uva-dare (digital academic repository) method development ... · john mommers uitnodiging voor het...
TRANSCRIPT
UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)
UvA-DARE (Digital Academic Repository)
Method development for comprehensive two‐dimensional gas chromatography
Mommers, J.H.M.
Link to publication
LicenseOther
Citation for published version (APA):Mommers, J. H. M. (2018). Method development for comprehensive two‐dimensional gas chromatography.
General rightsIt is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s),other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).
Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, statingyour reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Askthe Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam,The Netherlands. You will be contacted as soon as possible.
Download date: 23 Jan 2021
Method Development for ComprehensiveTwo-Dimensional Gas Chromatography
John Mommers
Metho
d D
evelop
ment fo
r Co
mp
rehensive Two
-Dim
ensional G
as Chro
mato
grap
hyJo
hn Mo
mm
ers
UITNODIGING
Voor het bijwonen van de openbare verdediging van mijn proefschrift
Method Development for Comprehensive
Two-Dimensional Gas Chromatography
Op donderdag 15 maart 2018 om 12.00 uur in de
AgnietenkapelOudezijds Voorburgwal 231
te Amsterdam
Receptie ter plaatse na afloop van de promotie
John [email protected]
Method Development for Comprehensive
Two‐Dimensional Gas Chromatography
ACADEMISCH PROEFSCHRIFT
ter verkrijging van de graad van doctor
aan de Universiteit van Amsterdam
op gezag van Rector Magnificus
Prof. dr. ir. K.I.J. Maex
ten overstaan van een door het College voor Promoties ingestelde
commissie, in het openbaar te verdedigen in de
in de Agnietenkapel
op donderdag 15 maart 2018, te 12:00 uur
door
Johannes Helena Michael Mommers
geboren te Meerssen
Promotiecommissie
Promotor: Prof. dr. S. Van der Wal
Universiteit van Amsterdam
Co‐Promotors: Prof. dr. ir. P. J. Schoenmakers
Universiteit van Amsterdam
Prof. dr. C. G. de Koster
Universiteit van Amsterdam
Overige leden: Prof. dr. J. Focant
Université de Liège
Prof. dr. T. Hankemeier
Universiteit Leiden
Prof. dr. ir. J.G.M. Janssen
Universiteit van Amsterdam
Prof. dr. R.A.H. Peters
Universiteit van Amsterdam
Prof. dr. W.P. de Voogt
Universiteit van Amsterdam
Dr. M. Camenzuli
Universiteit van Amsterdam
Faculteit der Natuurwetenschappen, Wiskunde en Informatica
Title: Method Development for Comprehensive Two‐Dimensional Gas Chromatography
ISBN: 978‐94‐6295‐871‐5
Printed by: ProefschriftMaken www.proefschriftmaken.nl
Cover: Designed in Blender 2.79 by John Mommers
TABLE OF CONTENTS
1 0B0BINTRODUCTION ......................................................................................................... 7
1.1 12B12BGENERAL INTRODUCTION TO GC×GC ................................................................................... 7
1.2 13B13BADVANTAGES OF GC×GC ................................................................................................... 8
1.3 14B14BGC×GC HISTORY, APPLICATIONS, AND TRENDS ...................................................................... 9
1.4 15B15BGENERAL OBSTACLES AND LIMITATIONS OF GC×GC .............................................................. 10
1.5 16B16BOVERVIEW OF THESIS ....................................................................................................... 13
1.6 17B17BREFERENCES ................................................................................................................... 14
2 1B1BMETHOD DEVELOPMENT FOR COMPREHENSIVE TWO‐DIMENSIONAL GAS
CHROMATOGRAPHY; A REVIEW ....................................................................................... 15
2.1 18B18BINTRODUCTION ............................................................................................................... 15
2.2 19B19BGC×GC METHOD DEVELOPMENT ...................................................................................... 16
2.3 20B20BCHOICE OF THE GC×GC SET‐UP ......................................................................................... 21
2.4 21B21BCHOICE OF THE GC×GC COLUMN‐SET ................................................................................ 28
2.5 22B22BGC×GC OPTIMIZATION .................................................................................................... 40
2.6 23B23BDATA PROCESSING .......................................................................................................... 46
2.7 24B24BCONCLUSIONS ................................................................................................................ 48
2.8 25B25BREFERENCES ................................................................................................................... 50
3 2B2BPREDICTION OF COMPREHENSIVE TWO‐DIMENSIONAL GAS CHROMATOGRAPHY
SELECTIVITY AND RANKING OF COLUMN‐SETS; APPLICATION TO BIO‐OIL SAMPLES ......... 61
3.1 26B26BINTRODUCTION ............................................................................................................... 61
3.2 27B27BEXPERIMENTAL ............................................................................................................... 63
3.3 28B28BRESULTS AND DISCUSSION ................................................................................................ 71
3.4 29B29BCONCLUSIONS ................................................................................................................ 86
3.5 30B30BACKNOWLEDGMENTS ...................................................................................................... 87
3.6 31B31BREFERENCES ................................................................................................................... 87
3.7 32B32BSUPPLEMENTARY DATA .................................................................................................... 89
4 3B3BRETENTION TIME LOCKING PROCEDURE FOR COMPREHENSIVE TWO‐DIMENSIONAL
GAS CHROMATOGRAPHY ................................................................................................. 99
4.1 INTRODUCTION ............................................................................................................... 99
4.2 34B34BEXPERIMENTAL ............................................................................................................. 101
4.3 35B35BRESULTS AND DISCUSSION .............................................................................................. 104
4.4 36B36BCONCLUSIONS .............................................................................................................. 117
4.5 37B37BREFERENCES ................................................................................................................. 118
5 4B4BA PROCEDURE FOR COMPREHENSIVE TWO‐DIMENSIONAL GAS CHROMATOGRAPHY
RETENTION TIME LOCKED DUAL DETECTION ................................................................... 121
5.1 38B38BINTRODUCTION ............................................................................................................. 121
5.2 39B39BEXPERIMENTAL ............................................................................................................. 123
5.3 40B40BRESULTS AND DISCUSSION .............................................................................................. 127
5.4 41B41BCONCLUSIONS .............................................................................................................. 142
5.5 42B42BREFERENCES ................................................................................................................. 142
5.6 43B43BSUPPLEMENTARY DATA .................................................................................................. 144
6 5B5BTEMPERATURE‐TUNABLE SELECTIVITY IN COMPREHENSIVE TWO‐DIMENSIONAL GAS
CHROMATOGRAPHY ....................................................................................................... 145
6.1 44B44BINTRODUCTION ............................................................................................................. 145
6.2 45B45BEXPERIMENTAL ............................................................................................................. 146
6.3 46B46BRESULTS AND DISCUSSIONS ............................................................................................. 148
6.4 47B47BCONCLUSIONS AND DISCUSSIONS ..................................................................................... 157
6.5 48B48BREFERENCES ................................................................................................................. 158
7 6B6BTUNABLE SECONDARY DIMENSION SELECTIVITY IN COMPREHENSIVE TWO‐
DIMENSIONAL GAS CHROMATOGRAPHY ........................................................................ 161
7.1 49B49BINTRODUCTION ............................................................................................................. 161
7.2 50B50BEXPERIMENTAL ............................................................................................................. 163
7.3 51B51BRESULTS AND DISCUSSIONS ............................................................................................. 166
7.4 52B52BCONCLUSIONS AND DISCUSSIONS ..................................................................................... 177
7.5 53B53BREFERENCES ................................................................................................................. 179
8 7B7BSUMMARY .............................................................................................................. 181
9 8B8BSAMENVATTING ..................................................................................................... 183
10 9B9BLIST OF AUTHOR’S PUBLICATIONS ........................................................................... 185
11 10B10BOVERVIEW OF AUTHOR’S CONTRIBUTIONS ............................................................. 189
12 11B11BDANKWOORD ......................................................................................................... 191
7
1 0B0BIntroduction
1.1 12B12BGeneral introduction to GC×GC
Comprehensive two‐dimensional gas chromatography (GC×GC), was introduced by Liu
and Phillips in 1991 [1]. A schematic overview of a standard GC×GC set‐up, showing its
main components, is given in figure 1. It consists of a primary dimension GC column
(1column) which is coupled in series to a secondary dimension GC column (2column),
having a different selectivity, by means of a GC column connector [2]. The 1column is
attached to the GC‐injector and the 2column is attached to the detector. The 1column
is positioned in the main GC oven (1oven) and the 2column is in installed in the 1oven
or in a separate, secondary GC oven (2oven). The most important component of a
GCGC instrument, often referred to as the ‘heart’ of a GCGC, is the modulator. In
most cases the modulation takes places on the 2column or on a retention gap, which
is a deactivated capillary column without a stationary‐phase, having the same internal
diameter as the 2column.
Figure 1 A schematic overview of a standard GCGC setup showing its main
components
GC Inlet
1Column
2Column
Modulator
1Oven
2Oven
Detector
8
The main function of the modulator is to periodically trap the 1column effluent,
refocusing analytes into a narrow chromatographic band and injecting this band into
the 2column [3]. The modulator must trap at least 2.5 to 3 fractions per eluting primary
peak [4, 5]. As a consequence the secondary separation must be very fast. For
example, obtaining 2.5 to 3 fractions of a primary peak with a peak width at base of,
say, 12 seconds implies that the secondary retention‐time must be 4.8 seconds or less.
Longer secondary retention‐times would lead to “undersampling”, thus compromising
the primary separation. The 1column is generally a GC column having a length (1L)
between 20 and 60 m and an internal diameter (1dc) of 0.25 mm. On the other hand,
the 2column is generally a short (2L between 0.5 and 2 m), narrow bore (2dc between
0.1 and 0.15 mm) GC column, providing a fast separation roughly between 2 and 20 s.
Both columns should have significantly different stationary‐phase selectivities to make
optimal use of the 2D separation space. This is also referred to as orthogonality [2]. A
GCGC analysis results in a 1D modulated chromatogram which is transformed by the
software into a 2D‐chromatogram. This transformation is performed by stacking all
the sequential second‐dimension chromatograms side by side to create a 2D‐
chromatogram representing the retention of the 1column on the x‐axis and the
retention of the 2column on the y‐axis, often visualized as a 2D color plot.
1.2 13B13BAdvantages of GC×GC
The three main reasons to use GC×GC instead of one‐dimensional GC are:
because of its highly increased separation power; the peak capacity in GC×GC
roughly equals the peak capacity of the 1column times the peak capacity of
the 2column (nGC×GC 1n × 2n);
because GC×GC may provide (highly) structured 2D‐chromatograms, where
compounds which are chemically related (e.g. aldehydes, degree of
unsaturation, isomers or homologous) elute in the same chromatographic
region or in a structured and recognizable manner. Structured chromatograms
enable group‐type analysis, assist in structure elucidation of unknown peaks,
or may assist in the verification of peaks identified by mass spectrometry (MS);
9
because of its increased sensitivity (5 to 10 times) due to trapping, refocusing
and fast injection of the narrow band into the 2column by the modulator,
which leads to an increased signal‐to‐noise ratio.
1.3 14B14BGC×GC history, applications, and trends
Since the introduction of GC×GC in 1991, the number of scientific publications on the
subject shows a steep annual increase in the early years and seem to stabilize in the
last few years (cf. figure 2). GC×GC technology is becoming more mature, more
accessible, and increasingly adopted by analytical chemists. Topics of published
GC×GC papers clearly shifted from instrumental developments, data processing
approaches and chemometrics, more and more towards real‐life applications.
Figure 2 Number of scientific publications involving GC×GC per year; information
obtained by searching “comprehensive two‐dimensional gas chromatography” or
“GC×GC” in title, abstracts, and keywords by using Scopus
GC×GC applications are found mainly in the fields of (1) characterization of oil and
petrochemicals, (2) food, food safety, flavor, and fragrances, (3) environmental
analysis, (4) circular economy (e.g. characterization of alternative green bio/pyrolysis
fuels and materials), (5) metabolomics, biomarker discovery, health, and diagnostics
and (6) forensic analysis (e.g. concerning oil spills and characterization of fire
10
accelerants). An overview of the main GC×GC application fields in 2016 is shown in
figure 3. In total 130 scientific papers concerning GC×GC were published in 2016.
Figure 3 Overview of GC×GC main topics published in 2016
1.4 15B15BGeneral obstacles and limitations of GC×GC
As stated, GC×GC provides more separation power, structured 2D‐chromatograms
and enhanced sensitivity when compared to one‐dimensional GC. However, GC×GC
also suffers from several obstacles and limitations.
To start, GC×GC method development is significantly more difficult than method
development for 1D‐GC. For GC×GC more choices need to be made, such as the
selection of the GC×GC column‐set (phase chemistries and dimensions) and the
modulator settings. Also, GC×GC method optimization is more difficult due to the
complex interplay of the many 1D and 2D parameters [6]. This complex interplay is
visualized in figure 4.
11
Figure 4 Complex interplay of GC×GC parameters of a standard GC×GC setup. The
parameters that can be chosen are indicated by the ‘pinion icons’. All parameters are
1Reten
‐tion (k)
Temp
Rate
Analysis
Time
Mass /
mod
Pm
1P/
P
Pmid
1 df
1dc
1L
1W
h
Chan
ce of
1overload
ing
Pin
info
2 df
2dc
T elution
2 RS
Chan
ce of
2Overload
ing
2u
2P
2 Reten
‐tion (k)
2 D space
2W
h
2L
Pout
1P 1 u
uopt
1RS
1Stat
Phase
2Stat
Phase
1 Selec‐
tivity (r)
2Selec‐
tivity (r)
1 Column
2Column
Modulator
Inlet
Detector
12
given in the colored boxes, primary dimension parameters in blue, secondary
dimension parameters green, modulator parameters in yellow and overall GC×GC
parameters in red. The arrows, pointed from the input to the output parameter,
indicate an increase (green), a decrease (red) or both possibilities (grey) in case the
input parameter is increased (Adapted from [6])
Furthermore, the method‐development choices and optimization are also restricted
by the need to maintain the first‐dimension separation and by column temperature
limitations. The upper temperature limit of the least stable column (often the most
polar column) determines the upper temperature limitation of the entire column‐set.
Moreover, due to the usual difference in the internal diameters of the primary and
secondary column (mostly 1dc > 2dc), both columns cannot be simultaneously operated
at their optimal linear gas velocities. This is referred to as the flow‐mismatch issue [7].
This complicates optimization and leads to compromise pressure/flow parameter
settings. Additionally, the difference in the internal diameters may also lead to mass‐
loadability issues [8, 9] for the secondary column given the fact that mass loadability
is proportional with the dc2 as the volume of stationary‐phase per plate is proportional
with dc2df [10] and focussing in the modulator worsen these loadability issues. Mass
overloading will lead to broader 2peaks and to a decrease of the overall separation
power.
Also, in GC×GC, a chromatographic peak has two retention‐times, one for each
dimension [11]. Thus, peaks may shift in two directions, which causes alignment issues
that are less straightforward to solve than retention‐time shifts in 1D‐GC. An
additional issue is that the secondary chromatograms are stacked (no
chromatographic continuum anymore), so a shift in the second dimension may lead
to peaks “jumping” from the top of the chromatogram to the bottom of the
chromatogram (wrap‐around) or visa‐versa. This aggravates alignment issues [12]. In
GC×GC, a shift in the primary retention‐time also causes a shift in the secondary
retention‐time, because the peak is eluting from the primary column at a different
temperature (due to the inevitable temperature programming) and is, therefore,
13
‘analyzed’ at a different (approximately isothermal) temperature on the 2column.
Peaks may shift in time due to column aging or upon replacing a column or the whole
column‐set. Peak alignment is especially important when comparing complex
chromatographic data, for example when comparing different samples with each
other or when following chromatographic data in time to investigate changes or
trends. Due to the complexity of the chromatograms, manual peak alignment is not
an option in GC×GC.
1.5 16B16BOverview of thesis
Chapter 2 ‐‐ in this chapter a review is given in which several approaches to GC×GC
method development are discussed and guidelines for GC×GC method development
are proposed. The focus is on selecting the GC×GC instrumental set‐up and the
column‐set, optimization of the GC×GC settings, and dealing with two‐dimensional
retention‐time shifts.
Chapter 3 ‐‐ In this chapter the selection of column‐sets is discussed. A procedure is
described for the global prediction of the best GC×GC column‐sets for the analysis of
complex samples. Also, two new orthogonality parameters (%FIT and %BIN5×5) are
described and evaluated. The GC×GC retention prediction is based on the adapted
Abrahams solvation parameter model, as published by Seeley et al. [13].
Chapter 4 ‐‐ In this chapter, a fast and easy to perform, two‐step retention‐time‐
locking procedure for GC×GC is proposed and its feasibility is demonstrated. It is
demonstrated that retention‐time shifts in both the primary and secondary
dimension, which may occur, for example, after replacing the column‐set, can be
reduced to less than half of the peak‐base width.
Chapter 5 ‐‐ In this chapter, a retention‐time‐locking procedure is proposed for locking
primary and secondary retention‐times corresponding to peaks observed in dual‐
detection GC×GC. Its advantages are demonstrated and discussed.
Chapter 6 ‐‐ In this chapter, a temperature‐tunable GC×GC setup is described and
discussed, which consists of three capillary columns with different selectivities. In this
setup, the selectivity of the primary dimension can be tuned by adjusting the
14
temperature offset of two capillary columns coupled in series. Both columns form part
of the primary dimension and they are positioned in two separate GC ovens.
Chapter 7 ‐‐ In this chapter, two tunable GC×GC setups are compared and described.
In both cases the secondary dimension consists of two different capillary columns
coupled in series. These setups can be used to optimize the overall GC×GC separation
by altering the second‐dimension column selectivity. Moreover, these set‐ups offer
enhanced possibilities for qualitative analysis.
Summary ‐‐ Includes a summary, conclusive notes and recommendations.
1.6 17B17BReferences
[1] Liu, Z., Phillips, J.B., (1991) J Chromatogr Sci, 29 (5), pp. 227‐231.
[2] Dallüge, J., Beens, J., (2003) J Chromatogr A, 1000 (1‐2), pp. 69‐108.
[3] Tranchida, P.Q., Purcaro, G., Dugo, P., Mondello, L., Purcaro, G., (2011) TrAC ‐
Trends in Anal Chem, 30 (9), pp. 1437‐1461.
[4] Beens, J., Janssen, H.‐G., Adahchour, M., Brinkman, U.A.Th., (2005) J Chromatogr
A, 1086 (1‐2), pp. 141‐150.
[5] Khummueng, W., Harynuk, J., Marriott, P.J., (2006) Anal Chem, 78 (13), pp. 4578‐
4587.
[6] Harynuk, J., Górecki, T., (2007) Am Lab, 39 (4), pp. 36‐39.
[7] Peroni, D., Janssen, H.‐G., (2014) J Chromatogr A, 1332, pp. 57‐63.
[8] Koek, M.M., Muilwijk, B., van Stee, L.L.P., Hankemeier, T., (2008) J Chromatogr A,
1186 (1‐2), pp. 420‐429.
[9] Harynuk, J., Górecki, T., De Zeeuw, J., (2005) J Chromatogr A, 1071 (1‐2), pp. 21‐27.
[10] Ghijsen, R.T., Poppe, H., Kraak, J.C., Duysters, P.P.E., (1989) Chromatographia, 27
(1‐2), pp. 60‐66.
[11] Mommers, J., Knooren, J., Mengerink, Y., Wilbers, A., Vreuls, R., van der Wal, S.,
(2011) J Chromatogr A, 1218 (21), pp. 3159‐3165.
[12] Weusten, J.J.A.M., Derks, E.P.P.A., Mommers, J.H.M., van der Wal, S., (2012) Anal
Chim Acta, 726, pp. 9‐21.
[13] J. Seeley, E. Libby, K. Hill Edwards, S. Seeley, J Chromatogr A, 1216 (2009), 1650‐
1657.
15
2 1B1BMethod development for comprehensive two‐dimensional gas
chromatography; a review
2.1 18B18BIntroduction
Method development for comprehensive two‐dimensional gas chromatography
(GC×GC) is significantly more difficult than method development for one‐dimensional
gas chromatography (1D‐GC).
Compared to 1D‐GC, more method‐development choices need to be made, such as
the selection of the GC×GC column‐set and the modulator settings. GC×GC method
optimization is complex, because the individual dimensions (1D and 2D) cannot simply
be optimized separately, due to a complex interplay of many 1D and 2D parameters
[1]. Additionally, GC×GC optimization is restricted by certain requirements, such as
the modulation criterion (2tr/1σ ≤1.5) [2, 3] and the temperature limits of the individual
columns.
In case the internal diameter of the primary column is larger than the internal
diameter of the secondary column, both columns cannot be simultaneously operated
under optimal separation conditions. This is referred to as the flow‐mismatch problem
[4, 5]. This issue complicates the optimization and results in a compromise of sub‐
optimal flow settings for both columns. This difference in column diameters may also
lead to issues with column loadability [6‐9] of the second‐dimension column and may,
therefore, indirectly contribute to a reduction of the overall separation efficiency.
Compared to 1D‐GC, GC×GC data processing can be judged as less complex, especially
for truly complex samples, because GC×GC provides more separation power, resulting
in fewer coelutions. GC×GC also provides structured chromatograms, which greatly
supports qualitative analysis. However, compared to 1D‐GC, GC×GC provides two‐
dimensional retention data, which implies that each peak contains two different
retention‐times, one for each dimension. Therefore, peak shifts may occur in two
dimensions, for example changes in time or after changing a column. This causes peak‐
16
alignment issues that complicate data processing, for example when comparing
chromatographic data of complex samples or when using GC×GC to follow complex
samples over a long period of time.
In this review, several papers concerning GC×GC method development are discussed
and guidelines for GC×GC method development are proposed. The focus is on
selecting the GC×GC instrumental set‐up and column‐set, optimization of the GC×GC
settings, and dealing with retention‐time shifts in two dimensions.
2.2 19B19BGC×GC Method development
The choices to be made for developing a GC×GC method mainly depend on (1) the
analytical question, (2) the sample characteristics and (3) the available equipment. A
schematic overview of important topics, parameters, and limitations, related to
GC×GC method development, is given in figure 1.
The analytical question can be seen as the objective of the analytical method to be
developed: which analytical question needs to be answered? The analytical question
can roughly be divided in (1a) type of analysis and (1b) requirements.
Ad 1a. Analytical question – type of analysis
The “types of analysis” can be (1) target analysis [10‐12], which entails the analysis of
an often limited set of known target analytes, (2) profiling or fingerprinting analysis
[13‐15], which implies the analysis of “all” known and unknown GC‐amenable analytes
(or all within a certain chemical group; e.g. profiling of carboxylic acids) in the given
sample and (3) group‐type analysis [16‐18], which indicates the analysis of groups of
chemically related analytes (e.g. homologues or isomers) instead of individual
analytes. Chemically related compounds show similar GC×GC retention behavior,
which leads to structured GC×GC chromatograms, rendering group‐type analysis
possible. For group‐type analysis structured chromatograms and well separated
signals for the different groups are more important than the separation of ‘all’
individual analytes.
17
Figure 1 Schematic overview of GC×GC method development
18
The “type of analysis” also includes whether the analytical method to be developed
must be quantitative, qualitative, both quantitative and qualitative (also referred to
as speciation), or will be used for differential analysis only. The latter implies
comparing (raw) chromatographic data to discern differences or trends, which may be
further identified and/or quantified later. The “type of analysis” also largely
determines how the data must be processed and how the results must be presented
or visualized in order to answer the original analytical question.
Ad 1b. Analytical question – requirements
The analytical question comes with requirements, such as the required limit of
detection (LOD), required concentration range in which the analytes should be
analyzed (related to the dynamic range), required accuracy, and required precision of
the analytical method. Also, speed of the analysis, or analysis cycle time (time between
two successive injections) may be important requirements.
Ad 2. Sample characteristics
Sample parameters (after possible sample pre‐treatment) are very important in
making method‐development choices. A sample contains the analytes of interest and
matrix compounds. Thus, method development may firstly aim at the separation of all
the analytes from the sample matrix compounds. As an example, the analysis of polar
carboxylic acids in hydrocarbon samples containing less‐polar aliphatic and aromatic
hydrocarbons as the matrix compounds. By applying, for example, a “polar non‐
polar” column‐set, the polar acids can be separated from the less‐polar matrix
compounds. The polarities of the analytes (and matrix compounds) are important in
selecting the stationary‐phase polarities and also the order of the columns (i.e. non‐
polar × polar or polar × non‐polar). The concentration ranges of the analytes and
matrix compounds (which might interfere with the target analytes) are important in
relation to, for example, the required limit of detection, specificity, dynamic range,
and column loadability. For example, matrix compounds with significantly higher
concentrations than the analytes of interest may cause loadability issues (mainly on
the second column) that may lead to co‐elutions, distortion of analyte peaks, and
19
reduced separation efficiencies. In this case column loadability issues may be
minimized by, for example, increasing the secondary column internal diameter, based
on the fact that the column mass loadability is proportional with dc2 [9]. However,
changing 2dc will affect the optimum flow rate, the second‐dimension separation
efficiency and the limit of detection.
Usually, a sample contains analytes and matrix compounds with different boiling
points. The analyte boiling‐point range is important in selecting the analytical column‐
set. The analysis of volatile analytes requires longer columns and/or thicker stationary‐
phase films, while the analysis of high‐boiling‐point analytes require thinner
stationary‐phase films and, especially, high‐temperature‐stable stationary‐phases.
The analysis of volatile analytes in samples containing high‐boiling‐point matrix
compounds may result in fouling of the column‐set, possibly causing bad peak shapes.
This problem may be prevented by using back‐flushing [19, 20] to prevent the high‐
boiling matrix compounds to enter the primary GC column. The thermal stability of
analytes of interest limits the temperature settings for the GC injector, GC ovens (both
the primary and secondary one), the GC×GC modulator (modulator temperature and
modulator hot‐jet flows), transfer tubings and detectors. Analytes which are thermally
less stable may require lower operating temperatures in combination with, for
example, shorter GC columns, thinner stationary‐phase films and higher column flows,
so as to lower elution temperatures, while maintaining enough separation power.
Usually, the complexity of a sample (related to the number of peaks) is important for
the choice of the column‐set in relation to its peak capacity; the higher the complexity,
the more separation power is required. More GC×GC separation power can be
achieved by using longer columns, optimized stationary‐phase film thickness, smaller
internal diameters, lower temperature‐programming rates and, especially, optimized
GC×GC orthogonality, providing appropriate selectivity, so as to make optimal use of
the two‐dimensional separation plane.
Ad 3. Hardware restrictions
Hardware restrictions can be divided in (3a) GC×GC equipment restrictions and (3b)
column‐set restrictions. The GC×GC equipment restrictions are mainly related to
20
temperature and flow settings, corresponding with the available injector type (e.g. hot
split/splitless injector or programmed‐temperature‐vaporizer injector), detector type
(e.g. flame‐ionization detector or mass spectrometer), and modulator type. The
column‐set restrictions are mainly related to temperature restrictions. For a given
column‐set the maximum temperature is determined by the primary or the secondary
column, whichever has the lowest upper temperature limit. An overview of several
commercially available GC stationary‐phases, including their polarities and maximum
operating temperatures, is given in table 1.
Table 1 Several commercially available GC stationary‐phases including their polarities
and maximum operating temperatures [21, 22]; the last three are ionic liquid
stationary‐phases.
GC Stationary‐phase composition Maximum
Operating
Temp. C
Polarity
100% dimethylpolysiloxane 340 low
5% phenyl 95% dimethylpolysiloxane 340 low
35% phenyl 65% dimethylpolysiloxane 310 mid
35% trifluoropropyl 65% dimethylpolysiloxane 310 mid
14% cyanopropylphenyl 86% dimethylpolysiloxane 290 low / mid
50% phenyl 50% dimethylpolysiloxane 290 mid
6% cyanopropylphenyl 94% dimethylpolysiloxane 290 low / mid
50% cyanopropyl 50% methylpolysiloxane 255 high
88% cyanopropyl 12% arylpolysiloxane 255 high
polyethyleneglycol 255 high
50% trifluoropropyl 50% dimethylpolysiloxane 250 high
acid modified polyethylene glycol 250 high
50% cyanopropylphenyl 50 dimethylpolysiloxane 240 mid / high
1,12‐Di(tripropylphosphonium)dodecane
bis(trifluoromethylsulfonyl)imide (SLB‐IL60)
300 high
Tri(tripropylphosphoniumhexanamido)triethylamine
bis(trifluoromethylsulfonyl)imide (SLB‐IL76)
270 high
1,5‐Di(2,3‐dimethylimidazolium)pentane
bis(trifluoromethylsulfonyl)imide (SLB‐IL111)
270 very high
21
2.3 20B20BChoice of the GC×GC set‐up
A standard GC×GC set‐up consists of (1) a GC injector (see 2.3.1), (2) a single or dual
GC oven, in which the primary and secondary columns are installed, (3) a modulator
(see 2.3.2) and (4) a detector (see 2.3.3). The choice of the complete set‐up will in
general depend on the analytical question, requirements, and sample characteristics.
Besides a standard GC×GC setup, alternative setups may be considered to minimize or
solve GC×GC flow mismatch and/or mass loadability issues and/or to provide multiple
detection options, as described in literature:
(1) using dual or multiple secondary columns in parallel, referred to as GC×GCn
[5, 23‐29];
(2) adjusting the midpoint pressure by, for example, splitting the column flow
before the secondary column, referred to as split‐flow GC×GC [30‐34];
(3) using equal internal diameters for both the primary and secondary column,
referred to as equal i.d. GC×GC [35];
(4) using GC×GC in stop‐flow mode [36‐38];
(5) using monolithic secondary columns [39];
(6) adjusting the outlet pressure by adding a restrictor at the end of the
secondary column [4].
Ad 1. By using multiple secondary columns (connection to primary column before the
modulator), it is possible to operate both GC×GC dimensions under optimal flow
conditions. By using multiple secondary columns, the volumetric flow rates difference
between the optima of both dimensions can be minimized or even eliminated. Two
different GC×GCn setups can be defined: (1) the multiple secondary columns are all
connected to the same detector [5] and (2) one of each secondary column is attached
to a different detector [23‐29]. The first setup requires that all secondary columns
have exactly the same column dimensions and requires that all secondary columns are
properly installed; small variations will lead to differences in retention between the
different secondary columns, causing peak splitting or peak broadening. In the second
GC×GCn setup the end of the primary column may be attached to a pressure‐
22
controlled splitter. The secondary columns are also attached to the splitter and led
through the modulator and the ends of the secondary columns are connected to
different detectors. For quantitative analysis, a pressure controlled splitter is required
for keeping the pressure at the split point constant during temperature programming,
so as to keep the split‐ratio constant during the analysis.
Ad 2. By splitting the column effluent of the primary column, towards the secondary
column and towards an uncoated capillary attached to a split‐valve (directed towards
waste), the split ratio and the GC×GC midpoint pressure can be regulated and primary
and secondary gas velocities can be optimized. A major disadvantage of this approach,
especially for trace analysis, is a decrease in sensitivity, in proportion with the primary
column effluent split‐ratio. However, split‐flow GC×GC enables the use of column‐sets
(under optimized conditions) with large internal‐diameter differences between the
primary and secondary column, such as a column‐set with a 0.25 mm 1dc and a 0.05
mm 2dc.
Ad 3. For the analysis of samples containing analytes and/or matrix compounds of
which the concentrations are unknown and possibly high, the use of larger secondary
column diameter or equal i.d. columns (both 0.25 µm) minimizes the chances and
consequences (loss of column efficiency) of overloading the second‐dimension
column. Especially for “chemical fingerprinting” or “profiling” analysis of moderately
complex samples, featuring large analyte concentration differences (requiring a large
dynamic range), primary and secondary columns having equal internal diameters may
successfully be used. Compared to conventional‐internal diameter GC×GC column‐
sets, equal diameter column‐sets will provide less efficiency. However, both columns
can be operated at near‐optimum carrier‐gas velocities and the mass loadability will
increase significantly. This will lead to more reliable qualitative and quantitative
“fingerprinting” results.
Ad 4. In stop‐flow GC×GC, the carrier‐gas flow in the primary column is stopped (with
pneumatic switching) for a given time, followed by injection of the trapped band into
the secondary column by the modulator. The carrier‐gas flow to the secondary
23
column, supplied from an auxiliary source, is not stopped. After the given time, the
flow in the primary column is restored, a new analyte band is collected in the
modulator and the process is repeated. The main advantage of this approach is that
the modulation period, and thereby the second‐dimension separation time, can be
varied independently of the first‐dimension peak width. However, increasing the stop‐
flow time will also increase peak broadening (caused by longitudinal diffusion) of the
primary dimension peaks.
Ad 5. The use of a monolithic second‐dimension column may allow optimum linear gas
velocities for both dimensions simultaneously. Macroporous polymer monoliths can
be tailored towards the required flow characteristics and analytical performance by
optimizing the polymerization recipe and conditions. Peroni et al. [39] demonstrated
the potential of monolithic 2D columns for real‐life applications.
Ad 6. Adding a restrictor at the end of the secondary column, causes an elevated outlet
pressure and slightly reduces the flow‐mismatch. Also, van Deemter curves become
flatter at higher outlet pressures resulting in a lower loss in efficiency at higher inlet
pressures. However, a major disadvantage of this approach is that analysis times
becomes much longer. This approach is only attractive if a slightly improved resolution
is required for a given column‐set.
2.3.1 54B54BGC×GC Injectors
As for 1D‐GC, the injector, and injection‐mode, in GC×GC is mainly determined by the
sample characteristics such, as the (thermal) stability, boiling point range, the sample
potential of fouling the injector (“dirty samples”) and the analytical requirements,
such as the required limit of detection (e.g. bulk or trace analysis), the required
“robustness”, and ease of operation of the analytical method. In table 2, the most
commonly used GC(xGC) injectors are summarized including their main characteristics
[40]. Representative sampling means that the sample plug entering the GC column has
the same composition as the sample itself, so that the injector causes no sample
discrimination. This is especially important when analyzing samples with a broad
24
boiling point range or samples containing analytes which are prone to adsorption or
which are thermally labile.
Table 2 Commonly used GC×GC injectors and their main characteristics
Hot split
Hot splitless
PTV split
PTV splitless
Cool On‐column
Type of analysis bulk trace Bulk or labile trace or labile Trace or labile
Representative sampling Poor Poor Moderate to
high Moderate to
high High
Inertness Medium Low Medium Medium High
Robustness for dirty samples Moderate Moderate High High Low
Sample volume flexibility Medium Medium High Medium
Including LVI1
Medium to high
Including LVI1
1 LVI=large volume injection
2.3.2 55B55BGC×GC secondary oven
A dual‐oven set‐up, in which the secondary column is installed in the secondary oven
and the primary column in the main GC oven, provides more flexibility in adjusting
and/or optimizing the secondary retention, for example to prevent wrap‐around by
applying a fixed temperature off‐set between the primary and secondary ovens. A
dual‐oven set‐up can also be used for selectivity tuning, installing a third column (with
a different selectivity) in the secondary oven [41, 42], and for dual‐detection
retention‐time locking to match primary and secondary retention‐times for both
detectors used [43].
2.3.3 56B56BGC×GC modulators
The modulator, often referred to as the ‘heart’ of the GC×GC instrument, periodically
traps condensable analytes in the primary column effluent, refocusing this to a narrow
band, and injecting it in the secondary column. Since the first GC×GC publication in
1991 by John Phillips [44], many developments in the field of GC×GC modulators
happened and have been published [45‐50]. At present, the three most used and
commercially available modulators are, (1) the thermal dual‐stage quad‐jet
modulator, (2) the thermal dual‐stage loop modulator and (3) the microfluidic
differential‐flow modulator (DFM).
25
Thermal dual‐stage modulators
The thermal dual‐stage modulators using liquid nitrogen are the most popular and are
considered the most effective modulators. However, the DFM are gaining in
popularity. Both thermal dual‐stage modulators can be operated by using liquid
nitrogen, which enables the efficient trapping of propane and higher‐molecular‐
weight compounds or by using cooled gas‐streams, referred to as consumable free
modulators, which enables trapping of heptane and higher‐molecular‐weight
compounds.
Optimization of modulator settings should aim for (1) efficient trapping to prevent or
minimize analyte breakthrough, (2) focusing the trapped analytes in a narrow band to
obtain high secondary column efficiencies and (3) efficient and fast remobilization of
the trapped analytes to the secondary column to avoid injection band broadening
and/or peak tailing in the second dimension.
In most cases the coupling of the primary and the secondary column, for example by
means of a universal press‐fit connector, is positioned before the modulator, so the
analyte trapping and remobilization processes occur in the secondary column. The
presence of the stationary‐phase will usually increase the analyte trapping efficiency,
capacity, and band focusing, but decrease the remobilization speed, in comparison
with a retention gap containing no stationary‐phase. In case remobilization is not
efficient and causes severe peak broadening and/or tailing, increasing the hot‐jet
temperature, hot‐jet time, lowering the cold‐jet flow or temperature, or trapping in a
retention gap may offer a solution. A disadvantage of this last solution is that an extra
column connector is required, positioned after the modulator, to couple the retention
gap to the secondary column. This may lead to extra band broadening of the second‐
dimension peaks.
Analytes with different volatilities require different cryogenic modulation conditions
to achieve optimum trapping and remobilization [51, 52]. To achieve efficient
trapping, the cold spot should be at a temperature 120 to 140°C lower than the elution
temperature. Thus, preferably, a constant temperature offset between the
26
chromatographic oven and the modulator should be set, to achieve efficient volatility‐
dependent modulation. The cold‐jet flow must be optimized for efficient trapping and
remobilization of all the analytes of interest. Insufficient cooling will lead to analyte
breakthrough and excessive cooling will lead to poor GC×GC performance, such as
peak tailing due to incomplete or slow analyte remobilization. The hot‐jet
temperature should be at least 40°C higher than the analyte elution temperature.
However, this temperature should not exceed the temperature limit of the secondary
column stationary‐phase. Otherwise, an uncoated thermally stable retention gap
should be used for trapping. The hot‐pulse time must be short enough so as to prevent
analyte breakthrough and long enough to fully remobilize the trapped analytes. A
good starting point for all GC×GC applications is a hot‐pulse time of 300 ms [52].
Loop‐type modulator
A critical parameter of a loop‐type modulator is the length of the delay loop. The
length should not be too long, to ensure that analytes are trapped and focused by the
downstream cold spot, and not be short, to avoid analyte breakthrough during the
hot‐pulse period. A good starting point is a delay‐loop length of 1 to 1.5 m. However,
the length should preferably be calculated based on the actual linear gas velocity in
the delay loop.
Microfluidic DFM
A microfluidic DFM can be used to analyze a wide range of compounds from
permanent gases up to high‐boiling compounds. The modulator can be operated
across a range exceeding 400C. Thereby it covers the largest range of all modulators.
The column stationary‐phases are the main reason for temperature limitations. The
secondary column flow is always high, typically 20 mL/min, thereby excluding the
direct use of a mass spectrometer. DFM modulation does not require additional
consumables, besides a higher carrier‐gas consumption, which makes it well suited for
routine implementation. DFM modulators produce similar sensitivity enhancements
as thermal modulators when using an FID. However, thermal modulation generates
higher second‐dimension resolution, due to more efficient cryogenic trapping,
focusing the analytes in a narrower band, compared to DFM modulation. Optimization
27
of DFM to achieve optimal performance, is more laborious than for thermal
modulators and strongly depends on the modulator dimensions and the ratio of the
primary and secondary flows. This flow ratio must be maintained throughout a
chromatographic run. The dimensions of the DFM device must be altered if
substantially different primary and secondary flows are desired. Fortunately,
modulator dimensions can easily be calculated by using flow‐resistance models.
2.3.4 57B57BGC×GC detectors
Ideally, the detector should not affect the result of the GC×GC chromatographic
process. It should cause negligible additional band broadening and it should be
capable of adequately measuring the profiles of very fast second‐dimension peaks. In
general, a detector acquisition rate of at least 100 Hz is required for quantitative
GC×GC applications, generating secondary peaks with peaks widths at base of about
100 ms.
Of all detectors, the universal FID shows the lowest peak broadening, due to its fast
ionization and acquisition process and negligible internal volume. The FID also has a
very large dynamic range of 6 to 7 decades and detection limits down to 2‐5 pg
carbon/s. Besides this, the FID is very robust, shows long term stability and is easy to
operate. The FID is mostly considered as the ideal detector for quantification
purposes.
Selective detectors are often used to increase the selectivity, and often also to obtain
lower detection limits for the target analytes by reducing or eliminating the signals
from interfering (coeluting) matrix compounds. Mass spectrometers can be used as
both universal (using full mass‐scan range) and selective (using selective m/z values,
e.g. in selected‐ion‐monitoring mode) detectors. A list of the most commonly used
detectors coupled to GC×GC, including their characteristics and selected applications,
is provided in table 7.
28
Table 7 Commonly used detectors for GC×GC including typical characteristics [87, 88]
Detector Acquisition
rate
Selectivity Detection
limit
Dynamic
range
Applications
TOFMS
≥100 Hz Universal or
selective
pg to ng
103 to 105 54‐57
High resolution
TOFMS
≥50 Hz Universal or
selective
pg to ng
103 to 105 58‐62
Fast scanning
Quadrupole MS
~10.000
AMU/s
Universal or
selective
pg to ng
103 to 105 63‐66
µECD
≥50 Hz Selective to
halogens
50 fg/mL 104 67‐69
NCD
≥100 Hz Selective to
nitrogen
3 pg N/s 104 70‐71
SCD
≥100 Hz Selective to sulfur 0.5 pg S/s 104 72‐75
NPD ≥100 Hz Selective to
nitrogen or
phosphorus
0.4 pg N/s
0.2 pg P/s
105 76‐77
FPD ≥100 Hz Selective to sulfur
or phosphorus
4 pg S/s
100 fg P/s
103 78‐81
FID
≥100 Hz Universal to carbon 2 pg C/s 107 82‐85
2.4 21B21BChoice of the GC×GC column‐set
The most important step in GC×GC method development is the choice of the column‐
set. The column‐set should provide adequate separation of the analytes of interest.
The degree of a chromatographic 1D separation of two peaks ‘j’ and ‘i’ can be
expressed by the resolution Rji and can be determined by:
∆ , (1)
Where ΔtR,ji is the retention difference between both peaks and j and i are the
standard deviations of the peaks. In case of two equally sized peaks, an R value of 1.5
29
corresponds to full peak separation. The resolution R can also be expressed by the
following equation:
(2)
Where L is the column length, H is the plate height, r is the selectivity factor (=kj/ki)
and k is the retention factor. To obtain an adequate separation, the selectivity factor
must be sufficiently large. Resolution is small in case of retention factors close to zero,
even if the selectivity is large. For an optimal resolution the plate height, which
depends on the linear gas velocity, should be as low as possible (Hmin). The resolution
increases with the square root of L, but increasing the length of the column also
increases the analysis time and the pressure drop across the column.
GC×GC is almost always used to separate complex mixtures, showing a large retention
range of the components. Therefore, a temperature program is used during the
1separation to limit the analysis time by decreasing retention of the later eluting
compounds, as it is clear from equation 2 that at k>2 little resolution gain is obtained
at the expense of increased analysis time.
The degree of a chromatographic separation in a two‐dimensional space, where one
peak is now surrounded by more than two peaks, can be described by the valley‐to‐
peak ratio, which is based on the concept of the ‘saddle point’, as proposed by Peters
et al. [86]. This allows the calculation of valley‐to‐peak ratio also for ‘real’ peaks that
are not Gaussian in shape. In this approach, prior to the calculations, criteria are
applied to determine which peaks can be considered as (meaningful) neighbors. This
resolution metric can also be used as an estimator of the quantification errors and to
estimate the separation quality of the entire chromatogram, which makes it also
suitable for optimization purposes.
30
2.4.1 58B58BStationary‐phase chemistry
The choice of the stationary‐phase chemistries of the column‐set mainly depends on
the sample properties which includes the properties of the analytes and the properties
of sample‐matrix compounds. For a given sample, a good GC×GC column‐set should
provide adequate retention, resolution, orthogonality (appropriate selectivity), group‐
type separation (in case of group‐type analysis), and good analyte peak shapes.
Retention in GC is based on the sum of polar, if polar functional groups are present in
either or both the sample molecules and the stationary‐phase, and non‐polar
interactions. Molecules interact with each other through intermolecular “van der
Waals” forces, which can be reduced by increasing the temperature. The “van der
Waals” forces include dipole‐dipole (Keesom), dipole‐induced dipole (Debye), and
induced dipole‐induced dipole (London dispersion) interactions [87‐89].
When using an apolar stationary‐phase in GC, induced dipole‐induced dipole, or
dispersive interactions, which are temporary dipole interactions, are the dominant
intermolecular attractive force. Dispersive interactions are non‐polar and exist
between all molecules. The magnitude of the dispersive interaction increases with the
size of the molecule; for this reason, boiling point and GC chromatographic retention
increases with the size of the molecule. When using a column with a non‐polar
stationary‐phase the sample molecules (polar and non‐polar compounds) will only be
retained by dispersive interactions and they will elute in the order of their boiling
points.
Selectivity in GC arises also from different strengths and types of interactions between
the analytes and the stationary‐phase molecules. Polarity of a compound is
determined by the nature of the polar groups and the molecular structure itself. In
case of polar interactions between the analyte and the stationary‐phase, the analyte
will be more retained compared to non‐polar (or less‐polar) compounds of similar
molecular weight.
31
Electronegative atoms in a molecule may cause a permanent dipole. The magnitude
of a molecule’s permanent dipole, or the distortion of the electron cloud of the
molecule, is expressed in its dipole moment. Two permanent dipoles attract each
other (dipole‐dipole interaction). The stronger the dipole moments of the analytes and
the stationary‐phase molecules, the higher the chromatographic retention.
An additional type of very strong interactions is hydrogen bonding. Atoms such as
nitrogen (N) and oxygen (O), present in organic molecules, act as strong hydrogen
(proton) acceptors. A hydrogen atom bound to O or N in an organic molecule becomes
a strong proton donor. A proton acceptor and a proton donor will strongly interact by
forming hydrogen bonds. Hydrogen bonds are the strongest interactions encountered
in GC.
The dipole‐induced dipole is the weakest polar interaction. It occurs when a
permanent dipole induces a temporary dipole in another molecule by distorting its
electron cloud. A permanent dipole can only distort the electron cloud of another
molecule when this molecule is polarizable (e.g. aromatic compounds). The stronger
the permanent dipole and the stronger the polarizability of the other molecule, the
stronger the interaction will be.
Thus, the choice of the stationary‐phase chemistries of the GC×GC column‐set should
be based on the expected analyte and stationary‐phase molecular‐interaction types
and strengths.
Poole et al. [89] describes the use of the Abraham solvation‐parameter model for the
classification of GC stationary‐phases [90‐95]. This model describes the transfer of an
analyte molecule from the gas phase to the stationary‐phase as a three‐step process:
(1) formation of a cavity in the stationary‐phase with the same size as the analyte
molecule, (2) reorganization of the analyte molecule and (3) insertion of the analyte
molecule in the cavity. Once in the cavity, analyte – stationary‐phase interactions can
occur, which include hydrogen bonding, induction, orientation, and dispersion
interactions. The model uses descriptors for the different interactions and is written
as follows:
32
(3)
Where K is the gas‐liquid partition coefficient and c is a constant. The uppercase letters
are descriptors of the analyte molecules and the lowercase letters are the descriptors
of the stationary‐phase. The letter combinations describe the contributions of the
specific analyte‐phase interactions where: eE is the contribution from lone‐pair
interactions between polarizable molecules; sS is the contribution from dipole‐dipole
and/or dipole – induced dipole interactions; aA is the contribution of hydrogen
bonding with an analyte behaving as a hydrogen donor (acid); bB is the contribution
of hydrogen bonding with an analyte behaving as an hydrogen‐bond acceptor (base)
and lL is the contribution of the opposing cavity formation and dispersion interactions.
By using the solvation parameter model to determine the system constants of the
most commonly used non‐ionic GC stationary‐phases and principal‐component
analysis (PCA), Poole et al. observed four distinct groups of stationary‐phases. Based
on these groups, five columns were selected, which roughly span the PCA selectivity
space (see table 3).
Table 3 System constants for columns selected from different selectivity groups at
100C; b is 0 for all columns [89]
System Constants
Stationary‐phase Temperature
Limit Polar e s a l
100% Dimethylpolysiloxane DB‐1
360C Non 0 0.22 0.18 0.51
(50%‐Phenyl)‐methylpolysiloxane DB‐17
300C Mid 0.05 0.85 0.38 0.57
(50%‐Trifluoropropyl)‐methylpolysiloxane DB‐210
260C Highly ‐0.46 1.38 0.20 0.46
(50%‐Cyanopropylphenyl)‐dimethylpolysiloxane DB‐225
260C Mid / Highly ‐0.06 1.33 1.32 0.51
(50%‐Cyanopropyl)‐methylpolysiloxane DB‐23
260C Highly 0.03 2.04 1.95 0.43
Poly(ethylene glycol) DB‐WAX
260C Highly 0.21 1.41 2.12 0.51
33
Based on these six stationary‐phases, 30 column combinations can be envisaged, if the
order of the stationary‐phases is not taken into account. This number of combinations
should be narrowed down by choosing the columns based on the expected analyte –
stationary‐phase molecular interactions and considering the limitations, such as the
maximum column temperature, in relation to the method requirements.
Furthermore, certain interactions may be enhanced or reduced by choosing a column
containing a higher or a lower percentage of the corresponding functional group. For
example, the retention of aromatics can be reduced by using a 35% diphenyl
polydimethylsiloxane column instead of 50% diphenyl polydimethylsiloxane.
A relatively novel type of commercially available GC stationary‐phases are ionic
liquids, or liquid salts, which are highly temperature stable, highly polar, and claimed
to have a broad application range [96]. Commercially available ionic‐liquids columns
for GC have also been characterized by using the solvation parameter model [97‐99].
A noticeable difference between ionic liquids and the commonly used poly(siloxane)
and poly(ethylene glycol) stationary‐phases is that ionic liquids show significant
hydrogen‐bond acidity. The non‐ionic liquid stationary‐phases typically show “b”
values between 0.5 and 1.6 instead of 0 [97].
Ionic liquids are said to show very low column bleed, due to their high thermal stability
and low vapor pressures. They can be operated at high temperatures. Non‐ionic polar
stationary‐phases show relatively high bleed at higher temperatures, resulting in
‘bleed lines’ throughout the 2D‐chromatogram, which may interfere with analytes of
interest.
However, ionic‐liquid GC column technology is not as mature yet as the conventional
poly(siloxane) and poly(ethylene) glycol column technology. As stated by Poole et al.
[99], column‐coating techniques, film stability, inertness, stationary‐phase chemistry
and column activity require further exploration. Despite this, the use of (commercially
available) ionic‐liquid columns in GC×GC is of interest and examples in literature are
increasing [100‐104].
34
Order of the stationary‐phases
The order of the stationary‐phases (primary and secondary column) in a column‐set is
also important to consider [105‐107]. Usually, the primary column, which in most
cases has standard column dimensions as used for 1D‐GC, provides a high‐resolution
separation in GC×GC. Also, the primary separation almost always involves
temperature programming. The secondary column, which in most cases is significantly
shorter and has a smaller internal diameter, provides significantly fewer
chromatographic plates. The secondary separation is significantly faster that the
primary separation and takes place under virtually isothermal conditions. The choice
of the order of the stationary‐phases should be based on the expected “analyte –
phase” molecular interactions and on the extent to which the order will affect
retention, separation, and chromatographic behavior of the analytes. As an example,
the method development for the analysis of a homologous series of polar carboxylic
acids in a complex hydrocarbon matrix is discussed. In this case, method development
could aim for a separation of all the carboxylic acids from the less‐polar hydrocarbon
matrix. Good peak shape and separation of the acids is important, while peak shapes
and separation of the matrix compounds is not important, provided that they do not
interfere with the analysis of the analytes. For the polar acids, a polar
polyethyleneglycol (PEG) stationary‐phase would provide significant retention,
separation, and good peak shapes. The matrix compounds, however, will show
significantly less retention and inferior peak shapes. For the paraffinic and aromatic
matrix compounds, a 50% diphenyl dimethylpolysiloxane would provide significant
retention and good peak shapes. However, the polar carboxylic‐acid analytes will
show significantly less retention and poor peak shapes. Thus, a good choice would be
to use the PEG column as the primary column, providing adequate separation and
good peak shapes for all the acid homologs, and to use the 50% diphenyl
dimethylpolysiloxane as the secondary column, providing adequate separation
between the acid analytes and the matrix compounds. For this column‐set, the polar
acid analytes, which enter the secondary column at higher temperatures compared to
the less‐polar matrix compounds with similar molecular weights, will show low
retention in the second dimension. The matrix compounds that co‐elute from the first
dimension will show significant retention, thus providing good separation between
35
the acids and the matrix compounds, while good peak shapes for the much‐less‐
retained polar acid analytes are maintained in both dimensions. In literature, many
examples can be found were “polar medium polar” column‐sets are used for the
analysis of polar analytes in less‐polar matrices in order to obtain good separation
between analytes and matrix, optimal analyte peak shapes, and adequate resolution
[108‐118].
Thus, available information concerning the chemical properties and possible analyte
– stationary‐phase interactions of the analytes and matrix compounds should be used
as a starting point for the column‐set selection. This selection may be based on the
expected molecular interactions and/or on information available in literature.
However, if sample information is not available or not adequate, the following general
guidelines may be followed:
Start with a “Non‐polar Medium polar” column‐set, such as a “100% dimethyl
polysiloxane” × “50% diphenyl dimethylpolysiloxane”. The analysis with this
column‐set will provide (more) information on the sample composition,
including the boiling point and polarity ranges (chemical group‐types) of the
GC‐amenable analytes and matrix compounds. This information can then be
used for a better considered column‐set selection (see table 3);
Use preferably the least polar stationary‐phases that provide adequate and
appropriate separation, retention, resolution, orthogonality, and satisfactory
primary and secondary analysis times. More polar columns are usually less
robust, less thermally stable and may cause more ‘column bleed lines’
throughout the 2D‐chromatogram. For high‐temperature applications, polar
ionic‐liquid columns may be considered [89, 97‐99].
Choose the order of the stationary‐phases of the column‐set based on the
expected analyte – stationary‐phase molecular interactions and on the extent
to which the order will affect retention, separation, and chromatographic
behavior (peak shapes) of the analytes [105‐107].
36
Prediction of GC×GC retention
Prediction of GC×GC retention may be useful for (1) selecting the primary and
secondary stationary‐phase chemistries to obtain optimal resolution, retention,
orthogonality, and/or group‐type separation, (2) optimizing a given GC×GC separation,
and (3) confirming identifications of components, which are well separated, but have
nearly identical mass spectra. Prediction of GC×GC retention requires input, such as
relevant chemical properties (e.g. molecular descriptors) of the sample analytes, the
primary and secondary stationary‐phase chemistries and/or relevant GC×GC
conditions (e.g. temperature, flow, pressure). As GC×GC is mainly used for the
separation of highly complex samples, containing often hundreds or even thousands
of compounds, prediction of the approximate retention‐times of a limited set of
analytes may be of limited use. However, predicting GC×GC retention, could be useful
for a first selection of primary and secondary stationary‐phase chemistries and/or for
GC×GC method optimization [119‐132].
2.4.2 59B59BColumn dimensions
The choice of the column dimensions, column length (L), column internal diameter (dc)
and stationary‐phase film thickness (df), for the primary (1L, 1dc, 1df) and secondary
columns (2L, 2dc, 2df) should be based on (1) the application requirements, which are
mainly the analysis time, the desired peak capacity (1n and 2n), and the required
column loadability or dynamic range for both dimensions, and on (2) the GC×GC
requirements, which mainly entail fulfilling the modulation criterion of having at least
2.5‐3 modulations [2, 3] per primary peak, implying that the secondary separation
must be fast enough to prevent wrap‐around. Also, acceptable flow regimes should
be realized in both dimensions, which often requires some kind of compromise.
However, flow‐mismatch and loadability issues may be solved or minimized by using
alternative GC×GC setups as discussed above.
For an optimized GC method, an average linear gas velocity ( ) should be used at
which the column plate height (Hmin) is at its minimum and the column generates its
maximum number of theoretical plates (Nmax).
37
(4)
The experimental number of theoretical plates N can be determined by:
5.54
. (5)
Where tR is the retention time, is the peak standard deviation and W0.5 is the peak
width at half height. More theoretical plates can be achieved by increasing L or by
decreasing Hmin. The plate height H can be described by the simplified Golay equation:
(6)
Where Dm is the diffusion coefficient of the analyte in the mobile phase, dc is the
column diameter, Df is the diffusion coefficient of the analyte in the stationary phase
and df is the stationary phase (film) thickness. The retention factor k can be calculated
by:
(7)
Where K is the partition coefficient, Vs/Vm is the phase ratio and t0 is the dead time.
Based on k the selectivity r12 of peak‐1 and peak‐2 can be calculated by:
(8)
The f(k) and g(k) factor are:
(9)
(10)
38
When introducing the following dimensionless parameters h (dimensionless plate‐
height), v (dimensionless gas velocity) and δf (dimensionless film thickness):
(11)
(12)
(13)
The Golay plate‐height equation can be simplified in
(14)
where the factor f(k) “describes” the mobile‐phase mass transfer and the
stationary‐phase mass transfer [164]. Based on this equation it can be concluded that
the reduced film thickness δf should best be less than 0.3 and the df not much larger
than 0.001dc. For higher δf values minimum reduced plate‐height (hmin) is large and
the slope of the h‐v curve steeply increases with increasing k values. Higher 1δf values
(thick films) are more favourable with respect to mass‐loadability of the first
dimension column and also increases 1retention and 1peak‐widths of the analytes
which may be favourable with respect to modulation performance and mass‐
loadability of the second dimension column; in case of more modulations (cuts) per
1peak less mass is transferred into the second column thereby reducing the chance of
mass‐overloading and also the modulation criterion is easier to fulfil in case of broader
primary peaks. More important, thick films are favourable for the analysis of volatiles
or low molecular weight compounds. Furthermore, the retention factors k should be
kept low, given the fact that the plate‐height usually increases rapidly with increasing
k. Low k values can generally be achieved by programmed temperature conditions as
is the case for the primary dimension in GC×GC.
39
Shorter columns and especially columns with thinner stationary‐phase films will
decrease the elution temperature of analytes, expanding the opportunity to analyse
less‐volatile and less‐thermally‐stable compounds, and permitting the use of less‐
thermally‐stable stationary phases. The advantage of thin‐film columns with small
internal diameters is that they provide high column efficiencies (low plate heights) at
high average linear gas velocities, thus providing fast and efficient separations.
However, disadvantages of such columns are their high pressure drop, low loadability
[6‐9], and ‐ related to this ‐ their limited dynamic range.
To fulfil the GC×GC modulation criterion, the secondary separation needs to be
significantly faster than the primary separation. For this reason, the secondary column
is often significantly shorter than the primary column (1L >> 2L). Also, the internal
diameter of the secondary column is usually chosen to be smaller than that of the
primary column, to increase the speed and efficiency of the secondary separation. The
commonly used GC×GC primary and secondary column dimensions are summarized in
table 4.
Table 4 Conventional GC×GC primary and secondary column dimensions
Column
dimension
Primary
column
Secondary
column
In general
Length L (m) 15 ‐ 60 0.5 – 2 1L >> 2L
Internal diameter dc (mm) 0.25 0.1 – 0.25 1dc > 2dc
Film thickness df (µm) 0.25 – 0.51 0.1 – 0.2 1df > 2df
1 thicker films may be used for the analysis of volatiles and or low molecular weight compounds or to increase mass loadability
Broad peaks, caused by mass overloading, are experienced especially with narrow,
thin‐film columns, because the loadability is proportional to dfdc2 [9]. The main
advantage of a narrow internal diameter, in case of no overloading (e.g. trace analysis,
no large interfering matrix peaks), is separation speed or peak production rate (peak
capacity per unit time). When speed of analysis is more important than secondary‐
dimension resolution thin‐film primary columns and narrow secondary columns are
40
preferred. When high resolution is more important than speed and the sample
features a broad analyte‐concentration range (large dynamic range), thicker film
primary columns and larger diameter secondary columns are better. An acceptable
compromise between speed and resolution, might be a 0.25 mm i.d., 0.25‐0.5 µm film
thickness primary column combined with 0.1 µm thin‐film, 150 µm i.d. secondary
column [134].
2.5 22B22BGC×GC optimization
After selecting the GC×GC set‐up, the injector and detector, the primary and
secondary stationary phase chemistries, and the corresponding (initial) column
dimensions, the GC×GC parameters need to be set and optimized. The most important
GC×GC parameters to optimize are the inlet pressure or column‐flow, the main oven
temperature program, and the modulation settings.
For method optimization and evaluation of the separation performance of a (two‐
dimensional) chromatographic system, performance metrics are necessary [135].
Optimization of the separation performance can focus on (1) the separation of specific
solutes (target analytes or target group types) or (2) on general separation
performance, such as the total number of peaks that can be resolved within a given
analysis time. Retention times (1tR, 2tR) and the corresponding peak widths (1, 2) of
chromatographic peaks are the basic parameters to asses separation performance.
Metrics of separation performance are:
‐ the resolution (see 2.4);
‐ the peak capacity (n) which describes the number of adjacent and equally
spaced ‘peaks’ (e.g. peak width of 4 required to fully resolve two peaks) that
may be accommodated in the chromatogram;
‐ the expected number of peaks that can be resolved in a given time window
which depends on the number of solutes and the peak saturation of the time
interval.
Optimization of a GC×GC separation towards several chromatographic performance
goals, such as resolution, peak capacity, orthogonality, dynamic range, and analysis
41
time, may be approached by multi‐criterion‐decision‐making (MCDM) methods, such
as Derringer’s desirability function [136], Pareto‐optimality [137], chromatographic
response functions (CRF) [162, 163], or combined‐threshold criteria [138]. Conflicting
chromatographic goals, such as maximizing peak capacity while minimizing analysis
time, lead to compromise settings. Balancing all goals against each other should result
in the most‐acceptable solution to the multi‐criterion chromatographic optimization
problem.
The Derringer desirability function is based on the (one‐ or two‐sided) transformation
of measured ‘performance values’ to a dimensionless desirability scale for each
criterion, so that these values can be combined to calculate the geometric mean
representing the overall‐quality D value. Bourguignon et al. [136] demonstrated the
use of this concept for the simultaneous optimization of three chromatographic
performance goals. It offers interesting opportunities for GC×GC optimization to
simultaneously optimize three or even more performance goals (such as “normalized
resolution product”, “group‐type resolution”, “total number of separated peaks”,
“orthogonality”, and “analysis time”), especially for complex samples.
For the Pareto‐optimality approach the minimum resolution among the target
compounds (as a measure of separation) and the maximum retention time was
explored. The minimum resolution and retention factors of each solute can be
calculated at every point in the ‘factor space’, for example for all values of the
parameters within their predefined ranges. This results in a series (“front”) of possible
Pareto‐optimal settings. Likewise, for complex samples the Pareto optimal front of the
peak capacity vs. analysis time can be determined.
Chromatographic response functions (CRF) contain factors related to separation and
analysis time, but include weighing factor (X; for example CRF = ‘separation metrics’ +
X* ”analysis time”) [162, 163]. For complex samples containing unknown numbers of
compounds, the number of detected peaks – which may vary during optimization –
may be added to the CRF. Thresholds criteria may be combined with CRF to define
‘solutions’ where, for example, resolution is considered to be sufficient.
42
2.5.1 60B60BInlet pressure or column‐flow
Beens et al. [2] describe a validated Microsoft Excel® spreadsheet for calculating the
efficiency of both columns in GC×GC‐FID and GC×GC‐MS as a function of the column‐
set inlet pressure. This Microsoft Excel® sheet may be used to assist in selecting
column dimensions and for inlet‐pressure or column‐flow optimization. The program
requires prospective column dimensions and carrier‐gas type (helium or hydrogen) as
the main input parameters. By using this program and published GC×GC data, Beens
et al. concluded that most column combinations reported in literature had been
operated far from the optimal flow settings, with negative effects on the obtained
resolution. Conventional column‐sets, using 0.1 mm i.d. second‐dimension columns,
were found to not necessarily be the best choice for GC×GC.
Although the dimensions suggested in table 4 are a good general starting point, the
optimal conditions may be different for the specific application being developed. As
an example, optimum conditions were calculated for a complex mixture of moderately
volatile compounds using the parameters in table 5 for GC×GC‐MS using helium as the
carrier gas.
43
Table 5 Parameters used for the calculation of optimum GC×GC‐MS conditions using
helium as the carrier gas
Primary column length 1L 15; 30; 60 m
Primary column internal diameter 1dc 250 µm
Primary column film thickness 1df 0.25 µm
Primary column coating efficiency 1CE 90 %
Secondary column length 2L 0.5; 1.0; 2.0 m
Secondary column internal diameter 2dc 100; 150; 200; 250 mm
Secondary column film thickness 2df 0.1 µm
Secondary column coating efficiency 2CE 90 %
Initial oven temperature Tb 100 C
Final oven temperature Te 300 C
Temperature gradient r 10C/t0 [139] C/s
Inlet (absolute) pressure Pi 50 ‐ 600 kPa
Outlet pressure Po 0 (MS) kPa
Primary retention factor at elution 1k 2 [140]
Secondary retention factor 2k Optimum (k 1)
Dm at 100C Dm 4*10‐5 m2/s
Ds for “silicone films” Ds 8*10‐10 m2/s
Modulator band broadening mod 0.008 [141] s
The total column‐set peak capacity (PC) is calculated by:
(15)
Where 1n and 2n are the primary and secondary column peak capacities, respectively.
The primary 1n peak capacity is estimated by:
.
(16)
44
Where Tb is the initial oven temperature, Te is the final oven temperature, 1k is the
primary‐column retention factor at elution [140], 1t0 is the primary‐column dead time,
and 1 is the primary‐column peak width. This calculation assumes the optimal
temperature rate of 10C per 1t0 void time [139] and isothermal elution during
1t0(1+1k) before starting the temperature program. The secondary 2n peak capacity is
roughly estimated by:
(17)
This calculation includes the time between 0 and 2t0 seconds as being part of the entire
usable secondary separation time which then equals 2tr. describes the secondary
peak broadening at 2t0 and is calculated by:
(18)
The extra peak broadening due to the modulator re‐injection ( ) is estimated to
be 0.008 seconds [141]. Additionally, describes the peak broadening at 2tR +
2t0 and is calculated by:
(19)
The results of the calculations, for GC×GC‐MS using helium as the carrier gas are given
in table 6. The results for the different column dimensions are given as peak capacities
(PC/1000, 1n and 2n) and the corresponding inlet pressures at three different analysis
times. Only results are shown for which 2tR/1σ is less than 1.5 (=modulation criterion).
45
Table 6 Calculated GC×GC‐MS settings using helium as the carrier gas for different
column dimensions and analysis times. The following dimensions were kept constant:
1dc=250µm, 1df=0.25µm and 2df=0.1µm.
analysis time
0.5 h
analysis time
1 h
analysis time
2 h
analysis time
3 h
1L 2L 2dc PC Pi 1n 2n PC Pi 1n 2n PC Pi 1n 2n PC Pi 1n 2n
m m µm 103 kPa
103 kPa
103 kPa
103 kPa
60 0.5 100 10 529 748 13 16 270 708 23 13 182 554 24
60 1 100 20 358 704 28 19 243 602 31
60 2 100 23 534 763 30 25 358 700 36
60 0.5 150 11 380 839 13 11 199 658 17 7 133 472 14
60 1 150 13 419 671 19 14 215 592 23 9 144 461 19
60 2 150 16 254 545 29 11 171 466 23
60 0.5 200 11 358 824 13 8 182 617 13 60 1 200 12 369 634 19 10 188 540 18 6 127 420 13
60 2 200 10 204 473 21 6 138 402 15
60 0.5 250 10 347 818 12 7 182 616 11 4 122 439 8
60 1 250 12 353 729 16 8 182 578 13 4 122 425 9
60 2 250 7 188 447 17 4 127 378 12
30 0.5 100 7 336 587 12 9 177 472 19 2 61 186 11
30 1 100 8 496 506 15 12 259 488 25 8 138 330 23 30 2 100 12 221 351 33 7 149 274 25
30 0.5 150 7 204 536 13 4 105 354 13 2 56 198 8 30 1 150 8 237 501 16 6 127 380 16 30 2 150 3 83 203 15 2 56 156 10
30 0.5 200 6 177 534 10 3 94 335 8 30 1 200 7 188 444 15 3 100 321 11 1 50 179 6 30 2 200 30 0.5 250 5 171 402 13 2 89 289 8 30 1 250 6 177 425 13 3 94 307 8 30 2 250 1 50 144 6 15 0.5 100 5 127 215 23 2 67 152 16 15 1 100 7 204 266 26 4 111 198 21 1 56 112 12 15 2 100 6 193 194 32 3 100 138 19 1 67 100 12
15 0.5 150 2 61 138 13 15 1 150 3 78 195 13 15 2 150 1 50 119 9 15 0.5 200 1 50 178 5 15 1 200 1 56 156 8 15 0.5 250 1 50 180 5 15 1 250 1 50 144 6
Table 6 may be used to select column dimensions and the corresponding optimum
inlet pressure for GC×GC‐MS using helium as the carrier gas, based on (1) the required
analysis time, (2) the required peak capacity and (3) the required dynamic range, as
the secondary column loadability is proportional with 2dc2 [9]. Caveat: the rough
estimates of peak capacities should be treated with care. In literature no experimental
confirmation for the peak capacity estimates could be found.
46
It should be stressed that the peak capacity signifies the potential of separating
compounds in a fully utilized two‐dimensional separation space. Although this is never
the case, it can be used to compare systems if “retention matched” columns are used
(showing peaks that span equal fractions of the entire two‐dimensional separation
space) and orthogonality is taken into account (cf. chapter 3). Additionally, for a given
sample 2df should be adjusted when 2dc is changed, to keep the phase ratio the same.
2.5.2 61B61BOven‐temperature programming
Often, a slow GC×GC programming rate of 2 to 5 C/min is used to produce relatively
broad primary peaks, which are required to fulfil the modulation criterion. As a good
starting point, the optimum programming rate, for the primary dimension may be
estimated as 10°C per primary void time [139].
When using two independent GC ovens, usually the main GC‐oven, with the primary
column installed and a second oven for the secondary column, the temperatures of
the two columns can, to some extent, be programmed independently. The secondary
retention can then be tuned and wrap‐around may be prevented. Usually, the same
programming rates are applied, but a temperature offset is introduced (e.g. a
secondary oven temperature offset of +5 to +30C). Chow et al. [142] demonstrated
the potential of temperature programming for the secondary dimension separation.
This is technically complicated, because of the short modulation time, during which
the column needs to be heated and cooled in a reproducible manner.
2.6 23B23BData processing
The data‐processing approach depends on the analytical question and on the related
type of analysis, as illustrated in figure 1. The type of analysis may be screening, target,
or group‐type and each could be quantitative, qualitative, both quantitative and
qualitative, or differential analysis.
The quality of GC×GC results strongly depends on the quality of the acquired raw data,
the data‐processing approach, and the settings. Key for obtaining good‐quality GC×GC
47
data are, as discussed previously, GC×GC method development, method optimization,
and good chromatography practice.
Compared to 1D‐GC, GC×GC provides more separation power and may provide
structured chromatograms. More separation power implies more resolution and
fewer co‐elutions, which should lead to enhanced qualitative and quantitative
analysis, for example because more single‐component mass spectra are obtained.
Structured chromatograms also greatly support qualitative analysis; the peak position
contains information on molecular structure (e.g. isomers, branching) and/or chemical
group types, which may be used for structure elucidation of unknowns.
The aim of data processing is to extract as much relevant information from the
obtained data as possible in order to answer the analytical question. Preferably,
irrelevant or interfering data should be filtered out, but no relevant information
should be missed. The data‐processing approach includes transformation of the
modulated raw “1D‐data” into 2D‐data, data pre‐treatment, data processing, and data
post‐treatment [143‐148].
Transformation is performed by stacking all the sequential second‐dimension
chromatograms side by side, so as to create a two‐dimensional chromatogram
representing the retention on the first column on the x‐axis and the retention on the
second column on the y‐axis [149]. Transformation and visualization the generated
2D‐chromatogram, for example as a 2D‐contour plot or a 3D‐plot, can be performed
by commercial GC×GC software [150‐152].
Data pre‐treatment may include baseline correction, smoothing and resampling of the
raw data. Resampling entails removing data points or certain mass traces (m/z values),
which do not contain relevant information This may speed up or simplify data
processing and/or improve the visualization.
Data processing mainly includes peak detection, peak integration, and calculation of
the peak volume. When using MS, mass deconvolution may be used for peak detection
and for obtaining (deconvoluted) mass spectra for identification, for example by
48
searching in an MS library. Peaks may also be identified based on their known primary
and secondary retention times, an analyte‐retention‐time database and retention‐
time windows or by using so‐called 2D templates. The final step of data processing is
in most cases the generation of a peak list containing, for examples, the names of
identified peaks, retention times (1tR and 2tR), and peak volumes.
Data post‐treatment may involve adjusting the peak list, for example by removing
non‐relevant peaks (such as column‐bleed peaks) or by summing areas of peaks which
belong to the same group in case of group‐type analysis.
In case the results of multiple complex samples must be compared, peak alignment
should be performed, preferably using chemometric tools [153‐160]. This is required,
because retention‐time shifts may occur in both dimensions. Retention‐time shifts
often occur after changing a column or column‐set, or in time due to column aging.
Retention‐time shifts are minimized by performing two‐dimensional retention time
locking before a series of analyses [161]. This improves the performance of post‐
treatment chemometric alignment tools, because only peaks with relatively small
shifts need to be aligned.
2.7 24B24BConclusions
It can be concluded that GC×GC method development is significantly more difficult
than method development for 1D‐GC. (1) more method‐development choices need to
be made, (2) optimization is complex, because of the complex interplay of many first‐
and second‐dimension parameters, (3) optimization is restricted by certain GC×GC
requirements, such as the modulation criterion and temperature restrictions, and (4)
the difference in the primary and secondary column diameters lead to flow‐mismatch
and column‐loadability issues.
GC×GC method development starts with a thorough understanding of the main
analytical question, since this determines the type of analysis (e.g. target, screening,
or group‐type) and the analytical requirements (e.g. limit of detection, speed of
49
analysis and dynamic range) for the method to be developed. Knowledge of the
sample characteristics (analytes and matrix compounds) is essential for determining
the required GC×GC setup, including a GC injector and injection mode that ensures
representative sampling, and a detector or multiple detectors that ensure appropriate
analyte detection. Alternative GC×GC setups may be chosen, such as a dual‐
secondary‐column‐setup (GC×GC2) to ensure more optimal flow conditions for both
columns and/or to facilitate dual‐detection.
The most‐important and most‐difficult task in GC×GC method development is the
choice of the column‐set. The choice should be based on the expected interactions
between the sample (analytes and matrix compounds) and the first‐dimension and
second‐dimension stationary phases. The column‐set must provide adequate
retention, resolution, selectivity (orthogonality), and good analyte peak shapes. In
case of group‐type analysis adequate separation between groups is required. The
choice of the column‐set also includes the selection of the primary and secondary
column dimensions (L, dc, df). This should be mainly based on (1) the application
requirements, such as required analysis time, required efficiency (peak capacity), and
required loadability and (2) the GC×GC requirements (or restrictions), such as the
modulation criterion and adequate retention.
A table is provided for an example in which GC×GC‐MS is employed, using helium as
the carrier gas. Optimal column dimensions are discussed. The results may be used to
select column dimensions and the corresponding inlet pressure based on (1) the
required analysis time, (2) the required peak capacity and (3) the required dynamic
range.
GC×GC data processing is in some respects less complex than 1D‐GC data processing,
especially for truly complex samples. However, compared to one‐dimensional GC,
peaks shifts may occur in two dimensions, causing peak alignment issues. Retention‐
time shifts can be minimized or eliminated by (pre‐analysis) retention‐time locking
and/or by applying chemometric (post‐analysis) peak‐alignment procedures.
50
2.8 25B25BReferences
[1] Harynuk, J., Górecki, T., (2007) Am. Lab., 39 (4), pp. 36‐39.
[2] Beens, J., Janssen, H.‐G., Adahchour, M., Brinkman, U.A.Th., (2005) J. Chromatogr.
A., 1086 (1‐2), pp. 141‐150.
[3] Khummueng, W., Harynuk, J., Marriott, P.J., (2006) Anal. Chem., 78 (13), pp. 4578‐
4587.
[4] Peroni, D., Janssen, H.‐G., (2014) J. Chromatogr. A., 1332, pp. 57‐63.
[5] Peroni, D., Sampat, A.A.S., van Egmond, W., de Koning, S., Cochran, J., Lautamo, R.,
Janssen, H.‐G., (2013) J. Chromatogr. A., 1317, pp. 3‐11.
[6] Koek, M.M., Muilwijk, B., van Stee, L.L.P., Hankemeier, T., (2008) J. Chromatogr. A.,
1186 (1‐2), pp. 420‐429.
[7] Harynuk, J., Górecki, T., De Zeeuw, J., (2005) J. Chromatogr. A., 1071 (1‐2), pp. 21‐
27.
[8] Kallio, M., Hyötyläinen, T., (2007) J. Chromatogr. A., 1148 (2), pp. 228‐235.
[9] Ghijsen, R.T., Poppe, H., Kraak, J.C., Duysters, P.P.E., (1989) Chromatographia, 27
(1‐2), pp. 60‐66.
[10] Dallüge, J., Van Rijn, M., Beens, J., Vreuls, R.J.J., Brinkman, U.A.Th., (2002) J.
Chromatogr. A., 965 (1‐2), pp. 207‐217.
[11] Xia, D., Gao, L., Zheng, M., Wang, S., Liu, G., (2016) Anal. Chim. Acta, 937, pp. 160‐
167.
[12] Muscalu, A.M., Edwards, M., Górecki, T., Reiner, E.J., (2015) J. Chromatogr. A.,
1391 (1), pp. 93‐101.
[13] Magagna, F., Cordero, C., Cagliero, C., Liberto, E., Rubiolo, P., Sgorbini, B., Bicchi,
C., (2017) Food Chem., 225, pp. 276‐287.
[14] Almstetter, M.F., Oefner, P.J., Dettmer, K., (2012) Methods in Molecular Biology,
815, pp. 399‐411.
[15] Schmarr, H.‐G., Bernhardt, J., (2010) J. Chromatogr. A., 1217 (4), pp. 565‐574.
[16] Jennerwein, M.K., Eschner, M., Gröger, T., Wilharm, T., Zimmermann, R., (2014)
Energy and Fuels, 28 (9), pp. 5670‐5681.
[17] Cordero, C., Bicchi, C., Rubiolo, P., (2008) J. Arg. Food Chem., 56 (17), pp. 7655‐
7666.
51
[18] Van Deursen, M., Beens, J., Reijenga, J., Lipman, P., Cramers, C., Blomberg, J.,
(2000) J. Sep. Sci., 23 (7‐8), pp. 507‐510.
[19] Edwards, M., Górecki, T., (2015) J. Chromatogr. A., 1402, pp. 110‐123.
[20] Nizio, K.D., Harynuk, J.J., (2014) Energy and Fuels, 28 (3), pp. 1709‐1716.
[21] www.Agilent.com Agilent J&W GC Column Selection Guide
[22] www.sigmaaldrich.com Column Selection Guide
[23] Sgorbini, B., Cagliero, C., Boggia, L., Liberto, E., Reichenbach, S.E., Rubiolo, P.,
Cordero, C., Bicchi, C., (2015) Flavour Frag. J., 30 (5), pp. 366‐380.
[24] Nicolotti, L., Cordero, C., Bressanello, D., Cagliero, C., Liberto, E., Magagna, F.,
Rubiolo, P., Sgorbini, B., Bicchi, C., (2014) J. Chromatogr. A., 1360, pp. 264‐274.
[25] Seeley, J.V., Kramp, F.J., Sharpe, K.S., (2001) J. Sep. Sci., 24 (6), pp. 444‐450.
[26] Seeley, J.V., Kramp, F.J., Sharpe, K.S., Seeley, S.K., (2002) J. Sep. Sci., 25 (1‐2), pp.
53‐59.
[27] Cordero, C., Rubiolo, P., Reichenbach, S.E., Carretta, A., Cobelli, L., Giardina, M.,
Bicchi, C., (2017) J. Chromatogr. A., 1480, pp. 70‐82.
[28] Reichenbach, S.E., Rempe, D.W., Tao, Q., Bressanello, D., Liberto, E., Bicchi, C.,
Balducci, S., Cordero, C., (2015) Anal. Chem., 87 (19), pp. 10056‐10063.
[29] Bieri, S., Marriott, P.J., (2006) Anal. Chem., 78 (23), pp. 8089‐8097.
[30] Tranchida, P.Q., Zoccali, M., Franchina, F.A., Dugo, P., Mondello, L., (2013) J. Sep.
Sci., 36 (1), pp. 212‐218.
[31] Tranchida, P.Q., Purcaro, G., Fanali, C., Dugo, P., Dugo, G., Mondello, L., (2010) J.
Chromatogr. A., 1217 (25), pp. 4160‐4166.
[32] Tranchida, P.Q., Purcaro, G., Conte, L., Dugo, P., Dugo, G., Mondello, L., (2009) J.
Chromatogr. A., 1216 (43), pp. 7301‐7306.
[33] Tranchida, P.Q., Purcaro, G., Conte, L., Dugo, P., Dugo, G., Mondello, L., 81 (20),
pp. 8529‐8537.
[34] Tranchida, P.Q., Casilli, A., Dugo, P., Dugo, G., Mondello, L., (2007) Anal. Chem.,
79 (6), pp. 2266‐2275.
[35] Cordero, C., Bicchi, C., Galli, M., Galli, S., Rubiolo, P., (2008) J. Sep. Sci., 31 (19),
pp. 3437‐3450.
[36] Oldridge, N., Panic, O., Górecki, T., (2008) J. Sep. Sci., 31 (19), pp. 3375‐3384.
52
[37] Harynuk, J., Górecki, T., (2006) J. Chromatogr. A., 1105 (1‐2 SPEC. ISS.), pp. 159‐
167.
[38] Harynuk, J., Górecki, T., (2004) J. Sep. Sci., 27 (5‐6), pp. 431‐441.
[39] Peroni, D., Vonk, R.J., van Egmond, W., Janssen, H.‐G., (2012) J. Chromatogr. A.,
1268, pp. 139‐149.
[40] Grob R, Barry E. Modern Practice of Gas Chromatography. Hoboken, N.J.: Wiley‐
Interscience; 2004.
[41] Mommers, J., Pluimakers, G., Knooren, J., Dutriez, T., van der Wal, S., (2013) J.
Chromatogr. A., 1297, pp. 179‐185.
[42] Mommers, J., Knooren, J., Dutriez, T., Ritzen, E., van der Wal, S., (2012) J.
Chromatogr. A., 1270, pp. 305‐309.
[43] Mommers, J., Ritzen, E., Dutriez, T., van der Wal, S., (2016) J. Chromatogr. A.,
1461, pp. 153‐160.
[44] Liu, Z., Phillips, J.B., (1991) J. Chrom. Sci., 29 (5), pp. 227‐231.
[45] Sharif, K.M., Chin, S.‐T., Kulsing, C., Marriott, P.J., (2016) TrAC ‐ Trends in Anal.
Chem., 82, pp. 35‐54.
[46] Seeley, J.V., (2012) J. Chromatogr. A., 1255, pp. 24‐37.
[47] Edwards, M., Mostafa, A., Górecki, T., (2011) Anal. Bioanal. Chem.., 401 (8), pp.
2335‐2349.
[48] Tranchida, P.Q., Purcaro, G., Dugo, P., Mondello, L., Purcaro, G., (2011) TrAC ‐
Trends in Anal. Chem., 30 (9), pp. 1437‐1461.
[49] Seeley, J.V., Micyus, N.J., McCurry, J.D., Seeley, S.K., (2006) Am. Lab., 38 (9), pp.
24‐26.
[50] Pursch, M., Sun, K., Winniford, B., Cortes, H., Weber, A., McCabe, T., Luong, J.
, (2002) Anal. Bioanal. Chem.., 373 (6), pp. 356‐367.
[51] Begnaud, F., Debonneville, C., Probst, J.‐P., Chaintreau, A., Morrison, P.D., Adcock,
J.L., Marriott, P.J., (2009) J. Sep. Sci., 32 (18), pp. 3144‐3151.
[52] Gaines, R.B., Frysinger, G.S., (2004) J. Sep. Sci., 27 (5‐6), pp. 380‐388.
[53] Mondello, L., Tranchida, P.Q., Dugo, P., Dugo, G., (2008) Mass Spectrom. Rev., 27
(2), pp. 101‐124.
[54] Pasikanti, K.K., Ho, P.C., Chan, E.C.Y., (2008) J. Chrom. B.: Analytical Technologies
in the Biomedical and Life Sciences, 871 (2), pp. 202‐211.
53
[55] Dallüge, J., Van Rijn, M., Beens, J., Vreuls, R.J.J., Brinkman, U.A.Th., (2002) J.
Chromatogr. A., 965 (1‐2), pp. 207‐217.
[56] Dallüge, J., Vreuls, R.J.J., Beens, J., Brinkman, U.A.Th., (2002) J. Sep. Sci., 25 (4),
pp. 201‐214.
[57] Li, X., Xu, Z., Lu, X., Yang, X., Yin, P., Kong, H., Yu, Y., Xu, G., (2009) Anal. Chim.
Acta, 633 (2), pp. 257‐262.
[58] Ochiai, N., Ieda, T., Sasamoto, K., Takazawa, Y., Hashimoto, S., Fushimi, A., Tanabe,
K., (2011) J. Chromatogr. A., 1218 (39), pp. 6851‐6860.
[59] Ieda, T., Ochiai, N., Miyawaki, T., Ohura, T., Horii, Y., (2011) J. Chromatogr. A.,
1218 (21), pp. 3224‐3232.
[60] Reichenbach, S.E., Tian, X., Tao, Q., Ledford Jr., E.B., Wu, Z., Fiehn, O., (2011)
Talanta, 83 (4), pp. 1279‐1288.
[61] Ubukata, M., Jobst, K.J., Reiner, E.J., Reichenbach, S.E., Tao, Q., Hang, J., Wu, Z.,
Dane, A.J., Cody, R.B., (2015) J. Chromatogr. A., 1395, pp. 152‐159.
[62] Veenaas, C., Haglund, P., (2017) Anal. Bioanal. Chem.., 409 (20), pp. 4867‐4883.
[63] Maciel, G.P.S., Machado, M.E., Da Cunha, M.E., Lazzari, E., Da Silva, J.M., Jacques,
R.A., Krause, L.C., Barros, J.A.S., Caramão, E.B., (2015) J. Sep. Sci., 38 (23), pp. 4071‐
4077.
[64] Schneider, J.K., da Cunha, M.E., dos Santos, A.L., Maciel, G.P.S., Brasil, M.C., Pinho,
A.R., Mendes, F.L., Jacques, R.A., Caramão, E.B., (2014) J. Chromatogr. A., 1356, pp.
236‐240.
[65] Mondello, L., Casilli, A., Tranchida, P.Q., Dugo, G., Dugo, P., (2005) J. Chromatogr.
A., 1067 (1‐2), pp. 235‐243.
[66] Adahchour, M., Brandt, M., Baier, H.‐U., Vreuls, R.J.J., Batenburg, A.M., Brinkman,
U.A.Th., (2005) J. Chromatogr. A., 1067 (1‐2), pp. 245‐254.
[67] Xia, D., Gao, L., Zhu, S., Zheng, M., (2014) Anal. Bioanal. Chem.., 406 (29), pp.
7561‐7570.
[68] da Silva, J.M., Zini, C.A., Caramão, E.B., (2011) J. Chromatogr. A., 1218 (21), pp.
3166‐3172.
[69] Bordajandi, L.R., Ramos, L., González, M.J., (2006) J. Chromatogr. A., 1125 (2), pp.
220‐228.
54
[70] Toraman, H.E., Franz, K., Ronsse, F., Van Geem, K.M., Marin, G.B., (2016) J.
Chromatogr. A., 1460, pp. 135‐146.
[71] Lissitsyna, K., Huertas, S., Quintero, L.C., Polo, L.M., (2013) J. Sep. Sci., 36 (11), pp.
1768‐1773.
[72] Ruiz‐Guerrero, R., Vendeuvre, C., Thiébaut, D., Bertoncini, F., Espinat, D., (2006)
J. Chrom. Sci., 44 (9), pp. 566‐573.
[73] Blomberg, J., Riemersma, T., Zuijlen, M.V., Chaabani, H., (2004) J. Chromatogr. A.,
1050 (1), pp. 77‐84.
[74] Hua, R., Wang, J., Kong, H., Liu, J., Lu, X., Xu, G., (2004) J. Sep. Sci., 27 (9), pp. 691‐
698.
[75] Hua, R., Li, Y., Liu, W., Zheng, J., Wei, H., Wang, J., Lu, X., Kong, H., Xu, G., (2003)
J. Chromatogr. A., 1019 (1‐2), pp. 101‐109.
[76] Nizio, K.D., Harynuk, J.J., (2014) Energy and Fuels, 28 (3), pp. 1709‐1716.
[77] von Mühlen, C., de Oliveira, E.C., Morrison, P.D., Zini, C.A., Caramão, E.B.,
Marriott, P.J., (2007) J. Sep. Sci., 30 (18), pp. 3223‐3232.
[78] Liu, X., Mitrevski, B., Li, D., Li, J., Marriott, P.J., (2013) Microchemical Journal, 111,
pp. 25‐31.
[79] Mitrevski, B., Amer, M.W., Chaffee, A.L., Marriott, P.J., (2013) Anal. Chim. Acta,
803, pp. 174‐180.
[80] Chin, S.‐T., Wu, Z.‐Y., Morrison, P.D., Marriott, P.J., (2010) Analytical Methods, 2
(3), pp. 243‐253.
[81] Liu, X., Li, D., Li, J., Rose, G., Marriott, P.J., (2013) Journal of Hazardous Materials,
263, pp. 761‐767.
[82] Freye, C.E., Fitz, B.D., Billingsley, M.C., Synovec, R.E., (2016) Talanta, 153, pp. 203‐
210.
[83] Blase, R.C., Patrick, E.L., Mitchell, J.N., Libardoni, M., (2015) Anal. Chem. Research,
3, pp. 54‐62.
[84] Amorim, L.C.A., Dimandja, J.‐M., Cardeal, Z.d.L., (2009) J. Chromatogr. A., 1216
(14), pp. 2900‐2904.
[85] Marsman, J.H., Wildschut, J., Mahfud, F., Heeres, H.J., (2007) J. Chromatogr. A.,
1150 (1‐2), pp. 21‐27.
55
[86] Peters S, Vivó‐Truyols G, Marriott P, Schoenmakers P., J. Chromatogr. A.,
2007;1146:232‐241.
[87] Poole C, Poole S. Chromatography Today, Oxford: Elsevier Science; 1991.
[88] Berezkin V. Poole, Coulin F., The Essence of Chromatography, Amsterdam,
Elsevier Science, 2002, 925 pp. Journal Of Anal. Chem., 2005;(2):193.
[89] Poole, C.F., (2012) Gas Chromatography, pp. 137‐159.
[90] Abraham, M.H., Poole, C.F., Poole, S.K., (1999) J. Chromatogr. A., 842 (1‐2), pp.
79‐114.
[91] Abraham, M.H., Ibrahim, A., Zissimos, A.M., (2004) J. Chromatogr. A., 1037 (1‐2),
pp. 29‐47.
[92] C.F. Poole, S.N. Atapattu, S.K. Poole, A.K. Bell, Anal. Chim. Acta. 652 (2009) 32‐53.
[93] M. Vitha, P.W. Carr, J. Chromatogr. A 1126 (2006) 143‐194.
[94] S.N. Atapattu, C.F. Poole, J. Chromatogr. A 1195 (2008) 136e145.
[95] Poole, C.F., Poole, S.K., (2002) J. Chromatogr. A., 965 (1‐2), pp. 263‐299.
[96] Nan, H., Zhang, C., O'Brien, R.A., Benchea, A., Davis, J.H., Jr., Anderson, J.L., (2017)
J. Chromatogr. A., 1481, pp. 127‐136.
[97] Poole CF, Poole SK., J. Sep. Sci. 34 (2011) 888‐900.
[98] C. Yao, J.L. Anderson, J. Chromatogr. A 1216 (2009) 1658‐1712.
[99] Poole, C.F., Lenca, N., (2014) J. Chromatogr. A., 1357, pp. 87‐109.
[100] Seeley, J.V., Seeley, S.K., Libby, E.K., Breitbach, Z.S., Armstrong, D.W., (2008)
Anal. Bioanal. Chem.., 390 (1), pp. 323‐332.
[101] Krupčík, J., Gorovenko, R., Špánik, I., Bočková, I., Sandra, P., Armstrong, D.W.,
(2013) J. Chromatogr. A., 1301, pp. 225‐236.
[102] Zapadlo, M., Krupčík, J., Kovalczuk, T., Májek, P., Špánik, I., Armstrong, D.W.,
Sandra, P., (2011) J. Chromatogr. A., 1218 (5), pp. 746‐751.
[103] Zapadlo, M., Krupčík, J., Májek, P., Armstrong, D.W., Sandra, P., (2010) J.
Chromatogr. A., 1217 (37), pp. 5859‐5867.
[104] Cappelli Fontanive, F., Souza‐Silva, É.A., Macedo da Silva, J., Bastos Caramão, E.,
Alcaraz Zini, C., (2016) J. Chromatogr. A., 1461, pp. 131‐143.
[105] Adahchour, M., Beens, J., Vreuls, R.J.J., Batenburg, A.M., Brinkman, U.A.Th.,
(2004) J. Chromatogr. A., 1054 (1‐2), pp. 47‐55.
56
[106] Hu, S.Z., Li, S.F., Cao, J., Zhang, D.M., Ma, J., He, S., Wang, X.L., Wu, M., A, (2014)
Petroleum Science and Technology, 32 (5), pp. 565‐574.
[107] Zhu, S., (2009) J. Chromatogr. A., 1216 (15), pp. 3312‐3317.
[108] Ristic, N.D., Djokic, M.R., Konist, A., Van Geem, K.M., Marin, G.B., (2017) Fuel
Processing Technology, 167, pp. 241‐249.
[109] Mcgregor, L.A., Gauchotte‐Lindsay, C., Daéid, N.N., Thomas, R., Daly, P., Kalin,
R.M., (2011) J. Chromatogr. A., 1218 (29), pp. 4755‐4763.
[110] Freye, C.E., Fitz, B.D., Billingsley, M.C., Synovec, R.E., (2016) Talanta, 153, pp.
203‐210.
[111] Li, S., Cao, J., Hu, S., Luo, G., (2015) Marine and Petroleum Geology, 66, pp. 791‐
800.
[112] Toraman, H.E., Dijkmans, T., Djokic, M.R., Van Geem, K.M., Marin, G.B., (2014) J.
Chromatogr. A., 1359, pp. 237‐246.
[113] Kehimkar, B., Hoggard, J.C., Marney, L.C., Billingsley, M.C., Fraga, C.G., Bruno,
T.J., Synovec, R.E., (2014) J. Chromatogr. A., 1327, pp. 132‐140.
[114] Li, S., Hu, S., Cao, J., Wu, M., Zhang, D., (2012) International Journal of Molecular
Sciences, 13 (9), pp. 11399‐11410.
[115] van der Westhuizen, R., Ajam, M., De Coning, P., Beens, J., de Villiers, A., Sandra,
P., (2011) J. Chromatogr. A., 1218 (28), pp. 4478‐4486.
[116] Dutriez, T., Borras, J., Courtiade, M., Thiébaut, D., Dulot, H., Bertoncini, F.,
Hennion, M.‐C., (2011) J. Chromatogr. A., 1218 (21), pp. 3190‐3199.
[117] Mahé, L., Dutriez, T., Courtiade, M., Thiébaut, D., Dulot, H., Bertoncini, F., (2011)
J. Chromatogr. A., 1218 (3), pp. 534‐544.
[118] Omais, B., Crepier, J., Charon, N., Courtiade, M., Quignard, A., Thiébaut, D.,
(2013) The Analyst, 138 (8), pp. 2258‐2268.
[119] Seeley, J.V., Libby, E.M., Edwards, K.A.H., Seeley, S.K., (2009) J. Chromatogr. A.,
1216 (10), pp. 1650‐1657.
[120] Seeley, J.V., Seeley, S.K., (2007) J. Chromatogr. A., 1172 (1), pp. 72‐83.
[121] Zhu, S., He, S., Worton, D.R., Goldstein, A.H., (2012) J. Chromatogr. A., 1233, pp.
147‐151.
[122] Vendeuvre, C., Bertoncini, F., Thiébaut, D., Martin, M., Hennion, M.‐C., (2005) J.
Sep. Sci., 28 (11), pp. 1129‐1136.
57
[123] Dorman, F.L., Schettler, P.D., Vogt, L.A., Cochran, J.W., (2008) J. Chromatogr. A.,
1186 (1‐2), pp. 196‐201.
[124] McGinitie, T.M., Harynuk, J.J., (2012) J. Chromatogr. A., 1255, pp. 184‐189.
[125] Zhu, S., Zhang, W., Dai, W., Tong, T., Guo, P., He, S., Chang, Z., Gao, X., (2014)
Analytical Methods, 6 (8), pp. 2608‐2620.
[126] Lu, X., Kong, H., Li, H., Ma, C., Tian, J., Xu, G., (2005) J. Chromatogr. A., 1086 (1‐
2), pp. 175‐184.
[127] Burel, A., Vaccaro, M., Cartigny, Y., Tisse, S., Coquerel, G., Cardinael, P., (2017) J.
Chromatogr. A., 1485, pp. 101‐119.
[128] Silva, A.C.A., Ebrahimi‐Najafadabi, H., McGinitie, T.M., Casilli, A., Pereira, H.M.G.,
Aquino Neto, F.R., Harynuk, J.J., 407 (14), art. no. 8627, pp. 4091‐4099.
[129] Barcaru, A., Anroedh‐Sampat, A., Janssen, H.‐G., Vivó‐Truyols, G., (2014) J.
Chromatogr. A., 1368, pp. 190‐198.
[130] Kulsing, C., Nolvachai, Y., Zeng, A.X., Chin, S.‐T., Mitrevski, B., Marriott, P.J.,
(2014) ChemPlusChem, 79 (6), pp. 790‐797.
[131] Noorizadeh, H., Noorizadeh, M., (2012) Medicinal Chemistry Research, 21 (8),
pp. 1997‐2005.
[132] D'Archivio, A.A., Incani, A., Ruggieri, F., (2011) Anal. Bioanal. Chem.., 399 (2), pp.
903‐913.
[133] Harynuk, J., Marriott, P.J., (2006) Anal. Chem., 78 (6), pp. 2028‐2034.
[134] Zhu, Z., Harynuk, J., Górecki, T., (2006) J. Chromatogr. A., 1105 (1‐2 SPEC. ISS.),
pp. 17‐24.
[135] Blumberg L., J. Chromatogr. A., January 1, 2011;1218:5375‐5385.
[136] Bourguignon B, Massart D., J. Chromatogr. A., 1991;586:11‐20.
[137] Vanhoutte D, Vivó‐Truyols G, Schoenmakers P., J. Chromatogr. A., 2012;1235:39‐
48.
[138] Debets H, Weyland J, Doornbos D., Anal. Chim. Acta, 1983;150:259‐265.
[139] Blumberg, L.M., Klee, M.S., (2000) Journal of Microcolumn Separations, 12 (9),
pp. 508‐514.
[140] Blumberg L, Klee M., J. Chromatogr. A., January 1, 2001;918:113‐120.
[141] Klee M, Cochran J, Merrick M, Blumberg L., J. Chromatogr. A., 2015;1383:151‐
159.
58
[142] Chow, H.‐Y.J., Górecki, T., (2017) Anal. Chem., 89 (16), pp. 8207‐8211.
[143] van Stee, L.L.P., Brinkman, U.A.T., (2016) TrAC ‐ Trends in Anal. Chem., 83, pp. 1‐
13.
[144] Zeng, Z., Li, J., Hugel, H.M., Xu, G., Marriott, P.J., (2014) TrAC ‐ Trends in Anal.
Chem., 53, pp. 150‐166.
[145] Murray, J.A., (2012) J. Chromatogr. A., 1261, pp. 58‐68.
[146] Pierce, K.M., Kehimkar, B., Marney, L.C., Hoggard, J.C., Synovec, R.E., (2012) J.
Chromatogr. A., 1255, pp. 3‐11.
[147] Pierce, K.M., Mohler, R.E., (2012) Separation and Purification Reviews, 41 (2),
pp. 143‐168.
[148] Synovec, R.E., Prazen, B.J., Johnson, K.J., Fraga, C.G., Bruckner, C.A., (2003)
Advances in Chromatography, 42, pp. 1‐42.
[149] Adahchour, M., Beens, J., Vreuls, R.J.J., Brinkman, U.A.Th., (2006) TrAC ‐ Trends
in Anal. Chem., 25 (5), pp. 438‐454.
[150] Chromatof, Leco: https://www.leco.com/products/separation‐
science/software‐accessories/chromatof‐software
[151] GC Image, Zoex: http://www.gcimage.com/index.html
[152] ChromSquare, Shimadzu: http://www.shimadzu.com/an/2d‐chrom/gc_cs.html
[153] Fraga, C.G., Prazen, B.J., Synovec, R.E., (2000) Anal. Chem., 72 (17), pp. 4154‐
4162.
[154] Kim, S., Fang, A., Wang, B., Jeong, J., Zhang, X., (2011) Bioinformatics, 27 (12),
art. no. btr188, pp. 1660‐1666.
[155] Parastar, H., Jalali‐Heravi, M., Tauler, R., (2012) Chemometrics and Intelligent
Laboratory Systems, 117, pp. 80‐91.
[156] Kim, S., Koo, I., Fang, A., Zhang, X., (2011) BMC Bioinformatics, 12, art. no. 235,
[157] Jeong, J., Shi, X., Zhang, X., Kim, S., Shen, C., (2012) BMC Bioinformatics, 13 (1),
art. no. 27,
[158] Bean, H.D., Hill, J.E., Dimandja, J.‐M.D., (2015) J. Chromatogr. A., 1394, pp. 111‐
117.
[159] Kim, S., Zhang, X., (2013) Computational and Mathematical Methods in
Medicine, 2013, art. no. 509761,
59
[160] Deng, B., Kim, S., Li, H., Heath, E., Zhang, X., (2016) Journal of Bioinformatics and
Computational Biology, 14 (6), art. no. 1650032,
[161] Mommers J, Knooren J, Mengerink Y, Wilbers A, Vreuls R, van der Wal S., J.
Chromatogr. A., 2011;1218, 3159‐3165
[162] Wright A.G., Fell A.F., Berridge J.C., Chromatographia, 24 (1987), 533‐590
[163] Berridge J.C., J. Chromatogr. A., 244 (1982), 1‐14
[164] Schoenmakers, P.J., J. High Res. Chrom. 11 (1988), 278‐282
60
61
3 2B2BPrediction of Comprehensive Two‐dimensional Gas
Chromatography Selectivity and Ranking of Column‐sets;
Application to Bio‐Oil Samples
In this chapter, a procedure is described for the global prediction of the most‐suitable
GC×GC column‐sets for the analysis of a particular complex sample. The GC×GC
retention prediction is based on the adapted Abrahams solvation parameter model as
published by Seeley et al. In our approach, we extended this model by incorporating
estimation of individual solute‐elution temperatures in order to re‐calculate the
temperature‐dependent column descriptors for each individual solute. A global
ranking of column‐sets based on three orthogonality parameters, is presented. The
new orthogonality parameters %FIT and %BIN5×5 are in agreement with the
orthogonality scores of 2D‐chromatograms of experts.
3.1 26B26BIntroduction
GC×GC is becoming more and more a routine analytical technique for solving all kinds
of analytical questions in a wide variety of application fields, including the chemical
characterization of pyrolysis bio‐oils [1‐7].
As described in chapter 2, GC×GC method development is not straightforward and the
most important and difficult task is the proper choice of the primary and secondary
stationary phase chemistries, in order to obtain optimal separation for a particular
complex sample [8].
Nowadays, more than 50 GC columns with different types of stationary phases are
commercially available, resulting in many possible GC×GC column‐sets [9‐10].
Theoretically, the number of column combinations for 50 different column types is
2450 (502 – 50), including the reversed combinations. For method development, from
a practical point of view, it is highly desirable to test only a limited set of column‐sets.
In table 3, chapter 2, six GC columns (stationary phases equivalent to DB‐1, DB‐17, DB‐
62
210, DB‐225, DB‐23, DB‐WAX) with different selectivities were recommended as a
good choice for starting method development. From these six columns, 30 column
combinations could be made, if the order of the columns is allowed to vary. In chapter
2 it was advised to narrow down the possible combinations based on the expected
molecular interactions between the analyte and the stationary phase and by
considering constraints, such as the maximum column temperature. The choice could
be facilitated by theoretically predicting the GC×GC retention for a particular set of
solutes (analytes and matrix compounds), which should be representative for the
entire sample. However, for truly complex samples, often containing hundreds or even
thousands of compounds, and especially for samples containing many unknown
solutes, the choice of a limited set of representative solutes may be an impossible task.
Several different models for predicting GC×GC retention have been developed, most
of which are based on single‐column retention data [11‐14]. Abraham et al. [15‐17],
developed a model for the prediction of retention factors for single‐column isothermal
separations. This model makes use of a linear combination of five solute descriptors
(L, S, A, B and E) and five stationary‐phase descriptors (l, s, a, b, e).
(1)
The five terms are (loosely) related to physical properties, i.e. solute size (lL), solute
dipolarity/polarizability (sS), hydrogen bond acidity (aA), hydrogen bond basicity (bB)
and excess polarizability (eE). Seeley et al. [18‐20], used the solvation‐parameter
model for predicting GC×GC retention by converting the predicted retention factors
first to indices and then to primary and secondary retention data for constructing
GC×GC retention diagrams.
In our approach the prediction of GC×GC retention is based on the adapted solvation‐
parameter model as published by Seeley [19]. We adapted this model as described in
section 3.2.3.
63
3.2 27B27BExperimental
3.2.1 62B62BEquipment and materials
All GC×GC analyses were performed using a Leco (St. Joseph, MI, USA) Pegasus 4D
GC×GC system, equipped with a secondary GC oven, a Combipal (CTC Analytics,
Zwingen, Switzerland) auto sampler, a hot split/splitless injector and a Leco Pegasus
time‐of‐flight mass spectrometer (TOFMS). Instrument control and data processing
were performed by Leco ChromaTOF software version 3.25 and NIST (NIST,
Gaithersburg, Maryland, USA) MS‐Search 2.0. All used GC×GC column‐sets are
summarized in table 1.
Table 1 GC×GC column‐sets. Dimensions are given as L(m) × i.d. (mm); film thickness.
Column
set nr.
Primary column
Primary column
dimensions
Secondary column
Secondary column
dimensions
1 VF1 30x0.25; 1.0 µm VF17ms 2x0.1; 0.2 µm
2 ZB‐FFAP 30x0.25; 0.5 µm VF17ms 2x0.1; 0.2 µm
3 VF1 30x0.25; 1.0 µm BPX90 2x0.1; 0.1 µm
4 ZB‐FFAP 30x0.25; 0.5 µm DB‐1 2x0.1; 0.2 µm
5 RTX‐200 30x0.25; 1.0 µm MG‐WAX‐HT 2x0.1; 0.1 µm
6 VF1 30x0.25; 1.0 µm DB‐1 2x0.1; 0.2 µm
7 RTX‐200 30x0.25; 1.0 µm DB‐1 2x0.1; 0.2 µm
8 DB‐1701 30x0.25; 1.0 µm DB‐1 2x0.1; 0.2 µm
9 RTX‐225 30x0.25; 0.5 µm VF17ms 2x0.1; 0.2 µm
The VF1, DB‐1, DB‐1701 and VF‐17ms columns were purchased from Agilent (Santa
Clara, CA, USA), the ZB‐FFAP column from Phenomenex (Torrance, CA, USA), and the
Rtx‐200, Rtx‐225 columns from Restek (Bellefonte, PA, USA). The BPX90 column was
obtained from SGE (Ringwood, Victoria, Australia) and the MG‐WAX‐HT column from
Mega (Milano, Italy).
A bio‐oil sample was kindly provided by the Bioboost consortium [21]. This bio‐oil
sample was prepared by flash pyrolysis of biomass material. For GC×GC analysis 100
64
mg of sample were dissolved in 1 mL acetone resulting in a clear, brown‐colored
solution.
3.2.2 63B63BAnalytical method
For all GC×GC analyses, the GC oven was held for 1 min at 40C and then programmed
at a rate of 3Cmin‐1 to the maximum temperature of the installed columns and held
there for 5 min. For each column combination, a modulation time was chosen at which
the degree of wrap‐around was “acceptable”. All separations were carried out using a
constant helium flow of 1.2 mL/min. The injector temperature was 250C, using a split
ratio of 1:30 and an injection volume of 1 µL. The TOFMS was operated in electron‐
ionization mode at 70 eV, a source temperature of 250C and a mass range of 15 to
550 amu.
3.2.3 64B64BGC×GC prediction model
The prediction of GC×GC retention is based on an adapted version of Abraham’s
solvation‐parameter model [15‐17], as published by Seeley et al. [19]. In our work, we
extend this model by incorporating the estimation of individual solute‐elution
temperatures (Te) to re‐calculate the temperature‐dependent column descriptors (l,
s, a, e) for each individual solute.
The column descriptors are assumed to depend logarithmically on temperature (e.g.
descriptor = d0 * ln(T) + d1) [9‐10]. The descriptors are first calculated at one
temperature (T), for example the final oven temperature for a given analysis. These
column descriptors are then used to compute the individual solute indices (i). After
determining the relationship between indices and elution temperatures of a series n‐
alkanes, the elution temperatures of the individual solutes can be estimated from
their computed descriptors. For each individual solute, the estimated elution
temperature is then used to re‐calculate the corresponding stationary‐phase
descriptors at a more relevant column temperature. The column descriptors for the
secondary dimension are calculated at the primary‐dimension elution temperatures
(2T=Te) of the solute. The primary column is temperature programmed, so that a solute
65
will interact with the stationary phase from the initial oven temperature up to its
elution temperature. The column descriptors for the primary dimension are calculated
at 0.9 times the solute elution temperature (1T=0.9*Te) given the fact that solutes will
mainly interact with the stationary phase toward the end of their elution trajectories.
The correlation between predicted primary indices and experimental solute elution
temperatures was established for the set of nine different column‐sets, featuring five
different stationary phases in the primary columns. The correlation between the
predicted primary indices of 31 selected components of bio‐oil (calculated at T=300C)
and their corresponding experimental elution temperatures was calculated. The linear
regression u‐, v‐ and r2 values (1Te = u * 1Index + v), are given in table 2.
Table 2 Correlation between predicted retention indices and their experimental
primary elution temperatures (1Te = u * 1Index + v) for a set of 9 different column‐sets
set nr. 1 2 3 4 5 6 7 8 9
1Column VF‐1 ZB‐FFAP VF‐1 ZB‐FFAP RTX‐200 VF‐1 RTX‐200 DB‐
1701
RTX‐225
2Column VF17ms VF‐
17ms
BPX90 DB‐1 MGWAX DB‐1 DB‐1 DB‐1 DB‐17
u 15.4 11.6 15.3 11.5 10.2 14.6 10.6 14.1 10.7
v ‐20.9 ‐71.0 ‐34.1 ‐84.4 ‐30.4 ‐29.2 ‐36.7 ‐38.6 ‐39.5
r2 0.92 0.82 0.92 0.90 0.82 0.89 0.80 0.91 0.91
As described by Seeley [18], a GC×GC retention diagram is defined as a plot of the
calculated primary “Seeley‐retention index” (1i, x‐axis) versus 1.6i (y‐axis), where i
is the difference between the calculated secondary index (2i) and the calculated
primary index (i=2i‐1i). In this paper, we scale the range of x‐axis and y‐axis values to
reasonable peak capacities 1n (200) and 2n (50), respectively. This choice of peak
capacities is based on experimental settings using which the bio‐oil sample was
analyzed by the nine different column‐sets (see table 3). The experimental peak
capacities, 1n and 2n, are given in table 3. It should be noted that these values are
obtained using standard (non‐optimized) GC×GC settings. The modulation time was
10 s for column‐sets 3 and 5 and 6 s for all other column‐sets. To roughly estimate the
66
peak capacities of the primary and secondary columns average peak widths of 18 and
0.1 s were used, respectively.
Table 3 Experimental peak capacities (1n and 2n) determined for nine different column‐
sets, using standard, non‐optimized GC×GC settings
set nr. 1 2 3 4 5 6 7 8 9
1 Col VF‐1 ZB‐FFAP VF‐1 ZB‐FFAP RTX‐200 VF‐1 RTX‐200 DB‐1701 RTX‐225
2 Col VF17ms VF‐17ms BPX90 DB‐1 MGWAX DB‐1 DB‐1 DB‐1 DB‐17
1n 164 204 174 203 163 173 163 182 168
2n 20 39 64* 44 81** 9 44 10 31
* Solutes 4‐hydroxy‐3,5‐dimethoxy‐benzaldehyde and hydroquinone were not included due to extensive wrap‐around
** Solutes 2‐naphthalenol and hydroquinone were not included due to extensive wrap‐around
Scaling is desired to compare and rank predicted retention diagrams based on
selectivity. Scaling of the x‐ and y‐axis is performed as given in equation 1 and 2,
respectively.
xnormalized = ( x ‐ xminimum ) * ( 1n / ( xmaxium ‐ xminimum ) ) (1)
ynormalized = ( y ‐ yminimum ) * ( 2n / ( ymaximum ‐ yminimum ) ) (2)
In order to perform a visual comparison of predicted versus experimental data, the
experimentally determined retention times, 1tR and 2tR, are normalized using the same
equations as used for x (x=1tR) and y (y=2tR), respectively.
Solute descriptors (A, B, L, S, E) are computed using ACD/Absolve software (Advanced
Chemistry Development, Toronto, Canada). The solutes and their computed
descriptors are given in supplementary data 1. It has to be noted that the computed
descriptors are estimates that may deviate from experimentally determined
descriptors. For prediction and ranking calculations 19 different GC columns were
selected, including the six columns summarized in table 3, chapter 2. The selected
columns and their corresponding temperature‐dependent column descriptors [9, 10]
are listed in supplementary data 2.
67
3.2.4 65B65BColumn‐set ranking parameters
For ranking column‐sets based on the predicted retention diagrams, a robust measure
for the degree of orthogonality is needed, since the orthogonalities of hundreds of
predicted diagrams (in our case 361 diagrams) need to be assessed automatically.
The first choice was to use only the parameter %Ao (based on the Asterisk equations)
as published by Camenzuli and Schoenmakers [27]. Compared to most other existing
orthogonality parameters [28], the %Ao shows good agreement between the
calculated orthogonality values and the visual, subjective assessment of the observed
two‐dimensional chromatograms for a wide variety of chromatogram “types”. Also,
%Ao is quite intuitive and easy to calculate. No settings (e.g. the number of bins) need
to be selected and the values do not strongly depend on the total number of peaks.
However, due to the fact that this parameter proved to be “insensitive” to (strong)
uneven clustering of peaks in the x‐ or y‐direction, and because such chromatograms
are predicted by our model (e.g. a diagram with most peaks at the bottom and only a
few at the top), it was decided to develop additional orthogonality parameters. These
are designed to be more sensitive to clustering in the x‐ or y‐direction and more
sensitive for local peak densities. For the ranking the following orthogonality
parameters are used (1) %Ao, (2) %FIT and (3) %BIN.
Ad 1. The %Ao is based on a set of “asterisk” equations [27]. In this approach, the
primary and secondary retention times are normalized to 1. The normalized
separation space is crossed (not divided as in case of the bin approach), through the
middle, by four lines, one horizontal, one vertical and two diagonals. The position of
the peaks around these lines is computed; in case of full orthogonality the peak
spreading around all lines is maximized and the %Ao will be close to 100%.
Ad 2. For the %FIT, two polynomials of degree 2 are calculated that provide least‐
squares fits of the scaled, according to equations (1) and (2), xy and yx data. For each
data point in the xy and the yx data space, the minimal distance to the corresponding
fitted curve is computed. From data, obtained in the xy and yx data space, the average
68
distances and standard deviations of all peaks below the curve (xy1_avg, xy1_SD,
yx1_avg, yx1_SD) and for all peaks above the curve (xy2_avg, xy2_SD, yx2_avg, yx2_SD) are
calculated.
In case of full orthogonality, the average distance of the peaks below and above the
fitted curve will be close to 2n/4, for the xy data, or 1n/4, for the yx data, and the
corresponding standard deviation will be approximately 2n/7, for the xy data, or 1n/7
for the yx data. The factor 7 is based on the second moment of a homogeneous block
function of duration C which equals C*12‐0.5 [30]. For example, normalized xy and yx
diagrams of data set 9 (100 peaks) are shown in figure 1. The fitted curves are shown
as the red lines. In case of full orthogonality, the absolute average distance of the
peaks to the fitted curve, in the xy‐diagram, is approximately 12.5 (=2n/4 = 50/4) and
the standard deviation approaches 7.1 (2n/(2*120.5)).
69
Figure 1 Diagrams of scaled xy (top) and yx (bottom) data of data‐set 9. The peaks are
shown as blue circles, the fitted curves are shown as the red lines and the bins are
shown as green rectangles. The x‐date is normalized to 1n=200 and y‐data is
normalized to 2n=50
2tr
scal
ed
0 5 10 15 20 25 30 35 40 45 50
2tr scaled
0
20
40
60
80
100
120
140
160
180
200
70
The %FIT orthogonality parameter is calculated as follows:
xyAVG = (1 ‐ 1‐(xy1AVG * 4 / 2n)) + (1 ‐ 1‐(xy2AVG * 4 / 2n)) / 2 (1)
xySD = (1 ‐ 1‐(xy1SD * 7 / 2n)) + (1 ‐ 1‐(xy2SD * 7 / 2n)) / 2 (2)
yxAVG = (1 ‐ 1‐(yx1AVG * 4 / 1n)) + (1 ‐ 1‐(yx2AVG * 4 / 1n)) / 2 (3)
yxSD = (1 ‐ 1‐(yx1SD * 7 / 1n)) + (1 ‐ 1‐(yx2SD * 7 / 1n)) / 2 (4)
%FIT = ( xyAVG + xySD + yxAVG + yxSD ) / 4 (5)
Note that equations 1 through 4 are normalized to 1, so that %FIT is insensitive to peak
capacities.
Ad 3. For calculating the %BIN, the normalized x‐axis is divided into 1n/m steps and
the y‐axis is divided into 2n/m steps, so the 2D‐space is divided into 1n/m × 2n/m
rectangles or bins, for example in figure 1 (left), for m=10, the bins (200/10 × 50/10 =
100 bins) are shown as green rectangles. If the number of peaks is known, the average
number of peaks per bin can be calculated.
AVGp/b = (total number of peaks) / (total number of bins) (6)
In case of ideal peak spreading the actual number of peaks in each bin is close to the
AVGp/b value. To calculate %BIN, the sum of the absolute deviation of the actual
number of peaks observed in each bin from the average number of peaks per bin
(dev) is calculated. The sum of the absolute deviation in case of full peak spreading
(devfs) and in case of no peak spreading (devns; one bin contains one peak, one bin
contains all but one peaks and all the others are empty) are also calculated. The %BIN
is then calculated from
71
%BIN = 100(1 ‐ ((dev – devfs) / (devns ‐ devfs)) (7)
%BIN can only be used if the total number of peaks is larger than the number of bins.
In our “bio‐oil” case a set of only 31 solutes is used for prediction and ranking of
column‐sets. Therefore, %BIN was also calculated and evaluated for 5×5 bins.
All calculations of prediction, orthogonality and ranking were performed using Matlab,
version R2017b, (MathWorks, Natick, MA, USA).
3.3 28B28BResults and discussion
3.3.1 66B66BTesting orthogonality parameters
For testing and comparing the three orthogonality parameters, 14 different types of
data sets (see table 4), containing 1000, 100, or 50 data points (peaks), were
constructed. Within each type ten data sets were generated to calculate the average
orthogonality values and the corresponding coefficients of variation (CV%). Data set 1
can be considered as fully orthogonal given the fact that the x‐ and y‐values of all data
points were generated randomly. All data‐set types are shown as xy plots in
supplementary data 3. The results of the three orthogonality parameters %FIT, %BIN
(%BIN20×5, %BIN8x2 and %BIN5×5) and %Ao, calculated for the 14 different types of data
sets with 1000 peaks, are given in table 4. All results for all types of data sets, for
various numbers of peaks, are given in supplementary data 4.
72
Table 4 Calculated orthogonality values (average of 10 replicates) for 14 different
types of data sets (see supplementary data 3) using 1000 peaks. The three highest
values are written in bold font and the highest is underlined.
1000 peaks %BIN20×5 %BIN8x2 %BIN5×5 %Ao %FIT
AVG CV% AVG CV% AVG CV% AVG CV% AVG CV%
1‐ORTHO 87 1,3 95 1,3 93 1,8 96 1,1 97 1,3
2‐DENSITY GRADIENT 76 0,5 79 0,6 78 0,6 91 0,6 89 1,1
3‐SLOPE 54 1,1 64 1,9 56 1,7 61 2,5 51 4,4
4‐CORRELATION 32 1,9 52 1,6 37 2,6 30 1,6 11 1,4
5‐OUTLIERS 22 5,0 48 0,0 19 5,8 46 0,7 56 1,7
6‐LOCAL DENSITY 68 2.5 96 0.7 80 1.7 89 0.9 89 0,8
7‐CLUSTERING x (5:5) 40 0,5 47 0,0 38 0,5 57 1,4 71 1,7
8‐CLUSTERING x (9:1) 30 1,0 31 0,0 27 0,9 88 0,7 58 3,3
9‐U‐SHAPE 51 0,7 72 1,3 50 0,4 73 3,2 83 1,3
10‐L‐SHAPE 36 0,7 59 1,1 34 0,8 66 1,4 61 2,3
11‐X‐SHAPE 36 1,0 61 2,1 34 1,1 54 3,4 79 0,6
12‐CLUSTERING y (5:5) 40 0,4 95 1,2 38 0,5 57 1,3 68 1,9
13‐CLUSTERING y (7:3) 40 0,8 79 0,0 38 0,6 67 1,8 67 3,2
14‐CLUSTERING y (9:1) 30 0,8 58 0,0 28 1,1 88 0,7 56 4,2
Based on table 4, it can be concluded that %FIT, %BIN and %Ao all show the highest
orthogonality value for data type 1 (fully orthogonal). %BIN using 5×5 and 20×5 bins
give very similar results and are most sensitive for peak‐density differences, as is
observed by the lower value for set 6, while 8x2 is not discriminative. %Ao is insensitive
for uneven x‐ or y‐clustering as is seen from the high values for sets 8 and 14. The
values increase from even (5:5) to uneven (9:1) peak clustering in the x‐ or y‐direction.
The %FIT is the most sensitive for strong correlations between x and y, as is evident
from the lowest value for set 4. In table 5, the calculated orthogonality values for the
14 different types of data sets, using 50 peaks, are summarized.
73
Table 5 Calculated orthogonality values (average of 10 replicates) for 14 different
types of data sets (see supplementary 3) using 50 peaks. The three highest values are
written in bold font and the highest is underlined.
50 peaks %BIN8x2 %BIN5×5 %Ao %FIT
AVG CV% AVG CV% AVG CV% AVG CV%
1‐ORTHO 82 6 74 5 84 7 88 5
2‐DENSITY GRADIENT 71 6 65 5 87 9 76 9
3‐SLOPE 66 5 50 7 63 9 58 14
4‐CORRELATION 52 6 34 12 27 6 11 10
5‐OUTLIERS 57 6 21 5 52 4 68 8
6‐LOCAL DENSITY 88 5 64 8 81 4 88 5
7‐CLUSTERING x (5:5) 47 5 38 0 49 4 71 7
8‐CLUSTERING x (9:1) 31 4 28 9 85 3 57 10
9‐U‐SHAPE 66 7 47 5 63 6 80 8
10‐L‐SHAPE 55 6 35 7 59 10 58 11
X‐SHAPE 60 6 42 11 63 9 76 4
11‐CLUSTERING y (5:5) 82 3 37 3 49 9 67 8
12‐CLUSTERING y (7:3) 73 5 36 5 60 8 58 18
13‐CLUSTERING y (9:1) 57 5 30 12 85 3 53 14
%BIN 20×5 could not be calculated since the number of peaks is smaller than the number of bins
Based on tables 4 and 5, it can be concluded that for set 1 (orthogonal) a decrease in
peak numbers from 1000 to 50 shows a decrease of 20, 12 and 9% in the orthogonality
values using %BIN, %Ao and for %FIT, respectively. For set 1 (orthogonal) the influence
of the number of peaks on the orthogonality values is the largest of all types of data
sets. Also for lower numbers of data points the %BIN8x2 parameter is less
discriminative towards clustering. It was decided to use only the %BIN5×5 parameter
for the evaluation of the 31‐solute ‘bio‐oil’ case.
The orthogonality values for the parameters %BIN5×5, %FIT and %Ao were also
calculated for two different data sets that were used and published by Schure et al.
[28]. They performed a comparison of 20 orthogonality parameters by evaluating 47
2D‐chromatograms. The authors, concluded that the specific orthogonality
parameters were related and that products of certain orthogonality parameters were
able to identify the chromatograms that were thought best by a panel of nine experts
74
in 2D chromatography. For the calculations the data were normalized to the maximum
value of 1tr (s) for x‐ and the maximum value of 2tR (s) for the y‐axis. The calculated
values are compared with the average “orthogonality” score (scale from A+ down to
F, or from 98 down to 55) of the expert panel. For the first data set, containing 21
chromatograms, each with 196 peaks, our newly developed parameters %BIN5×5 and
%FIT, as well as the product of both show excellent agreement with the experts’
orthogonality scores for best and worst chromatograms. A high correlation is observed
between %FIT and %BIN5×5 and the experts’ orthogonality scores, with r‐squared
values of 0.95 and 0.94, respectively. The results are show in table 6 and the
chromatograms ‘D14’, ‘D12’ and ‘D26R’ are given supplementary data 5A.
Table 6 Orthogonality values %BIN5×5, %Ao, %FIT and product “%BIN5×5 × %FIT” for a
data set of 21 2D‐chromatograms [29]. The 4 best chromatograms for each parameter
and for the reviewer’s scores are marked in yellow and the 4 worst are marked in grey.
Chromatogram %BIN 5×5 %Ao %FIT PRODUCT EXPERTSAVG
'D14' 60 50 71 42 90,1
'D35' 56 53 65 36 89,2
'D24' 59 53 71 42 88,1
'D47' 56 46 66 37 88,1
'D36R' 54 58 60 33 87,2
'D34' 54 52 55 30 84,9
'D16R' 51 55 50 26 84,4
'D67R' 50 51 58 29 84,2
'D46R' 52 44 47 24 84,0
'D23' 51 55 40 20 83,3
'D13' 52 53 40 21 82,8
'D26R' 49 58 50 24 82,6
'D15' 53 47 39 21 82,1
'D57' 53 44 51 27 81,8
'D25' 51 49 37 19 79,9
'D56R' 49 46 45 22 79,1
'D45' 50 38 35 17 76,5
'D37' 43 45 24 10 75,7
'D27' 38 36 15 6 67,9
'D17' 37 32 14 5 67,1
'D12' 27 20 6 2 58,0 # scale from 55 (lowest orthogonality) to 98 (full orthogonal)
For the other data set, containing 26 chromatograms and numbers of peaks ranging
from 13 to 192, the correlation is not as good. However three of the experts’ top 5 are
75
also in the top 5 of values obtained for %FIT, %BIN5×5, %Ao and the product of %FIT ×
%BIN5×5. Also, four of the experts’ bottom 5 are also in the bottom 5 of the %FIT and
%Ao scores. The results are given in supplementary data 5B. Remarkable is the high
%Ao value (in top 5 for %Ao and in experts’ bottom 5) for chromatogram ‘YExp13’, due
to clustering in x and y. For %FIT much poorer correlations are observed for
chromatograms containing fewer than 40 peaks.
76
3.3.2 67B67BSelection of representative sample solutes
First the bio‐oil sample was analyzed using column‐set 1 (VF1 × VF17ms). The 2D‐
chromatogram, including the identification of 31 bio‐oil solutes, is given in figure 2.
These 31 identified solutes are used for the prediction and column‐set‐ranking
calculations. They are considered to be relevant and representative for the whole bio‐
oil sample based on their chemical diversity. The solutes and their calculated solute
descriptors are given in supplementary data 1.
Figure 2 GC×GC chromatogram of the bio‐oil sample analyzed using column‐set 1 (VF1
× VF17ms) under standard GC×GC settings; peak identifications are given in
supplementary data 1
3.3.3 68B68BPrediction and ranking results
For the 31 representative bio‐oil solutes and for 361 different column‐sets (19
different stationary phases), the retention diagrams have been computed. For each
retention diagram the 3 different ranking parameters, %FIT, %BIN5×5, and Ao%, are
77
calculated. The ranking parameters for each column‐set (shown as column‐set
numbers) are plotted as %FIT versus %Ao and as %BIN5×5 versus %FIT in figure 3 and 4,
respectively. For each ranking parameter, the top 20 column‐sets are summarized in
table 6.
Table 6 The predicted top‐20 column‐sets for the different ranking parameters
%BIN5×5, %Ao and %FIT, calculated for 31 solutes representative of bio‐oil. The column‐
sets in bold font are present in all three parameters’ top‐20 lists
%BIN5×5 %Ao %FIT
305 ‐ Rtx‐440 × DB‐5 74 150 ‐ DB‐624 × Rtx‐440 92 7 ‐ DB‐5 × DB‐1301 80
40 ‐ DB‐35 × DB‐1 71 134 ‐ DB‐624 × DB‐5 92 8 ‐ DB‐5 × DB‐624 77
7 ‐ DB‐5 × DB‐1301 70 148 ‐ DB‐624 × DB‐XLB 92 273 ‐ DB‐XLB × DB‐1301 77
306 ‐ Rtx‐440 × DB‐1 70 66 ‐ DB‐17ms × DB‐1701 91 17 ‐ DB‐5 × Rtx‐440 77
15 ‐ DB‐5 × DB‐XLB 69 57 ‐ DB‐35 × Cyclosil B 91 283 ‐ DB‐XLB × Rtx‐440 77
58 ‐ DB‐17ms × DB‐5 68 131 ‐ DB‐1301 × Rtx‐440 91 274 ‐ DB‐XLB × DB‐624 76
286 ‐ Rtx‐50 × DB‐5 68 129 ‐ DB‐1301 × DB‐XLB 90 134 ‐ DB‐624 × DB‐5 76
66 ‐ DB‐17ms × DB‐1701 67 115 ‐ DB‐1301 × DB‐5 90 115 ‐ DB‐1301 × DB‐5 76
115 ‐ DB‐1301 × DB‐5 67 7 ‐ DB‐5 × DB‐1301 90 150 ‐ DB‐624 × Rtx‐440 75
59 ‐ DB‐17ms × DB‐1 67 346 ‐ Cyclosil B × DB‐17ms 88 148 ‐ DB‐624 × DB‐XLB 75
116 ‐ DB‐1301 × DB‐1 67 273 ‐ DB‐XLB × DB‐1301 88 66 ‐ DB‐17ms × DB‐1701 74
135 ‐ DB‐624 × DB‐1 67 156 ‐ DB‐1701 × DB‐17ms 88 311 ‐ Rtx‐440 × DB‐1301 74
60 ‐ DB‐17ms × DB‐35 66 155 ‐ DB‐1701 × DB‐35 87 155 ‐ DB‐1701 × DB‐35 74
311 ‐ Rtx‐440 × DB‐1301 66 76 ‐ DB‐17ms × Cyclosil B 87 129 ‐ DB‐1301 × DB‐XLB 74
8 ‐ DB‐5 × DB‐624 66 19 ‐ DB‐5 × Cyclosil B 87 131 ‐ DB‐1301 × Rtx‐440 74
27 ‐ DB‐1 × DB‐624 66 345 ‐ Cyclosil B × DB‐35 86 47 ‐ DB‐35 × DB‐1701 74
39 ‐ DB‐35 × DB‐5 66 26 ‐ DB‐1 × DB‐1301 85 116 ‐ DB‐1301 × DB‐1 73
45 ‐ DB‐35 × DB‐1301 66 17 ‐ DB‐5 × Rtx‐440 85 343 ‐ Cyclosil B × DB‐5 73
150 ‐ DB‐624 × Rtx‐440 66 305 ‐ Rtx‐440 × DB‐5 85 26 ‐ DB‐1 × DB‐1301 73
344 ‐ Cyclosil B × DB‐1 66 319 ‐ Rtx‐440 × DB‐XLB 85 135 ‐ DB‐624 × DB‐1 73
Of which 5 in %Ao
Of which 8 in %FIT
Of which 5 in %BIN
Of which 12 in %FIT
Of which 12 in %Ao
Of which 8 in %BIN
78
Figure 3 Ranking parameters %Ao versus %FIT for the bio‐oil sample of Figure 2 for all
computed column‐sets
Figure 4 Ranking parameters %BIN5×5 versus %FIT for the bio‐oil sample of Figure 2 for
all computed column‐sets
Based on table 6, it can be concluded that all of the “top‐20” column‐sets feature only
non‐polar to medium‐polar columns. Within the top‐10, the low/mid polarity DB‐624,
DB1701 and DB‐1301 stationary phases contain cyanopropyl‐phenyl
dimethylpolysiloxane stationary phases. The mid polarity DB‐35 contains 35% phenyl
65% dimethyl arylene siloxane and the Rtx‐50 contains 50% phenyl, 50%
%FIT
%FIT
79
dimethylpolysiloxane. The medium‐polar stationary phase of the Cyclosil‐B column
consists of 30% heptakis‐B‐cyclodextrin in a DB‐1701 stationary phase. The stationary
phases of both the non‐polar DB‐1 and DB‐5 columns are 100% dimethyl polysiloxane
and 95% dimethyl / 5% diphenyl polysiloxane, respectively. The low polarity DB‐XLB
and mid polarity Rtx‐440 columns contain proprietary stationary phases.
Column‐sets containing a high and a low or medium polarity column show lower
orthogonality values. The retention diagram of DB1 × DB‐FFAP is shown in figure 5.
The main reason for the lower orthogonality values is that the solute set contains
several highly polar compounds (e.g. 2‐naphthalenol). In comparison with the other
solutes, these show high retention on (highly) polar stationary phases, thereby
lowering the overall orthogonality. Since 2‐naphthalenol represents the group of
hydroxylated polycyclic aromatic hydrocarbons that is known to be present in real bio‐
oils, this solute cannot simply be discarded or ignored.
Figure 5 Retention diagram of the 31 representative solutes from the bio‐oil sample of
Figure2 analyzed with column‐set 33 “DB1 × DB‐FFAP”. 2‐naphthalenol elutes in the
upper right corner, thereby significantly lowering the overall orthogonality
The results obtained with the best column‐sets according to the ranking parameters
%FIT, %BIN5×5 and %Ao are shown in figures 6, 7 and 8, respectively.
y norm
alized
80
Figure 6 Retention diagram of the 31 representative solutes from the bio‐oil sample of
Figure2 analyzed on column‐set “DB‐5 × DB‐1301” (best column‐set based on %FIT).
Figure 7 Retention diagram of the 31 representative solutes from the bio‐oil sample of
Figure2 analyzed on column‐set “Rtx‐440 × DB‐5” (best column‐set based on %BIN5×5).
y norm
alized
y norm
alized
81
Figure 8 Retention diagram of the 31 representative solutes from the bio‐oil sample of
Figure2 analyzed on column‐set “DB‐624 × Rtx‐440” (best column‐set based on %Ao).
3.3.4 69B69BPredicted versus experimental data
3.3.4.1 103B103BColumn‐set ranking
In order to compare the predicted and experimental ranking, the bio‐oil sample was
analyzed using nine different column‐sets, under standard GC×GC conditions. For each
column‐set the primary and secondary retention times of the 31 selected solutes were
determined and used to calculate the ranking parameters. In table 7, the predicted
and experimental ranking parameters are summarized for each column‐set.
y norm
alized
82
Table 7 Predicted and experimental ranking parameters, including the sum and
product of these three values, based on 31 selected solutes and calculated for nine
different column‐sets.
RANKING BASED ON
PREDICTED DATA
%BIN5×5 %Ao %FIT SUM PRODUCT INDEX 1i‐2i
154 ‐ DB‐1701 × DB‐1 62 66 60 1.9 2.5 2
78 ‐ DB‐200 × DB‐1 64 68 59 1.9 2.6 2
31 ‐ DB‐1 × BPX90 2 25 68 54 1.5 0.9 ‐16
175 ‐ DB‐225 × DB‐17ms 55 54 51 1.6 1.5 4
89 ‐ DB‐200 × DB‐WAXetr 2 31 53 51 1.3 0.8 ‐6
249 ‐ DB‐FFAP × DB‐1 47 50 42 1.4 1.0 9
23 ‐ DB‐1 × DB‐17ms 57 59 40 1.6 1.4 ‐2
251 ‐ DB‐FFAP × DB‐17ms 44 47 40 1.3 0.8 7
21 ‐ DB‐1 × DB‐11 51 54 38 1.4 1.0 ‐0.1
RANKING BASED ON
EXPERIMENTAL DATA
%BIN5×5 %Ao %FIT SUM PRODUCT
154 ‐ DB‐1701 × DB‐1 74 86 78 2.4 5.0
175 ‐ DB‐225 × DB‐17ms 63 86 77 2.3 4.2
78 ‐ DB‐200 × DB‐1 79 87 76 2.4 5.2
251 ‐ DB‐FFAP × DB‐17ms 71 65 64 2.0 3.0
249 ‐ DB‐FFAP × DB‐1 65 59 62 1.9 2.4
23 ‐ DB‐1 × DB‐17ms 64 63 45 1.7 1.8
21 ‐ DB‐1 × DB‐1 37 37 14 0.9 0.2
1) Very low secondary peak capacities (n9) leading to unreliable orthogonality values
2) RTX‐200 × MG‐WAX‐HT and VF1 × BPX90 are not shown in the Experimental listing, due to significant wrap‐around of polar
solutes
The parameters %BIN5×5 and %Ao, predicted two and %FIT predicted all three of the
experimental top‐3 column‐sets. %BIN5×5 also predicted low orthogonality for “RTX‐
200 × DB‐WAXetr” and “DB‐1 × BPX90”. Ao% shows a relatively high value for the DB1
× BPX90 column‐set (most peaks at the bottom of the diagram), which can be
explained by the fact that %Ao is insensitive for uneven clustering in x or y.
The sum and the product of the three parameters predict the entire experimental top‐
3. The observation that combining orthogonality parameters may lead to better
assessment of orthogonality was also made by Schure et al. [28].
In case of the VF1 DB1 column‐set the available separation space is inefficiently
used due to lack of retention on DB1. The ‘Seeley’ Index (1i – 2i), calculated by using
83
the predicted indices for the primary and secondary dimension, may be used as a
rough indication of the degree of the (required) secondary retention. A positive
‘Seeley’ Index indicates a “reversed phase” column‐set and very low absolute values
are observed for column‐sets having equal or similar stationary phases for example
DB‐1 × DB‐1 has an Index of ‐0.1 and the reversed phase column‐set DB1 × BPX90
has a value of ‐16 which indicates significant retention of polar solutes (on the polar
secondary column), leading to wrap‐ around.
The use of retention‐matched systems and conditions may lead to a more correct
comparison, possibly resulting in a better correlation between predicted and
experimental results.
3.3.4.2 104B104BPrimary and secondary retention data
The degree of correlation between the predicted (x, y values), calculated with and
without the temperature correction, and experimental (1tR, 2tR) primary and secondary
retention data is summarized in table 8. The r‐squared and standard‐error values are
used to characterize the extent of correlation.
Table 8 Correlation between predicted (x, y) and experimental (1tR, 2tR) primary and
secondary retention data. Correlation were based on predicted values with and
without temperature correction.
Temperature correction for re‐calculation of column
descriptors
No temperature
correction
column‐set 1tR
2tR
1tR 2tR
r2 Standard
error (PW)
r2 Standard
error (PW)
r2 r2
DB‐FFAP × DB‐17 0.93 13 0.67 9 0.90 0.69
DB‐1 × DB‐17 0.97 12 0.77 5 0.92 0.63
DB‐1 × BPX90 0.96 15 0.61 5 0.92 0.72
DB‐FFAP × DB‐1 0.93 13 0.75 8 0.92 0.75
DB‐200 × DB‐WAX 0.88 18 <0.5 9 0.85 <0.5
DB‐1 × DB‐1 0.89 17 0.6 9 0.89 #
DB‐200 × DB‐1 0.91 14 0.61 7 0.89 0.56
DB‐1701 × DB‐1 0.94 14 0.57 9 0.91 0.6
DB‐225 × DB‐17 0.92 14 <0.5 12 0.91 <0.5
# cannot be calculated since the primary and secondary indices are identical (y1 = 1,6^(Index2 – index1))
84
Based on these data it can be concluded that the correlation between the predicted
and experimental primary dimension data is reasonable, with an average correlation
coefficient of 0.9 and standard error of 14 (CV ca. 7%). However, the correlation
between the predicted and experimental secondary retention data is much poorer,
with an average r‐squared value of 0.6 and a standard error of 8 (CV ca. 16%).
When applying temperature correction to the primary‐column descriptors the
average r2 value for the primary columns of the nine different column‐sets increased
from 0.90 to 0.93 and the standard error decreased from 17 to 14 peak widths (PW).
As an example, the predicted retention diagram of 31 bio‐oil solutes using a DB‐FFAP
× DB‐1 column‐set is plotted against the experimentally obtained GC×GC retention
diagram using a ZB‐FFAP × DB‐1 column‐set in figure 8. Despite the fact that the
prediction accuracy (mainly for the secondary dimension separation) is very limited,
global elution information and/or trends can be observed. In figure 8b and, especially,
figure 8a the solutes eluting early from the first‐dimension column (left‐hand side of
the diagrams) show good secondary retention and peak spreading. The intermediate
solutes (mainly phenols) show less retention and poorer peak spreading in the second
dimension. The right upper corner of the separation space is completely empty.
85
Figure 8 A. Predicted retention diagram of 31 bio‐oil solutes using a DB‐FFAP × DB‐1
column‐set and B. measured GC×GC retention diagram for the analysis of these
analytes using a ZB‐FFAP × DB‐1 column‐set
Higher correlations between predicted and observed retention data were observed by
Seeley et al. [19], perhaps mainly because for each column combination the
temperature program was optimized, while for our separations of the bio‐oil sample
a standard temperature program was utilized. Seeley et al. established unique
temperature programs, containing three different ramps, for each tested column
combination. The temperature program was adjusted in such a way that each
homologous series appeared as an approximate horizontal band, with the objective of
maximizing the visibility of the correlation between predicted indices and observed
y norm
alized
y norm
alized
8A.
8B.
86
retention data. Perhaps more importantly, Seeley et al. use practically determined
solute descriptors, while in our example the solute descriptors were estimated using
the Absolv/ACD labs software (see Experimental). Comparison of the experimental
and calculated descriptor values for a few compounds [9, 10] indicates that the
descriptor ‘A’ may be 30% higher and ‘S’ 10% lower than the calculated values.
3.3.5 70B70BInfluence of choice of 1n and 2n on column‐set ranking
The column‐sets were ranked on the basis of orthogonality, which does not take into
account that the peak capacities of the second‐dimension separation are usually much
smaller than those of the first dimension separation. Because there appears to be no
correlation between the judgement of quality of the 2D‐chromatograms (Experts’
view) and the retention‐time ratios (see table 5B in supplementary data) there is as
yet no indication that the 1n/2n ratio plays an important role in column‐set ranking.
However, this inequality may have a major effect on the actual separation power of a
column‐set, i.e. the potential of separating “all” compounds in a mixture. For
estimating separability of a specific sample (e.g. by the NND method [26]) all relevant
components of the sample have to be taken into account. For a complex mixture
containing many compounds that are not known on beforehand this is not feasible.
Thus, we limited ourselves to an overall assessment of the suitability of the collection
of potential column‐sets. The initially obtained separation can be optimized (see
chapter 2) and the actual separability of the sample on the optimized system may be
compared (in hindsight) with that of other optimized systems.
3.4 29B29BConclusions
The extended Abraham model can be used to roughly predict the most appropriate
column‐sets for GC×GC. The predicted ranking of the best and worst column‐sets is in
reasonable agreement with experimental observations. However, retention‐matched
columns and conditions in GC×GC are essential for an optimized system. The
prediction model can be slightly improved by incorporating temperature‐corrected
column descriptors. The new orthogonality parameters %FIT and %BIN5×5 are in
87
agreement with the orthogonality scores for 2D‐chromatograms provided by nine
experts [28]. Therefore, these parameters can be used for ranking column‐sets.
3.5 30B30BAcknowledgments
The authors would like to thank the Bioboost consortium [21] for kindly providing the
bio‐oil sample.
3.6 31B31BReferences
[1] M. Moraes, M. Migliorini, F. Damasceno, F. Georges, S. Almeida, C. Zini, R. Jacques,
E. Caramão, J Anal Appl Pyrol, 98 (2012), 51‐64.
[2] B. Onorevoli, M. Machado, C. Dariva, E. Franceschi, L. Krause, R. Jacques, E.
Caramão, Ind Crop Prod, 52 (2014), 8‐16.
[3] N.Tessarolo, L. dos Santos, R. Silva, D. Azevedo, J Chromatogr A, 1279 (2013), 68‐
75.
[4] M. Djokic, T. Dijkmans, G. Yildiz, W. Prins, K. Van Geem, J Chromatogr A, 1257
(2012), 131‐140.
[5] R. Silva, A. Casilli, A. Sampaio, B. Ávila, M. Veloso, D. Azevedo, G. Romeiro, J Anal
Appl Pyrol, 106 (2014), 152‐159.
[6] T. Sfetsas, C. Michailof, A. Lappas, Q. Li, B. Kneale, J Chromatogr A, 1218 (2011),
21, 3317‐3325.
[7] J. Marsman, J. Wildschut, F. Mahfud, H. Heeres, J Chromatogr A, 1150 (2007), 1‐2,
21‐27.
[8] A. Mostafa, M. Edwards, T. Górecki, J Chromatogr A, 1255 (2012), 38‐55.
[9] S. N. Atapattu, C.F. Poole, J Chromatogr A, 1195 1‐2, (2008) 136‐145
[10] C. Poole, S. Poole, J Chromatogr A, 1184 (2008), 254‐280.
[11] X. Lu, H. Kong, H. Li, C. Ma, J. Tian, G. Xu, J Chromatogr A, 1086 (2005), 175‐184.
[12] J. Beens, R. Tijssen, J. Blomberg, J Chromatogr A, 822 (1998), 233‐251.
[13] R. Western, P. Marriott, J Chromatogr A, 1019 (2003), 3‐14
[14] R.J. Western, P.J. Marriott, J. Sep. Sci. 25 (2002), 832.
[15] M.H. Abraham, Chem Soc Rev, 22 (1993), 73.
[16] M. Abraham, C. Poole, S. Poole, J Chromatogr A, 842 (1999), 79‐114.
88
[17] M. Abraham, A. Ibrahim, A. Zissimos, J Chromatogr A, 1037 (2004), 29‐47.
[18] J. Seeley, S. Seeley, J Chromatogr A, 1172 (2007), 72‐83.
[19] J. Seeley, E. Libby, K. Hill Edwards, S. Seeley, J Chromatogr A, 1216 (2009), 1650‐
1657.
[20] J. Seeley, C. Bates, J. McCurry, S. Seeley, J Chromatogr A, 1226 (2012), 103‐109.
[21] Bioboost http://bioboost.eu/
[22] G. Semard, V. Peulon‐Agasse, A. Bruchet, J‐P. Bouillon, P. Cardinaël, J Chromatogr
A, 1217 (2010), 33, 5449‐5454.
[23] S. Rutan, J. Davis, P. Carr, J Chromatogr A, 1255 (2012), 267‐276.
[24] Z. Zeng, J. Li, H. Hugel, G. Xu, P. Marriott, TrAC‐Trend Anal Chem, 53 (2014), 150‐
166.
[25] Z‐D. Zeng, H. Hugel, P. Marriott, Anal Chem, 85 (2013), 6356‐6363.
[26] W. Nowik, M. Bonose, S. Héron, M. Nowik, A. Tchapla, Anal Chem, 85 (2013),
9459‐9468.
[27] M. Camenzuli, P.J. Schoenmakers, Anal Chim Acta, 838 (2014), 93–101
[28] M. R. Schure, J. M. Davis, J Chromatogr A, 1414 (2015), 60‐76.
[29] J.C. Sternberg, in J.C. Giddings, R.A. Keller (eds), Adv. in Chromatography Vol. 2,
Marcel Dekker (NY), (1966) 205‐270.
89
3.7 32B32BSupplementary Data
Supplementary data 1
Solute descriptors (A, B, L, S, E) computed using ACD/Absolv
Solute name
A B L S E
ethylbenzene 0 0.119 3.9243 0.6388 0.5790
propylbenzene 0 0.1221 4.4187 0.6431 0.5772
1,2,3‐trimethylbenzene 0 0.1173 4.3731 0.5193 0.6295
o‐xylene 0 0.1166 3.9015 0.5768 0.6051
styrene 0 0.1699 3.8645 0.696 0.7029
methyl‐styrene 0 0.1706 4.3362 0.6385 0.7273
indane 0 0.0961 4.6844 0.6682 0.8145
5‐methyl‐indane 0 0.0968 5.1560 0.6106 0.8389
cyclopentanon 0 0.3193 3.2290 0.7658 0.4244
indene 0 0.1594 4.6766 0.7669 0.9704
me‐indene 0 0.1785 5.0355 0.7144 0.9474
tri‐me‐indene 0 0.2283 5.7743 0.6505 0.9324
2‐cyclopenten‐1‐one 0 0.3826 3.2212 0.8645 0.5802
benzofuran 0 0.2086 4.4110 0.8935 1.0572
benzaldehyde 0 0.4037 4.2709 1.1832 0.8142
naphthalene 0 0.1734 5.3319 1.0158 1.2748
me‐naphthalene 0 0.1741 5.8036 0.9583 1.2992
2,6‐di‐me‐naphthalene 0 0.1748 6.2752 0.9007 1.3236
furfural 0 0.4389 3.3500 1.0608 0.5965
2‐methoxy‐phenol 0.2746 0.4713 4.5475 0.9647 0.7659
2‐methoxy‐4‐me‐phenol 0.2746 0.4719 5.0192 0.9072 0.7902
2‐methoxy‐4‐et‐phenol 0.2746 0.4750 5.5136 0.9116 0.7885
2,3‐dihydro‐1H‐inden‐1‐one 0 0.3999 5.3899 1.1601 1.0267
phenol 0.4990 0.3889 3.7766 0.9026 0.7840
4‐me‐phenol 0.4990 0.3896 4.2483 0.8450 0.8083
2,6‐di‐me‐phenol 0.3117 0.3902 4.7200 0.7874 0.8327
2‐me‐phenol 0.4990 0.3896 4.2483 0.8450 0.8083
3‐et‐phenol 0.4990 0.3927 4.7428 0.8494 0.8065
2,6‐di‐methoxy‐phenol 0.1624 0.5629 5.2647 1.0306 0.7684
1,2,4‐trimethoxybenzene 0 0.7512 6.0014 1.5856 0.7425
2‐naphthalenol 0.4990 0.4470 6.1504 1.2265 1.5024
90
Supplementary data 2
Column descriptors (e, s, a, l) for 19 selected GC stationary phases obtained from
literature [10, 11], transformed to ln‐functions (column descriptor = d0 * ln(T) + d1).
Descriptor:
e
s
a
l
Column
e0
e1
s0
s1
a0
a1
l1
l2
DB‐5 0.1089 ‐0.4872 ‐0.1242 0.8572 ‐0.1443 0.8661 ‐0.2485 1.6509
DB‐1 0.1013 ‐0.4835 ‐0.0688 0.5460 ‐0.2747 1.4846 ‐0.2794 1.8430
DB‐35 0.1379 ‐0.5889 ‐0.2788 1.9761 ‐0.2276 1.3699 ‐0.2387 1.6721
DB‐17ms 0.1995 ‐0.8290 ‐0.3906 2.6464 ‐0.2305 1.4307 ‐0.2275 1.6455
DB‐200 0.2662 ‐1.6092 ‐0.4060 2.9872 ‐0.2485 1.3417 ‐0.2357 1.5952
DB‐210 0.3340 ‐1.9927 ‐0.3157 2.8251 ‐0.1982 1.1154 ‐0.1619 1.2071
DB‐1301 0.1905 ‐0.9633 ‐0.2334 1.5865 ‐0.3206 1.9861 ‐0.2522 1.7469
DB‐624 0.1964 ‐0.9928 ‐0.2455 1.6571 ‐0.3651 2.1604 ‐0.2609 1.7720
DB‐1701 0.1918 ‐1.0142 ‐0.3200 2.2436 ‐0.4704 2.8730 ‐0.2842 1.8794
DB‐225 0.2303 ‐1.1177 ‐0.6870 4.4854 ‐0.6859 4.4242 ‐0.2794 1.7839
DB‐23 0.1545 ‐0.7698 ‐0.3662 3.2751 ‐0.7312 4.9671 ‐0.2216 1.5143
BPX90 0.1091 ‐0.4726 ‐0.1383 2.6999 ‐0.7581 5.4512 ‐0.1774 1.2470
DB‐WAXetr 0.0421 0.0276 ‐0.4640 3.6123 ‐1.1273 7.4005 ‐0.2145 1.5069
DB‐FFAP 0.0078 0.1894 ‐0.5651 4.1070 ‐1.2503 7.9858 ‐0.2455 1.6312
DB‐XLB 0.1853 ‐0.8769 ‐0.2198 1.4507 ‐0.2388 1.3701 ‐0.2689 1.8482
Rtx‐50 0.1613 ‐0.6873 ‐0.3437 2.4332 ‐0.2358 1.4577 ‐0.2315 1.6303
Rtx‐440 0.1618 ‐0.7548 ‐0.1872 1.3259 ‐0.2913 1.6727 ‐0.2519 1.7675
HP‐88 0.1189 ‐0.5437 ‐0.3803 3.6252 ‐0.7351 5.2442 ‐0.1994 1.3761
Cyclosil B 0.3303 ‐1.6326 ‐0.4051 2.5639 ‐1.1752 6.6820 ‐0.2708 1.8407
91
Supplementary data 3
Datasets for testing orthogonality parameters.
All data sets, from 1 to 14, are shown as xy plots (x‐axis = ‘x‐normalized’ and y‐axis =
’y‐normalized’). The x‐axes are normalized to 1n=200 and the y‐axes are normalized to
2n=50. The 2D‐space is divided in 20×5 rectangles or bins, bordered by yellow lines.
The red line in each plot indicates a quadratic curve fitted through the xy data points.
The following sets are shown: 1‐orthogonal, 2‐density gradient, 3‐slope, 4‐correlation,
5‐outliers, 6‐local peak density, 7‐clustering x (5:5), 8‐clustering x (9:1), 9‐U‐shape, 10‐
L‐shape, 11‐X‐shape, 12‐clustering y (5:5), 13‐clustering y (7:3) and 14‐clustering y
(9:1)
1 2
3 4
5 6
92
7 8
9 10
11 12
13 14
93
Supplementary data 4
Calculated orthogonality parameters for all 14 types of data set using 1000, 100 and
50 peaks. Results are given as average (AVG) and variation of coefficient (CV%) of 10
generated data sets. The colors indicate the highest values in blue and the lowest
values in red.
Set nr. Data set type %BIN 20×5 %BIN 8x2 %BIN 5×5 %Ao %FIT
1000 peaks AVG CV% AVG CV% AVG CV% AVG CV% AVG CV%
SET ‐ 1 ORTHO 87 1,3 95 1,3 93 1,8 96 1,1 97 1,3
SET ‐ 2 DENSITY GRADIENT 76 0,5 79 0,6 78 0,6 91 0,6 89 1,1
SET ‐ 3 SLOPE 54 1,1 64 1,9 56 1,7 61 2,5 51 4,4
SET ‐ 4 CORRELATION 32 1,9 52 1,6 37 2,6 30 1,6 11 1,4
SET ‐ 5 OUTLIERS 22 5,0 48 0,0 19 5,8 46 0,7 56 1,7
SET ‐ 6 LOCAL DENSITY 68 2,5 96 0,7 80 1,7 89 0,9 89 0,8
SET ‐ 7 CLUSTERING x (5:5) 40 0,5 47 0,0 38 0,5 57 1,4 71 1,7
SET ‐ 8 CLUSTERING x (9:1) 30 1,0 31 0,0 27 0,9 88 0,7 58 3,3
SET ‐ 9 U‐SHAPE 51 0,7 72 1,3 50 0,4 73 3,2 83 1,3
SET ‐ 10 L‐SHAPE 36 0,7 59 1,1 34 0,8 66 1,4 61 2,3
SET ‐ 11 X‐SHAPE 36 1,0 61 2,1 34 1,1 54 3,4 79 0,6
SET ‐ 12 CLUSTERING y (5:5) 40 0,4 95 1,2 38 0,5 57 1,3 68 1,9
SET ‐ 13 CLUSTERING y (7:3) 40 0,8 79 0,0 38 0,6 67 1,8 67 3,2
SET ‐ 14 CLUSTERING y (9:1) 30 0,8 58 0,0 28 1,1 88 0,7 56 4,2
100 peaks
SET ‐ 1 ORTHO 62 5,6 85 3,5 79 3,6 89 3,7 91 4,5
SET ‐ 2 DENSITY GRADIENT 59 5,7 77 3,9 73 3,5 90 1,3 84 4,7
SET ‐ 3 SLOPE 45 10,4 66 5,4 53 6,9 60 5,3 52 11,1
SET ‐ 4 CORRELATION 30 5,4 54 5,8 36 6,7 29 4,0 11 4,6
SET ‐ 5 OUTLIERS 24 3,0 58 1,9 21 3,4 74 4,8 62 4,8
SET ‐ 6 LOCAL DENSITY 54 7,1 91 3,5 71 5,4 86 2,8 88 2,5
SET ‐ 7 CLUSTERING x (5:5) 36 3,9 47 3,1 38 1,2 53 4,7 74 5,0
SET ‐ 8 CLUSTERING x (9:1) 27 7,9 32 0,8 28 9,3 87 2,1 60 6,7
SET ‐ 9 U‐SHAPE 44 5,3 71 3,0 50 3,3 68 5,0 84 5,5
SET ‐ 10 L‐SHAPE 34 5,8 56 5,0 34 3,8 65 5,2 62 5,9
SET ‐ 11 X‐SHAPE 37 7,7 63 3,8 39 8,5 59 6,5 77 1,5
SET ‐ 12 CLUSTERING y (5:5) 36 5,7 86 4,7 37 4,5 54 5,1 64 14,0
SET ‐ 13 CLUSTERING y (7:3) 34 8,1 78 2,4 37 3,6 64 3,8 64 7,8
SET ‐ 14 CLUSTERING y (9:1) 28 6,8 58 2,0 29 9,1 88 3,1 51 14,8
94
50 peaks
SET ‐ 1 ORTHO 79 7 82 6 74 5 84 7 88 5
SET ‐ 2 DENSITY GRADIENT 74 8 71 6 65 5 87 9 76 9
SET ‐ 3 SLOPE 63 6 66 5 50 7 63 9 58 14
SET ‐ 4 CORRELATION 48 11 52 6 34 12 27 6 11 10
SET ‐ 5 OUTLIERS 40 12 57 6 21 5 52 4 68 8
SET ‐ 6 LOCAL DENSITY 73 7 88 5 64 8 81 4 88 5
SET ‐ 7 CLUSTERING x (5:5) 57 9 47 5 38 0 49 4 71 7
SET ‐ 8 CLUSTERING x (9:1) 44 7 31 4 28 9 85 3 57 10
SET ‐ 9 U‐SHAPE 62 6 66 7 47 5 63 6 80 8
SET ‐ 10 L‐SHAPE 55 5 55 6 35 7 59 10 58 11
SET ‐ 11 X‐SHAPE 59 11 60 6 42 11 63 9 76 4
SET ‐ 12 CLUSTERING y (5:5) 57 9 82 3 37 3 49 9 67 8
SET ‐ 13 CLUSTERING y (7:3) 54 4 73 5 36 5 60 8 58 18
SET ‐ 14 CLUSTERING y (9:1) 48 8 57 5 30 12 85 3 53 14
95
Supplementary data 5A
Reconstructed chromatograms ‘D12’, ’D14’ and ‘D26R’ from “data set 1” published by
Schure et al. [29]. The red line indicates the fitted quadratic curve, and the bins (55)
are shown as blue rectangles.
Chromatogram D12
Chromatogram D14
y norm
alized
y norm
alized
96
Chromatogram D26R
Supplementary data 5B
Calculated orthogonality parameters and corresponding average orthogonality scores
of nine experts in the field of 2D chromatography, for “data set 2” published by Schure
et al [28]. The colors indicate the top 5 chromatograms in yellow, the bottom 5 in grey,
orange indicates chromatograms containing less than 100 and red less than 50 peaks.
The 1tr/2tr range is calculated from (1trmax(s) – 1trmin(s)) / (2trmax(s) – 2trmin(s)). The
reconstructed chromatograms YExp10, YExp13, Yexp2 and YExp15 are also shown.
y norm
alized
y norm
alized
97
0 0.2 0.4 0.6 0.8 1x normalized
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1x normalized
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1x normalized
0
0.2
0.4
0.6
0.8
1
y norm
alized
98
Number of pea
ks
Expert's Score
chromatogram
code
%BIN 5×5
%Ao
%FIT
%BIN × %FIT
Ran
ge 1tr(s) / 2tr(s)
119 90 YExp10 75 75 77 5,7 221
192 87 YExp15 59 57 61 3,6 193
67 86 YExp6 67 72 76 5,1 104
152 86 HExp4 77 75 71 5,5 83
147 86 HExp1 74 68 65 4,8 80
138 85 YExp11 70 63 65 4,5 88
157 85 HExp2 69 71 67 4,7 84
166 85 YExp14 63 59 62 3,9 447
153 85 HExp5 69 67 66 4,5 81
154 84 HExp3 71 67 65 4,6 87
109 84 YExp12 69 62 62 4,3 43
144 83 YExp16 68 57 61 4,1 101
136 83 HExp6 73 70 67 4,9 83
73 80 YExp7 67 61 62 4,1 40
55 79 YExp8 63 65 66 4,2 22
32 78 YExp2 57 77 85 4,9 38
32 77 YExp3 57 64 76 4,4 19
30 75 YExp4 57 62 83 4,7 8
25 75 YExp5 56 65 74 4,1 289
125 74 XExp1 59 45 40 2,4 87
37 73 YExp9 59 65 59 3,4 653
13 73 YExp1 52 48 73 3,8 191
60 73 YExp13 65 77 58 3,8 1485
109 72 XExp2 57 46 38 2,2 80
54 69 XExp4 63 40 39 2,5 79
43 66 XExp3 60 44 37 2,2 98
99
4 3B3BRetention Time Locking Procedure for Comprehensive Two‐
Dimensional Gas Chromatography
In gas chromatography (GC) reproducible retention times are in many cases highly
favorable or in some cases even required. In 1D‐GC, retention time shifts can be
eliminated or minimized by using a procedure called retention time locking (RTL). This
procedure is based on adjusting the (constant) column head pressure. Unfortunately,
this RTL procedure cannot be used in GC×GC given the fact that peaks will shift in both
dimensions. Adjusting the column head pressure in GC×GC will only minimize or
eliminate the primary retention time shifts.
In this chapter, a fast and easy to perform, two‐step retention time locking procedure
for GC×GC (2D‐RTL) is proposed and its feasibility is demonstrated. This 2D‐RTL
procedure involves adjustment of the column head pressure or constant column flow,
followed by the adjustment of the effective secondary column length (2Leff). The 2Leff is
increased or decreased, simply by moving it stepwise through the modulator. It is
demonstrated that retention time shifts in both the primary and secondary dimension,
which may occur after for example replacing the column‐set, can be minimized to less
than half peak base width. The proposed 2D‐RTL procedure is used successfully for
already 5 years in our laboratory at DSM.
4.1 Introduction
One of the main advantages of GC×GC is its high separation power making this
technique ideal for unraveling complex mixtures. Another main advantage is that
GC×GC provides structured chromatograms in which compounds with similar chemical
properties appear as distinct groups in the two‐dimensional chromatogram.
Nowadays, GC×GC is used to solve all kinds of real‐life analytical problems in a wide
variety of fields such as food [1, 2], biological [3, 4], environmental [5, 6] and
petrochemical [7, 8] areas.
As in 1D‐GC, retention time shifts in GC×GC are in many cases undesired. Reproducible
retention times are highly favorable or even required for visually comparing 2D‐
100
chromatograms, when using 2D templates for group‐type analysis, when using 2D‐
chromatograms as chemical fingerprints, or when applying all kinds of chemometric
operations.
The problem of retention time shifts in 1D‐GC can be solved by a procedure called
retention time locking (RTL), introduced by Blumberg and Klee [9]. RTL allows one to
maintain equal retention times for the same or different columns as long as both
columns have the same type of stationary phase and equal nominal phase ratio. By
using RTL, chromatograms can be reproduced accurately from one column to another
or from one GC to another. RTL is achieved simply by adjusting the column head
pressure. Since the introduction of RTL many applications can be found in the literature
[10‐13].
However, in GC×GC retention times may or will shift in both the primary and the
secondary dimensions. Locking both dimension retention times in GC×GC cannot be
achieved by only adjusting the column head pressure. Given the fact that no retention
time locking tools exists for GC×GC, only post‐analysis alignment techniques for
eliminating retention time shifts in both dimensions have been reported in literature
[14‐17].
In this paper, a GC×GC retention time locking procedure is proposed and its feasibility
is demonstrated. The proposed 2D‐RTL procedure involves two main steps. The first
step is locking the primary retention times by adjusting the column head pressure or
the constant column flow. The second step is locking the secondary retention times by
adjusting the effective secondary column length (2Leff). The 2Leff, which can be defined
as the length measured from the modulator to the detector, can be adjusted by
stepwise moving the second column through the modulator. The main idea of this
procedure is that the part of the secondary column which is positioned in front of the
modulator does not contribute to the secondary dimension separation and does not
have a significant influence on the primary dimension separation.
101
4.2 34B34BExperimental
4.2.1 71B71BChemicals
Grob test mixtures were purchased from Restek® (Restek Corporation, Bellefonte, PA).
The naphtha sample was purchased from Sigma‐Aldrich®.
4.2.2 Instrumental
All GC×GC‐FID analyses were carried out on a Leco (St. Joseph, MI, USA) GC×GC system
equipped with an Agilent 7683 autosampler, a hot split/splitless injector and a flame
ionization detector (FID). Three VF‐1MS columns (50m × 0.25mm; 0.4µm film
thickness) and three VF‐17MS columns (10m × 0.10mm; 0.2µm film thickness) were
purchased from Varian B.V. (Middelburg, The Netherlands).
4.2.3 73B73BSoftware
GC×GC instrument control and data processing was performed by Leco ChromaTOF®
software (St. Joseph, MI, USA) version 3.25. For all calculations Microsoft® Office Excel
2003 (Redmond, WA, USA), was used.
4.2.4 74B74BChromatographic conditions
In all experiments using a Grob test mixture a non‐polar VF1‐MS column was used for
the first dimension separation and a medium‐polar VF17‐MS column (variable length)
was used for the second dimension separation. The primary and secondary columns
are attached by means of a pressfit (Varian, Palo Alto, CA, USA) or Meltfit® (Nlisis
Chromatography BV, Veldhoven, Netherlands) connector. The GC×GC instrument was
operated under temperature‐programmed conditions from 40C, held for 0.2 minutes,
to 280C for the primary GC oven and from 45C, held for 0.2 minutes, to 285C for
the secondary GC oven; both at a temperature rate of 5Cmin‐1. The secondary oven
was only used to connect the secondary column from the modulator directly to the
primary GC oven; so, both columns are situated in the primary GC oven. The
modulation time was 3 seconds. The temperature of the modulator hot jets was 15C
higher than the actual primary oven temperature, and the pulse time was set to 1
102
second. Helium was used as the carrier gas. All separations were carried out using a
constant head pressure or constant column flow. The injection volume was 1µL. The
injector temperature was 280C. A split injection with a split ratio of 100:1 was applied
for all analyses. The FID was operated at a temperature of 300C, using a data‐
acquisition rate of 200 Hz.
A naphtha sample was used in order to demonstrate the 2D‐RTL procedure with a real‐
life sample. For these experiments two different column‐sets were used. A non‐polar
50m × 0.25mm × 0.4µm VF1‐MS column was used for the first‐dimension separation
and a medium‐polar 1.5m × 0.10mm × 0.2µm VF17‐MS for the second‐dimension
separation. The GC×GC instrument was operated under temperature‐programmed
conditions from 50C, held for 0.5 minutes, to 320C for the primary GC oven and from
55C, held for 0.5 minutes, to 325C for the secondary GC oven; both at a temperature
rate of 3Cmin‐1. The secondary oven was only used to connect the secondary column
from the modulator directly to the primary GC oven; so, both columns are situated in
the primary GC oven. The modulation time was 4 seconds.
4.2.5 75B75BOriginal column‐set, method, and retention times
Column‐set A is defined as the original column‐set. The analysis method using a
constant column head pressure of 41.75 PSI and a secondary column length of 1.50
meters is defined as the original method. The retention times obtained by using the
original column‐set (set A) and the original method are defined as the original
retention times.
For the experiment with the naphtha sample, both constant pressure and constant
flow modes were used. For these experiments, column‐set A is defined as the original
column‐set. The constant pressure method uses a constant column head pressure of
55 PSI and the constant flow method uses a constant column flow of 1 ml/min. Both
methods are defined as the original methods. The retention times obtained by using
the column‐set A, and the original methods are defined as the original retention times.
103
4.2.6 76B76BRun‐to‐run repeatability
In order to determine the repeatability a Grob mixture was analyzed four times by
using the original column‐set A, and the original analysis method.
4.2.7 77B77BRetention time shifts due to differences in column‐sets
A Grob mixture was analyzed to determine retention time shifts due to differences in
column‐sets on three different column‐sets (A, B, and C) using the original analysis
method.
4.2.8 78B78B2D‐RTL procedure
In order to demonstrate the feasibility of the 2D‐RTL procedure a new column‐set, in
which the secondary column length was approximately 15 cm longer than in the
original secondary column length, was installed. The extra 15 cm was situated after
the modulator so contributing to the second‐dimension separation. Before installing,
the first 25 cm of the secondary column (modulator side) was graduated by marking
the column every centimeter by using a heat resistant paint. The extra 15 cm can be
required in case the new second‐dimension retention times are significantly lower
compared to the original retention times.
The first step of the 2D‐RTL procedure is locking the first dimension. For this a Grob
mixture is analyzed at five different column head pressures or at five different constant
column flows, in the range of the column head pressure or column flow as used in the
original method +/‐ 20%. From the dependence of the retention time of a target
compound on column head pressure or column flow, the new column head pressure
or column flow, at which the primary retention of the target compound matches its
original primary retention time, is calculated, and has to be set into the analysis
method to lock the primary retention time.
The second step of the 2D‐RTL procedure is locking the second‐dimension. For this a
Grob mixture is analyzed, using the locked primary‐dimension method, at five different
effective secondary column lengths: the effective secondary column length as installed
104
+/‐ 15 cm. Shortening the effective secondary column length has to be done by sliding
the secondary column through the modulator making use of the painted markings.
Next, the delta second‐dimension retention times (original retention time of the target
compound minus the new obtained retention time) of the target compound is plotted
against the sliding distance measured in centimeters. From this plot, the sliding
distance at which the secondary retention of the target compound matches its original
secondary retention time, is calculated. Next a Grob mixture is analyzed again in order
to check the 2D‐RTL result.
This procedure is mainly suitable for modulator‐types in which it is possible to lengthen
of shorten the effective secondary column length by sliding the secondary column
through the modulator; these types can be referred to as so‐called pass‐through
modulators. A similar approach could also be used for single‐stage loop‐type
modulators in which the position of the loop is displaced across the secondary column
length.
The part of the secondary column length situated before and after the modulator
should preferably reside in the same thermal zone, more specifically the length of
secondary column that resides in each thermal zone (column part situated before
modulator, in modulator, after modulator and in transfer line or in detector) must stay
the same before and after sliding the column through the modulator. Therefore, the
current setup precludes the use of a separate thermal zone (secondary oven) for the
second dimension separation.
4.3 35B35BResults and discussion
4.3.1 79B79BRun‐to‐run repeatability
The run‐to‐run repeatability was determined by analyzing a Grob mixture four times
by using the original column‐set A, and the original analysis method. The results are
given in table 1. The results for the primary and secondary retention time repeatability
are given in table 1 and 2, respectively.
105
Table 1 Grob mix run‐to‐run repeatability of the first dimension retention times
Compound name
Average
Peak
width
Analysis
#1
Analysis
#2
Analysis
#3
Analysis
#4
s s s s s
butanediol 9 591 591 591 591
n‐decane 9 1092 1095 1095 1095
octanol 9 1212 1212 1212 1212
nonanal 9 1275 1275 1275 1275
dime‐phenol 9 1287 1287 1287 1287
n‐undecane 9 1299 1299 1299 1299
ethylhexanoic acid 12 1302 1302 1302 1302
dime‐aniline 9 1410 1410 1410 1410
Me‐decanoate 12 1689 1689 1689 1689
Me‐undecanoate 9 1857 1857 1857 1857
dicyclohexylamine 12 1899 1899 1899 1899
Me‐dodecanoate 9 2016 2016 2016 2016
106
Table 2 Grob mix run‐to‐run repeatability of the second dimension retention times
component
Average
Peak
Width
(Wb) #1 #2 #3 #4 average CV%
ms ms ms ms ms ms %
butanediol 143 1835 1815 1815 1820 1821 0.5
n‐decane 89 1350 1350 1345 1350 1349 0.2
octanol 88 1780 1775 1770 1775 1775 0.2
nonanal 88 1845 1840 1840 1840 1841 0.1
Dime‐phenol 109 2425 2410 2415 2420 2418 0.3
n‐undecane 76 1415 1415 1415 1415 1415 0.0
ethylhexanoic acid 108 1810 1805 1805 1805 1806 0.1
Dime‐aniline 117 2705 2700 2700 2700 2701 0.1
Me‐decanoate 82 1870 1870 1870 1875 1871 0.1
Me‐undecanoate 81 1915 1915 1910 1915 1914 0.1
dicyclohexylamine 92 2125 2120 2120 2125 2123 0.1
Me‐dodecanoate 81 1965 1965 1955 1960 1961 0.2
The retention time in the first dimension is determined by the modulated peak, which
has the largest peak area of all modulated peaks belonging to a single compound. The
first dimension retention time is not a continuum but is expressed as the product of
the number of the second‐dimension chromatogram and the modulation time. The
run‐to‐run primary retention time variation for the 12 compounds in the Grob mixture
is better than the modulation period of 3 seconds (n=4). As a consequence, no
differences in the primary retention times, from run‐to‐run could be determined. The
peak widths at peak base (Wb) in the primary‐dimension have been estimated by
multiplying the number of modulated peaks, belonging to the single compound, with
the modulation period of 3 seconds.
The run‐to‐run second‐dimension retention time variation of the 12 compounds in the
Grob mixture is on average better than 10 ms for peaks having an average peak base
width (Wb) of approximately 100 ms.
107
4.3.2 80B80BRetention time shifts due to differences in column‐sets
In order to get a rough idea about the influence of (small) manufacturing differences
in GC columns, including small differences in the positioning of the column‐set, a Grob
mixture was analyzed using three different column‐sets (column‐set A, B, and C). It has
to be noted that all columns are new and were ordered at the same supplier at the
same time, so large variations in the column parameters and stationary phase
properties by manufacturing variations or usage are not expected. For each analysis,
the original analysis method was used. Installation of the column‐sets and analysis was
performed by one person. The results are summarized in table 3.
Table 3 Retention time shifts determined by analysis of a Grob mixture by three
different column‐sets (column‐set A, B and C) using the original analysis method.
Compounds Wb Wb
Original
Column‐set A Column‐set B Column‐set C
1tr 2tr 1tr 2tr 1tr 2tr 1tr 2tr
s ms s ms s ms s ms
butanediol 9 143 591 1815 27 ‐55 0 110
n‐decane 9 89 1095 1345 30 ‐20 3 80
octanol 9 88 1212 1770 33 ‐45 6 100
nonanal 9 88 1275 1840 36 ‐55 6 100
dime‐phenol 9 109 1287 2415 36 ‐90 6 135
n‐undecane 9 76 1299 1415 33 ‐25 6 75
ethylhexanoic acid 12 108 1302 1805 36 ‐60 9 105
dime‐aniline 9 117 1410 2700 36 ‐105 6 155
me‐decanoate 12 82 1689 1870 39 ‐50 9 105
me‐undecanoate 9 81 1857 1910 42 ‐45 12 105
dicyclohexylamine 12 92 1899 2120 42 ‐55 12 115
me‐dodecanoate 9 81 2016 1955 45 ‐50 12 110
Analysis on column‐set B shows a shift of all peaks to higher primary retention times
and to lower secondary retention times. Analysis on column‐set C shows a shift of all
peaks to higher primary and secondary retention times. Additionally, a correlation
108
between the retention time and the absolute retention time shift is clearly visible; the
absolute peak shift increases at higher primary and secondary retention times.
It’s obvious that shifts in the primary retention times can be caused by small changes
in the primary column dimensions (1L, 1dc, 1df), however these shifts may also be
caused by changes in the secondary column dimensions (2L, 2dc) given the fact these
changes also influence the pressure drop across both the secondary and primary
column. The same is true for shifts in the secondary retention times, these can be
caused by small changes in the secondary column dimensions (2L, 2dc, 2df) or by small
changes in the primary column dimensions (1L, 1dc). In case of temperature‐
programming, a compound eluting at a different primary retention time, caused by a
change in the primary and/or the secondary column dimension, will enter the
secondary column at a different oven temperature, which will lead to a secondary
retention time shift. In summary, peaks may shift in both directions and the direction
and degree of shift cannot be predicted.
4.3.3 81B81B2D retention time locking procedure
Locking the first dimension
After installing a new column‐set, in which the secondary column length was
approximately 15 cm longer, a Grob mixture was analyzed at five different column head
pressures. In table 4 the original primary retention times (measured by using column‐
set A) and the retention time shifts measured at the different constant column head
pressures are given for all Grob mix compounds. The results clearly indicate a primary
retention time shift of 60 to 80 seconds, for all compounds when analyzing the Grob
mixture using the original analysis method having a constant head pressure of 41.75
PSI.
109
Table 4 Differences in retention times, compared to the original retention times,
measured at different constant column head pressures
Column Head Pressure
(psi) 41.75 33.40 37.58 41.75 45.93 50.10
Original
(column‐set A)
1tr 1tr 1tr
1tr 1tr 1tr
s s s s s s
1,4‐butanediol 591 ‐150 ‐99 ‐57 ‐21 9
n‐decane 1095 ‐171 ‐114 ‐63 ‐24 15
Octanol 1212 ‐174 ‐117 ‐66 ‐27 12
Nonanal 1275 ‐177 ‐120 ‐69 ‐27 12
dime‐phenol 1287 ‐183 ‐123 ‐69 ‐27 12
n‐undecane 1299 ‐177 ‐117 ‐66 ‐24 12
ethylhexanoic acid 1302 ‐174 ‐117 ‐66 ‐27 12
dime‐aniline 1410 ‐189 ‐126 ‐72 ‐27 12
me‐decanoate 1689 ‐183 ‐123 ‐72 ‐27 9
me‐undecanoate 1857 ‐183 ‐126 ‐75 ‐30 9
dicyclohexylamine 1899 ‐192 ‐129 ‐78 ‐30 12
me‐dodecanoate 2016 ‐186 ‐126 ‐75 ‐30 9
An overlay of the 2D‐chromatograms of the Grob mixture analyzed by the original
column‐set A and the original analysis method and the 2D‐chromatogram of the Grob
mixture analyzed by the newly installed column‐set and the original (not locked)
analysis method is given in figure 1.
110
Figure 1 Overlay of the 2D‐chromatogram of a Grob mixture obtained by using column‐
set A (ellipses) and a new installed column‐set (rectangles) both obtained using the
original analysis method
The results of methyl decanoate were used to calculate the constant column head
pressure at which the retention time difference compared to the original retention
time of methyl decanoate (1689 sec) is zero. The plot of the constant column head
pressure versus the primary retention time shift is given in figure 2.
111
Figure 2 Primary retention time shift of methyl decanoate compared to its original
retention time as a function of the constant column head pressure
By using the plot function as given in figure 2, the column head pressure at which the
retention time of methyl decanoate matches its original retention time can be
calculated. The calculated locked constant head pressure is 48.99 PSI. This pressure
was set into the analysis method used for locking the secondary dimension.
Locking the second‐dimension
In order to lock the second‐dimension a Grob mixture was analyzed, using the locked
primary‐dimension method, at five different effective secondary column lengths. In
table 5 the original secondary retention times and the retention time shifts measured
with different effective secondary column lengths are given for all Grob mix
compounds.
The results of methyl decanoate were used to calculate the secondary column length
shift at which the retention time difference compared to the original retention time of
methyl decanoate (1670 sec) is zero. The plot of the secondary column shift versus the
secondary retention time shift is given in figure 3.
112
Figure 3 Secondary retention time shift of methyl decanoate compared to its original
retention time as a function of the secondary column length shift
By using the plot function as given in figure 3, the secondary column length shift at
which the retention time of methyl decanoate matches its original retention time can
be calculated. The calculated secondary retention time shift is ‐4 cm. The secondary
column was positioned to ‐4 cm.
In table 6, the differences in primary retention times, compared to the original
retention times, measured at different effective secondary column lengths, are given.
These results clearly show that there is no significant correlation of shifting the
secondary column through the modulator, thereby lengthening or shortening its
effective length, on the primary retention times.
113
Table 6 Differences in primary retention times, compared to the original retention
times, measured at different effective secondary column length
Secondary Column
Length shift (cm) 0 ‐2 ‐5 ‐10 ‐15
Original
(column‐set A)
1tr
ms
1tr
s
1tr
s
1tr
s
1tr
s
1tr
s
1,4‐butanediol 591 3 3 3 3 3
n‐decane 1095 6 6 6 6 3
octanol 1212 3 3 3 6 0
nonanal 1275 3 3 3 6 0
dime‐phenol 1287 3 3 3 6 0
n‐undecane 1299 3 3 6 6 3
ethylhexanoic acid 1302 3 3 3 6 0
dime‐aniline 1410 3 3 6 6 3
me‐decanoate 1689 0 0 3 6 0
me‐undecanoate 1857 0 0 0 3 ‐3
Dicyclohexylamine 1899 0 0 3 6 ‐3
me‐dodecanoate 2016 ‐3 0 0 3 ‐3
After completing the locking procedure, a Grob mixture was analyzed again. The
results are summarized in table 7. Both the primary and secondary retention time
shifts are, on average, minimized to less than 0.5 Wb.
114
Table 7 Summarized results of the 2D‐RTL locking procedure
Compounds
Original
Original
not
locked
locked
primary
dimension
locked
primary and
secondary
dimension
1tr 2tr 1Wb 2Wb 1tr 2tr 1tr 2tr 1tr 2tr
s ms s ms s ms s ms s ms
1,4‐butanediol 591 1815 9 143 ‐57 ‐205 3 ‐85 3 ‐5
n‐decane 1095 1350 9 89 ‐63 ‐195 6 ‐40 6 15
octanol 1212 1775 9 88 ‐66 ‐215 3 ‐65 3 15
nonanal 1275 1840 9 88 ‐69 ‐215 3 ‐70 3 15
dime‐phenol 1287 2410 9 109 ‐69 ‐250 3 ‐105 3 5
n‐undecane 1299 1415 9 76 ‐66 ‐205 3 ‐45 3 20
ethylhexanoic acid 1302 1805 12 108 ‐66 ‐215 3 ‐65 3 10
dime‐aniline 1410 2700 9 117 ‐72 ‐255 3 ‐105 6 5
me‐decanoate 1689 1870 12 82 ‐72 ‐240 0 ‐70 3 15
me‐undecanoate 1857 1915 9 81 ‐75 ‐245 0 ‐70 0 20
dicyclohexylamine 1899 2120 12 92 ‐78 ‐265 0 ‐80 3 15
me‐dodecanoate 2016 1965 9 81 ‐75 ‐255 ‐3 ‐60 0 25
An overlay of the 2D‐chromatograms of the Grob mixture analyzed by the original
column‐set A and the original analysis method and the 2D‐chromatogram of the Grob
mixture analyzed by the new installed column‐set and the locked analysis method is
given in figure 4.
115
Figure 4 Overlay of the 2D‐chromatogram of a Grob mixture obtained by using column‐
set A and the original analysis method (ellipses) and a new installed column‐set
(rectangles) obtained by using the locked primary and secondary dimension analysis
method
4.3.4 82B82BReal life sample
The naphtha sample was analyzed using column‐set A, and both the original methods,
utilizing constant pressure and constant flow. The 2D‐chromatogram of the sample
obtained by using the column‐set A, and the original constant pressure method
(pressure is 55 PSI) is given in figure 5.
116
Figure 5 2D‐chromatogram of a naphtha sample analyzed by using column‐set A and
the original constant pressure method.
In figure 5, 5 peaks are indicated which are used to check the performance of the 2D‐
RTL procedure. After installing column‐set B, the naphtha sample was analyzed again
using both unlocked methods. Next, the 2D‐RTL procedure was applied and the
naphtha sample was analyzed once more using both locked methods. The constant
pressure, used in the constant pressure method was, was changed from 55.00 to 50.47
PSI in order to lock the primary‐dimension. The constant column flow, used in the
constant column flow method, was changed from 1.50 to 1.30 ml/min, in order to lock
the primary‐dimension. For both methods, the secondary column length was
shortened 4.5 cm in order to lock the secondary‐dimension. In table 8, the 2D‐RTL
results of 5 different peaks are summarized. The 5 peaks are well spread over the
whole 2D‐chromatogram and are therefore assumed to be representative for the
whole chromatogram.
117
Table 8 Results of the 2D‐RTL procedure
Constant pressure mode
Column‐set A
Column‐set B
Not locked
Column‐set B
Locked
Column‐set
A vs. B
Peak 1tr 2tr 1tr 2tr 1tr 2tr 1tr 2tr
nr. s s s s s s s ms
1 432 1.18 400 1.15 428 1.16 ‐4 ‐20
2 1876 3.04 1816 3.23 1868 3.07 ‐8 20
3 2992 3.72 2932 3.91 2988 3.72 ‐4 0
4 3400 0.371 3344 0.571 3400 0.351 0 ‐20
5 5536 2.74 5492 2.65 5540 2.70 4 ‐30
Constant column flow mode
Column‐set A
Column‐set B
Not locked
Column‐set B
Locked
Column‐set
A vs. B
Peak 1tr 2tr 1tr 2tr 1tr 2tr 1tr 2tr
nr. s s s s s s s ms
1 532 1.38 496 1.40 528 1.39 ‐4 10
2 1948 2.97 1892 3.14 1944 3.00 ‐4 40
3 2988 3.50 2932 3.67 2984 3.53 ‐4 30
4 3368 0.081 3316 0.231 3368 0.081 0 0
5 5508 2.25 5468 2.12 5512 2.25 4 0
1 these peaks are wrapped around
The results given in table 8 clearly indicate a significant retention time difference
between column‐set A and B when using the original, non‐locked, methods. After
performing the 2D‐RTL procedure, retention time shifts are on average minimized to
less than 0.5 Wb for both constant pressure and constant column flow methods.
4.4 36B36BConclusions
A fast and easy two step 2D‐RTL procedure is proposed and its feasibility is
demonstrated. The results show that significant primary and secondary retention time
shifts (Grob mixture compounds) can be minimized to less than 0.5Wb when applying
118
this procedure. The two‐step procedure consists of locking the first dimension by
adjusting the constant head pressure or constant column flow, followed by locking the
second‐dimension by adjusting the effective secondary column length. Locking the
second‐dimension, by moving the second‐dimension column trough the modulator,
thereby shortening or lengthening only its effective length, does not have a significant
effect on the already locked primary retention times. The 2D‐RTL procedure is mainly
suitable for to so‐called pass‐through modulators, however the applicability to single‐
stage loop‐type modulators will be part of our future research.
4.5 37B37BReferences
[1] M. Adahchour, J. Wiewel, R. Verdel, J. Beens, R.J.J. Vreuls, A.M. Batenburg, U.A.Th.
Brinkman, J. Chromatogr. A 1086 (2005) 99.
[2] R. Mayadunne, T.‐T. Nguyen, P.J. Marriott, Anal. Bioanal. Chem. 382 (2005) 836.
[3] S.M. Song, P.J. Marriott, P. Wynne, J. Chromatogr. A 1058 (2004) 223.
[4] K.M. Pierce, J.L. Hope, J.C. Hoggard, R.E. Synovec, Talanta 70 (2006) 797.
[5] R.K. Nelson, B.M. Kile, D.L. Plata, S.P. Sylva, L. Xu, C.M. Reddy, R.B. Gaines, G.S.
Frysinger, S.E. Reichenbach, Environ. Forens. 7 (2006) 33.
[6] M. Kallio, M. Jussila, T. Rissanen, P. Anttila, K. Hartonen, A. Reissell, R.J.J. Vreuls, M.
Adahchour, T. Hyötyläinen, J. Chromatogr. A 1125 (2006) 234.
[7] C. von Muhlen, C.A. Zini, E.B. Caramao, P.J. Marriott, J. Chromatogr. A 1105 (2006)
39.
[8] R. Hua, J. Wang, H. Kong, J. Liu, X. Lu, G. Xu, J. Sep. Sci. 27 (2004) 691.
[9] L.M. Blumberg, M.S. Klee, Anal. Chem. 70 (1998) 3828.
[10] N. Etxebarria, O. Zuloaga, M. Olivares, L. J. Bartolome, P. Navarro, J. Chromatogr.
A 1216 (2009) 1624.
[11] N. Ochiai, K. Sasamoto, H. Kanda, T. Yamagami, F. David, B. Tienpoint, P. Sandra, J.
Sep. Sci. 28 (2005) 1083.
[12] T. Ishida, K. Kudo, H. Inoue, A. Tsuji, T. Kojima, N. Ikeda, J. Anal. Toxicol. 30 (2006)
468.
[13] I. Rasanen, I. Kontinen, J. Nokua, I. Ojanpera, E. Vuori, J. Chromatogr. B 788 (2003)
243.
119
[14] T. Skov, J.C. Hoggard, R. Bro, R.E. Synovec, J. Chromatogr. A 1216 (2009) 4020.
[15] J. Vial, H. Noçairi, P. Sassiat, S. Mallipatu, G. Cognon, D. Thiébaut, B. Teillet, D.N.
Rutledge, J. Chromatogr. A 1216 (2009) 2866.
[16] M. Kallio, M. Kivilompolo, S. Varjo, M. Jussila, T. Hyötyläinen, J. Chromatogr. A
1216 (2009) 2923.
[17] D. Zhang, X. Huang, F.E. Regnier, M. Zhang, Anal. Chem. 80 (2008) 2664.
120
121
5 4B4BA Procedure for Comprehensive Two‐dimensional Gas
Chromatography Retention Time Locked Dual Detection
In this chapter, a novel, and easy to perform, retention time locking procedure for
locking primary and secondary retention times of detector signals in comprehensive
two‐dimensional gas chromatography (GC×GC) dual‐detection is proposed and its
advantages are demonstrated and discussed. The dual detection retention time
locking procedure is a 2‐step process for a GC×GC system in which the effluent of the
primary column is split, by using a pressure regulated splitter, towards the GC×GC
modulator using two identical secondary GC columns of which one is installed in the
main GC oven and the other is installed in a secondary GC oven. The first step of the
locking procedure is to minimize the secondary retention time difference between
both detectors of a compound, which has a retention factor (k) close to 0. This is done
by stepwise altering the effective secondary column length, simply by sliding the
secondary column, which is installed in the main oven, forwards or backwards through
the modulator. The second step is to minimize the secondary retention time
difference of a compound which has a significant retention in both dimensions. This is
done by stepwise altering the secondary oven temperature rate. This locking
procedure was successfully demonstrated for the analysis of a diesel sample by GC×GC
coupled to a time of flight mass spectrometer (TOFMS) and a nitrogen
chemiluminescence detector (NCD) and by GC×GC coupled to a TOFMS and a flame
ionization detector (FID). For all compounds, the average absolute secondary
retention time differences between the NCD or the FID and the TOFMS detectors were
0.03, and 0.07 seconds, respectively, which are significantly less than the average peak
widths at half heights, which was 0.2 seconds.
5.1 38B38BIntroduction
GC×GC is becoming more and more a routine analytical technique for solving all kinds
of analytical questions in a wide variety of application fields, for example the nitrogen,
sulfur and oxygen speciation of complex petrochemical and bio‐oil samples [1‐5].
122
For particular GC×GC applications, using two different detectors simultaneously (dual
detection) can offer a significant advantage compared to single detection. The main
benefits of GC×GC dual‐detection are speed and efficiency of the particular analysis,
obtaining two signals per analysis while there is no need to alter the chromatographic
system [6‐15]. For example, when performing a dual‐detection GC×GC analysis with a
nitrogen chemiluminescent detector (NCD) and a mass spectrometer (MS), in one
single analysis the nitrogen containing compounds could be detected with high
selectivity and quantified by the NCD and directly be identified and/or verified by the
MS detector.
However, practically, GC×GC dual‐detection is not that straightforward. The column
effluent of the primary or the secondary column(s) needs to be split towards both
detectors.
The split ratio during a temperature programmed GC analysis may vary, which
complicates quantitation, especially for detectors operating at different outlet
pressures. To overcome this problem, the pressure at the split point needs to be kept
constant [16] in order to keep the split ratio constant; this can be performed by using
a so‐called electronic pressure regulated splitter. Splitting after the primary column
and using two secondary columns is highly preferred since it leads to more optimal
linear velocity operation in both dimensions [17]. Besides this, splitting after the
secondary column introduces extra‐column band broadening which could be critical
for the narrow secondary dimension peaks which have a typically peak width at half
height in the order of 0.1 seconds.
Another issue of GC×GC dual‐detection is that the secondary retention times of both
detectors are generally not equal. Retention time differences are caused by
differences in the column or retention gap dimensions towards both detectors and/or
differences in the capillary flow regimes towards both detectors during a temperature
programmed GC×GC analysis, again especially for detectors operating at different
outlet pressures. When using two detectors which may show non‐similar 2D‐
chromatograms, for instance NCD/FID, NCD/MS, SCD/FID, SCD/MS (FID/MS show
similar 2D‐chromatograms), secondary retention time differences between both
detectors may complicate the data processing. For example, for identifying the
123
nitrogen containing compounds, at low concentration levels, in a complex
petrochemical sample matrix, the NCD 2D‐chromatogram must be compared with the
completely different complex MS chromatogram in order to locate the corresponding
nitrogen containing compound peaks.
In this chapter we proposed a novel, and easy to perform, retention time locking
procedure for locking primary and secondary retention times of detector signals in
GC×GC dual‐detection.
5.2 39B39BExperimental
5.2.1 83B83BEquipment and materials
All GC×GC analyses were performed using a Leco Pegasus 4D (St. Joseph, MI, USA)
GC×GC system, equipped with a secondary GC oven, an Agilent Technologies (Santa
Clara, CA, USA) capillary flow technology splitter, a hot split/splitless injector, a Leco
Pegasus time‐of‐flight mass spectrometer (TOFMS), an Agilent Technologies Nitrogen
Chemiluminescent Detector (NCD 255) and an Agilent flame ionization detector (FID).
Instrument control and data processing was performed by Leco ChromaTOF software
version 3.25 and NIST MS‐Search 2.0. A VF1ms column (50m × 0.25mm; 0.4µm film
thickness) and a VF17ms column (10m × 0.10mm; 0.2µm film thickness) were
purchased from Agilent Technologies (Santa Clara, CA, USA).
5.2.2 84B84BTest mixtures and samples
Two different test mixtures containing nitrogen containing compounds were prepared
in toluene. All compounds were purchased from Sigma‐Aldrich. For test mixtures 1
and 2, the compounds, compound purities and compound concentrations, are
summarized in table 1 and 2, respectively. A real‐life diesel sample was purchased
from a local gas station.
124
Table 1 Composition of test mixture 1
Compounds Purity
%
Conc.
ppm
acetonitrile 99 449
propanenitrile 99 398
isobutyronitrile 99 438
Table 2 Composition of test mixture 2
Compounds Purity
%
Conc.
ppm
hydantoine 99 439
acridine 97 344
phenanthridine 98 362
2,5‐dimethyl‐pyrrole 98 438
quinaldine 99 413
isobutyronitrile 99 459
4‐picoline 98 452
propionitrile 99 442
N,N‐dimethyl‐cyclohexylamine 98 583
3‐ethyl‐4‐methyl‐pyridin >95 451
1‐ethyl‐2‐pyrrolidon 99 558
4‐pentennitrile 98 444
pyrrolidine 99 518
5.2.3 85B85BChromatographic conditions
For all analyses, the non‐polar VF1ms column was used for the first‐dimension
separation and two, in parallel coupled, 2 meters medium‐polar VF17ms columns
were used for the second‐dimension separations. The end of the primary column and
both the beginnings of the secondary columns were attached to the Agilent electronic
pressure regulated (EPC) splitter. Both secondary columns were directed through the
cryogenic modulator. The end of one secondary column was connected to the TOFMS
125
retention gap (0.5m × 0.10mm i.d.), by means of a Siltite® coupling (SGE Analytical
Science, Ringwood Victoria, Australia), and the end of the other secondary column was
connected directly into the NCD (NCD/TOFMS) detector or directly into the FID
(FID/TOFMS) detector. The secondary column leading to the NCD or FID was installed
in the main GC oven and the secondary column leading to the TOFMS was installed in
the secondary GC oven. A schematic overview of the GC×GC‐NCD/TOFMS or GC×GC‐
FID/TOFMS dual detection set‐up is given in figure 1.
Figure 1 Schematic overview of the GC×GC‐NCD/TOFMS or the GC×GC‐FID/TOFMS dual
detection set‐up
The GC×GC instrument was operated under temperature programmed conditions
from 40C, held for 1 minute, to 330C, held for 5 minutes, for the primary GC oven
and from 45C, held for 1 minute, to 335C, held for 5 minutes, for the secondary GC
oven. The temperature rate for the main oven was 2C/min. For retention time
locking, the temperature rate for the secondary oven was 2+ΔT C/min. The
modulation time was 10 seconds. A hot pulse of 2 seconds, having a temperature of
20C higher than the actual primary GC oven, was used. Helium was used as the carrier
gas. All separations were carried out using a constant head pressure of 35 PSI. A
constant split pressure of 18 PSI and 23 PSI was used for the NCD/TOFMS and the
FID/TOFMS setup, respectively. The injector temperature was 300C. An injection
volume of 1µL and a split ratio of 1:50 was applied for all analyses. The NCD was
operated at a burner temperature of 925C, with an oxidizer flow of 10 sccm, a
126
hydrogen flow of 5 sccm and using a data‐acquisition rate of 100Hz. The FID
temperature was 280C, using an air flow of 450 ml/min and a hydrogen flow of 40
ml/min. The TOFMS was operated in electron impact mode using 70eV, a source
temperature of 250C and a mass range of 15 to 550 amu.
5.2.4 86B86BDual detection retention time locking procedure
The dual detection retention time locking procedure is an easy to perform 2‐step
process. The first step in the locking procedure is to minimize the secondary retention
time difference of a compound, which has a retention factor (k) preferably close to 0,
between both the NCD or FID and TOFMS detector signals. For this, the compound
acetonitrile, present in the test mixture 1, was used. In order to minimize this
retention time difference, the effective secondary column length of the NCD or FID
was stepwise (using equal steps) increased or decreased, by sliding the secondary
column through the modulator. The effective secondary column length is the length
of the column after the modulator, from the cryogenic modulator to the detector, so
the part of the secondary column which is effectively used for the second dimension
separation. Adjusting the secondary retention time, by stepwise altering the effective
secondary column length, was already described by Mommers et al, for
comprehensive two‐dimensional gas chromatography retention time locking [18].
The effective secondary column length of the TOFMS was kept constant. Also, both
the main and secondary oven temperature rates were kept equal and set to 2C/min.
When knowing the relationship of the differences in secondary retention time
between the NCD or FID and TOFMS signals (delta) of acetonitrile as function of the
secondary column length shift, the column length shift at which the delta is zero can
be calculated, set and checked.
The second step in the locking procedure is to minimize the delta of a compound which
has a significant retention in both dimensions, between the NCD or FID and TOFMS
detector signals. For this, the compound phenanthridine, present in the test mixture
2, was used. In order to minimize this delta, the secondary oven temperature rate of
the TOFMS secondary column was stepwise increased or decreased. The oven
127
temperature rate of the secondary column (2C/min), installed in the main oven, was
not altered. Also, here, when knowing the relationship of the delta of phenanthridine
as function of the TOFMS secondary oven rate, the secondary oven rate at which the
delta is zero can be calculated, set, and checked.
5.3 40B40BResults and discussion
5.3.1 87B87BGC×GC‐NCD/TOFMS
Locking step 1
For the initial analysis (no shift of the NCD secondary column), using a primary and
secondary oven temperature rate of 2C/min, the secondary retention time of
acetonitrile was 1.89 seconds for the NCD signal and 1.60 seconds for the TOFMS
signal. In order to minimize this delta of ‐0.29 seconds, the effective secondary column
length of the NCD was stepwise (equal steps of approximately 3 cm) decreased by
sliding the column through the modulator. The effective secondary column length of
the TOFMS was not changed. In figure 2, the delta of acetonitrile as function of the
NCD secondary column length shift, is given.
Figure 2 Delta secondary retention time of acetonitrile (TOFMS – NCD signal) as a
function of the NCD secondary column length shift
128
Figure 2 shows a clear linear relationship (r2=0.994) between the effective secondary
column length and the delta secondary retention time of acetonitrile. When
shortening the effective secondary column length of the NCD by 13.8 cm (=intercept),
the secondary retention time difference between the NCD and the TOFMS is 0.02
seconds. This retention time difference is significantly smaller than the peak width at
half height for acetonitrile which is 0.15 seconds.
Locking step 2
For the initial analysis (secondary retention times of both detector signals are locked
for acetonitrile), the secondary retention time of phenanthridine is 7.59 seconds for
the NCD signal and 7.83 seconds for the TOFMS signal. In order to minimize this
retention time difference of 0.24 seconds, the secondary oven temperature rate of
the TOFMS secondary column was altered stepwise. An overlay of two GC×GC
chromatograms obtained with different secondary oven temperature rates is given in
figure 3. In figure 4, the delta secondary retention time of phenanthridine as function
of the TOFMS secondary oven temperature rate, is given.
129
Figure 3 Overlay of GC×GC chromatograms of test mixture 2 obtained with a TOFMS
secondary oven temperature rate of 1.90C/min (peaks indicated with yellow ovals)
and 2.10C/min. The last indicated peak is phenanthridine.
Figure 4 Delta secondary retention time of phenanthridine (TOFMS – NCD signal) as
function of the TOFMS secondary oven temperature rate (C/min)
130
Figure 4 shows a quadratic relationship (r2=0.997) between the secondary oven
temperature rate and the delta of phenanthridine. When analyzing the test mixture 2
with a secondary oven temperature rate of 2.02C/min (=intercept) the secondary
retention time difference of phenanthridine between the NCD and the TOFMS is 0.01
seconds.
After performing the 2‐step locking procedure, thereby adjusting the effective
secondary column length of the secondary column to the NCD (locking step 1) and
adjusting the secondary oven temperature rate of the secondary column to the
TOFMS (locking step 2), the test mixture 2 was analyzed 4 times. For all compounds,
the results of the two‐step dual‐detection retention time locking process are
summarized in table 3.
Table 3 Result of the two‐step dual detection retention time locking procedure for the
analysis of the test mixture 2 (n=4)
TOFMS
Detector
NCD
Detector
Name AVG
1tr
(s)
AVG
2tr
(s)
AVG
1tr
(s)
AVG
2tr
(s)
AVG
PW1/2
(s)
AVG
1tr
(s)
AVG
2tr
(ms)
SD
2tr
(ms)
propanenitrile 360 1,95 360 1,97 0,19 0 15 6
isobutyronitrile 420 2,09 420 2,13 0,24 0 33 25
4‐pentennitrile 810 3,44 810 3,41 0,20 0 35 13
4‐Picoline 1160 3,96 1160 3,91 0,20 0 50 8
2,5‐dimethyl‐pyrrole 1440 4,29 1440 4,27 0,20 0 20 21
N,N‐dimethyl‐cyclohexanamine 1780 2,99 1780 2,95 0,18 0 45 6
3‐ethyl‐4‐methyl‐pyridine 2240 4,68 2240 4,63 0,18 0 57 15
1‐Ethyl‐2‐pyrrolidinone 2260 6,39 2260 6,27 0,23 0 22 10
quinaldine 3290 5,78 3290 5,72 0,18 0 55 24
acridine 5210 7,29 5210 7,28 0,21 0 13 10
phenanthridine 5290 7,46 5290 7,46 0,22 0 10 8
The results in table 3 show that, by performing the two‐step dual detection locking
procedure, the secondary retention time differences between the NCD and the
131
TOFMS detectors are minimized to an absolute average of 0.03 seconds which is, for
all compounds, significantly less than the average peaks widths at half heights, which
is 0.20 seconds. The standard deviation of the absolute secondary retention time
differences is 0.04 seconds. These results also show that the time difference observed
for the primary dimension is acceptable.
Additionally, differences in the observed Δ2tr for all compounds are mainly caused by
differences in chromatographic selectivity (between different types of compounds)
when changing the secondary oven temperature rate. However, these observed
differences are significantly smaller than the corresponding observed peaks widths at
half height.
5.3.2 88B88BGC×GC‐FID/TOFMS
Locking step 1
For the initial analysis, using a primary and secondary oven temperature rate of
2C/min, the secondary retention time of acetonitrile was 2.03 seconds for the TOFMS
signal and 2.48 seconds for the FID signal. In order to minimize this delta of ‐0.45
seconds, the effective secondary column length of the FID was stepwise (equal steps)
decreased by sliding the column through the modulator. The effective secondary
column length of the TOFMS was not changed. In figure 5, the delta of acetonitrile as
function of the FID secondary column length shift, is given.
132
Figure 5 Delta secondary retention time of acetonitrile (TOFMS – FID signal) as a
function of the FID secondary column length shift
When shortening the effective secondary column length of the FID by 3.4 cm
(=intercept), the secondary retention time difference between the FID and the TOFMS
is 0.03 seconds which is significantly less than 0.15 seconds which is the peak width at
half height for acetonitrile.
Locking step 2
For the initial analysis, the secondary retention time of phenanthridine is 7.23 seconds
for the TOFMS signal and 9.32 seconds for the FID signal. In order to minimize this
retention time difference of ‐2.09 seconds, the secondary oven temperature rate of
the TOFMS secondary column was altered stepwise. In figure 6, the delta of
phenanthridine as function of the TOFMS secondary oven temperature rate, is given.
133
Figure 6 Delta of phenanthridine (TOFMS – FID signal) as function of the TOFMS
secondary oven temperature rate (C/min)
When analyzing the test mixture 2 with a secondary oven temperature rate of
1.85C/min (=intercept) the secondary retention time difference of phenanthridine
between the FID and the TOFMS is 0.01 seconds.
After performing the 2‐step locking procedure, the test mixture 2 was analyzed. The
secondary retention time differences between the FID and the TOFMS detectors are
minimized to an absolute average of 0.07 seconds which is, for all compounds,
significantly less than the average peaks widths at half heights, which is 0.19 seconds.
The results of the two‐step locking procedure for all compounds are given in
supplementary 1. The FID chromatograms, before and after performing the locking
procedure, are given in figure 7A and 7B, respectively. The corresponding TOFMS
peaks are indicated by the yellow ovals.
134
Figure 7 FID chromatograms of test mixture 2, before (A) and after (B) performing the
locking procedure; the corresponding TOFMS peaks are indicated by the yellow ovals.
It also has to be noted that the locking procedure is easy and fast to perform; usually,
only 3 or 4 analyses per locking step are required to calculate the required effective
secondary column length and the required secondary oven temperature rate,
respectively. In our lab, this procedure is already, successfully, and routinely, applied
for more than 5 years for the analysis of a wide variety of petrochemical samples such
135
as naphtha cracker feedstocks, diesels and also (pyrolysis) bio‐oils by GC×GC‐
NCD/TOFMS and GC×GC‐SCD/TOFMS. In general, the locking procedure is performed
in less than 1 day and the retention times are stable for several weeks.
Split ratios
By using a constant split pressure, during the whole temperature program, the column
flows of both secondary columns, and thereby the corresponding split ratios, can
easily be calculated by using for instance the Agilent pressure flow calculator. For the
TOFMS/NCD and the TOFMS/FID setups the constant split ratios are 1.0 and 1.3,
respectively. Besides calculating, the split ratios can also be determined by 2
sequential analysis of a test mixture by single (e.g. FID) and dual detection (e.g. FID
and TOFMS), respectively. The ratio of the corresponding peak areas of the single and
dual detection analysis, equal the split ratio when performing dual‐detection.
5.3.3 89B89BSplitting after the secondary column
As described by Nicolotti et al [8], GC×GC dual detection can also be performed by
splitting the secondary column effluent towards two different detectors. In order to
keep the split ratio constant a constant split pressure must be maintained during the
whole analysis by using a pressure regulated splitter. As already mentioned in the
introduction, splitting after the secondary column, so after the modulator, has several
disadvantages compared to splitting before the modulator. Splitting after the primary
column and using two secondary columns is highly preferred since it leads to more
optimal linear velocity operation in both dimensions and also higher column
loadability, as was demonstrated by Peroni et al [17]. Splitting after the secondary
column introduces extra‐column band broadening which could be critical for the
narrow secondary dimension peaks which have a typically peak width at half height in
the order of 0.2 seconds. Retention gaps are required to obtain a certain required split
ratio and to ensure flows to be within certain system requirements especially for MS
systems. These extra retention gaps (restrictions) will lead to less optimal linear gas
velocities.
136
Another approach, of locking GC×GC dual detection MS/FID signals, was described by
Nicolotti et al [8] where the effluent of the primary column was split before the
modulator, into two secondary columns, towards an FID and a MS detector. Their
approach requires two pressure controllers, one situated before the modulator,
needed to keep the pressure at the split point toward both secondary columns
constant to ensure a constant split ratio, needed for quantitative analysis. The other
EPC, situated after the modulator, is needed to change the pressure at the secondary
column outlet towards the MS detector in order to adjust the retention times. The
pressure controller situated after the modulator could lead to some extra band
broadening. Adjusting the retention times by means of altering the outlet pressure is
done by supplying helium to the column effluent; this is limited to the pump capacity
of the MS, which should not be exceeded. Retention time differences between the MS
and FID signals were slightly larger than their measured peak width at half heights.
In table 4 the retention time results are summarized of the GC×GC‐NCD/TOFMS
analysis of the test mixture by splitting the secondary column effluent towards both
detectors. Splitting was performed by using a pressure regulated splitter, at a constant
pressure of 18 PSI, and 2 identical (methyl deactivated 1m × 0.1mm) retention gaps.
137
Table 4 Result of GC×GC‐NCD/TOFMS dual detection by splitting the secondary
column effluent using a constant split pressure of 18 PSI
TOFMS
Detector
NCD
Detector
Name AVG
1tr
(s)
AVG
2tr
(s)
AVG
1tr
(s)
AVG
2tr
(s)
AVG
PW1/2
(s)
AVG
1tr
(s)
AVG
2tr
(ms)
propanenitrile 650 5,00 650 4,78 0,2 0 220
isobutyronitrile 760 5,19 760 4,97 0,2 0 220
4‐pentennitrile 1010 5,56 1010 5,28 0,3 0 280
4‐Picoline 1780 7,08 1770 6,81 0,2 10 270
2,5‐dimethyl‐pyrrole 2090 7,28 2090 7,03 0,2 0 250
N,N‐dimethyl‐cyclohexanamine 2490 6,11 2490 5,85 0,2 0 260
3‐ethyl‐4‐methyl‐pyridine 2980 7,95 2980 7,67 0,2 0 280
1‐Ethyl‐2‐pyrrolidinone 2990 9,51 2990 9,23 0,3 0 280
quinaldine 4060 9,64 4060 9,36 0,3 0 280
acridine 6000 3,78 6000 3,46 0,3 0 320
phenanthridine 6080 4,17 6080 3,86 0,3 0 310
The results in table 4 show that splitting the secondary column effluent result in an
average secondary retention time difference of 270 ms (standard deviation = 30 ms)
which is approximately equal to the average measured peak width at half height.
Adjusting these retention time differences between both NCD/TOFMS detector
signals could only be established by stepwise physically shortening the length of the
retention gap towards the TOFMS detector. Additionally, compared to splitting after
the primary column, this setup shows an increase of the secondary peak widths from
approximately 20 to 30 ms.
5.3.4 90B90BReal‐life samples
A real‐life diesel sample was analyzed 3 times by the GC×GC‐TOFMS/NCD dual‐
detection retention time locked system. The TOFMS chromatogram is given in figure
8a and the corresponding NCD chromatogram is given in figure 8b. The TOFMS data
was processed, which resulted in about 12000 chromatographic features. The
deconvoluted mass spectra were searches against the NIST library. The NCD data were
also processed, which resulted in about 100 chromatographic features.
138
An overlay of the GC×GC NCD chromatogram and the TOFMS peak markers are shown
in figure 9. For 31 nitrogen containing compounds, the NCD and TOFMS primary and
secondary retention times are given in table 5. This table also includes the
identification of the nitrogen containing compounds and the primary and secondary
retention time differences between the NCD and TOFMS signals (1tr and 2tr). The
primary retention time difference is less than 10 seconds for most compounds and the
absolute secondary retention time difference is 0.04 seconds which is well within the
average peak width at half height. The standard deviation (n=3) for the absolute
secondary retention time differences between the NCD and the TOFMS detectors is
0.02 seconds.
139
Figure 8 Analysis of a real‐life diesel samples with dual detection retention time locked
GC×GC‐TOFMS/NCD; the TOFMS TIC GC×GC chromatogram is given in 8A and the NCD
GC×GC chromatogram is given in 8B.
8a
Nitro‐
compounds
Indoles
Carbazoles
Anilines
Unknowns
8b
140
Figure 9 Overlay of the GC×GC NCD chromatogram and the TOFMS peak markers of
the analyzed diesel sample.
From this overlay, it is clear that most nitrogen containing group‐types coelute with
many non‐nitrogen containing compounds. In this case, dual detection retention time
locking significantly supports the fast and efficient identification of nitrogen
containing compounds in a very complex hydrocarbon matrix. The TOFMS and NCD
GC×GC chromatograms are complex and not similar which makes localization of a
particular compound in both chromatograms, if not retention‐time locked, a difficult
and time consuming task. After dual‐detection retention time locking, the
corresponding TOFMS peak markers (if detected by the TOFMS) are located within the
corresponding peak width of the NCD peaks, making identification of NCD peaks a
much easier and more efficient task.
Nitro‐
compounds
Indoles
Carbazoles
Anilines
Unknowns
141
Table 5 Result of the two‐step dual detection retention time locking procedure for the
analysis of a real‐life diesel sample by GC×GC‐TOFMS/NCD; 31 identified nitrogen
containing compounds
TOFMS
detector
NCD
detector
Identified nitrogen
containing compound
1tr
(s)
2tr
(s)
1tr
(s)
2tr
(s)
1tr
(s)
2tr
(s)
nitro‐methane 350 1,96 350 1,86 0 ‐0,10
nitro‐ethane 460 2,51 460 2,52 0 0,01
nitro‐propane 660 3,18 670 3,12 10 ‐0,06
nitro‐butane 1020 3,81 1020 3,75 0 ‐0,06
unknown nitro‐compound 2100 3,78 2100 3,74 0 ‐0,04
unknown nitro‐compound 2200 3,91 2200 3,86 0 ‐0,05
dimethyl‐aniline 2630 5,39 2630 5,37 0 ‐0,02
Indole 3200 6,84 3200 6,78 0 ‐0,06
methyl‐indole (1) 3500 6,77 3500 6,73 0 ‐0,04
methyl‐indole (2) 3610 6,54 3610 6,46 0 ‐0,08
methyl‐indole (3) 3630 6,75 3630 6,72 0 ‐0,03
dimethyl‐indole (1) 3920 6,34 3920 6,36 0 0,02
dimethyl‐indole (2) 4000 6,52 4010 6,43 10 ‐0,09
dimethyl‐indole (3) 4020 6,20 4020 6,18 0 ‐0,02
dimethyl‐indole (4) 4030 6,43 4030 6,35 0 ‐0,08
trimethyl‐indole 4340 5,91 4340 5,83 0 ‐0,09
carbazole 5310 7,90 5310 7,90 0 0,00
methyl‐carbazole (1) 5570 7,31 5570 7,31 0 0,00
methyl‐carbazole (2) 5660 7,52 5660 7,54 0 0,02
methyl‐carbazole (3) 5680 7,46 5680 7,48 0 0,02
methyl‐carbazole (4) 5720 7,88 5720 7,90 0 0,02
dimethyl‐carbazole (1) 5790 6,50 5790 6,49 0 ‐0,01
dimethyl‐carbazole (2) 5940 6,87 5940 6,88 0 0,01
dimethyl‐carbazole (3) 5980 7,18 5980 7,19 0 0,01
dimethyl‐carbazole (4) 6020 7,11 6020 7,14 0 0,03
dimethyl‐carbazole (5) 6080 7,34 6080 7,39 0 0,05
trimethyl‐carbazole (1) 6130 6,22 6130 6,22 0 0,00
trimethyl‐carbazole (2) 6180 6,43 6180 6,44 0 0,01
trimethyl‐carbazole (3) 6260 6,49 6260 6,51 0 0,02
trimethyl‐carbazole (4) 6280 6,73 6280 6,79 0 0,06
trimethyl‐carbazole (5) 6320 6,71 6320 6,74 0 0,03
142
5.4 41B41BConclusions
A novel, and easy to perform, two‐step retention time locking procedure for locking
primary and secondary retention times of detector signals in GC×GC dual‐detection is
proposed and its feasibility and added value for data processing is demonstrated. The
results of this locking procedure show that secondary retention time differences
between TOFMS/NCD and between TOFMS/FID detectors can be minimized to an
absolute average of 0.03 and 0.07 seconds, which is significantly less than the average
corresponding peak widths at half height. Primary retention time differences are for
most peaks less than the applied modulation time.
Especially when using two detectors which may show non‐similar GC×GC
chromatograms, GC×GC dual detection retention time locking may significantly
improve and speed up the data processing of complex samples. After locking,
comparing complex GC×GC chromatograms and/or localization of specific peaks
becomes a much easier, less time consuming and therefore more efficient task.
Besides this, locking the retention times of both detectors will also reduce the risk of
wrong peak localizations which could lead to wrong peak identifications.
5.5 42B42BReferences
[1] Adam F., Bertoncini F., Dartiguelongue C., Marchand k., Thiébaut D., Hennion M‐
C., Fuel, 88 (2009), 5, 938‐946
[2] Adam F., Bertoncini F., Brodusch n., Durand E., Thiébaut D., Espinat D., Hennion M‐
C., J Chromatogr A, 1148 (2007), 1, 55‐64
[3] Omais B., Courtiade M., Charon N., Thiébaut D., Quignard A., Hennion M‐C., J
Chromatogr A, 1218 (2011), 21, 3233‐3240
[4] Flego C., Zannoni C., Fuel, 90 (2011), 9, 2863‐2869
[5] Mondello L., Casilli A., Tranchida P.Q., Dugo G., Dugo P., J Chromatogr A, 1067
(2005), 1–2, 235‐243
[6] Cordero C., Bicchi C., Joulain D., Rubiolo P., J Chromatogr A, 1150 (2007), 1–2, 37‐
49
[7] Dijkmans T., Djokic M.R., Van Geem K.M., Marin G.B., Fuel, 140 (2015), 15, 398‐406
143
[8] Nicolotti L., Cordero C., Bressanello D., Cagliero C., Liberto E., Magagna F., Rubiolo
P., Sgorbini B., Bicchi C., J Chromatogr A, 1360 (2014), 264‐274
[9] Bressanello D., Liberto E., Collino M., Reichenbach S.E., Benetti E., Chiazza F., Bicchi
C., Cordero C., J Chromatogr A, 1361 (2014), 265‐276
[10] Tranchida P.Q., Salivo S., Bonaccorsi I., Rotondo A., Dugo P., Mondello L., J
Chromatogr A, 1313 (2013), 194‐201
[11] Krupčík J., Gorovenko R., Špánik I., Sandra P., Armstrong D.W., J Chromatogr A,
1280 (2013), 104‐111
[12] Vallejo A., Olivares M., Fernández L.A., Etxebarria N., Arrasate S., Anakabe E.,
Usobiaga A., Zuloaga O., J Chromatogr A, 1218 (2011), 20, 3064‐3069
[13] Tran T.C., Marriott P.J., Atmos. Environ., 42 (2008), 32, 7360‐7372
[14] Cordero C., Bicchi C., Joulain D., Rubiolo P., J Chromatogr A, 1150 (2007), 1–2, 37‐
49
[15] Ochiai N., Ieda T., Sasamoto K., Fushimi A., Hasegawa S., Tanabe K., Kobayashi S.,
J Chromatogr A, 1150 (2007), 1–2, 13‐20
[16] Luong J., Gras R., Shellie R.A., Cortes H.J., J Sep Sci, 36 (2013), 1, 182–191
[17] Peroni D., Sampat A.A.S., van Egmond W., de Koning S., Cochran J., Lautamo R.,
Janssen H‐G., J Chromatogr A, 1317 (2013), 3‐11
[18] Mommers J., Knooren J., Mengerink Y., Wilbers A., Vreuls R., van der Wal Sj., J
Chromatogr A, 1218 (2011), 21, 3159‐3165
144
5.6 43B43BSupplementary data
Result of the two‐step dual detection retention time locking procedure for the analysis
of the test mixture 2 by GC×GC TOFMS / FID
TOFMS
Detector
FID
Detector
Name 1tr
(s)
2tr
(s)
1tr
(s)
2tr
(s)
PW1/2
(s)
1tr
(s)
2tr
(ms)
propanenitrile 490 2,493 490 2,553 0,14 0 ‐60
isobutyronitrile 570 2,662 570 2,745 0,20 0 ‐83
4‐Picoline 1410 4,898 1410 4,992 0,19 0 ‐94
2,5‐dimethyl‐pyrrole 1700 5,297 1700 5,345 0,19 0 ‐48
N,N‐dimethyl‐cyclohexanamine 2070 3,739 2070 3,746 0,18 0 ‐7
3‐ethyl‐4‐methyl‐pyridine 2560 6,005 2560 5,909 0,17 0 96
1‐Ethyl‐2‐pyrrolidinone 2570 8,139 2570 8,066 0,18 0 73
quinaldine 3630 7,473 3630 7,405 0,19 0 68
acridine 5620 9,469 5620 9,407 0,21 0 62
phenanthridine 5700 9,725 5700 9,655 0,21 0 70
145
6 5B5BTemperature‐Tunable Selectivity in Comprehensive Two‐
dimensional Gas Chromatography
A temperature‐tunable two‐dimensional gas chromatography setup, consisting of
three capillary columns with different selectivities, is described and evaluated. In this
setup, the selectivity of the primary dimension can be tuned by adjusting the
temperature offset of two in series‐coupled capillary columns, both columns being
part of the primary dimension and positioned in two separate GC ovens. The overall
GC×GC separation can be optimized by altering the selectivity of the primary
dimension. Besides tuning selectivity, in order to achieve optimal separation, this 2D‐
GC setup also offers enhanced opportunities for qualitative analysis. Sequentially
altering the selectivity of the primary dimension enables one to identify groups of
compounds which show similar chromatographic retention behavior.
6.1 44B44BIntroduction
Comprehensive two‐dimensional gas chromatography (GC×GC), introduced in 1991 by
Liu and Phillips [1], is a powerful analytical technique providing structured
chromatograms and high separation power, making this technique ideal for analyzing
complex samples. Despite the high separation power of this technique, coelutions of
target analytes still may occur.
In this chapter we describe a temperature‐tunable two dimensional gas
chromatography setup in which the column selectivity of the primary dimension can
be altered for solving critical coelutions of target analytes. In our approach, the
primary dimension consists of two columns coupled in series in which the second
column is positioned in a separate GC oven. This second in series‐coupled primary
column should have different retention characteristics, in terms of selectivity and
retention mechanism, compared to the other two columns. The contribution of the
second in series‐coupled primary column, which is positioned in the second oven, can
be altered by adjusting the temperature offset of this oven compared to the main oven
and thereby the overall selectivity of the primary dimension can be tuned. The
contribution of the second in series‐coupled primary column to the overall first
146
dimension retention time can be increased by lowering the temperature offset and
can be decreased or minimized by increasing the temperature offset.
Tuning of column selectivity for one dimensional gas chromatography has extensively
been described in literature. In most publications column selectivity is tuned by means
of controlling the pressure at the junction point of two different in series‐coupled
capillary columns [2‐14]. This approach, however, requires a T‐connector and an extra
carrier gas supply. Another approach for tuning selectivity is optimizing the
temperatures for series‐coupled columns in dual‐oven gas chromatographic systems
[15‐18]. Column temperature has a significant effect on selectivity, especially for polar
phases and compounds.
To our knowledge, tuning strategies for comprehensive two‐dimensional gas
chromatography, based on temperature or pressure tuning the selectivity of the
primary dimension by using series‐coupled columns has not been described in the
literature yet.
Besides optimizing the resolution by tuning the selectivity of the primary dimension,
our tunable 2D‐GC setup also offers enhanced opportunities for qualitative analysis.
Altering stepwise the contribution of the second in‐series coupled primary column,
which has different retention characteristics compared to the first in series‐coupled
primary column and second dimension column, enables the opportunity to identify
groups of compounds which show similar chromatographic retention behavior (similar
interactions) with the stationary phase of the second in series‐coupled primary
column.
6.2 45B45BExperimental
6.2.1 91B91BTest mixtures and samples
All experiments were carried out with a homemade test mixture containing 38
compounds and two different industrial plant samples. The test mixture contained 500
µL of each of the following compounds: diethylketone, 3,3‐dimethyl‐2‐butanone, 2,4‐
147
dimethyl‐3‐pentanone, 3‐methyl‐2‐pentanol, 3‐methyl‐2‐pentanone, 3‐methyl‐2‐
butanone, 2‐methoxyethanol, 2‐pentanone, methyl‐cyclohexane, 3‐methoxy‐
propionitrile, isobutyronitrile, propionitrile, 1‐pentanal, 3‐methyl‐2‐butanol, butanal,
1‐hexanol, isobutanol, 1‐butanol, benzene, cyclohexane, 1‐propanol, hexane,
dimethylsulfoxide, ethanol, heptane, n‐butylacetate, chloroform, ethylacetate,
diethylether, 2‐cyano‐ethylether and also 100 µL of each of the following compounds:
allylcyanide, methyl‐cyclopentane, ethyl‐cyclopentane, crotononitrile, 2‐pentanol, 2‐
hexen‐1‐ol, 3‐hexanol, 2‐hexanol.
6.2.2 92B92BInstrumental
All GC×GC‐FID analyses were carried out on a Leco (St. Joseph, MI, USA) GC×GC
system, equipped with a secondary GC‐oven which is positioned inside the main GC‐
oven, an Agilent 7683 autosampler, a hot split/splitless injector and a flame ionization
detector (FID). Instrument control and data processing was performed by Leco
ChromaTOF® software (St. Joseph, MI, USA) version 3.25. For all calculations
Microsoft® Office Excel 2007 (Redmond, Washington, VWA, USA), was used.
A schematic overview of the tunable two‐dimensional gas chromatographic setup is
given in figure 1. In order to be able to tune the overall GC×GC selectivity, all three
capillary columns must have different retention characteristics.
Column 1 is a non‐polar 25m × 0.25mm, 1.2µm film thickness CPSil8CB column,
purchased from Varian (Middelburg, The Netherlands). This column is coupled in
series with a polar 5m × 0.25mm, 0.5µm film thickness Stabilwax®‐DB column (column
2, Restek, Bellefonte, PA, USA). The Stabilwax®‐DB‐column is installed in the second
GC‐oven (oven‐2). The other end of this column is coupled to a 2m × 0.1mm, 0.08µm
film thickness polar ionic liquid SLB®‐IL59 column (column 3), purchased from Sigma‐
Aldrich (St. Louis, MO, USA). The ionic liquid column is installed in oven‐1 and passes
through the modulator. All capillary column connections were made by using SilTite™
μ‐Unions, purchased from SGE Europe (United Kingdom). When performing two‐
dimensional gas chromatography, the in series coupled columns 1 and 2 make up the
tunable primary dimension and column 3 the secondary dimension. Due to
148
temperature controlling restrictions of this GC×GC system, oven‐2 must be
programmed at least 5C higher than the temperature of oven‐1.
Figure 1 Schematic overview of the tunable two‐dimensional gas chromatography set‐
up
For all experiments, oven‐1 was held for 1 minute at 50C and next programmed to
180C. Oven‐2 was programmed having a positive temperature offset, for both the
initial and final oven temperature, of 5, 10, 15, 20, 25, 30, 35 or 40C compared to
oven‐1. The temperature rates, for both ovens, were 3Cmin‐1. A modulation time of
3.5 seconds was used. All separations were carried out using a constant helium flow
of 1.2 ml/min. The injector temperature was 280C, using a split ratio of 1:50 and an
injection volume of 1µL. The FID was operated at a temperature of 300C, using a data‐
acquisition rate of 200 Hz.
6.3 46B46BResults and discussions
6.3.1 93B93BInfluence of 2oven temperature offset on primary dimension separation
In figure 2, part of the primary dimension 1D‐chromatograms of the test mixture,
measured at different oven‐2 temperature offsets, are given. The lower the oven‐2
149
temperature offset the more the Stabilwax®‐DB column will contribute to the primary
dimension separation. In figure 2 it can clearly be seen that the most polar compounds
1, 3, 4, 7, 9 and 10, shift to higher retention times when lowering the oven‐2
temperature offset. The relation between the oven‐2 temperature offset and the
retention of the polar compounds can be described by a quadratic function. In this
example, the primary dimension separation can be tuned by adjusting the
temperature offset to 10C, in order to achieve a full baseline separation for all 10
compounds.
Figure 2 Influence of oven‐2 temperature offset on the primary dimension separation;
analysis of a test mixture at different oven‐2 temperature offsets (1=ethanol; 2=ethyl
ether; 3=1‐propanol; 4=propanenitrile; 5=hexane; 6=ethylacetate; 7=chloroform;
8=cyclohexane; 9=isobutyronitrile; 10=2‐methyl‐1‐propanol
6.3.2 94B94BInfluence of the temperature of 2oven offset on two‐dimension separation
In figure 3 an overlay is given of the two‐dimensional chromatograms of the analysis
of the test mixture measured at different oven‐2 temperature offsets. The highlighted
and colored peaks are measured with an oven‐2 temperature offset of 40C, the
150
dark/grey peaks are measured sequentially with an oven‐2 temperature offset of 35,
30, 25, 20, 15, 10 and 5C.
In figure 3 it clearly can be seen that the influence of lowering the oven‐2 temperature
offset is not the same for all compounds. The compounds in box A, all nonpolar
compounds, do not shift when lowering the offset temperature indicating that these
nonpolar compounds have zero affinity with the polar Stabilwax®‐DB column. The
compounds in the boxes B have significant retention on the polar ionic liquid SLB®‐
IL59 column but almost no retention on the polar Stabilwax®‐DB column. However,
the compounds in the boxes C have significant retention on the polar ionic liquid SLB®‐
IL59 column and also significant retention on the polar Stabilwax®‐DB column. The
compounds in the boxes C are not susceptible to hydrogen bonding with the
polyethylene glycol stationary phase of the Stabilwax®‐DB column. The compounds in
the boxes B however are susceptible to hydrogen bonding. This result clearly
demonstrates
the different retention characteristics of the polar Stabilwax®‐DB‐column and the
polar ionic liquid SLB®‐IL59 column: compounds with similar retention on the ionic
liquid SLB®‐IL59 column, for example peak B` (3‐methyl‐2‐pentanone) and C` (1‐
pentanol) in figure 3, show different retention behavior on the Stabilwax®‐DB column.
So, by changing the oven‐2 temperature offset, the overall selectivity of the primary
dimension column can be altered. Partly or fully coeluting peaks, which have adequate
different selectivity’s on the Stabilwax®‐DB‐column and the polar ionic liquid SLB®‐
IL59 column may be separated simply by tuning the oven‐2 temperature offset.
151
Figure 3 Influence of oven‐2 temperature offset on the two‐dimensional separation;
analysis of a test mixture at different oven‐2 temperature offsets; highlighted colored
peaks are measured with a positive oven‐2 offset of 40C, colorless peaks are
sequentially measured with a positive offset of 35, 30, 25, 20, 15, 10 and 5C
In figure 4 the second dimension retention times of the test mix compounds are
plotted against the primary retention time differences measured when analyzing the
test mix applying a temperature oven‐2 offset of 40C and 5C. These measured
retention time differences are related to the retention of the compounds on the polar
Stabilwax®‐DB column (column‐2).
In this plot three different groups, A, B, and C can be identified. The compounds in
group A (e.g. cyclohexane, hexane, heptane, diethylether, methylcyclopentane,
ethylcyclopentane) show almost no retention on the second dimension column and
their retention is not influenced by altering the temperature offset of the polar
Stabilwax®‐DB column (column‐2). Compounds in group A are all nonpolar compounds
152
showing low retention on both polar columns. The compounds in group B (e.g. diethyl
ketone, 3,3‐dimethyl‐2‐butanone, 2,4‐dimethyl‐3‐pentanone, 3‐methyl‐2‐pentanone,
3‐methyl‐2‐butanone, 2‐pentanone, 1‐pentanal, butanal) show significant retention
on the second dimension column and low retention on the polar Stabilwax®‐DB
column (column‐2). All these compounds are significantly polar however are not able
to undergo strong hydrogen bonding with the stationary polyethylene glycol phase of
the polar Stabilwax®‐DB column. The compounds in group C (3‐methyl‐2‐pentanol, 2‐
methoxyethanol, 3‐methyl‐2‐butanol, 1‐hexanol, isobutanol, 1‐butanol, 1‐propanol,
ethanol, 2‐pentanol, 2‐hexen‐1‐ol, 3‐hexanol, 2‐hexanol) show significant retention
on the polar Stabilwax®‐DB and the ionic liquid SLB®‐IL59 column. All these
compounds are polar and able to undergo hydrogen bonding with the stationary
phase of the polar Stabilwax®‐DB column.
This example clearly demonstrates that the selectivity of the primary dimension in this
GC×GC setup can be tuned and also that this tunable GC×GC setup can be used for the
identification of compound groups which have different retention behavior on both
the Stabilwax®‐DB and the ionic liquid SLB®‐IL59 column.
153
Figure 4 Analysis of test mixture by tunable GC×GC applying an oven‐2 temperature
offset of 40C and 5C; second dimension retention times plotted against the delta
primary retention time. Group A are nonpolar compounds; group B are polar
compounds, not able to undergo hydrogen bonding; group C are polar compounds,
able to undergo hydrogen bonding
6.3.3 95B95BAnalysis of real‐life industrial plant samples
In figure 5, part of the two‐dimensional chromatogram the industrial plant sample A,
measured at different oven‐2 offset temperatures, is given. A few coelutions or critical
separations, measured with an oven‐2 temperature offset of 40C, are indicated by
the yellow circles. When lowering the temperature offset, so increasing the
contribution of the Stabilwax®‐DB column, the separation of the critical pairs
improves. For the critical pairs in the yellow circle B, it can clearly be seen than one
peak shows relatively more retention on the Stabilwax®‐DB column compared to the
other peaks, leading to a full separation at lower oven‐2 temperature offsets. The
critical separations indicated in the yellow circles A and C are coelutions of compounds
which are wrapped‐around with peaks which are non‐polar and not wrapped‐around.
154
At lower oven‐2 temperature offsets the wrapped‐around compounds have more
retention on the Stabilwax®‐DB column, compared to the non‐wrapped‐around peaks.
Lowering the oven‐2 temperature offset has no or little effect on the non‐polar
compounds, while the more polar coeluting compounds show more retention on the
Stabilwax®‐DB column.
Of course, in complex chromatograms concurrently a decrease of resolution of other
compounds will occur, so careful optimization is required. Temperature‐tuning GC×GC
thus will offer an alternative and convenient screening‐strategy for very complex
samples.
155
Figure 5 Analysis of the industrial plant sample B by tunable GC×GC applying an oven‐
2 temperature offset of 40C, 30C, 20C, 10C and 5C; Yellow circles A, B and C
indicate coelutions or critical separations
156
In figure 6, part of the two‐dimensional chromatogram of the industrial plant sample
B, measured at an oven‐2 offset temperature of 40C and 10C, is given. At an offset
of 40C, the low intensity peaks A and B coelute with the large intensity peak C. When
lowering the temperature offset to 10C, peak A and B can be fully separated from
peak C. Compounds A and B undergo more retention on the Stabilwax®‐DB column
than compound C.
Figure 6 Analysis of the industrial plant sample B by tunable GC×GC applying an oven‐
2 temperature offset of 40C and 10C; A, B and C indicate the target analytes
157
6.4 47B47BConclusions and discussions
By using the temperature‐tunable two dimensional gas chromatography setup
described in this paper it is possible to tune the selectivity of the first dimension
separation. In this setup, the primary dimension exists of two in series‐coupled
columns, a nonpolar CPSil8CB and a polar Stabilwax®‐DB column, in which the latter
is positioned in a separate GC‐oven and a polar ionic liquid SLB®‐IL59 as the second
dimension column. By altering the temperature offset of the separate GC‐oven the
contribution of the second, primary dimension, column (Stabilwax®‐DB) can be
changed and thereby changing the overall primary dimension selectivity.
It is essential that all three columns have different retention characteristics in order
to optimize the overall two‐dimensional separation. However, the type of stationary
phase and also the column dimensions of the in series‐coupled second primary
dimension column should be selected carefully; the contribution of this second
column to the overall GC×GC separation may not lead to a significant decrease in
orthogonality.
Moreover, besides optimizing the resolution by tuning the selectivity of the overall
primary dimension, this temperature tunable GC×GC setup also offers enhanced
opportunities for qualitative analysis. By changing stepwise the contribution of the
second in series‐coupled primary column (Stabilwax®‐DB), simply by changing the
oven‐2 temperature offset, groups of compounds which have different retention
behavior on a polar Stabilwax®‐DB and a polar ionic liquid SLB®‐IL59 can be identified.
In fact, in this setup the extra column can be seen as a virtual third dimension. For
example, it was demonstrated that polar compounds which are susceptible to
hydrogen bonding, for instance alcohols, are relatively more retained when lowering
the oven‐2 temperature offset, so thereby increasing the Stabilwax®‐DB retention
contribution, than polar compounds which are not susceptible to strong hydrogen
bonding for example ketones.
158
Besides the described temperature‐tunable GC×GC setup also other related
configurations could be applied for tuning the overall two‐dimensional selectivity. A
similar approach could be used to tune the selectivity of the secondary dimension. For
this, two in series‐coupled secondary columns having similar internal diameters and
in which one of the two columns is positioned in a separate GC oven, the overall
selectivity of the second dimension could be tuned by optimizing the temperature
offset. However, the main disadvantage of this setup is the fact that the extra
secondary dimension column, including a required column connector, which are
positioned after the GC×GC modulator, would lead to extra band broadening and also
to a higher, in general less favorable, midpoint pressure.
6.5 48B48BReferences
[1] Liu, Z., Phillips, J. B., J. Chromatogr. Sci., 29 (1991), 227‐233.
[2] Maštovská, K., Lehotay, S.J., J Chromatogr A, 1000 (2003), 1‐2, 153‐180.
[3] Smith, H., Sacks, R., J Sep Sci, 25 (2002), 1‐2, 37‐44.
[4] McGuigan, M., Sacks, R., Anal Chem, 73 (2001), 13, 3112‐3118.
[5] Sacks, R., Coutant, C., Veriotti, T., Grail, A., J Sep Sci, 23 (2000), 3, 225‐234.
[6] Grall, A.J., Sacks, R.D., Anal Chem, 72 (2000), 11, 2507‐2513.
[7] Sacks, R., Coutant, C., Veriotti, T., Grall, A., J High Res Chromatog, 23 (2000), 3, 225‐
234.
[8] Leonard, C., Sacks, R., Anal Chem, 71 (1999), 22, 5177‐5184.
[9] Smith, H., Sacks, R., Anal Chem, 69 (1997), 24, 5159‐5164.
[10] Akard, M., Sacks, R., Anal Chem, 68 (1996), 9, 1474‐1479.
[11] Akard, M., Sacks, R., Anal Chem, 66 (1994), 19, 3036‐3041.
[12] Sacks, R., Akard, M., Environ Sci Technol, 28 (1994), 9, 428A‐433A.
[13] Engewald, W., Maurer, T., J Chromatogr, 520 (1990), 3‐13.
[14] Maurer, T., Engewald, W., Steinborn, A., J Chromatogr, 517 (1990), 77‐86.
[15] Sandra P., David F., Proot M., Diricks G., Verstappe M., Verzele M., J High Res
Chromatog, 8 (1985), 782.
[16] Benická E., Krupčík J., Répka D., Sandra P., Chromatographia, 33 (1992), 9‐10, 463‐
466.
159
[17] Repka D., Krupčík J., Benická E., Maurer T., Engewald W., J High Res Chromatog,
13 (1990), 5, 333‐337.
[18] Repka D., Krupčík J., Benická E., J Chromatogr, 463 (1989), 243‐251.
160
161
7 6B6BTunable secondary dimension selectivity in comprehensive two‐
dimensional gas chromatography
In this chapter two tunable two‐dimensional gas chromatography setups are
compared and described in which the secondary dimension consists of two different
capillary columns coupled in series. In the first setup, the selectivity of the second
dimension can be tuned by adjusting the effective column length of the first secondary
dimension column, simply by sliding it stepwise back or forward through the GC×GC
modulator. In the second setup, in which the first secondary dimension column is
installed in a separate GC‐oven (oven‐2), the overall selectivity of the second
dimension can be tuned by adjusting the oven‐2 temperature offset with respect to
the main oven. The contribution of the first secondary dimension column to the
overall secondary dimension separation can be decreased by applying a higher
temperature offset. A real‐life sample, the headspace of a coffee powder, was used to
demonstrate the added value of tunable GC×GC by solving coelutions of some specific
aroma compounds. Besides optimizing the overall GC×GC separation, by altering the
second dimension column selectivity, these set‐ups also offer enhanced possibilities
for qualitative analysis. By stepwise altering the selectivity of the second dimension,
classes of compounds showing similar retention behavior could be discriminated.
7.1 49B49BIntroduction
GC×GC method optimization is far more complicated compared to 1D‐GC. An
important reason for this complexity is the fact that the two columns are linked and
that changes in, for instance the column dimensions, flow or temperature of the first
dimension affects both the primary and secondary dimension separation. Besides the
difficult choice of the most optimal stationary phases and column dimensions for both
the primary and secondary columns, also the modulator parameters, carrier flow,
oven temperature and detector settings have to be optimized in order to optimize
overall separation power and sensitivity [1–4]. Despite the high separation power of
GC×GC, coelutions of target analytes or coelutions of target groups, in case of group‐
type analysis, still may occur especially for truly complex samples.
162
In this chapter, we describe two different tunable two‐dimensional gas
chromatography setups in which the overall selectivity of the second dimension can
be tuned. Coelutions of target analytes which occur in the second dimension
separation could be solved by applying these setups. In our tunable GC×GC setups
three different types of capillary GC columns, different in terms of selectivity and
retention mechanisms, are used in which the second dimension consists of two
columns coupled in series. The contribution of the first in series‐coupled second
dimension column can be altered in two different ways; in the first setup by altering
its effective column length, simply by sliding it stepwise back or forward through the
modulator; in the second setup, in which the first second dimension column is
installed in a separate GC‐oven (oven‐2), by altering the oven‐2 temperature offset
with respect to the temperature of the main oven. So, by adjusting the contribution
of the first second dimension column, the overall selectivity of the second dimension
can be altered or tuned.
Tuning column selectivity, for one‐dimensional gas chromatography, by using two
different in series‐coupled columns has already been described in the literature. In
most cases the selectivity can be altered by adjusting the pressure at the mid‐point of
the two in series‐coupled columns by using a T‐connector and an extra carrier gas
supply [6–10]. Also, tuning of selectivity by optimizing the individual column
temperatures for series‐coupled columns in dual‐oven one‐dimensional gas
chromatographic systems has been described [11–14].
To the best of our knowledge, tuning of comprehensive two‐dimensional gas
chromatography separations, based on optimizing the overall second dimension
selectivity of two in series‐coupled columns by altering the effective lengths or the
temperature offset of one of the two in series coupled second dimension columns has
not been described in the literature. Tuning of GC×GC separations by optimizing the
primary dimension selectivity of two in series‐coupled columns by altering the
temperature offset of one of the two columns has been published in literature by the
authors of this paper [5].
163
7.2 50B50BExperimental
7.2.1 96B96BTest mixtures and real‐life samples
A homemade test mixture, containing 60 compounds, was used for all experiments.
For analysis, the test mix was diluted 50 times in acetonitrile. The test mixture
contained 25 μL of each of the following compounds: n‐hexane, n‐heptane, n‐octane,
n‐nonane, n‐decane, 2‐methylbutane, 2‐methylpentane, 2,3‐dimethylbutane, 2,3,4‐
trimethylpentane, 1‐hexene, 1‐heptene, 1‐octene, methanol, ethanol, 1‐propanol, 1‐
butanol, 1‐pentanol, 1‐hexanol, 1‐heptanol, isobutanol, 2‐heptanol, 3‐methyl‐1‐
butanol, 3‐methyl‐2‐butanol, 2‐methyl‐2‐butanol, propionaldehyde, heptanal,
nonanal, decanal, acetone, 2‐butanone, 2‐pentanone, 2‐hexanone, formic acid, acetic
acid, propionic acid, pentanoic acid, benzyl acetate, ethyl acetate, ethyl propionate,
butyl acetate, dipropylether, t‐butylmethylether, diethylether, n‐butyronitrile,
valeronitrile, propanenitrile, heptanenitrile, octanenitrile, toluene, o‐xylene, m‐
xylene, benzene, butylamine, hexylamine, diethylamine, dibutylamine, triethylamine,
cyclohexane, methylcyclohexane, 1,3‐dimethylcyclohexane. Coffee powder (Senseo®)
was used as a real‐life sample for demonstrating the use of tunable GC×GC.
7.2.2 97B97BInstrumental and materials
All GC×GC‐FID and GC×GC‐TOFMS analyses were carried out on a Leco Pegasus 4D
GC×GC system (St. Joseph, MI, USA), equipped with a secondary GC‐oven (oven‐2)
which was positioned inside the main GC‐oven, a Combipal autosampler (CTC
Analytics AG, Switzerland) equipped with a solid phase micro extraction (SPME)
option, a hot split/splitless injector, a time‐of‐flight mass spectrometer (ToFMS) and a
flame ionization detector (FID). Instrument control and data processing was
performed by Leco ChromaTOF® software (St. Joseph, MI, USA) version 3.25. For all
calculations Microsoft® Excel version Office Professional Plus 2010 (Redmond,
Washington, VWA, USA), was used. All tuning experiments were carried out using the
FID. The TOFMS was only used for identifying compounds.
A schematic overview of the two different, tunable GC×GC setups, A and B, are given
in Figure 1. Column 1, the primary dimension column, is a nonpolar 30 m × 0.25 mm,
164
1 μm film thickness VF1ms column, purchased from Agilent Technologies (Santa Clara,
CA, USA). The second dimension consists of two columns which are coupled in series;
column 2 is a polar 1 m × 0.1 mm, 0.1 μm film thickness Wax‐HT® column (Mega,
Legnano, MI, Italy) and column 3 is a medium polar 2 m × 0.1 mm, 0.2 μm film
thickness VF17ms column (Agilent Technologies, Santa Clara, CA, USA). All column
connections were made by using SilTite™ μ‐Unions, purchased from SGE Europe Ltd
(Buckinghamshire, United Kingdom).
Figure 1 Schematic overview of the two tunable two‐dimensional gas chromatography
setups.
In setup A, all three columns are installed in the main GC‐oven. Column 2, the Wax‐
HT®, passes through the modulator. The sum of the length of both columns 2 and 3,
positioned after the last cold jet of the modulator can be defined as the effective
secondary column length; only this part of the column will contribute to the secondary
dimension separation. In figure A1, column 2 is fully positioned in front of the
modulator so column 2 does not contribute to the second dimension separation; the
second dimension separation will be fully based on interactions of the compounds
with the VF17ms stationary phase. In figure A2, column 2 is now fully positioned after
the modulator so column 2 does fully contribute to the second dimension separation;
in this case the separation will be based on interactions of the compounds with both
the VF17ms and the Wax‐HT® stationary phases. Tuning of secondary dimension
165
separation can be performed by sliding the column 2 stepwise back or forwards
through the modulator decreasing or increasing its contribution to the overall
selectivity. Sliding the column through the modulator is straightforward and easy to
perform. In order to determine the final effective column length with precision, the
column is marked every 5 cm with a heat resistant marker.
In setup B, column 2, the Wax‐HT® column is installed in a separate GC‐oven (oven‐2),
the other two columns are positioned in the main GC oven. The contribution of column
2 can be altered by adjusting the temperature offset of oven‐2 with respect to the
temperature of the main GC oven.
In setup C, only column 1 (30 m × 0.25 mm, 1 μm film thickness VF1ms) and column 2
(2 m × 0.1 mm, 0.1 μm film thickness Wax‐HT®) are used. The front part of column 2
(1 m) is positioned in the Oven‐2 and the rear part (1 m) is positioned in the main GC
oven. Setup C is used to demonstrate the added value of adding a third column, with
different selectivity, to a GC×GC setup by comparing the test mixture results obtained
with setups B and C.
7.2.3 98B98BChromatographic conditions
For all the GC×GC experiments, the oven was held for 2 min at 50°C and next
programmed to 280°C and held for 1 min. The temperature rate was 3°C min−1. A
modulation time of 5 s was used. Helium was used as the carrier gas. All analyses were
carried out using a constant flow of 1.5 ml/min. The injector temperature was 250°C,
using a split ratio of 1:50 and an injection volume of 1 μL. The FID was operated at a
temperature of 280°C, using a data‐acquisition rate of 200 Hz. For the setup B
experiments, a positive temperature offset (5 to 40°C in steps of 5°C) with respect to
the main oven temperature program was used. The main oven temperature was
programmed to 260°C.
For the headspace analysis of the coffee powder sample by SPME‐GC×GC a Supelco
(Bellefonte, PA, VS) PDMS/DVB, 65 μm thickness, SPME fiber was used. Approximately
5 g of the coffee powder sample was transferred into a headspace vial and capped.
The SPME headspace extraction was performed during 15 min at 50°C. The fiber was
166
desorbed in the GC×GC injector at a temperature of 270°C for 1 min under splitless
conditions. The modulation time was set to 7 s.
7.3 51B51BResults and discussions
7.3.1 99B99BResults of the tunable GC×GC setup A
Influence of the effective secondary dimension column length on retention times
As expected, when increasing the effective secondary length of column 2 all peaks
shift to higher secondary retention times due to the fact that the total secondary
column length increases as does the dead volume time. In Figure 2, the two‐
dimensional chromatograms of the test mixture, measured at 0 and 90 cm effective
lengths of column 2 (Wax‐HT® column) are given, corrected for the dead volume time
increase. In order to compensate for dead volume increase when increasing the
effective secondary column length, each chromatogram was vertically shifted in the
secondary dimension by its increase of the dead volume time as measured for octane.
167
Figure 2. Results of the tunable GC×GC setup A; two‐dimensional chromatograms of
the test mixture, measured at 0 and 90 cm effective secondary lengths of column 2
(Wax‐HT® column). Data are corrected for the second dimension dead time increase.
Peak identifications: #1 = 3‐methyl‐1‐butanol), #2 = 1‐pentanol, #3 = toluene, #4 = 2‐
hexanone, #5 = valeronitrile, #6 = butyl acetate, and #7 = methylcyclohexane.
These results clearly show that the selectivity of the second dimension can be altered
simply by adjusting the effective length of the first secondary dimension column. For
example, peak 1 (3‐methyl‐1‐butanol), 2 (1‐pentanol) and 5 (valeronitrile), shift
significantly more than peak 3 (toluene), 4 (2‐hexanone) and 6 (butyl acetate), when
increasing the effective secondary column length of column 2 from 0 cm to 90 cm;
even the second dimension elution order of peak 1 and 2 compared to peak 4 is
reversed. Peak 1 and 2 are alcohols and peak 5 is a nitrile; alcohols and nitriles can
Prim. dimension (s)
Sec. dim
ension (s)
168
undergo strong hydrogen‐bonding with the stationary phase of column‐2 (Wax‐HT®
column). So, increasing the effective length of the Wax‐HT® relatively increases the
second dimension retention of compounds which can undergo hydrogen bonding.
Additionally, it is observed that the primary retention times do not shift when varying
the effective secondary column length. This can be explained by the fact that the sum
of the lengths of columns 2 and 3 are for all experiments constant, thus the pressure‐
drop across the primary column 1 and therefore also the primary retention times, are
constant. This also proves that the residence time of the column 2 segment upstream
of the modulator is negligible.
For all different compounds in the test mixture an approximately linear function is
observed for the secondary retention time increase versus the effective length of
column 2. The increase is different for different types of compounds and is also related
to the secondary retention time.
In Figure 3, the secondary retention times of the test mixture compounds, measured
at an effective column‐2 (Wax‐HT®) length of 0 cm, are plotted versus the secondary
retention time measured at an effective column‐2 length of 90 cm. The dashed line in
Figure 3 represents the theoretical situation when the influence of altering the
effective secondary length of the first secondary column is zero. Also, the linear plot
of the alkanes is extrapolated. For the nonpolar alkanes, it is expected that the
interaction with the stationary phase of the polar Wax‐HT® column (column‐2) is
virtually zero; so the increase of the second dimension retention times of the alkanes,
when increasing the effective length of the Wax‐HT® column, could be fully attributed
to the corresponding increase of the dead volume times. The increase of the dead
volume time going from 0 cm to 90 cm effective length of the first second dimension
column is approximately 1 s for all alkanes.
169
Figure 3. Secondary retention time of the test mixture compounds measured using an
effective secondary length of column‐2 (Wax‐HT®) of 0 cm versus 90 cm.
Based on Figure 3 and the assumption that the alkanes have no significant interaction
with stationary phase of the Wax‐HT® column, it can be concluded that alkenes, ethers
and cycloalkanes also show no significant interaction with the Wax‐HT® column; all
compounds of these groups are positioned on the linear alkane plot. The medium
polar acetates, ketones and aromatics undergo significant interaction with the Wax‐
HT® and the most interaction is observed for the nitriles and the primary alcohols,
both able to undergo strong hydrogen bonding with the stationary phase of the Wax‐
HT® column. These observations are in good agreement with data published
previously concerning the analysis of homologous series on “Non‐polar × Wax”
column‐sets [15]. In figure 3, all groups showing significant interaction with the Wax‐
HT® column are positioned above the linear alkane plot; each group having a higher
slope compared to the slope of the alkane plot.
Based on these results it is clear that compounds within one chemical group undergo
similar retention behavior on the Wax‐HT® column; all compounds within one
chemical group are approximately positioned on the same linear plot having a certain
170
slope. For example, the 9 primary alcohols: methanol, ethanol, 1‐propanol, 2‐methyl‐
1‐propanol (isobutanol), 1‐butanol, 3‐methyl‐1‐butanol, 1‐pentanol, 1‐hexanol and 1‐
heptanol having a carbon range from C1 to C7 and a primary retention time range
from 5 to 27 min are all approximately positioned on the same linear plot, as given in
figure 3, having an r‐squared value of 0.99.
7.3.2 100B100BResults of the tunable GC×GC setup B
In figure 4, the two‐dimensional chromatograms of the test mixture, measured at a
positive oven‐2 temperature offset of 40 and 5°C, compared to the main oven, are
given. For setup B, the test mixture results, with respect to selectivity, are similar with
the results obtained for setup A. When decreasing the oven‐2 temperature offset the
contribution of the Wax‐HT® column increases and thereby increases the second
dimension retention of compounds which can undergo hydrogen bonding, for
example peak 1 (3‐methyl‐1‐butanol), 2 (1‐pentanol) and 5 (valeronitrile), with the
stationary phase of the Wax‐HT® column.
171
Figure 4. Results of the tunable GC×GC setup B; two‐dimensional chromatograms of
the test mixture, measured at an oven‐2 temperature offset of 40 and 5°C (Wax‐HT®
column is installed in the oven‐2).
In contrast to setup‐A, altering the oven‐2 temperature offset has nearly no effect on
the second dimension dead volume time, given the fact that the effective secondary
column length is kept constant. So, the increase of the second dimension retention
times, when lowering the oven‐2 temperature offset, could be fully attributed to the
interaction of the compounds with the stationary phase of column‐2 (Wax‐HT®). For
all test mixture compounds a quadratic function is observed for the secondary
retention time decrease versus the oven‐2 temperature offset.
Besides this, it is observed that the primary retention times slightly decrease when
lowering the oven‐2 temperature offset. The primary retention time shifts are
Prim. dimension (s)
Sec. dim
ension (s)
172
calculated for each test mixture compound by subtracting its retention time measured
at an oven‐2 offset of 35, 30, 25, 20, 15, 10 or 5°C minus its retention time measured
at an oven‐2 offset of 40°C. A linear function (r2 = 0.999) is observed for the influence
of the oven‐2 temperature offset on the primary retention time shift. This could be
explained by the fact that the viscosity of the carrier gas helium increases when
increasing the oven‐2 temperature. So, when increasing the oven‐2 temperature, the
viscosity of the helium gas increases, therefore the pressure at the GC×GC midpoint
also will increase leading to a lower pressure drop across the primary column resulting
in an increase of the primary retention times. However, it has to be noted that the
observed primary retention time shift is not larger than 5 modulations and no
noticeable primary separation selectivity changes were induced.
In figure 5 the secondary retention times of the test mixture compounds, measured
at an oven‐2 temperature offset of 40°C are plotted versus the secondary retention
times measured at an oven‐2 temperature offset of 5°C. The dashed line in Figure 5
represents the theoretical situation when the influence of altering the oven‐2
temperature offset is zero. Furthermore, the linear plot of the nonpolar alkanes, which
are expected to have virtually no interaction with the stationary phase of the Wax‐
HT® column, is extrapolated. However, as can be observed in Figure 5, the nonpolar
alkanes show a slight retention time increase when lowering the oven‐2 temperature
offset from 40°C to 5°C.
173
Figure 5. Secondary retention time of test mixture compounds measured at an oven‐2
temperature offset of 40°C versus 5°C.
Additionally, it is observed that the influence of the tuning experiments on the second
dimension retention time shifts is larger for setup A than for setup B. The influence of
setup B could be significantly increased by using, for instance a negative temperature
oven‐2 offset, which was not possible with our current GC×GC setup.
As for setup A, it is clear that compounds within one chemical group undergo similar
retention behavior on the Wax‐HT® column. For setup A, the 9 different primary
alcohols having a carbon range from C1 to C7 and a primary retention time range from
5 to 27 min are all approximately positioned on the same linear plot, as given in Figure
5, having an r‐squared value of 0.99.
7.3.3 101B101BResults of the GC×GC setup C
In Figure 6, the two‐dimensional chromatograms of the test mixture, measured at a
positive oven‐2 temperature offset of 40 and 5°C, compared to the main oven, are
given. In contrast to setups A and B, no noticeable differences in selectivity in the
secondary dimension separation are observed for setup C when lowering the oven‐2
174
temperature offset from 40°C to 5°C. The secondary retention time shift for all peaks
is approximately equal and proportional to the relative retention in the second
dimension.
Figure 6. Results of the GC×GC setup C; two‐dimensional chromatograms of the test
mixture, measured at an oven‐2 temperature offset of 40 and 5°C.
In Figure 7 the secondary retention times of the test mixture compounds, measured
at an oven‐2 temperature offset of 40°C are plotted versus the secondary retention
times measured at an oven‐2 temperature offset of 5°C. The dashed line in Figure 7
represents the theoretical situation when the influence of altering the oven‐2
temperature offset is zero. As can be seen in Figure 7, all test mixture compounds from
different chemical classes can be positioned on one linear plot having an r‐squared of
0.99. Based on these results it can be concluded that altering the temperature offset
Prim. dimension (s)
Sec. dim
ension (s)
175
of part of the second dimension column (Wax‐HT®), so without using a different third
column, will not introduce noticeable selectivity differences. In order to alter the
selectivity of the second dimension, a different third column, different than the other
two columns with respect to selectivity, should be used as demonstrated using setup
A or B.
Figure 7. Secondary retention time of test mixture compounds measured at an oven‐2
temperature offset of 40°C versus 5°C.
The main advantage of setup B compared to setup A is the ease of use; for setup B,
tuning by altering the oven‐2 temperature offset can be programmed using the GC×GC
acquisition software, for setup A, tuning must be performed by physically increasing
or decreasing the effective length of the first column of the second dimension.
7.3.4 102B102BReal‐life example
The headspace of a coffee powder sample has been analyzed by SPME‐GC×GC using
setup‐A with an effective secondary column length (Wax‐HT® column) of 0, 20 and
90 cm. In Figure 8 the two‐dimensional chromatograms are given. The chromatograms
176
are corrected for the second dimension dead time increase when increasing the
effective secondary column length.
Figure 8. Two‐dimensional chromatograms of the headspace of a coffee powder
sample, measured at 0, 20 and 90 cm effective secondary lengths of column 2 (Wax‐
HT® column). Data are corrected for the second dimension dead time increase.
Sec. dim
ension (s)
Prim. dimension (s)
177
This real‐life example clearly demonstrates the added value of the tunable‐GC×GC. For
example the coelution of the two aroma compounds 2‐methyl‐3‐hydroxy‐4‐pyrone
(maltol) and 1‐methylpyrrole‐2‐carboxaldehyde (see Figure 8, number 1), observed at
an effective secondary column length of 0 cm is solved when increasing the effective
column length to 20 or 40 cm. In this case the 1‐methylpyrrole‐2‐carboxaldehyde
undergoes more interaction with the stationary phase of the WAX‐HT® column and is
therefore more retained when increasing the effective length than the 2‐methyl‐3‐
hydroxy‐4‐pyrone.
However, coelutions also may be introduced when altering the selectivity of the
second dimension. For example the two aroma compounds 4‐hydroxy‐2,5‐dimethyl‐
3(2H)‐furanone (Furaneol) and 2‐acetylpyrrole (see Figure 8, number 2) are baseline
separated using an effective secondary column length of 0 cm and show full coelution
when increasing the effective secondary length to 40 cm. In this case the Furaneol
undergoes more interaction with the WAX‐HT® stationary phase and is therefore more
retained when increasing the effective length of column 2, compared to 2‐
acetylpyrrole.
Besides tuning the selectivity in order to optimize separation or solve coelutions, this
setup also can be used to discriminate different types of chemical compounds or
chemical classes in order to support identification. In this setup compounds which are
able to undergo strong hydrogen bonding with the stationary phase of the Wax‐HT®
column can easily be discriminated from compounds which show no or significantly
less interaction with the Wax‐HT® column. For example, peak 3 (3‐pyrrole
carboxaldehyde), show significantly more retention when increasing the effective
secondary column length of the Wax‐HT® column, than peak 4 (2‐furanmethyl
acetate), which has similar retention times, indicating that peak 3 is able to undergo
strong hydrogen bonding with the stationary phase of the Wax‐HT® column.
7.4 52B52BConclusions and discussions
The two‐dimensional gas chromatography setups A and B, in which the secondary
dimension consists of two different capillary columns coupled in series, can both be
used to alter or tune the overall column selectivity of the second dimension.
178
In setup A, the selectivity of the second dimension can be tuned by adjusting the
effective column length of the first secondary dimension column, simply by sliding it
stepwise back or forward through the GC×GC modulator. In setup B, in which the first
secondary dimension column is installed in a separate GC‐oven (oven‐2), the overall
selectivity of the second dimension can be tuned by adjusting the oven‐2 temperature
offset with respect to the main oven. Increasing the effective secondary column length
of column 2 or lowering the temperature offset of column‐2 compared to the main
oven lead to similar results with respect to the second dimension selectivity changes.
For both setups, the influence of tuning on the second dimension retention times is
the most for compounds which are able to undergo strong hydrogen bonding with the
stationary phase of the Wax‐HT® column.
Both setups also show that compounds within one chemical group undergo similar
retention behavior on the Wax‐HT® column; all compounds within one chemical group
are approximately positioned on the same linear plot of second dimension retention
time measured at an effective length of 0 cm or an oven‐2 temperature offset of 5°C
plotted versus second dimension retention time measured at an effective length of
40 cm or an oven‐2 temperature offset of 40°C, respectively. Analogous to the
structure in two‐dimensional GC×GC chromatograms tunable two‐dimensional GC×GC
leads additionally to structured‐difference GC×GC chromatograms. To conclude, this
approach could be used to support the identification or discrimination of different
chemical groups which show different retention behavior.
For setup A, increasing the secondary effective length of column‐2 also leads to a
proportional increase of the secondary dimension dead volume time. Altering the
second dimension effective length has no significant effect on the primary retention
times due to the fact that the total secondary column length is constant. For setup B,
changing the oven‐2 temperature offset of column‐2 from 40°C to 5°C has a minor
effect on the primary dimension retention times (this effect is not larger than 5
modulations) which all slightly decrease. This influence could be explained by the fact
that the viscosity of the carrier gas helium increases when increasing the oven‐2
temperature offset.
Future research will mainly focus on the use of tunable GC×GC to support the
identification or discrimination of different chemical groups (or homologous series) by
179
developing chemometric data processing tools. When altering the selectivity of one of
both GC×GC dimensions, these chemometric tools should be able to predict for each
individual compound, known or unknown, its primary and secondary retention times
but also its retention time shift induced by the change of the GC×GC selectivity. This
retention time shift corresponds to the contribution of the extra third column to the
overall selectivity; this retention time shift could be defined as a “pseudo third
dimension” retention time. The ease of determining the retention time shifts
accurately for each individual compound, by correlating two or preferably more
GC×GC chromatograms, will strongly depend on the complexity of the GC×GC
chromatograms.
7.5 53B53BReferences
[1] A. Mostafa, M. Edwards, T. Górecki, J. Chromatogr. A, 1255 (2012), pp. 38‐55
[2] J. Dallüge, R.J.J. Vreuls, J. Beens, U.A.Th. Brinkman, J. Sep. Sci., 25 (2002), pp. 201‐
214
[3] K. Banerjee, S.H. Patil, S. Dasgupta, D.P. Oulkar, S.B. Patil, R. Savant, P.G. Adsule, J.
Chromatogr. A, 1190 (1/2) (2008), pp. 350‐357
[4] J. Omar, I. Alonso, M. Olivares, A. Vallejo, N. Etxebarria, Talanta, 88 (2012), pp. 145‐
151
[5] J. Mommers, J. Knooren, T. Dutriez, S. van der Wal, J. Chromatogr. A, 1270, (2012),
pp. 305‐309
[6] K. Maštovská, S.J. Lehotay, J. Chromatogr. A, 1000 (1/2) (2003), pp. 153‐180
[7] M. McGuigan, R. Sacks, Anal. Chem., 73 (13) (2001), pp. 3112‐3118
[8] A.J. Grall, R.D. Sacks, Anal. Chem., 72 (11) (2000), pp. 2507‐2513
[9] W. Engewald, T. Maurer, J. Chromatogr., 520 (1990), pp. 3‐13
[10] T. Maurer, W. Engewald, A. Steinborn, J. Chromatogr., 517 (1990), pp. 77‐86
[11] P. Sandra, F. David, M. Proot, G. Diricks, M. Verstappe, M. Verzele
J. High Resolut. Chromatogr., 8 (1985), p. 782
[12] E. Benická, J. Krupčík, D. Répka, P. Sandra, Chromatographia, 33 (9/10) (1992), pp.
463‐466
180
[13] D. Repka, J. Krupčík, E. Benická, T. Maurer, W. Engewald, J. High Resolut.
Chromatogr., 13 (5) (1990), pp. 333‐337
[14] D. Repka, J. Krupčík, E. Benická, J. Chromatogr., 463 (1989), pp. 243‐251
[15] J.V. Seeley, E.M. Libby, K.A. Hill Edwards, S.K. Seeley, J. Chromatogr. A, 1216 (10)
(2009), pp. 1650‐1657
181
8 7B7BSummary
Method Development for Comprehensive Two‐Dimensional Gas Chromatography
Comprehensive two‐dimensional gas chromatography (GC×GC) is a powerful
analytical technique capable of proving high separation power, enhanced sensitivity,
and structured chromatograms. Although, GC×GC technology is becoming more
mature, more accessible, and increasingly adopted by analytical chemists, GC×GC
method development is significantly more difficult than for 1D‐GC. For GC×GC more
method development choices need to be made and optimization is difficult due to a
complex interplay of the many 1D and 2D parameters. Method development is also
restricted by the modulation criterion and column‐set temperature limitations. Also,
the usual difference in the primary and secondary internal diameters prevent that
both columns can be operated at their optimal linear gas velocities (flow mismatch
issue) and the usually smaller secondary internal diameter may often lead to mass
loadability issues. If that’s not all, peaks in GC×GC may shift in two directions which
causes alignment issues which are less straightforward to solve compared to 1D‐GC
retention time shifts.
In order to make optimal use of GC×GC, a systematic approach of method
development is essential. The choices should be based on a thorough understanding
of the main analytical question (and its requirements) and knowledge of the sample
analytes and matrix compounds. Based on this, the GC×GC set‐up, the initial column‐
set and all other parameters could be chosen, tested and optimized. In chapter 2, a
review is given in which several papers concerning GC×GC method development are
discussed and guidelines for GC×GC method development are proposed. The most
important and also most difficult task in GC×GC method development is the choice of
the column‐set. In chapter 3, a procedure is described for the global prediction of best
GC×GC column‐sets, based on the adapted Abraham solvation parameter model as
published by Seeley et al. This prediction procedure may assist in roughly selecting
best column‐sets as a starting point. In chapter 4 the issue of GC×GC retention time
shifts is addressed. A fast and easy to perform, two‐step retention time locking
procedure for GC×GC is proposed and its feasibility is demonstrated. This 2D‐RTL
procure is routinely used for already 5 years in our Lab at DSM. In chapter 5, a
182
retention time locking procedure for locking primary and secondary retention times
of detector signals in GC×GC dual‐detection is proposed and its practical advantages
are demonstrated and discussed. Chapter 6 and 7, both deal with optimizing GC×GC
selectivity by tuning selectivity in the first and/or the second dimension. These
approaches could be used to further optimize or fine‐tune certain (difficult) target or
group‐type separations especially for truly complex chromatograms. These set‐ups
also offer enhanced possibilities for qualitative analysis.
The actual analytical power of GC×GC for the analysis of a particular complex sample
(application) strongly depends on the method development choices and optimization,
but also on good chromatography practice and the data processing approach; all of
this to be able to answer the analytical question.
183
9 8B8BSamenvatting
Methodeontwikkeling voor “comprehensive” tweedimensionale gaschromatografie
“Comprehensive” (Alomvattende) tweedimensionale gaschromatografie (GC×GC) is
een krachtige analysetechniek en biedt in vergelijking tot eendimensionale
gaschromatografie (1D‐GC) mogelijkheden tot een groter scheidend vermogen,
betere gevoeligheid en gestructureerde chromatogrammen.
GC×GC is in de afgelopen 20 jaar uitgegroeid tot een relatief volwassen en
toegankelijke analysetechniek die steeds vaker wordt toegepast voor het oplossen van
complexe analytische vraagstukken. GC×GC‐methodeontwikkeling is echter niet zo
eenvoudig als die voor 1D‐GC. Voor GC×GC‐methodeontwikkeling moeten veel meer
keuzes gemaakt worden en optimalisatie is moeilijk door een complexe wisselwerking
tussen de vele eerste‐ en tweede‐dimensionale parameters.
Methodeontwikkelingskeuzes zijn ook gelimiteerd door het modulatie criterium en
door de maximale toelaatbare temperatuur van de kolommen. Ook het feit dat de
diameter van de tweede dimensie kolom in de meeste gevallen kleiner is dan die van
de eerste dimensie kolom, verhindert dat beide kolommen kunnen worden bedreven
onder hun optimale lineaire gassnelheden, dit wordt ook wel het “flow‐mismatch”
probleem genoemd. Het verschil in diameters kan ook leiden tot
massabelaadbaarheidsproblemen van de tweede dimensie kolom. Een andere
moeilijkheid is het feit dat pieken in GC×GC kunnen verschuiven in twee richtingen
hetgeen kan leiden tot problemen met onder andere het onderling vergelijken van
chromatogrammen of problemen met het gebruik van 2D‐sjablonen (2D‐templates).
In 1D‐GC kunnen piek verschuivingen eenvoudig worden opgelost door aanpassing
van de kolom inlaat druk (retentietijd locking), in GC×GC is dit echter niet zo
eenvoudig.
Om optimaal gebruik te kunnen maken van GC×GC is een systematische aanpak voor
methodeontwikkeling essentieel. Belangrijk bij het maken van de juiste
methodeontwikkelingskeuzes is het grondig begrijpen van de analytische vraagstelling
en kennis van de te analyseren monsters. Op basis hiervan kan de GC×GC‐opstelling
en een eerste kolommen‐set worden gekozen en geoptimaliseerd. In hoofdstuk 2
wordt een overzicht gegeven met betrekking tot GC×GC‐methodeontwikkeling,
184
gebaseerd op verschillende wetenschappelijke artikelen, waarin deze worden
bediscussieerd en richtlijnen worden voorgesteld voor methodeontwikkeling. De
belangrijkste en tevens moeilijkste taak bij GC×GC‐methodeontwikkeling is de keuze
van de kolommen‐set. In hoofdstuk 3 wordt een procedure beschreven voor het
globaal voorspellen van de beste GC×GC‐kolommensets, voor een specifieke set van
geselecteerde componenten. De procedure is gebaseerd op een model dat
gepubliceerd is door Seeley en gebruik maakt van Abraham’s
oplosbaarheidsparameters. Deze procedure kan gebruikt worden voor het selecteren
van een eerste kolommen‐set als startpunt van de methodeontwikkeling. In
hoofdstuk 4 wordt een eenvoudige twee‐staps procedure beschreven voor het
“locken” (=vastleggen) van GC×GC‐retentietijden; met deze procedure is het mogelijk
om piek verschuivingen te minimaliseren. Deze 2D‐locking procedure wordt al meer
dan vijf jaar succesvol toegepast binnen DSM. In hoofdstuk 5 wordt een opstelling en
procedure beschreven voor het “locken” van retentie tijden in GC×GC met
tweevoudige detectie. De praktische voordelen van een dergelijke opstelling en
procedure worden in dit hoofdstuk gedemonstreerd en bediscussieerd.
In hoofdstuk 6 en 7 worden opstellingen en procedures beschreven voor het
optimaliseren van GC×GC‐selectiviteit door het afstellen (“tunen”) van de selectiviteit
van de eerste en/of de tweede dimensie kolom. Deze opstellingen en procedures
kunnen worden gebruikt voor het verder optimaliseren van met name moeilijke (doel
of groep‐type analyses) scheidingen in zeer complexe chromatogrammen. Tevens
bieden deze opstellingen en procedures extra mogelijkheden voor kwalitatieve
analyse.
De werkelijke analytische kracht van GC×GC voor de analyse van een bepaald complex
monster hangt sterk af van de keuzes die worden gemaakt bij de
methodeontwikkeling, de methode optimalisatie, goede chromatografische
praktische vaardigheden (“good chromatography practice”) en de keuze van de
dataverwerkings aanpak. Dit alles gericht op het beantwoorden van de analytische
vraagsteling.
185
10 9B9BList of Author’s publications
Mommers, J., Ritzen, E., Dutriez, T., van der Wal, S., A procedure for comprehensive
two‐dimensional gas chromatography retention time locked dual detection, (2016) J
Chromatogr A, 1461, pp. 153‐160.
Mommers, J., Pluimakers, G., Knooren, J., Dutriez, T., van der Wal, S.
Tunable secondary dimension selectivity in comprehensive two‐dimensional gas
chromatography, (2013) J Chromatogr A, 1297, pp. 179‐185.
Mommers, J., Knooren, J., Dutriez, T., Ritzen, E., van der Wal, S.
Temperature‐tunable selectivity in comprehensive two‐dimensional gas
chromatography, (2012) J Chromatogr A, 1270, pp. 305‐309.
Mommers, J., Knooren, J., Mengerink, Y., Wilbers, A., Vreuls, R., van der Wal, S.
Retention time locking procedure for comprehensive two‐dimensional gas
chromatography, (2011) J Chromatogr A, 1218 (21), pp. 3159‐3165.
Patent WO2011147974 A1, Retention time locking for multi‐dimensional gas
chromatography, Johannes Helena Michael Mommers, Jeroen Albert Angelicus
Knooren, DSM IP Assets B.V.
Mommers, J., et al., Method development for comprehensive two‐dimensional gas
chromatography; a review, to be published.
John Mommers, Peter Tummers, Thomas Dutriez, Jelle Bos, Erik‐Jan Altink, Sjoerd van
der Wal, The use of Curve Fitting as new method for measuring orthogonality in
multidimensional‐chromatography, to be published in Journal of Chromatography A
Weusten, J.J.A.M., Derks, E.P.P.A., Mommers, J.H.M., van der Wal, S., Alignment, and
clustering strategies for GC×GC‐MS features using a cylindrical mapping, (2012) Anal
Chim Acta, 726, pp. 9‐21.
186
Other publications (not related to GC×GC)
Mommers, J., Mengerink, Y., Ritzen, E., Weusten, J., van der Heijden, J., van der Wal,
S.
Quantitative analysis of morphine in dried blood spots by using morphine‐d3 pre‐
impregnated dried blood spot cards, (2013) Anal Chim Acta, 774, pp. 26‐32.
Mommers, J.H.M., de Wildeman, S.M.A., Koolen, W.A.F., Duchateau, A.L.L.
Enantioselective gas chromatographic analysis of aqueous samples by on‐line
derivatization. Application to enzymatic reactions, (2008) J Chromatogr A, 1182 (2),
pp. 215‐218.
Mengerink, Y., Mommers, J., Qiu, J., Mengerink, J., Steijger, O., Honing, M.
A new DBS card with spot sizes independent of the hematocrit value of blood
(2015) Bioanalysis, 7 (16), pp. 2095‐2104.
Peters, R., Tonoli, D., van Duin, M., Mommers, J., Mengerink, Y., Wilbers, A.T.M., van
Benthem, R., de Koster, Ch., Schoenmakers, P.J., van der Wal, Sj., Low‐molecular‐
weight model study of peroxide cross‐linking of ethylene‐propylene (‐diene) rubber
using gas chromatography and mass spectrometry. I. Combination reactions of
alkanes
(2008) J Chromatogr A, 1201 (2), pp. 141‐150.
Sonke, T., Kaptein, B., Wagner, A.F.V., Quaedflieg, P.J.L.M., Schultz, S., Ernste, S.,
Schepers, A., Mommers, J.H.M., Broxterman, Q.B., Peptide deformylase as
biocatalyst for the synthesis of enantiomerically pure amino acid derivatives
(2004) Journal of Molecular Catalysis B: Enzymatic, 29 (1‐6), pp. 265‐277.
de Vries, A.H.M., Mulders, J.M.C.A., Mommers, J.H.M., Henderickx, H.J.W., de Vries,
J.G.
187
Homeopathic ligand‐free palladium as a catalyst in the heck reaction. A comparison
with a palladacycle, (2003) Organic Letters, 5 (18), pp. 3285‐3288.
de Vries, A.H.M., Parlevliet, F.J., Schmieder‐van De Vondervoort, L., Mommers, J.H.M.,
Henderickx, H.J.W., Walet, M.A.M., De Vries, J.G., A Practical Recycle of a Ligand‐Free
Palladium Catalyst for Heck Reactions, (2002) Advanced Synthesis and Catalysis, 344
(9), pp. 996‐1002.
188
189
11 10B10BOverview of author’s contributions
It has to be noted that the contribution of each co‐author was of great and essential
value for all chapters/papers, which is difficult to briefly summarize in just a few
words.
Chapter 1
John Mommers – performed literature search and wrote the chapter.
Sjoerd van der Wal – provided discussions and corrections.
Peter Schoenmakers – provided discussions and corrections.
Chapter 2
John Mommers – performed literature search and wrote the chapter.
Sjoerd van der Wal – provided discussions concerning theoretical aspects of
chromatography and method development, corrections, pointed out to useful papers
and checked all calculations.
Peter Schoenmakers – provided discussions and corrections.
Chapter 3
John Mommers – performed literature search, Matlab coding and calculations,
practical work and wrote the chapter.
Peter Tummers – provided valuable help in Matlab coding.
Thomas Dutriez – provided discussions concerning the theoretical approach.
Jelle Bos – performed descriptor calculations, gathered all column descriptor data and
performed practical work.
Erik‐Jan Altink – performed descriptor calculations and performed practical work.
Sjoerd van der Wal – provided discussions and corrections, pointed out to valuable
papers and information, discussed and checked all calculations.
Peter Schoenmakers – provided discussions and corrections.
Chapter 4
John Mommers – performed literature search, calculations, practical work and wrote
the chapter.
Jeroen Knooren – provided discussions, performed calculations, practical work, and
processing the data.
190
Ynze Mengerink – provided discussions / challenging ideas and initiated the filing of a
US Patent for retention time locking for multi‐dimensional gas chromatography.
Arno Wilbers – provided discussions concerning mathematical approaches.
Rene Vreuls – provided discussions / challenging ideas.
Sjoerd van der Wal – provided discussions and corrections and checked all
calculations.
Chapter 5
John Mommers – performed literature search, calculations, practical work and wrote
the chapter.
Erik Ritzen – provided discussions and performed practical work.
Thomas Dutriez – provided discussions and performed practical work.
Sjoerd van der Wal – provided discussions and corrections and checked all
calculations.
Chapter 6
John Mommers – performed literature search, calculations, GC×GC analysis and wrote
the chapter.
Erik Ritzen – provided discussions and performed practical work.
Thomas Dutriez – provided discussions / challenging ideas.
Sjoerd van der Wal – provided discussions and corrections and checked all
calculations.
Chapter 7
John Mommers – performed literature search, calculations, practical work and wrote
the chapter.
Erik Ritzen – provided discussions and performed practical work.
Thomas Dutriez – provided discussions / challenging ideas.
Jeroen Knooren – provided discussions, calculations and practical work.
Giulia Pluimakers ‐ provided discussions, calculations and practical work.
Sjoerd van der Wal – provided discussions and corrections and checked all
calculations.
191
12 11B11BDankwoord
Dit proefschrift is mede tot stand gekomen door de hulp, steun en motivatie van
velen. Bij deze wil ik iedereen, die op welke manier dan ook een bijdrage heeft
geleverd, heel erg bedanken.
Mijn promotor Prof. dr. S. Van der Wal. Beste Sjoerd, enorm bedankt voor je grote
betrokkenheid (soms bezorgdheid), gedrevenheid, hulp, motivatie en het (eindeloos)
blijven stellen van kritische vragen. Ik weet, ik was niet altijd even makkelijk. Ik ken je
al meer dan 20 jaar (vanaf het moment dat ik als stagiair bij DSM begon op de
chromatografie afdeling) en heb altijd grote bewondering voor je gehad, voor je
chromatografie kennis, deskundigheid en inzichten maar zeker ook als mens. Ik heb
zeer veel van je geleerd, je was voor mij altijd een groot voorbeeld en het was dus
voor mij ook een hele eer dat je mijn promotor bent geworden. Nogmaals zeer
bedankt, en geniet van je pensioen samen met Martien en je familie.
Mijn co‐promotor Prof. dr. ir. P. J. Schoenmakers. Beste Peter, ik wil je bedanken
voor het kritisch nakijken van mijn proefschrift en regelen van zaken rondom mijn
promotie.
De overige leden van de promotiecommissie, Prof. dr. C. G. de Koster, Prof. dr. J. F.
Focant, Prof. dr. T. Hankemeier, Prof. dr. ir. J.G.M. Janssen, Prof. Dr. R.A.H. Peters,
Prof. Dr. W.P. de Voogt en Dr. M. Camenzuli, dank ik zeer voor het kritisch lezen en
beoordelen van mijn proefschrift.
Ik wil ook mijn werkgever DSM zeer bedanken voor de mogelijkheid die me werd
geboden om te kunnen promoveren en het sponseren van het drukken van dit
proefschrift. Naast het gebruik van de beschikbare GCxGC apparatuur heb ik ook de
ondersteuning kunnen krijgen van een aantal stagiaires die ik binnen DSM heb
mogen begeleiden. Beste Jelle, Erik‐Jan en Giulia, bedankt voor jullie enthousiasme
en inzet, jullie hebben een belangrijke bijdrage geleverd aan het tot stand komen
van dit proefschrift.
192
Mijn DSM (ex) collega’s, Peter Tummers, Thomas Dutriez, Ynze Mengerink, Arno
Wilbers, Erik Ritzen, Rene Vreuls en Jeroen Knooren die allen een directe bijdrage
hebben geleverd aan een of meerdere van de gepubliceerde artikelen. Beste (ex)
collega’s bedankt voor jullie deskundige inbreng. Ik kon bij jullie terecht om te
sparren, om advies of hulp te vragen als ik er zelf niet uitkwam. Zonder jullie was dit
proefschrift niet tot stand gekomen. Daarnaast wil ik al mijn DSM (ex) collega’s van
de chromatografie afdeling (Wilma, Math, Sonja, Marianne, Jos, Huub, Ilse, Jan, Tim,
Henk, Wil, Esra, Marcel, Ward, Gerard, Benjamin, Remco, Erwin) en van de wiskunde
en statistiek afdeling (Eduard, Jos) enorm bedanken voor jullie steun en hulp.
De promotie periode is niet altijd even makkelijk geweest en dit was dan ook zeker
niet mogelijk zonder de hulp en steun van mijn ouders. Mijn ouders hebben ons al
die jaren enorm geholpen met onder andere ons huishouden, Linde naar school
brengen en weer ophalen, klussen in huis, tuinieren en noem maar op. Pap en mam,
zonder jullie hulp was dit zeker niet mogelijk geweest, bedankt voor alles.
Mijn vrienden en familie. Bedankt voor jullie steun en interesse en dat jullie hebben
gezorgd voor de nodige afleiding.
Last but not least wil ik mijn vrouw Estelle en mijn dochter Linde bedanken. Lieve
schat, je hebt me al die jaren geweldig gesteund. We hebben door de jaren heen
talloze gesprekken gevoerd over mijn promotie. Je hebt me meer dan eens goed
advies gegeven (soms zelfs inhoudelijk). Mijn promotietraject kende pieken en dalen.
Bij elk dal wist je mij weer te motiveren en sleepte je mij er doorheen. Je hebt al die
jaren gezorgd voor een goede balans in ons gezin. Zonder jou had ik dit uiteraard
nooit kunnen doen. Lieve Linde, we hebben vaak samen huiswerk gemaakt. Het was
niet altijd even leuk dat pappa vaak moest werken aan zijn promotieonderzoek.
Binnenkort gaan we vaker leuke dingen doen. Beloofd!