Theoretical model for flow modulated comprehensive two-dimensional gas chromatography

MSc Chemistry

Analytical Sciences

Master Thesis

Theoretical model for Flow

Modulated Comprehensive

Two-Dimensional Gas

Chromatography

by

Larissa Ram

11056355

July 2018

48 EC

September 2017 – July 2018

Supervisor/Examiner:

Examiner:

Dr. W.T. Kok

Prof. P.J. Schoenmakers

Van’t Hoff Institute for Molecular

Sciences

2

Acknowledgements

This thesis is written to fulfill the master research project for the master Chemistry study at the University of Amsterdam. This project is carried out outside the University of Amsterdam at Da Vinci Laboratory Solutions. During my research I was supervised by Paul van den Engel from Da Vinci Laboratory Solutions and prof. Hans-Gerd Janssen from the University of Amsterdam.

I would like to express my specials thanks to Paul van den Engel and Prof. Hans-Gerd Janssen for their support, by giving me guidance and suggestions throughout this project. Their advice helped me throughout this project and the opportunity to finish the project successfully.

Furthermore, I would like to thanks Da Vinci Laboratory Solutions to give me the opportunity to fulfill my master research project and the possibility to make use of the facilities. I would like to thank the application team at Da Vinci for their support, advice, patience and great time during the project.

Larissa Ram

3

Abstract

In chromatography, the specialist is often confronted with the problem of optimizing the Gas Chromatography system. In one-dimensional GC, method development is already a difficult task, but with well-known rules and formulas optimal conditions for column geometries and pressure settings optimization is possible. Method development in GC×GC becomes even more complex and problematic due to the multiple dependent variables.

In this research a theoretical model has been developed for method development and optimization for reversed fill/flush modulation comprehensive two-dimensional gas chromatography (RFF-GC×GC). The theoretical model consists of a combination of equations for optimizing the instrumental parameters in RFF-GC×GC and optimization of columns and carrier gas flows to improve the RFF-GC×GC separation. The theoretical model needs column dimensions, carrier gas type and first-dimension column flow as input parameters and calculates all instrumental parameters, including dimensions of the collection channel and restriction capillary using resistance diagrams similar to Ohms law. It also calculates velocities and plate numbers in RFF-GC×GC separation using flow rate and plate height equations.

The theoretical model is validated against practical data and interesting conclusions can be drawn. For example, with the theoretical model the optimal flows of the first and second dimension column, the dimension of the collection channel and restriction capillary can be calculated when the dimensions of the first and second dimension column are known. Besides, the model calculates the maximum fill time and the minimal flush time needed to completely flush the collection channel. These theoretical predictions drastically reduced the method optimization time.

The model described the configuration accurately enough to be a helpful tool in optimizing RFF-GC×GC analysis. With the newly developed model construction of the flow modulator is significantly simplified. In particular selection of the collection channel in RFF-GC×GC in a way that is does not cause extra band broadening RFF-GC×GC system is now much easier.

4

Table of Contents

2.1 Comprehensive two-dimensional gas chromatography (GC×GC) ... 6

2.2 Cryogenic modulators... 7

2.3 Flow Modulation... 8

2.3.1 Forward Flow Modulation ... 9

2.3.2 Reversed Fill Flush Flow Modulator (RFF) ... 11

2.4 RFF – GC×GC – FID optimization at ambient outlet condition ... 13

2.4.1 Fill Position... 13

2.4.2 Flush Position ... 16

2.5 Plate height Model for isothermal Flow Modulated GC×GC ... 18

2.5.1 Peak broadening in RFF GC×GC – FID ... 20

3. Experimental ... 22

3.1 Chemicals ... 22

3.2 Materials ... 22

3.3 Analytical Procedure ... 23

4. Results ... 26

4.1 Validation of the computer model H/U curves ... 26

4.2 Collection channel optimization ... 28

4.3 Restriction capillary optimization ... 31

4.4 Modulation Time optimization ... 43

4.4.1 Collection channel fill time ... 43

4.4.2 Collection channel flush time ... 46

4.5 Remaining challenges and recommendations ... 47

4.5.1 Flow modulated – GC×GC – MS ... 47 5. Conclusion ... 49 6. Literature ... 50 7. Appendix ... 53 7.3 List of Figures ... 53 7.4 List of Tables ... 54

5

1. Introduction

Comprehensive two-dimensional gas chromatography (GC×GC) was introduced and pioneered by Philips et al.1–3 _{using a dual stage thermal desorption (TD) modulator. The nowadays mostly used and commercially}

available modulators use cryogenic cooling to trap the primary column effluent 4,5_{. The optimization and use}

of cryogenic modulators has been extensively studied so far6_{. Besides the cryogenic modulator, Bueno and}

Seeley7 _{demonstrated in 2004 a differential flow modulator based on Deans switching for GC×GC without}

the use of cryogenic cooling. In the second generation flow modulators a simpler version is developed by Seeley8_{. This setup uses two tee unions and a three-way solenoid valve to control the modulator.}

Subsequent papers of La Clair et al.9 _{, Micyus et al.}10_{, Gu et al.}11 _{and Zoccali et al.} 12 _{demonstrated the}

usefullness of flow modulation. In 2011, Griffith et al. 13 _{modified the Agilent FF modulator to obtain a}

reversed fill flush modulation (RFF). The performance differences between the FF and RFF are discussed in detail by Duhamel et al. 14 _{and Krupčík et al.} 15 _.

In chromatography, the specialist is often confronted with the problem of optimizing the Gas Chromatography system. In one-dimensional GC, method development is already a difficult task. Fortunately, with well-known rules and formulas optimal conditions for column geometries and pressure settings can be selected. This method development task becomes even more complex and problematic in two-dimensional gas chromatography. This is due to the multiple dependent variables affecting the system (e.g., column dimensions, pressure settings, modulation times, temperatures etc.). Therefore, theoretically describing the GC×GC procedure and using these calculation methods for establishing the best instrumental settings is attractive and can safe time. Beens et al.16,17 _{developed calculation programs for cryogenic GC×GC}

systems that predict the outcome of the GC×GC separation and can be used to optimize the GC×GC conditions.

The current drawbacks of flow modulation include the long method development and optimization time needed. Method development via trial-and-error is a long procedure, especially for flow modulated GC×GC systems. The calculators developed by Beens et al. cannot be used in combination with flow modulation. This is due to the differential flow in flow modulation; a high second-dimension column flow is added to flush the collection channel in a fraction of the collection time. Besides, the collection channel, restriction capillary and the two states in flow modulation causes optimization in the flow modulated system to be more complex compared to cryogenic GC×GC systems.

The goal of this study is to develop a theoretical model for describing Reversed Fill Flush Flow modulated comprehensive two-dimensional chromatography (RFF-GC×GC) by addressing the above mentioned parameters. This means that besides the separation optima, flow management and H/U curves, also the instrumental parameters such as loop volume and dimensions, restriction capillary length and diameter and modulation time are described. These calculations are performed in Microsoft Excel and require column dimensions, carrier gas type and flows to calculate the optimal conditions. The calculations are applicable for ambient outlet conditions, but are easily converted to vacuum outlet conditions.

6

2. Theoretical Background

2.1 Comprehensive two-dimensional gas chromatography (GC×GC)

Comprehensive two-dimensional gas chromatography (GC×GC) was introduced and pioneered by Philips et al.1–3 _{using a dual stage thermal desorption (TD) modulator. GC×GC consists of two columns connected}

serially such that all components are passing through the first column and elute as separate fractions onto the second-dimension column. The hyphenation produces a three dimensional chromatogram where two axis represent the retention times of the eluting compound on the two columns and the third axis represents the detector response, which corresponds to the peak height16_{. A GC×GC retention plane has much more}

peak capacity than a one-dimensional retention lane. Phillips et al. described that the total peak capacity is approximately the product of the peak capacities of both columns (1_{n ×} 2_n)1_{. Therefore, GC×GC is a powerful}

technique for the analysis of complex samples of volatile and semi-volatile compounds.

Giddings18,19 _{defined the conditions that a system has to fulfill in order to be comprehensive; (i) every part}

of the sample is subjected to two different separations, (ii) equal percentages of all sample components pass through both columns and reach the detector and (iii) the separation of the first-dimension column is maintained while injecting into the second-dimension column and reaching the detector. For example, if a peak that elutes from the first-dimension is sampled in slices of about ¼ of the total peak width (four cuts over a peak), the separation achieved in the first-dimension is preserved during the second-dimension separation20,21_.

In GC×GC mainly two types of separation applications are described, namely a separation of target compounds22,23 _{and a group-type separation}24,25_{. Two columns of different polarities are coupled in series}

via a modulator. If the separations performed on the two columns are independent of each other, the chromatograms obtained are orthogonal. In conventional GC×GC applications, the first-dimension column is a long nonpolar column and the second-dimension column is a short polar column to increase the orthogonality26,27_{. The heart of the GC×GC system is the modulator. The main functions of the modulator}

are to cut, re-concentrate and rapidly re-inject narrow zones of first-dimension effluent into the short second-dimension. The modulator is capable of repeatedly collecting first-dimension effluent and injecting it onto the second-dimension. The time required to fulfill this collect and reinject process is called the modulation time. In GC×GC the sample is introduced and separated on the first-dimension column. Primary column effluent is cut in regular series of small fractions, focused and rapidly reinjected into the second fast separation column. The second-dimension column must operates sufficiently fast for its separation to be finished before the compounds from the next modulation will start to elute onto the second-dimension, to minimize the occurrence of so-called wraparounds19_{. If wraparound occurs, the second separation is not}

finished before the next fraction is already reinjected in the second-dimension. These wraparounds cause data interpretation and quantification problems.

7

2.2 Cryogenic modulators

The nowadays mostly used and commercially available modulators use cryogenic cooling to trap primary column effluent 4,5_{. In a cryogenic modulator, compounds eluting from the first-dimension are trapped in a}

cold zone with liquid nitrogen (LN2) or carbon dioxide during a selected modulation time. The cold zone is rapidly heated to reinject the trapped compounds via the carrier gas flow onto the second-dimension separation column, see Figure 1.

The applicability of cryogenic modulators is described in a wide range of articles6_{. One of the drawbacks of}

the cryogenic modulator is the high consumption of liquid nitrogen, 30 liters of LN2 per day when measuring continuously28 _{. With cryogenic focusing it is difficult to trap the very volatile compounds below C4}29_{, also}

the chance of leaks in the column connector or press fits due to the continuous change between hot and cold is high, resulting in efficiency loss. However, with cryogenic cooling very narrow input bands can be obtained. The second-dimension peak widths generated are around 50 to 120 ms 24,29,30_{, resulting in a gain}

in dimension peak capacity. With cryogenic modulators the flow through the first and second-dimension is equal and therefore relatively easy to optimize, however the disadvantage is that both columns cannot be operated at optimal conditions which limits the dimensional flexibility and the performance of the second column.

8

2.3 Flow Modulation

Due to the disadvantages of cryogenic modulators, Bueno and Seeley7 _{developed a differential flow}

modulator based on Deans switching for GC×GC without the use of cryogenic cooling (Figure 2). The applicability of differential flow modulation based on Deans switching was also demonstrated by Seeley et al10,31_{. Differential flow modulation employs a high second-dimension column flow to forward flush the}

collection channel in a fraction of the collect time, in practice a 1:20 primary to secondary flow rate is often used32_{. Under these conditions the loop fill time is 1.0 s while the reinjection into the second-dimension is}

done within 50 ms.

Figure 2: diagram of the flow switching device based on Deans switching used for differential flow modulation. Exiting primary column flow (F1) enters the device at T- union A. Union A is connected to two T- unions B with equal dimensions of tubing. An

auxiliary flow F2 exits the 3-way valve and flows either through union BL or Br. In this setup the right loop is filled with primary

column effluent and F2 flushes the left loop. Switching the 3-way valve changes the position of the fill and flush loop. The right loop is flushed and injected into the second dimension column and the left loop is filled with primary column effluent. This cycle is repeated continuously.

9

2.3.1 Forward Flow Modulation

A simplified flow modulator design was developed by Seeley8 _{in 2006. This setup uses only two tee unions}

and a three-way solenoid valve to control the modulator. Agilent commercialized this differential flow modulator using Capillary Flow Technology (CFT) that facilitates the GC×GC set up (Figure 3). In the Agilent flow modulator, the loop volume is fixed in the CFT plate, thus the modulation time is limited by the column dimensions and the loop volume applied.

In the fill state, primary column effluent flows into the collection channel, while a high auxiliary flow is directed to the second-dimension column. The solenoid valve is connected to an auxiliary programmable control module (PCM) to control the flow. When the valve is switched to the flush state, the auxiliary flow forward flushes the collection channel and injects the content into the second-dimension column where it undergoes a very fast elution at a very high flow rate. After the fast-flush time, the valve returns to the fill state and the process is repeated. The loop is flushed in the same direction as the fill cycle, which is called Forward (Flush) Flow Modulation (FF). A drawback of FF is that the flush flow is in the same direction as the fill flow. This results in a continuous first-dimension column flow to fill the collection channel. When a component elutes from the first-dimension during the flush period, this portion of first-dimension column effluent is passing slowly through the collection channel into the second-dimension resulting in unmodulated peaks and tailing13,15,33

The modulation time (fill + flush time) determines the maximum retention time of the second-dimension column. Increasing the modulation time results in more time being available for the separation in the second-dimension, however too long a fill time results in breakthrough from (unmodulated) primary column effluent directly into the second-dimension column, which is called streaking. Tailing and streaking in the second-dimension reduces the number of peak that can be separated in the second-dimension11,34_.

1st dimension Column Sample Loop 2nd dimension Column 3-Way Solenoid Valve

A

10

B

Figure 3: Diagram of the Forward Flow Modulator, A: Fill position, B: Flush Position. Exiting primary column effluent enters the collection channel for a defined fill time. At the same time, a high (± 20 mL/min) auxiliary flow exiting the 3-way solenoid valve and is passing through the second-dimension column. Switching the 3-way valve changes the position from the fill to the flush position. In the flush position, the high auxiliary flow from the PCM is forward flushing the collection channel and injecting in the second- dimension column. After the short flush time the 3-way solenoid valve is switched to the fill cycle and the setup is repeated.

11

2.3.2 Reversed Fill Flush Flow Modulator (RFF)

Griffith et al. 13 _{modified the Agilent FF modulator to obtain Reversed Fill Flush (RFF) modulation, which}

means that the loop filled with primary column effluent is injected into the second-dimension column in the opposite flow direction (Figure 4). The aim of the RFF modulator is to minimize unmodulated peaks and reduce peak width in the transfer from the first to the dimension. Reversed injection in the second-dimension reduces the tailing at the base of the second-second-dimension peak, decreases the peak widths of the modulated peaks and increases the number of peaks that can be separated in the second-dimension13,15_.

The RFF modulator consist of two different CFT plates which are connected via a variable collection channel. The collection channel is an extra variable that is adjustable in RFF. This allows more flexibility in the selection of the modulation time and maximum second-dimension retention time. The end of the loop is connected to a restriction capillary column, which in turn is connected to an FID. The purpose of the restriction capillary is to provide an outlet for the carrier gas passing through the collection channel during the fill cycle and to allow a reversal of the flow direction during the flush cycle. Besides, the restriction column enables a restriction and is a tool to optimize the pressure and flow conditions during the fill and flush position. The restriction capillary column is connected to an FID for monitoring the fill position e.g., an overfilled collection channel shows signal on the monitoring back FID.

The performance of the FF modulator and RFF modulator are discussed by Duhamel et al. 14 _{and by Krupcik}

et al.15_{. The peak widths generated by RFF modulation are around 100 ms at second-dimension column}

outlet. This is in the same order as the peak widths generated with cryogenic cooling.

3-Way Solenoid Valve Restriction Capillary Collection channel 2nd dimension Column 1st dimension Column

A

12

B

Figure 4: Diagram of the RFF modulator. A: Flow path of the fill cycle, B: Flow path of the flush cycle. The end of the first-dimension column is connected to the collection channel, the collection channel is the link between the two CFT plates. The end of the collection channel is connected to the restriction capillary. In the fill state the auxiliary flow is continuously flushing the second dimension column and a small part is filling the loop. In the flush state the valve is switched, flushing the loop and transferring its content onto the second- dimension column. A small part of the auxiliary flow is passing through the restriction column.

13

2.4 RFF – GC×GC – FID optimization at ambient outlet condition

In this chapter the equations and calculation methods to describe and improve the FM – GC×GC separation are presented. These calculations are performed in Microsoft Excel and require column dimensions, carrier gas type and flows to calculate the optimal conditions. The calculations are applicable for ambient outlet conditions. However, these calculations can be easily converted to vacuum outlet conditions.

In RFF there are two states affecting the GC×GC separation, the fill and the flush state. In both situations the optimal parameters are described. The calculations are based on resistor diagrams similar to Ohms law; ΔP = F * R whereby the compressibility of the gas (ΔP) is described as function of in-and outlet pressure Pin and

Pout Respectively (Eq. 1). Fout is the flow rate at column atmospheric outlet conditions, where the flow

through the column is calculated using the Poiseuille equation (Eq. 2). The resistance (R) to flow is calculated using equation 3. The resistance of the connections between the columns in the CFT plates are assumed to be negligible and therefore not included in the model.

ΔP = Pin2 − Pout2 Pout Eq. 1 𝐹𝑜𝑢𝑡 = Π∗ r4 16 ∗n∗L ∗ Pin2− Pout2 Pout Eq. 2 𝑅 =16nL πr4 Eq. 3

Here, the total flow (Fout) is dependent on the radius of the column (r), the viscosity (ƞ), length of the column

(L) and the in-and outlet pressures of the column. The resistance of the column (R) is a function of the radius (r) and length (L) of the column and the viscosity of the carrier gas (ƞ)

2.4.1 Fill Position

The calculations in the fill position are based on the fill resistor diagram, shown in Figure 5. In the fill position, primary column effluent (F1) flows through the first-dimension column (R1) towards the collection channel (R2), where the end of the collection channel is connected to the restriction capillary (R3). The flow though the restriction capillary is defines as F3. The flow through the collection channel (F2) is a sum of primary column effluent (F1) and a leak flow from the PCM (F’). F4 is the flow from the PCM module through the second-dimension column (R4). The total flow (Ftot) over R2, R3 and R4 is a combination of F1 and F’ whereby F1 and F’ are adjustable (Eq. 4). The flow through the second-dimension column depends on the resistance generated by the dimensions of the second-dimension column (Eq. 5), where the inlet pressure of the second-dimension column is defined as midpoint pressure (Pa). The flow through R4 also depends on the resistance of R2 + R3. Ftot = F1 + F′(PCM) Eq. 4 Eq. 5 F4 = 𝑅4 ∗Pa 2_{− Pout}2 Pout

14

The inlet pressure of the second-dimension column, defined as midpoint pressure (Pa) is the same as the inlet pressure of the collection channel (R2). The flow through the collection channel is defined according to equation 6.

Eq. 6

Whereby the inlet pressure of the collection channel is defined as Pa and the outlet pressure of the collection channel is defined as ²Pa. The outlet flow of the collection channel (F2out) is the same as the inlet flow of the restriction capillary (F3in). Equation 7 explains the outlet flow through the restriction capillary (F3). The calculated flows are volumetric flows and are calculated at the observed pressure. This means that the flow is not calculated to atmospheric conditions.

F3 = 𝑅3 ∗²Pa2− 𝑃𝑜𝑢𝑡2

Pout Eq. 7

The flow through the restriction capillary is almost equal to the flow passing through the collection channel. This is due to the negligible resistance of the collection channel (R2<< R3).

The midpoint pressure (Pa) is crucial in optimizing the flow conditions in flow modulation. The

midpoint pressure defines the exact flow from the first-dimension column towards the collection

channel and the flow from the PCM through the second-dimension column. For the midpoint

pressure calculations, first ΔP is calculated using equations 8, 9 and 10:

Ftot = Pa2− Puit2 Puit ∗ ( 1 𝑅4+ 1 R2+ R3) Eq. 8 Rtot = 1 𝑅4+ 1 R2+ R3 Eq. 9 Pa2− Puit2 Puit (ΔP) = Ftot Rtot Eq. 10

The collection channel and restriction capillary are connected in series and therefore under the same denominator. Normally R2 is much lower than R3 and can hence be neglected. Here however, for reasons of completeness, the resistance of collection channel is included in the calculation. The midpoint pressure (Pa) is calculated using equation 11. From the midpoint pressure, the inlet pressure (1_{Pin) or flow (F1) from}

the first-dimension column is calculated using the Poiseuille equation (Eq. 12), where the outlet pressure of the first-dimension column is defined as Pa.

Pa = √FtotPout

Rtot + Pout Eq. 11

F2 = 𝑅2 ∗Pa

2_{− ²Pa}2

15

F1 = 𝑅1 ∗¹Pin2− Pa2

Pa Eq. 12

Based on the equation above, all parameters are known and the total fill system is described based on flows, pressure and column dimensions.

FID

FID

F

¹P

P

²P

PCM C

H2

Figure 5: Resistance diagram of the fill position in RFF modulation. The PCM flow is defined as F’, the resistors define the 4 different columns, R1: first-dimension column, R2: collection channel, R3: restriction capillary, R4: second-dimension column. The inlet pressure of the second- dimension column and collection channel is the first midpoint pressure (Pa). The outlet pressure of the

collection channel and inlet pressure of the restriction capillary is defined as second midpoint pressure ( 2_{Pa). R4 is connected in}

16

2.4.2 Flush Position

In the flush position the resistor diagram is more complex, see Figure 6. In this situation, the collection channel is parallel with the restriction capillary, but in series with the second-dimension column. The outlet pressure of the first-dimension (Pa) column is between the collection channel and second-dimension column. As in the fill position, the first-dimension column flow (F1) and the PCM flow (F’) are adjustable. However, the direction of the PCM flow has changed in the flush position. In this position the 3-way valve switched the PCM flow towards the collection channel (R2) and restriction capillary (R3). In this flush situation the inlet pressure of the collection channel and the inlet pressure of the restriction capillary is the same (2_{Pa). The flow through the restriction capillary is calculated as follow,}

𝐹3 = 𝑅3 ∗ ²Pa2_Pout −Pout2 Eq. 13

Whereby the outlet pressure of the restriction capillary is at atmospheric conditions.

In the flush position the outlet pressure of the collection channel (R2) is the same as the inlet pressure of the second-dimension column (R4) and outlet pressure of the first-dimension column (R1), this pressure point is defined as midpoint pressure (Pa).

The flow through the second-dimension column is calculated using equation 14.

𝐹4 = 𝑅4 ∗ Pa2−Pout2

Pout Eq. 14

Here the flow through the second-dimension column is dependent on the resistance of the column, inlet pressure of the second-dimension column (Pa) and outlet pressure of the column (Pout).

The total flow through the system can be practically measured at the end of both FID detectors. The total flow (Ftot) is the sum of the flow through the restriction capillary (R3) and second-dimension column (R4), see equation 15.

Ftot = F3 + F4 Eq. 15

Ftot = R3 ∗ ²Pa2− Pout2

Pout + R4 ∗

Pa2− Pout2

Pout Eq. 16

In equation 16 both midpoint pressures are unknown. The flow from the PCM (F’) is the sum of the flow passing through the restriction capillary and collection channel. The second-midpoint pressure (2_{Pa) is}

calculated using the resistance and flow of the collection channel and restriction capillary. The flow through the collection channel is defined as follow,

F2 = R2 ∗ ²Pa2−Pa2

Pa Eq. 17

17

In equation 18 The PCM flow (F’) is calculated as the sum of the flow from the collection channel (F2) and restriction capillary (F3). This equation is rewritten as function of second midpoint pressure (2_{Pa) , whereby}

the 2_{Pa is depended on, PCM flow (F’), resistance of R2 and R3, first midpoint pressure (Pa) and outlet}

pressure of the restriction capillary ( equation 19), ²Pa2_{= F}′_{+ R3 ∗ Pout + R2 ∗ Pa ∗} Puit∗Pa

R3∗Pa+R2∗Puit Eq. 19

The above Equation 19 is substituted into equation 16, this results in one final equation for Ftot, where only the first midpoint pressure (Pa) is unknown and is calculated.

Ftot = R3 ∗ (F′_{+ (R3 ∗ Pout) + (R2 ∗ Pa) ∗ (} Pa

R3∗Pa+R2∗Pout)) + (R4 ∗ ( Pa2

Puit) − R3 ∗ Pout − R4 ∗ Pout) Eq. 20

Once the midpoint pressure (Pa) is known, all the instrumental parameters can be calculated using equations 19 and 20.

FID

FID

F

¹P

P

²P

PCM C

H2

Figure 6: Resistance diagram of the flush position in RFF modulation. The PCM flow is defined as F’, the resistors define the 4 different columns, R1: first-dimension column, R2: collection channel, R3: restriction capillary, R4: second dimension column. The

inlet pressure of the restriction capillary and collection channel are the second midpoint pressure ( 2_{Pa). The outlet pressure of the}

collection channel and first-dimension column and inlet pressure of the second-dimension column is defined as first midpoint pressure (Pa). In this situation, R3 and R2 are connected in parallel. R2 and R4 are connected in series and R1 and R4 are connected in parallel.

18

2.5 Plate height Model for isothermal Flow Modulated GC×GC

In this section the equations used to calculate plate heights in isothermal GC×GC are described. An objective of this research is to optimize flow modulated GC×GC with the aim to maximize peak capacity and minimize the second-dimension peak broadening. A model is constructed that relates both objectives including column dimensions, flows and modulation times.

The width of the first-dimension column peaks (1_{σ) under isothermal conditions is given by:}

¹σ = √1H∗ 1tR2

1L Eq. 21

Where 1_{H is the plate height of the first-dimension column,} 1_t

R is the retention time of an eluting compound

from the first-dimension column and 1_{L is the length of the first-dimension column. Here} 1_{H is given by}

equation 22 ¹H = 1 ¹CE[( 2¹Dm ¹Uo + 11¹k2 +6¹k+1 96(1+¹k)2 ∗ dc2¹Uo ¹Dm ) ∗ ¹f1 + ( 2¹k 3(1+¹k)2∗ ¹d_f2Uo ¹Ds ) ∗ ¹f2 ] Eq. 22

Where CE is the column efficiency, Dm the diffusion coefficient of the analyte in the mobile phase, Uo the

linear velocity at outlet conditions, k the retention time factor, dc the column diameter, df the film thickness

of the column and Ds the diffusion coefficient of the analyte in the stationary phase (a realistic estimation is

that Ds is 50.000 times smaller than Dm). f1 and f2 are pressure correction factors due to the compressibility

of ideal gases according to Giddings et al.35 _{and James and Martin}36_,

¹𝑓1= 9(P4−1)∗(P2−1) 8(P3₋₁₎2 Eq. 23 ¹𝑓2= 3 2 P2₋₁ P3₋₁ Eq. 24

19

In equation 25, the linear velocity at outlet conditions of the first-dimension column is related to the flow rate and internal diameter of the column,

¹Uout = ¹F

πr2 Eq. 25

Since the flow rate, in – and outlet pressure and linear velocity at outlet conditions are known, the average velocity and the retention time of the compound can be calculated using equation 26 and 27:

¹u

̅̅̅ = ¹u_o¹f2 Eq. 26

¹tR =

¹L (1+¹k)

¹u̅ Eq. 27

To calculate the diffusion coefficient used in equation 22, the equation described by schettler et al.37 _for

binary gas mixtures is used. The diffusion coefficient is calculated from the temperature (T), Molar mass of the carrier gas (hydrogen) and compound (Mm and Mo), the pressure (P) and the molar volumes of the carrier gas and compound (Vm and Vo).

²Dm= 1.00 x 10−3_T1.75₍ 1 Mm+ 1 Mo)1/2 P[(∑ Vm)1/3_{+ (∑ Vo)}1/3_]2 Eq.28

From equation 28, the diffusion coefficient of the first-dimension column can be established (if the temperatures of the first and second-dimension column are the same):

¹Dm= 2Dm,o 2Pout 1Pout Eq. 29 ¹Ds= Dm 50.000 Eq. 30

For the second-dimension column and collection channel the same equations are used to obtain the plate height and peak broadening.

20

2.5.1 Peak broadening in RFF GC×GC – FID

As described in chapter 2.5, the chromatographic peak broadening of the column is described by equation 21. The total peak broadening of the first-dimension is the sum of injection peak broadening and chromatographic peak broadening,

¹σtot2 = σinj2 + σcol2 Eq. 31

The peak broadening due to the injection band width can be calculated from: σ_inj2 = V

12∗F Eq. 32

Where V is defined as the volume of the injection band, F is the total flow for transfer of the sample onto the column (column flow plus split flow) and 12 is the factor for a block-shaped injection. The equations above are analogous for describing the total peak broadening generated in the fill state of the collection channel,

σtotLoop2 = σinj(fill)2 + σLoop2 Eq. 33

σ_{inj Loop} 2 =Vloop

12∗F Eq. 34

VLoop= Fill time ∗ ¹Floop Eq. 35

Here the volume of the collection channel filled with primary column effluent (Vloop) is defined as the product

of the first-dimension column flow in m3_{/s (}1_{F) and the fill time in seconds. The σ²}

tot loop is the sum of the

peak broadening in the loop generated in FF modulation. However, In RFF modulation the injection band width into the second-dimension column is reduced due to the high backflush flow through the loop into second-dimension column. The final peak broadening in the second-dimension column is defined as: ²σtot2 = σinj(flush)2 + ²σcol2 Eq. 36

Where the σ²inj (flush) of the loop can be calculated from:

σ_inj(flush)2 = Vflush

12∗²Fflush Eq. 37 V flush = ¹Floop ∗ σtot loop Eq. 38

The volume of the injection band of the collection channel in the flush position (Vflush) which is injected into

the second-dimension column (Eq. 38) is defined as the product of the first-dimension column flow in m3_/s

through the loop in the fill position (1_F

Loop) and the total peak broadening expressed in seconds in the

collection channel which is calculated in equation 33 (σtot loop). The injection band broadening in the

second-dimension column (Equation 37) is defined as the volume of the collection channel filled with primary column effluent (Vflush)) and the PCM flow to flush the collection channel (2Fflush). The

chromatographic band broadening in the second-dimension column is defined using equation 21. As a note, the assumption of the peak broadening generated in the collection channel is not completely correct,

21

because the first part of the band entering the collection channel is more broadened than the later part in the same fill cycle. Front end and back end of the band experience a different broadening.

22

3. Experimental

3.1 Chemicals

n-Octane and o-Xylene were obtained from Sigma Aldrich Chemie GMbH (Steinheim, Germany) and used for validation of the RFF-GC×GC-FID system. Cyclohexane was obtained from Biosolve B.V. (Valkenswaard, the Netherlands) and used as (washing) solvent.

Samples from the proficiency test on Jet Fuel (sample #0959, iis09J02) and the proficiency test on Gasoil (sample #11014, iis11G01) from the Institute for Interlaboratory Studies (Spijkenisse, The Netherlands) were applied for RFF-GC×GC-FID analysis.

3.2 Materials

The theoretical model is validated and tested on an Agilent 7890B GC (Agilent Technologies, Little Falls, DE, USA). The system was equipped with a split/splitless inlet and two flame ionization detectors. The reverse fill/flush modulator consist of a 2-way purged and 2-way non-purged splitter CFT plates (Agilent Technologies, Little Falls, DE, USA). A programmable logic controller (OMRON 10C1DR-D-V2 24VDC, Kyoto, Japan) was used to control the 3-way solenoid valve (Parker model 009-933-900 24VDC, Cleveland OH, USA) for switching between the fill and flush state. The reversed fill flush modulator was set to a fill time of 4 s and a flush time of 0.1 s.

The sample collection loops and restriction capillary columns were constructed from deactivated fused silica (Agilent Technologies, USA). The 8 m x 0.180 mm I.D. first-dimension column was coated with 0.40 µm film of DB-1 (Agilent Technologies, USA). The 2 m x 0.32 mm I.D. second-dimension column was coated with 0.5 µm film of HP-INNOWAX (Agilent Technologies, USA). The carrier gas was hydrogen (99.999% Purity; Stoutjesdijk, Hoogvliet, The Netherlands). The flow for the second-dimension column was supplied by an auxiliary pressure control module (PCM) with no internal restriction.

Agilent Openlab EZchrom version, C.01.08 was used for data acquisition. For data transformation and visualization MSMetrix GC×GC-AnalyzerTM _{(Utrecht, The Netherlands) software was used to convert the raw}

23

3.3 Analytical Procedure

The theoretical model described in chapter 2 is valid within realistic parameter ranges. It only requires first and second-dimension column dimensions, carrier gas type, coating efficiencies, molecular structure of the compound and retention factors as input. In order to check if the theoretical model gives reasonable results, the practical H/u curves of the first and second-dimension column are compared with the theoretical model. The workflow for predicting H/U curves and retention times is shown in Figure 7. For the practical determination of the H/U curves of the first and second-dimension column, isothermal runs at 60 °C were performed. For the second-dimension H/U curve, 0.1 µl of gaseous o-xylene was injected and the flow rate was varied from 0.5 – 30 mL/min. The H/U curve was obtained using the o-xylene peak width and methane elution time as t0 time. For the H/U curve of the first-dimension column, the flow from the auxiliary PCM was set constant in the fill position (modulator off). In this position the auxiliary flow is constantly passing through the second-dimension column. 0.1 µl of n-octane was injected in the first-dimension column and the flow was varied from 0.2 – 4.0 mL/min. The generated n-octane peak widths were measured at the back-FID.

For the GC×GC experiments the setup is listed in Table 1. The first-dimension column flow is set to a flow rate of 0.2 mL/min at constant flow mode. The second-dimension column flow is controlled by the auxiliary PCM and is set to 20 mL/min. The sample collection loop was constructed at two times the loop volume; 20 cm x 0.45 mm I.D. deactivated fused silica. The collection loop fill time was set to 4 s and the loop was flushed within 0.1 s. The restriction capillary length (1.3 m x 0.1mm I.D. deactivated fused silica) was constructed such that primary column effluent + 10 % from the auxiliary PCM flows through the collection channel. The temperature of the GC columns was programmed from 40 °C (1 min) to 240 °C (10 min) with a rate of 10 °C/min or for some measurements set to 60 °C isothermal.

Experimental Conditions

First-dimension column DB-1 8 m x 0.18 mm x 0.40 µm

First-dimension column flow 0.2 mL/min @ constant flow mode

Second-dimension column HP-INNOWAX 2 m x 0.32 mm x 0.50 µm

Second-dimension column flow 20 mL/min @ constant flow mode

Collection channel dimensions 20 cm x 0.45 mm ID (31.81 µl)

Restriction capillary column 1.3 m x 0.10 mm ID

Fill time 4.0 s

Flush time 0.1 s

Oven temperature program 40 °C (1min) – 10 °C/min – 240 °C (10min)

Injector 250 °C 0.1 µl split 200:1

FID Front Flows H2 20 mL/min, Air 450 mL/min, N2 10 mL/min

FID Back Flows H2 40 mL/min, Air 450 mL/min, N2 10 mL/min

24

For studying the influence of the collection channel in the RFF modulator on the second-dimension peak width, the collection channel is varied in length and diameter. The volume of the collection channel was kept constant, but changes in length and diameter are made (Table 2). The analyses were performed under isothermal conditions (60 °C). The other parameters were kept constant, see Table 1 above.

Loop I.D. (mm) Loop Length (m) Loop Volume (mL)

0.25 0.65 31.81

0.32 0.40 31.81

0.45 0.20 31.81

0.53 0.14 31.81

The same setup was also used for optimizing the restriction column in flow modulated GC×GC. The restriction capillary is varied from 0.5 m to 1.9 m at 0.1 mm I.D. Increasing the length of the restriction capillary creates more resistance for the primary column effluent and PCM flow to pass through the collection channel. Decreasing the length or increasing the I.D. of the restriction capillary results in less resistance towards the collection channel and more primary column effluent plus PCM flow passing through the collection channel. Table 3 shows the lengths of the restriction capillary used with a 0.10 mm I.D. The third column in table 3 shows the theoretically calculated flow (first-dimension column flow + PCM flow) passing through the collection channel at a certain restriction capillary length and diameter. A restriction capillary of 1.9 m generates more restriction than a 0.4 m restriction capillary. This results in a factor 4-5 reduction in flow through the collection channel. The practical tests are performed using a 0.1 µl gaseous n-octane injection at isothermal conditions (60 °C) or using a kerosene sample. For the kerosene injections the temperature of the GC columns was set to 40 °C (1 min) to 240 °C (10 min) with a temperature programming rate of 10 °C/min.

Restriction Capillary I.D. (mm) Restriction Capillary Length (m)

Flow through collection channel (mL/min)

0.10 0.40 0.756

0.10 1.10 0.282

0.10 1.90 0.164

Table 2: Loop dimensions used for practical peak width determination.

25

Column dimensions: 1_L, 2_L, 1_ID, 2_ID, 1_Df, 2_Df

Loop dimensions: LLoop, IDLoop

Restriction capillary dimensions: Lrs, IDrs

Oven Temperature: K Column coating efficiencies: 1_CE, 2_CE

Retention factors: 1_k, 2_k

Molecular formula of compound Range of first-dimension column flows: ¹F Range of second dimension column flows: ²F

Range of PCM flows: F’

Initialization

Viscosity: ŋ39

Total Flow (Fill): Ftot (Eq.4) Total resistance of system: Rtot (Eq. 9) First midpoint pressure (Fill): Pa (Eq.11) Flow second dimension column (Fill): F4 (Eq. 5)

Flow collection channel (Fill): F2 (Eq. 6) Second midpoint pressure (Fill): 2_{Pa (Eq.6)}

Flow restriction capillary (Fill): F3 (Eq.7) Total Flow (Flush): Ftot (Eq. 15) First midpoint pressure (Flush): Pa (Eq. 20) Second midpoint pressure (Flush): 2_{Pa (Eq. 19)}

Flow second dimension column (Flush): F4 (Eq. 14) Flow collection channel (Flush): F2 (Eq. 17) Flow restriction capillary (Flush): F3 (Eq. 13)

Calculate 1_f 1, 2f1 (Eq.23) 1_f 2, 2f2 (Eq. 24) 1_U

out, 2Uout (Eq. 25) 2_Dm ,o (Eq. 28) 1_Dm ,o (Eq. 29) 1_Ds, 2_{Ds (Eq. 30)} 1_H, 2_{H (Eq. 22)} 1_N, 2_N 1_t r,2tr (Eq. 27 ) 1_σ, L_σ 2_{σ, (Eq. 31, 33, 36)}

Figure 7: Work flow scheme for the calculation of the H/U curves of the first and second dimension column and other parameters in FM-GC×GC where the flow is varied.

26

4. Results

4.1 Validation of the computer model H/U curves

The theoretical model calculates the plate heights of the first and second-dimension column at a specific first and second-dimension flow rate (1_F, 2_{F). Figure 7 depicts the workflow of the computer model. First the}

model calculates the resistances and flow rates through the collection channel and restriction capillary in the fill position using equations 4 – 11. Next, equations 13 – 20 provide the flow rates and pressure in the flush position of the GC×GC system. For the H/U curve calculations only the equations in the fill positions 0are used. After the pressure calculations the model calculates the pressure correction factors. The corresponding average velocities are calculated through equations 25 and 26. Next, the diffusion coefficients in the mobile phase of the second-dimension column and first-dimension column are calculated using equations 28 and 29, respectively. All the above values are substituted into equation 22 to give the plate height for the first and second-dimension column. These calculations yield one specific point in the H/U curve. The theoretical model automatically calculates a series of points within a user-specified flow range. Next, the theoretical plate heights of the first and second-dimension column are plotted against the average linear velocities, as is demonstrated in Figures 8 and 9.

The theoretically obtained H/U curve for the first-dimension column is based on a practically determined k-value of 2.5 and a column efficiency factor (1_{CE) of 0.8. For the experimentally obtained H/U curve of the}

first-dimension column the n-octane peak widths were observed at flow rates ranging from 0.3 – 4.0 mL/min. The details of the GC×GC settings are described in experimental.

0 50 100 150 200 250 300 350 400 450 500 0 20 40 60 80 100 120 140 H [µm]

Average linear velocity (cm/s)

Figure 8: Comparison of experimental (dots) and theoretical (line) data for the H/U curve of the first-dimension column in GC×GC with n-octane as analyte and hydrogen as carrier gas.

27

For the H/U curve of the second-dimension column, the column was directly connected to the front inlet. The theoretical H/U curve is based on a k-value of 2.5 and a column efficiency factor (2_{CE) of 0.8. Due to the}

high polarity of the second-dimension column o-xylene is used for peak width determination. The flow rate studied range from 0.5 – 30 mL/min. Figure 9 depicts the theoretically calculated and practically determined H/U curve of the second-dimension column.

Figure 9: Comparison of experimental (dots) and theoretical (line) data for the H/U curve of the second-dimension column in GC×GC with o-xylene as analyte and hydrogen as carrier gas.

The results in Figure 8 and 9 show that the calculated and experimental H/U curves are similar for both the first and second-dimension column. For the first-dimension column the optimum linear velocity is around 64 cm/s, or in a flow of 1.5 mL/min. However, it is advisable to lower the first-dimension column to get broader peaks which is favorable for the number of modulations across a peak. The optimum linear velocity of the dimension column is 74 cm/s which corresponds to a flow of 4 mL/min. This low second-dimension column flow cannot be used to completely flush the collection channel in a fraction of the fill cycle. 0 500 1000 1500 2000 2500 3000 3500 4000 0 100 200 300 400 500 600 H (µm) Average Velocity (cm/s)

28

4.2 Collection channel optimization

Duhamel et al. 14 _{concluded that the contribution of differential forward flow modulation to the peak}

broadening was more important with wider I.D. collection channel columns. In RFF modulation this has never been tested, therefore a theoretical calculation is made. In paragraph 2.5.1 the theoretical approach for calculating the second-dimension peak width is discussed. In table 4 the theoretical results of peak broadening per dimension (first-dimension column, collection channel and second-dimension column) are shown. Four different loop dimension are selected which all have the same volume of 31.81 µL. As discussed in the theoretical background chapter the resistance of the collection channel is negligible compared to that of the restriction capillary (R2<<R3). Therefore, the contribution of the collection channel to the flows passing through the first-dimension column and collection channel is also negligible. This can be seen in the first-dimension column band broadening. The total band broadening in the first-dimension column is for all the selected collection channels the same (σ2_{tot = 4.81 s). After the first-dimension column, primary column}

effluent passes through the collection channel. In the collection channel the total band broadening is dependent on the dimensions of the collection channel. The injection band broadening in the collection channel is the same for all the collection channels (σ2_{inj Loop = 1.33 s). This is due to the fact that the flow}

passing through the collection channel is determined by the restriction column and the same for all the collection channels in RFF. The column band broadening in the collection channel (σ2_{col) is different for the}

four collection channels. A 0.53 mm I.D. collection channel generates a factor 2 wider peaks than a 0.25 mm I.D. collection channel. The total peak broadening in the collection channel is the sum of the generated band width due to injection and the chromatographic band broadening. Based on these theoretical results, it can be concluded that the collection channel does have an influence on the generated peak widths in forward flow modulation. However, in RFF modulation, the collection channel is reversed injected into the second-dimension column which reduces the peak broadening generated in the collection channel. In the reversed injection the injection band width of the second-dimension column is dependent on the generated peak widths in the fill state of the collection channel and the flush flow. The injection band broadening in the flush position of a 0.53 mm I.D. collection channel is slightly larger than for a 0.25 mm I.D collection channel. This is due to the larger chromatographic band broadening in the fill state of the collection channel using a 0.53 mm I.D. collection channel (Vflush 0.53 mm I.D. > Vflush 0.25 mm I.D.). The chromatographic band

broadening in the second-dimension column is the same for all collection channels. The injection band broadening into the dimension column is 100 times smaller than the chromatographic second-dimension column broadening. Based on these observations it can be concluded that the band broadening in RFF modulation is generated by the second-dimension column. The theoretically calculated difference in peak broadening in the second-dimension column using a 0.53 mm I.D. collection channel or a 0.25 mm I.D is negligible. Based on these theoretical results the conclusion is that the dimensions of the collection channel do not affect the total band broadening in RFF.

29

Table 4: Theoretical calculations of the generated peak width per column at different channel lengths and I.D’s in RFF. .

0.25 mm x 65 cm 0.32 mm x 40 cm 0.45 mm x 20 cm 0.53 mm x 14 cm

Volume Loop (µl) 31.81 31.81 31.81 31.81

1st column broadening

σ²inj [s] 3.00E-10 3.00E-10 3.00E-10 3.00E-10

σ²col [s] 4.81 4.81 4.81 4.81 σ²tot [s] 4.81 4.81 4.81 4.81 Loop broadening σ²inj [s] 1.33 1.33 1.33 1.33 σ²col [s] 0.07 0.18 0.69 1.30 σ²tot [s] 1.40 1.51 2.03 2.63 2nd column broadening

σ²inj [s] 1.74E-05 1.87E-05 2.49E-05 3.23E-05

σ²col [s] 1.28E-03 1.28E-03 1.28E-03 1.28E-03

σ²tot [s] 1.30E-03 1.30E-03 1.31E-03 1.31E-03

Based on these theoretical results, three different collection channels described in the experimental part are used to determine the second-dimension peak width of n-octane in RFF GC×GC. The details of the GC×GC settings can be found in chapter 3, Experimental. The experimental results summarized in table 5 and Figure 10 confirm the results of the theoretical calculations. They clearly show that the second-dimension peak widths are in the same order and differ less than 0.5 % in peak width, concluding that the collection channel has little or no influence on the total peak width. The band broadening is almost exclusively generated by the second-dimension separation.

Table 5: Measured peak widths of n-octane at half height using three different collection channels.

Loop Dimensions Loop volume (µl) W1/2 Measured (s) W1/2 calculated (s)

0.32 mm x 40 cm 31.81 0.0834 0.0847

0.45 mm x 20 cm 31.81 0.0876 0.0849

30

Figure 10: Overlay of chromatograms of the n-octane peak with three different loop dimensions. Blue chromatogram:40 cm x 0.32 mm I.D, Green chromatogram: 20 cm x 0.45 mm I.D, Pink Chromatogram: 14 cm x 0.53 mm I.D.

31

4.3 Restriction capillary optimization

The dimensions of the restriction capillary are highly critical in optimizing the flow modulated GC×GC system. A small change in the length of the restriction capillary has a large influence on the midpoint pressure (Pa). This means that the optimization of the restriction capillary via practical trial-and-error is a time-consuming route to take.

The restriction capillary is important in both the fill and flush position. In the fill position the resistance of the restriction capillary should be large enough to ensure that the majority of the PCM flow is passing through the second-dimension column, otherwise the PCM flow is passing through the collection channel resulting in breakthrough of the collection channel towards the restriction capillary. On the other hand, the restriction should be low enough to ensure that first-dimension column effluent is flowing into the collection channel and not directly to the second-dimension column This latter phenomenon will result in unmodulated peaks in the second-dimension (second-dimension breakthrough). In the fill positions there are two requirements for filling the collection channel. First the flow rate through the collection channel should be at least the same as the first-dimension column flow rate. Secondly, the flow rate through the loop should not be greater than the maximum volume that can be stored in the collection channel in the selected fill time, otherwise the loop is overfilled and part of first-dimension column effluent leaves the system via the restriction capillary.

In the flush position, the restriction capillary has one requirement. In the flush position the flow from the PC through the collection channel should be high enough to completely empty the collection channel and reinject its content into the second-dimension column within the set flush time. Only a small part of the PCM flow is passing through the restriction capillary. An excessively long flush time has no direct impact on the second-dimension separation, but too a long flush time can result in unmodulated primary column effluent. This latter is not included in the theoretical model.

Based on the restriction capillary considerations described above, the work range of the restriction capillary is defined as a function of the capillary I.D’s. The work range is defined for the settings described in the experimental part.

32

The work range of the restriction capillary in the fill and flush position are shown separately in Figure 11 and 12 with the same configuration described in Table 1. The work range per restriction column I.D. in the fill position is described in Figure 11. The different color lines define the different restriction column I.D.’s used. On the x-axis the length of the restriction capillary is displayed. On the y-axis the flow rate through the collection channel is displayed with a minimum and maximum flow rate as requirements. The minimal flow through the collection channel is the flow rate from the first-dimension column (green dotted line). If the flow through the collection channel is below the flow rate of the first-dimension column, this means that part of the primary column effluent flows directly into the second-dimension column. The maximum requirement is the maximum flow through the collection channel to ensure the collection channel is not overfilled in the fill cycle (red dotted line). In the fill position, a short or wide restriction capillary results in less resistance towards the collection channel. This means that part of the PCM flow is passing through the collection channel (instead of into the second-dimension column). This increases the total flow rate through the collection channel. If the total flow through the collection channel is above 0.44 mL/min, the collection channel is overfilled and breakthrough through the restriction capillary occurs. The lengths and I.D.’s of the restriction capillary between the green and red dotted lines are advisable in RFF-GC×GC with the current flow settings described in Table 1. Figure 11 shows that the work range for a 0.25 mm I.D. column is much wider compared to a 0.100 mm I.D. column. However, using a 0.25 mm I.D. column a length of 40 – 60 m is required. This is not always favorable.

Figure 11: Working range of restriction capillary in the Fill position. The length and I.D. of the restriction capillary should be between the green and red dotted line to ensure that all primary column effluent flows into the collection channel and that the collection channel is not overfilled in the fill cycle. The flow through the collection channel should be at least the same as the first-dimension column flow rate (0.2 mL/min). The maximum flow rate through the collection channel to avoid collection channel breakthrough is 0.44 mL/min. 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 0 10 20 30 40 50 60 70 80 Flow th ro u gh colle ctio n ch an n el (m L/mi n )

Restriction Column Length [m]

100 µm 150 µm 180 µm 250 µm min max

Excessive flow through loop →

Breakthrough through restriction column

Partial transfer to loop →

Breakthrough to 2D column

33

Figure 12 depicts the work range of the restriction capillary in the flush position. In this position the requirement is that the collection channel is completely flushed during the very short flush cycle. This can be achieved by using a high flow rate through the collection channel or increasing the flush time. However, a long flush time decreases the fill time, which is unfavorable. Besides, in the flush position the high PCM flow is passing through the collection channel and pushes back primary column effluent into the first-dimension column for a very short period of time. Increasing the flush time will eventually lead in unmodulated primary column effluent leaving the system via the second-dimension column. Therefore, the flush time should be as short as possible to completely flush the collection channel and to avoid second-dimension breakthrough. The theoretical results are shown in Figure 12, the selected flush time is set to 0.1 s, which represents the green dotted line. Above this green dotted line means that the required flush time is longer than 0.1 s and that the loop is not completely flushed during the flush cycle. When the restriction capillary is too short or too wide, the PCM flow is passing through the restriction capillary instead of flushing the loop. This will eventually result in breakthrough through restriction capillary or ghost peaks in the second-dimension column. The work range of a wide I.D. restriction capillary is larger compared to smaller I.D restriction capillaries. This is the same as in the fill position.

Figure 12: Working range of restriction capillary in the Flush position. In the flush position there is only a minimal length needed. If the restriction capillary is too short, the PCM flow will flow mostly through the restriction capillary instead of flushing the collection channel. Below the dotted green line, flushing of the collection channel is finished within 0.1 s.

Longer flush time means that the restriction of the restriction capillary is too low. This will eventually result in breakthrough through restriction capillary or ghost peaks in the second-dimension column.

0 0,1 0,2 0,3 0,4 0 10 20 30 40 50 60 70 80 90 100 Re q u ie rd Flush time [ s]

Restriction column length [m]

100 µm 150 µm 180 µm 250 µm 0.1 s restriction

Collection channel not completely

flushed

34

Figure 13 depicts the work range of the restriction capillary where the fill and flush position are combined in one Figure. A proper function of the RFF-GC×GC is achieved when I.D and length of the restriction capillary are selected in the area between the green lines. Area under the light green line means that the length is too short or the I.D of the restriction capillary is too wide, resulting in breakthrough towards the restriction capillary (too high flow in collection channel). Above the dark green line means that the length is too long, or I.D. is too small, resulting in more restriction towards the collection channel, this translates to unmodulated primary column effluent into the dimension column (breakthrough toward second-dimension column). The work range of a 100 µm restriction capillary is smaller compared to a 250 µm column, also a small difference in length using a 100 µm may cause problems with loop overfilling or under filling.

Figure 13: Theoretically determined work range for the restriction capillary column. In this work range the loop is nether overfilled or under filled during the fill cycle and is completely flushed in the flush cycle.

Based on the Figure 13, the theoretical model provides a complete overview of the behavior of the restriction capillary in RFF modulation. To validate the restriction-capillary model, three different restriction capillary lengths are selected and tested on a n-octane standard and a kerosene sample. As shown in Table 3, the length of the restriction capillary is decisive for the flow rate passing through the collection channel. 40 cm x 100 µm restriction capillary is too short, resulting in too little resistance (lower red part of Figure 11). 1.9 m x 100 µm restriction capillary gives too much resistance, resulting in under filling the loop (upper part in Figure 11). A 1.30 m x 100 µm restriction capillary is between the green lines and is used as reference point. Smaller I.D. columns requires shorter lengths to obtain the same restriction. At this point one remark should be made. Column I.D. values stated by the manufacturer are not always reliable. Hence, the resistance of short column lengths may be affected by the variances in column I.D.

0 10 20 30 40 50 60 70 80 100 120 140 160 180 200 220 240 260 Re strictio n colu m n le n gth (m) Restriction column ID (µm) Minimal Length [m] Maximal Length [m]

Partial transfer to loop →

Breakthrough to 2D column

Excessive flow through loop →

35

Figure 14 shows the experimental results of the 40 cm x 100 µm restriction capillary. The blue chromatogram shows the correctly modulated octane peak; however the green back-FID signal below shows also the n-octane peak. The back FID is connected to the restriction capillary. A signal on the back FID means that the loop is overfilled. This is also tested for the kerosene sample shown in Figure 15. This Figure consists of two chromatograms, the top chromatogram is the full chromatogram on the front-FID and the breakthrough visible on the back-FID. The bottom chromatogram of Figure 15 shows a zoomed-in region of 10-15 min. the breakthrough is broad and measured over the full chromatogram. From the kerosene sample a 3D contour plot is shown in Figure 16. This Figure depicts that the kerosene sample is modulated correctly, but for quantitation purposes, this restriction capillary cannot be used.

Figure 14: Measured GC×GC chromatogram of n-octane with 40 cm x 100 µm I.D. restriction capillary. The top blue chromatogram is the signal from the front-FID connected to the second-dimension column. The green chromatogram is the breakthrough signal from the loop and back-FID connected to the restriction capillary.

36

Figure 15: Top: Measured GC×GC chromatogram of kerosene with 40 cm x 100 µm I.D. restriction capillary. Bottom: zoomed in GC×GC chromatogram (10 – 15 min) of kerosene with 40 cm x 100 µm restriction capillary. The top blue chromatogram is the signal from the front-FID connected to the second-dimension column. The green chromatogram is the breakthrough signal from the loop and back-FID connected to the restriction capillary.

37

38

For the 1.9 m x 100 µm restriction capillary, there is a flat baseline on the back-FID, which means that there is no breakthrough through the loop during the fill cycle (Figures 17 and 18). However the front FID shows a high baseline between the modulations of n-octane. The high baseline rise between the two modulations is unmodulated primary column effluent which has directly flowed into the second-dimension column. For primary column effluent the resistance towards the same loop is higher than the resistance towards the second-dimension column resulting in the transfer of unmodulated primary column effluent into the second-dimension column. in the GC×GC countour plot in Figure 19 the transfer of unmodulated primary column effluent into the second-dimension can be seen as streaking/ tailing in the second-dimension of the contour plot. The light blue vertical lines in the contour plot of the 1.9 x 0.1 mm I.D. are a result of streaking.

Figure 17: Measured GC×GC chromatogram of n-octane with 190 cm x 100 µm I.D. restriction capillary. The top blue chromatogram is the signal from the front-FID connected to the second-dimension column. The green chromatogram is the breakthrough signal from the loop and back-FID connected to the restriction capillary.

39

Figure 18: Measured GC×GC chromatogram of kerosene with 190 cm x 100 µm I.D. restriction capillary. The top blue chromatogram is the signal from the front-FID connected to the second-dimension column. The green chromatogram is the breakthrough signal from the loop and back-FID connected to the restriction capillary.

40

For the final check on the work range of the restriction capillary, a 1.30 m x 100 µm column is selected, which is in between the green lines of the working range of the restriction capillary. The results in Figures 20 and 21 show correctly modulated n-octane and kerosene peaks, which means that all primary column effluent is injected in the collection channel. Besides, the back FID shows a flat baseline, which means that the loop is not overfilled with primary column effluent during the fill cycle. It also shows that the collection channel is completely flushed during the flush cycle. When the collection channel is not completely flushed during the flush cycle, some part of primary column effluent remains in the beginning of the collection channel. During the next fill cycle, this part is passing through the collection channel and enters the restriction capillary in the next fill cycle. This results in ‘’ghosts peaks’’ on the back FID. Another phenomenon which occurred if the collection channel is not completely flushed during the flush is tailing of the second-dimension peaks. This happened when part of primary column effluent remains in the beginning of the collection channel and is flushed with the PCM flow towards the second-dimension column.

Figure 20: Measured GC×GC chromatogram of n-octane with 130 cm x 100 µm I.D restriction capillary. The top blue chromatogram is the signal from the front-FID connected to the second-dimension column. The green chromatogram is the breakthrough signal from the loop and back-FID connected to the restriction capillary.

41

Figure 21: Measured GC×GC chromatogram of kerosene with 130 cm x 100 µm I.D. restriction capillary. The top blue chromatogram is the signal from the front-FID connected to the second-dimension column. The green chromatogram is the breakthrough signal from the loop and back-FID connected to the restriction capillary.