 1 ng, respectievelijk). De kolom had echter een schotelgetal van12.500 en is daarom voornamelijk geschikt voor relatief eenvoudige scheidingen.Tenslotte worden enkele voorbeelden gegeven van zeer snelle scheidingen van fattyacid methyl esters (FAMES) en n-alkanen.

Het gebruik van vacuüm condities als methode om GC analyses te versnellenwordt beschreven in hoofdstuk 5. Een lage druk in een capillaire kolom heeft eenverhoogde diffusiecoëfficiënt van de componenten in de mobiele fase tot gevolg. Omdeze lage drukken te bereiken werd een kolom met een grote diameter gekoppeld aaneen massaspectrometer. Om de kolom flow te reduceren en om vacuüm condities in degehele kolom te verkrijgen werden twee methoden toegepast: een nauwe voorkolomen een SFC restrictie. Bovendien werden injecties bij subatmosferische drukkenuitgevoerd door gebruik te maken van een snelle injectie kraan. Op deze manierkonden de analyses van een aantal industriële monsters 3 tot 4 keer sneller wordenuitgevoerd dan met een analyse onder conventionele atmosferische condities. Metdeze methode werd een schotelgetal van 20.000 verkregen. De methode is daaromgeschikt voor het versnellen van relatief eenvoudige monsters.

In GC is het gebruik van een massaspectrometer voor de directe identificatie vanonbekende componenten heel belangrijk. Tevens dient de data acquisitie snelheid vaneen MS in snelle GC hoog te zijn om de nauwe pieken nauwkeurig te kunnen

Samenvatting 165

beschrijven. In hoofdstuk 6 wordt de prestatie van een TOF-MS getest voor extreemsnelle scheidingen op een zeer korte kolom (1 m × 50 µm). Door gebruik te makenvan een cryogeen injectie systeem, werden zeer nauwe injectie band breedtes (σi) vanenkele milliseconden breed verkregen. Op deze manier werd een scheiding van 10componenten in 500 milliseconden bewerkstelligd. In dit hoofdstuk wordt beschrevendat een snelle TOF-MS nodig is voor een nauwkeurige identificatie van componentenin milliseconden scheidingen. De kwaliteit van de verkregen MS-spectra is hoog.Door gebruik te maken van automatische deconvolutie werden tevens co-eluerendecomponenten geïdentificeerd.

Een van de meest belangrijke toepassingen van snelle GC is in twee dimensionaleGC (GC×GC). Met deze methode (beschreven in hoofdstuk 7) kan een heel hoogscheidend vermogen worden verkregen. De methode is daarom uitermate geschiktvoor de karakterisering van olie monsters. De scheiding in de eerste dimensie vindtplaats op basis van de kookpunten van de componenten en wordt uitgevoerd op eenapolaire capillaire kolom (10 m × 250 µm). Elke piek die uit de eerste kolom elueertwordt 2 à 3 keer opnieuw geïnjecteerd op een zeer korte polaire tweede dimensiekolom (70 cm × 100 µm), waar de scheiding plaatsvindt op basis van de polariteit ofde chemische structuur van de componenten. De scheiding in de tweede dimensie iszeer snel (seconden scheiding) waardoor ook een zeer snelle detectie methode vereistis. In dit onderzoek is GC×GC-TOFMS gebruikt voor de karakterisering van oliemonsters. Met deze methode kan een zeer hoge selectiviteit worden verkregen voor deanalyse van groep-typen in olie, die elk apart kunnen worden gepresenteerd in contourplots door voor elk groep-type een karakteristiek ion-fragment te selecteren. Op dezemanier kunnen ook zwavelhoudende componenten worden aangetoond. Door gebruikte maken van speciale deconvolutie-software kunnen zelfs overlappende pieken zoalsnafthalenen en benzothiofenen of alkanen en olefinen worden geïdentificeerd. Van hetelutie-patroon van een twee dimensionaal contour plot van een olie analyse kan veeleenvoudiger het type olie worden vastgesteld dan van een complex eendimensionaalGC-MS chromatogram van een olie analyse. De concentratie van aromatischecomponenten (mono-, di- en tri-aromatisch) en van zwavelhoudende componenten inolie bepaald met GC×GC-TOFMS lag tussen de 2 en 8000 mg/kg olie.

In hoofdstuk 8 wordt de ontwikkeling van een “comprehensive” LC×GC- systeemvoor de karakterisering van complexe triglyceriden in voedsel monsters beschreven.Een belangrijke stap in de koppeling van vloeistofchromatografie (LC) met GC is debemonstering van grote volumina van het LC effluent in het GC-systeem. In ditonderzoek werd voor dit doel een nieuw type interface ontwikkeld: een hogetemperatuur split-injector. Met dit type injector zal, in tegenstelling tot conventionele

166 Samenvatting

groot volume injectie technieken, een snelle injectie kunnen worden uitgevoerd in hetGC systeem. Een snelle injectie is vereist om extra bandverbreding in de snelle tweededimensie GC analyse zoveel mogelijk te voorkomen. Vanwege de hoge concentratievan triglyceriden in het LC effluent was het niet noodzakelijk om alle componenten inhet GC systeem te injecteren. Door gebruik te maken van een split-injectie wordtslechts een gedeelte van het monster in het GC-systeem geïnjecteerd. De eerstedimensie LC scheiding in het comprehensive LC×GC systeem werd uitgevoerd op eenzilver fase LC kolom waar de triglyceride moleculen werden gescheiden op basis vanhet aantal dubbele bindingen in het molecuul. Daaropvolgend vond de tweededimensie GC scheiding op een korte capillaire kolom plaats op basis van dekookpunten van de componenten. Met dit systeem werden analyses uitgevoerd vantriglyceriden in boter monsters. Resultaten van het on-line comprehensive LC×GC-systeem werden vergeleken met resultaten verkregen met een conventioneel off-line“comprehensive” LC×GC systeem. In vergelijking met het conventionele systeem,kon de totale analyse tijd worden verkort van 24 naar 11 uur.

DANKWOORD

Een promotie-onderzoek is niet iets dat je alleen doet. Graag wil ik via deze weg devele mensen bedanken met wie ik de afgelopen vier jaar heb samengewerkt, en diehebben bijgedragen aan de zeer goede werksfeer op de universiteit.

In de eerste plaats wil ik mijn promotor prof.dr.ir. Carel Cramers bedanken voorhet bieden van de mogelijkheid dit promotie-onderzoek uit te voeren. Speciale dankgaat uit naar Hans-Gerd Janssen. Jouw enthousiasme en kennis van de chromatografiewaren een enorme stimulans bij het uitvoeren van dit onderzoek. De velewerkbesprekingen die we hebben gevoerd hebben altijd weer geleid tot eenverbetering en verdere ontwikkeling van chromatografische technieken eninstrumentatie en opstellingen op het lab. Bedankt dat je me de mogelijkheid hebtgeboden het laatste stukje werk van dit onderzoek bij Unilever uit te voeren. JanBeens wil ik graag bedanken voor de goede begeleiding. Samen met Jan Blomberghebben we met succes als allereerste analytische groep ter wereld een comprehensiveGC×GC-systeem gekoppeld aan een time-of-flight massaspectrometer. En eigenlijkwas een massaspectrometer overbodig: jij kon elk piekje met het blote oogidentificeren. Bedankt voor de vele malen dat je helemaal vanuit Amsterdam naarEindhoven bent gereisd voor besprekingen. Peter Lipman, heel erg bedankt voor dekeren dat je mijn metingen hier op het lab hebt overgenomen als ik weer eens opcongres was. Roy Reinierkens, bedankt voor de prettige samenwerking tijdens jeafstudeerwerk. Je bent een ster in het oplappen van GC’s en je hebt aangetoond dateen FID toch echt niet bestand is tegen water. Erik Vonk, altijd enthousiast over jegrootste hobby: de sport. Gerard Rutten, jouw jarenlange ervaring in degaschromatografie was onmisbaar bij het uitvoeren van dit onderzoek. Marion vanStraten, heel erg bedankt voor je hulp bij het opzetten van ons comprehensiveLC×GC-systeem. Piet Leclercq heeft mij geleerd massaspectra te interpreteren. Helaasheeft onze samenwerking maar twee jaar mogen duren. Verder wil ik DeniseTjallema, Jetse Reijenga, Henk Claessens, Harrie Maathuis, Hans van Rijsewijk, GiusRongen, Huub van Leuken, Anton Bombeeck, Mark van Lieshout, Anja van Ysacker,alle collega’s van CAS (Unilever) en voor de rest iedereen van onze “oud”-vakgroepInstrumentele Analyse die ik nog niet vermeld heb, bedanken voor de prettigesamenwerking. De gezellige lunches, borrels en uitstapjes hebben deze tijdonvergetelijk gemaakt.

ICI (England) is gratefully acknowledged for the financial support of thisresearch. Nigel Wilson, thank you for visiting our university several times to discuss

168 Dankwoord

my work. Dow Chemicals is gratefully acknowledged for sponsoring this researchwith a University Development Fund. Shell Research and Technology CentreAmsterdam (SRTCA) wil ik graag bedanken voor de samenwerking in het GC×GC-TOFMS project.

Heel graag wil ik ook mijn ouders bedanken voor hun steun en belangstellingtijdens mijn onderzoek. Echt geweldig, jullie vertrouwen in mij heeft ervoor gezorgddat ik deze promotie met goed gevolg heb afgerond. Saskia en Gerard, erg fijn datjullie altijd zo geïnteresseerd waren in de vorderingen van mijn onderzoek. Verder wilik familie, vrienden en kennissen bedanken voor hun getoonde interesse.

Arjan, natuurlijk was jij mijn grote steun en toeverlaat bij het schrijven van ditproefschrift. Jouw positieve instelling, enthousiasme en belangstelling voor mijnonderzoek waren onontbeerlijk. Ik beloof je dat ik vanaf nu weer heel vaak mee gafietsen.

CURRICULUM VITAE

Marieke van Deursen werd op 21 oktober 1969 geboren te Weert. Aanscholengemeenschap Bisschoppelijk College te Weert werd in 1988 het diplomagymnasium β behaald. In hetzelfde jaar werd begonnen met de studie ScheikundigeTechnologie aan de Technische Universiteit Eindhoven. Het afstudeeronderzoek“Concentrerings-technieken van luchtverontreinigingen” werd uitgevoerd bij devakgroep Instrumentele Analyse onder leiding van prof.dr.ir. C.A. Cramers. In 1994werd het ingenieursdiploma behaald. Aansluitend hierop startte zij in juli 1994 alsprojectmedewerkster Luchtverontreiniging en Externe veiligheid bij de MilieudienstAmsterdam. Van november 1996 tot januari 1998 was zij werkzaam bij PTRL(Pharmacology and Toxicology Research Laboratory) te Ulm (Duitsland). Het in ditproefschrift beschreven onderzoek werd gestart in februari 1998 in de groep van prof.Cramers. Vanaf februari 2002 is zij werkzaam bij het Nederlands Forensisch Instituutte Rijswijk.

170 Curriculum vitae

BIBLIOGRAPHY

1. M.M. van Deursen, M. van Lieshout, R. Derks, H.-G. Janssen and C.A. Cramers,Theoretical design considerations for multi-capillary columns in fast gaschromatography, Journal of High Resolution Chromatography, 22(2) (1999) 119-122.

2. M.M. van Deursen, J. Beens, H.-G. Janssen and C.A. Cramers, Possibilities andlimitations of fast temperature programming as a route towards fast GC, Journalof High Resolution Chromatography, 22(9) (1999) 509-513.

3. M.M. van Deursen, H.-G. Janssen, J. Beens, P. Lipman, R. Reinierkens, G. Ruttenand C.A. Cramers, Fast gas chromatography using vacuum outlet conditions,Journal of Microcolumn Separations, 12(12) (2000) 613-622.

4. M.M. van Deursen, J. Beens, J. Reijenga, P. Lipman, J. Blomberg and C.A.Cramers, Group-type identification of oil samples using comprehensive two-dimensional gas chromatography coupled to a time-of-flight mass spectrometer(GC×GC-TOFMS), Journal of High Resolution Chromatography, 23(7/8) (2000)507-510.

5. J. Beens and M. van Deursen, Trends in gas chromatography. Separation ofcomplex samples by super high-speed GC, Chemisch2Weekblad, 96(11) (2000)33- 35.

6. M.M. van Deursen, J. Beens, H.-G. Janssen, P.A. Leclercq and C.A. Cramers,Evaluation of time-of-flight mass spectrometric detection for fast gaschromatography, Journal of Chromatography A, 878(2) (2000) 205-213.

7. M.M. van Deursen, H.-G. Janssen, J. Beens, G. Rutten and C.A. Cramers, Designconsiderations for rapid-heating columns applied in fast capillary gaschromatography, Journal of Microcolumn Separations, 13(8) (2001) 337-345.

8. M.M. van Deursen, J. Beens, J. Blomberg and C.A. Cramers, Quantification ofaromatic and sulphur compunds in petroleum products using comprehensive two-dimensional gas chromatography coupled to time-of-flight mass spectrometry,submitted for publication in Journal of Chromatography A.

9. M.M. van Deursen, H.-G. Janssen, J. Beens, G. Rutten and C.A. Cramers, The useof a split/splitless injector as a vaporisation-interface for comprehensive LC×GCseparations of triglycerides in food samples, paper in preparation.

172 Bibliography

10. M. van Lieshout, M.M. van Deursen, R. Derks, H.-G. Janssen and C.A. Cramers,The influence of liner dimensions on injector band broadening in split injections infast capillary gas chromatography, Journal of High Resolution Chromatography,22(2) (1999) 116-118.

11. M. van Lieshout, M.M. van Deursen, R. Derks, H.-G. Janssen and C.A. Cramers,A practical comparison of two recent strategies for fast gas chromatography:packed capillary columns and multicapillary columns, Journal of MicrocolumnSeparations, 11(2) (1999) 155-162.

12. C.A. Cramers, H.-G. Janssen, M.M. van Deursen and P.A. Leclercq, High-speedgas chromatography: an overview of various concepts, Journal of ChromatographyA, 856 (1/2) (1999) 315-329.

Stellingen behorende bij het proefschrift


van Marieke van Deursen 1) In “comprehensive” twee-dimensionale GC verdient een her-injectie systeem zonder bewegende

delen de voorkeur.

J. Beens, M. Adahchour, R.J.J. Vreuls, K. van Altena en U.A.Th. Brinkman, J. of Chromatogr. A, 919, 2001, 127-132.

2) In werk van Amirav et al. wordt ten onrechte beweerd dat het gebruik van een “time-of-flight”

massa spectrometer in combinatie met zeer snelle GC-scheidingen niet noodzakelijk is.

A. Amirav, N. Tzanani, S.B. Wainhaus and S. Dagan, Eur. Mass Spectrom., 4, 7-13, 1998.

3) Het gebruik van snelle temperatuur programmering voor kortere analyse tijden verdient de voorkeur boven toepassing van korte kolommen.

Dit proefschrift, hoofdstuk 3

4) Door gebruik te maken van twee-dimensionale analyse technieken kan de samenstelling van complexe mengsels beter worden bepaald.

Dit proefschrift. hoofdstuk 7 en 8

5) Snelle temperatuur programmering toegepast voor een snelle elutie van hoogkokende triglyceriden vereist de ontwikkeling van meer stabiele stationaire fasen.

Dit proefschrift, hoofdstuk 8

6) In de publicatie van Wheeler et al. over het gebruik van weerstand verwarmde GC kolommen zijn hoogkokende componenten ten onrechte buiten beschouwing gelaten.

S.J. MacDonald en D. Wheeler, International Laboratory News, 13C, 1998.

7) De beperkende factor in snelle GC is niet de scheidingssnelheid.

8) Ook uit milieu-overwegingen dient de productie van XTC te worden bestreden. 9) Het gebruik van presentatie-software staat een beknopte weergave van

onderzoeksresultaten in de weg. 10) Een snellere aanpak bij ongevallen kan het ontstaan van incidentele files aanzienlijk

verminderen.

11) De veiligheid in illegale laboratoria laat vaak te wensen over.

Novel concepts for fast capillary gas chromatography · The most important breakthrough in gas...

Documents

Transcript of Novel concepts for fast capillary gas chromatography · The most important breakthrough in gas...