Multicapillary gas chromatography - ACS Publications - American

Subscriber access provided by - Access paid by the | UCSB Libraries

Article

Multicapillary gas chromatography - temperature modulated metal oxide semiconductor sensors array detector for monitoring of volatile organic compounds in closed atmosphere using Gaussian apodization factor analysis Amir Hossein Alinoori, and Saeed Masoum Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.8b00426 • Publication Date (Web): 14 May 2018 Downloaded from http://pubs.acs.org on May 14, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Multicapillary gas chromatography - temperature modulated metal oxide semiconductor sensors array detector for monitoring of volatile organic compounds in closed atmosphere using Gaussian apodization factor analysis Amir Hossein Alinoori, Saeed Masoum * Department of Analytical Chemistry, Faculty of Chemistry, University of Kashan, Kashan, Iran. E-mail address: [email protected] ABSTRACT: A unique metal oxide semiconductor sensor (MOS) array detector with eight sensors was designed and fabricated in a PTFE chamber as an interface for coupling with multicapillary gas chromatography. This design consists of eight transfer lines with equal length between the multicapillary columns (MCC) and sensors. The deactivated capillary columns were passed through each transfer line and homemade flow splitter to distribute the same gas flow on each sensor. Using eight ports flow splitter design helps us to equal the length of carrier gas path and flow for each sensor, minimizing the dead volume of the sensor’s chamber and increasing chromatographic resolution. In addition to coupling of MCC to MOS array detector and other consideration in hardware design, modulation of MOS temperature was used to increase sensitivity and selectivity and data analysis was enhanced with adapted Gaussian apodization factor analysis (GAFA) as a multivariate curve resolution algorithm. Continues air sampling and injecting system (CASI) design provides fast and easily applied method for continues injection of air sample with no additional sample preparation. Analysis cycle time required for each run is less than 300 sec. The high sample load and sharp injection with the fast separation by MCC decrease the peak widths, and improve detection limits. This homemade customized instrument is an alternative to other time consuming and expensive technologies for continuous monitoring of outgassing in air samples.

In the closed atmosphere (like airline, spacecraft and space station cabins) the health and safety of the crew depend on real time monitoring of trace gas contaminants.1 The signs of air contamination must be regularly monitored for finding outgassing of volatile organic compounds (VOCs) from different items such as plastics, boards, structures, etc., which could be a threat to the health of the crew.2,3 Therefore, measuring the amount and type of outgassing VOCs in closed environments is of particular importance.4,5 Popular standard VOCs detection methods include costly, difficult maintenance and time consuming devices such as mass spectrometry,1,2,6 thus developing effective and inexpensive systems for detection of trace gas contaminants present in the closed atmosphere is necessary. Due to low cost, good sensitivity, easy maintenance and appropriate response time, electronic nose gas sensors are considered as promising alternatives that used in VOCs detection applications.7 Among the different variety of electronic nose sensors, MOS sensors are popular in gas monitoring systems.8–10 In spite of advantages of simplicity, portability and reduced cost, weight and power, the performance of these sensors including sensitivity, selectivity and reliability should be promoted more to get the necessary criteria to determine the VOCs pollutant in the closed atmosphere. Furthermore, in real sample analysis one of the problems of MOS array alone is the matrix effects, because interpretation of the MOS array response of the target com-

pound is complicated, especially in the presence of interferences 11. For this purpose, extensive researches have been done, which can be categorized to: 1) development in sensing material of MOS sensor like applying different metal oxide materials 12,13 , modification of metal oxide materials 14–16 and fabrication of nanostructures 17–19, 2) development of sensing instrumentation, such as modulation 20–23, the use of sensor arrays 7,24,25 and hyphenated techniques 26–28 and 3) development of chemometric data processing and discrimination algorithms 29–32 like pattern recognition, machine learning 33. One of the effective parameters to improve the sensitivity and selectivity of the MOS sensor is temperature of metal oxide surface34,35. It is clear that every chemical has an optimal oxidation temperature at the metal oxide surface. In the temperature modulation, with the increase in the surface temperature of the metal oxide, the reaction kinetic between chemical species and metal oxide surface changes, and subsequently the metal oxide resistance changes. Temperature modulation by changing kinetic of transient phenomena (adsorption, desorption, diffusion, and reaction) between metal oxide surface and gas molecules, can affect on electron transduction processes and metal oxide resistance.36 Thus, by using temperature modulation, a particular resistancetemperature profile for a chemical specie in each MOS sensor can be acquired.37 This particular conductance-temperature profile can be considered as a sensitivity profile (oxidation spectrum) of analyte on the surface of metal oxide sensor.

1 ACS Paragon Plus Environment

Analytical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 9

Figure 1. Schematic diagram of the customized CASI-MCC-MOS array detector, a) sampling mode b) injection mode.

By applying array of sensors, the number of sensitivity profile points increases, which leads to selectivity increase, but it is still not enough to analyze the complex matrix 29. Despite the increasing data order, it is not possible to use the multivariate curve resolution (MCR) technique for only response profiles and we also need to change the elution profiles of the constituents of the chemical mixture for resolution and enhancement in linearity of system 38. Gas chromatography (GC) is one of the several techniques in VOCs analysis. Coupling of GC to a MOS array detector is an operative way to overcome MOS alone limitations 39. While two compounds have similar MOS response, typically have different retention times, and are separated before entrance to a MOS array detector. Furthermore, the separation by the GC can minimize the number of interference compounds introduced to a MOS array detector 40,41. On the other hand, linearity of output signals increases. As a result, GC can hyphenate to a MOS array detector for analysis of VOCs in air samples. The coupling of GC to a MOS array detector has proven for VOCs detection with better resolution, extending dynamic range and preventing from detector saturation 42. For continuous monitoring of VOCs, the separation time in GC should be minimized, so the traditional GC is not suitable for this task, therefore, fast GC methods should be used. The MCC can withdraw the inherent limitations of conventional GC column include slow separation, low sample load, and high carrier gas pressure. The MCC with a large number of thin parallel capillaries in a small length tube can solve conventional GC column weaknesses by enabling fast separation of relatively larger volumes of sample load with higher carrier gas flow rate 43. To achieve a powerful tool for fast effective chromatographic separation an MCC was coupled to temperature modulated MOS array detector to decrease the run time for each analysis 43. However, overlapping in chromatographic peaks may occur, and can really complicate the interpretation of the fast GC-MOS array detector data. Appropriate and quick data analysis tools, based on finding the number of pure components in overlapped areas, can enable resolution of such data. Therefore, MCR method with high precision peak purity detection based on modified factor analysis was developed for resolution and discrimination of fast GC-MOS array detector data. The monitoring of VOCs in a closed atmosphere requires an operative injection system with continuous sampling, injecting and cleaning ability, fast GC with high data acquisition rate and fast response time detector to reach analytical cycle times less than 300 sec. For increasing the capabilities of the MCC-temperature modulated MOS array detector as an air quality monitoring instrument, a continues air sampling and injecting system (CASI)

module is needed for injection the real time collected air samples to MCC without any sample preparation. In this study, for continuous monitoring of outgassing in air samples, a homemade CASI-MCC GC was constructed and hyphenated to a modified design of temperature modulated MOS array detector with low dead volume, high selectivity and fast cleaning times and combined with Gaussian apodization factor analysis algorithm.

EXPERIMENTAL SECTION Materials. Standard mixture solutions were prepared from ethanol, acetone, n_hexane, benzene and toluene that were purchased from Merck in analytical grade. The 13X molecular sieve and charcoal were purchased from Fluka and are used for filtration of ambient air and produce pure air as a carrier gas. Instrumentation. CASI-MCC-MOS array detector has the ability to meet the needs of the actual air monitoring tool. This customized design makes a fast and direct sampling and detection of ambient air samples by using MCC- MOS array detector. The configuration of the CASI-MCC-MOS array detector is shown in Figure 1. Pure air was used as carrier gas in CASIMCC -MOS array detector. As shown in Figure 1 carrier gas was made by filtering of ambient air by molecular sieve and charcoal trap to eliminate contaminations and vapors, and there is no need to other carrier gas supplies like helium or nitrogen that used in conventional gas chromatography. The CASIMCC-MOS array detector can be divided into four main parts: continues air sampling and injecting system, multicapillary columns gas chromatography, temperature modulated MOS array detector, and signal processing using multivariate curve resolution based on Gaussian apodization factor analysis (GAFA). Continues air sampling and injecting system. CASI works as a simple, fast and versatile auto air sample collection and injection tool to introduce air samples into MCC GC or directly to the MOS array detector. It can be applied as a fast efficient modern replacement for complex air injection technology. CASI is capable for introducing air samples from real matrices to the fast-response detector. CASI consists of three major parts: air sampling pump, 2.5 ml air collector chamber and a six-port valve. When the analysis cycle starts, as shown in Figure 1a, firstly air samples are collected by suction with air sampler pump into the air collector chamber. For the sample analysis, the six-port valve position turns to injection mode and the carrier gas (filtered pure air) injects 2.5 ml collected air sample from the air collector chamber into the MCC, at the same time the air collector chamber is cleaned with filtered pure air for the next analysis (Figure 1b). After that, the valve switches back


Page 3 of 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry into the sampling position (Figure S1). One of the noticeable aspects of designing of this part is removal or reduction of the dead volume from the air collector chamber to the MCC. In the presence of dead volume, band broadening and tailing occur and the efficiency of chromatographic separation decreases, thus the detection limit increases. Dead volume also decreases concentration of the analytes when transferred to the MCC and MOS array detector 44. As CASI is coupled to the MCC with the internal diameter of 4 mm, internal diameter of the six-port valve, transfer lines and connections should be the same or near the diameter of the MCC to reduce dead volume in this part. One of the important points that are considered in construction of this part is deactivation of tubing and fitting of CASI. In order to accurately measure the analytes, the compounds must pass safely through the injection system to the MCC and then MOS array detector without any adsorption or surface degradation. Chemicals may have reversible interaction with internal surfaces of tubing and fittings of instrument, which causes the tailing and also can exhibit irreversible interactions (adsorption or reaction to a surface or catalytic decomposition), or a combination of both possible effects, which reduces the efficiency. These interactions are affected by various factors such as surface reactivity, time contact of surface with the analyte, the amount and type of analyte and surface temperature. Therefore, internal surfaces of transfer lines and fittings in CASI and interfaces of the MCC and MOS array detector were deactivated for preventing from reaction between reactive compounds such as alcohols or highly polar compounds with them 45.(See supporting information) Multi capillary columns gas chromatography. When sensitivity profiles (conductance-temperature profiles) of chemical species were indistinguishable from each other for each temperature modulated MOS array sensors, or when mixture of VOCs existed in air samples, an efficient fast separation is necessary. However, the separation efficiency, sample load capacity and analysis cycle time of the single capillary column need to be further improved. The first idea of MCC was introduced by Golay in 1988 46. MCC have higher sample load capacity than single capillary column and reduce analysis cycle time without losses of separation efficiency 47. In this study, rapid chromatographic separation was done by the MCC (OV-1, Multichrom, Ltd, Novosibirsk, Russia), made by 1200 parallel capillaries with a 0.2 µm film thickness and inner diameter of 40 µm for each capillary and was placed in a 100 mm long straight protective stainless steel tube. Coupling interface consists of two flow splitter. A flow splitter was used at the inlet of MCC for dividing carrier gas flow equally between the capillaries of the MCC and the other one was used in outlet of the MCC to divide carrier gas flow equally between the MOS sensor. This configuration with short length MCC was used for rapid, isothermal separation of VOCs mixtures in the 2.5ml air samples at 30 ºC and carrier gas flow rate of 70ml/min, within 300 sec. Temperature modulated MOS array detector. The customized designed temperature modulated MOS array detector involves of eight commercial MOS sensors mounted on PTFE chamber, an interface sensors board, a micro controller unit (MCU) board and a solid state fast switching board. MCU board is designed to be an Arduino compatible general board with fourteen 8-bit pulse width modulated (PWM) output pins for pulse generating and sixteen 10-bit resolution analog - digital

converter (A/D) pins for data acquisition and USB-serial interface. PWM parameters and data-acquisition algorithm were programed in the ATmega2560 16au microprocessor. The MCU is connected to the computer via a USB port. Sensitivity profiles were plotted and saved by specially developed software in LABVIEW 2009. The 5V DC 2800 mA was applied to eight MOS heaters for 0.01 sec. pulse width with 0.10 sec. duty cycle by solid state fast switching board. In conclusion, total power consumption for eight MOS sensors is 1.4W. The temperature increased during pulse width and it returned to room temperature at the end of duty cycles. Waveforms of the pulse voltage applied to the MOS heaters and MOS array output signal are presented in Figure S2 and Figure S3. Selection of sensors is significant impact on performance of MOS array detector and should have enough stability and sensitivity to cover detection of wide range of air contaminants 30. According to the dynamic range and type of detectable gases, eight gas sensors for the MOS array detector are selected (Table S1). The MOS array detector comprises of MiCS-5914, MiCS-5524, MiCS-2714 from SGX SENSORTECH, TGS2602, TGS2620 from FIGARO, SB-53-00, SP3S-AQ2-01 from FIS Inc, and MS1100-P111 from HS Electonics. Multivariate curve resolution based on Gaussian apodization factor analysis (GAFA). Getting the appropriate data analysis tools, which are based on chemometrics processing of data to identify overlapping / embedded areas and finding the number of pure components that are hidden in these areas, are problems of hyphenated chromatographic techniques. In this work, multivariate curve resolution (MCR) method based on Gaussian apodization factor analysis is proposed to determine the sensitivity profile S and the corresponding elution profile C of each constituent in MCC-temperature modulated MOS array data. This method can extract the sensitivity profile and the elution profile of each component from the data without prior learning steps. The extraction of the pure sensitivity profile and elution profile of each constituent in the data has many advantages, because it can help us to identify any existing chemical species and concentration of it in the data. Gaussian weighting as a Gaussian apodization function can be applied on factor analysis. Firstly, for data preprocessing, augmentation in both row and column wise modes, was done as it was previously described by Tauler 29. In Gaussian apodization factor analysis (GAFA), submatrices were extracted from a data matrix along of the elution time direction by moving a fixed size window with a default window size and each submatrix is weighted by the Gaussian window (Figure 2a). The Gaussian window gives more weight to the central row of moving window and then gradually decreases the weights of rows as distance from central row increases according to the Gaussian formula. This window can improve temporal resolution to obtain eigenvalues of each elution profile point and utilize to find appearing and disappearing of every component (Figure 2b). By using this strategy, the GAFA algorithm can estimate the rank of each submatrix that mostly indicates to central position in a data matrix. GAFA exhibits higher accuracy in the determination of the start or end points of rising and falling of each component because the weighted moving window characterizes almost one sensitivity profile and the other sensitivity profiles have negligible weights.



Figure 2. Diagram of GAFA algorithm. (a) submatrices are extracted by a Gaussian window from a data matrix along of the elution time direction; (b) plot of GAFA that illustrates log(eigenvalues) versus the row number of the Gaussian moving in a four-component data.

Thus, GAFA performance can be independent from window size and the problem of time shift and reduction of temporal resolution in fixed size moving window factor analysis (FSMWFA) by increasing window size can be improved in GAFA. Multivariate curve resolution of MCC- temperature modulated MOS array data achieved by GAFA algorithm is based on the second Manne's resolution theorem. This theorem was previously stated for model-free resolution of two-dimensional data in hyphenated chromatography48. This algorithm works based on this theorem by finding sensitivity profile of pure component one by one and elimination of obtained components from a data matrix until all of the components are determined. For species where a sensitivity profile of pure component is not present, purest sensitivity profile with maximum distance between first eigenvalue and second eigenvalue is used and the errors in the initial estimation can be significantly improved by applying constraints. This algorithm performs by the following steps: 1- Determination of the total rank (number of components) of a data matrix X (nc) by singular value decomposition (SVD). 2- Find a local rank map of X by GAFA and discriminate rows of X that have rank of one (selectivity region). 3- One row of X with rank one or for species where a selective region is not present, purest row with maximum distance between the first eigenvalue and the second one is used as ith pure component spectrum i: 1, 2, 3, …, nc and added to the estimated pure sensitivity matrix Ŝ as ith column. 4- Least squares calculation of the concentration matrix Ĉ from the estimated pure sensitivity matrix Ŝ and X: (1) Ĉ = (X. Ŝ). (ŜT. Ŝ)-1 5- Elimination of estimated pure components data matrix Ĉ.ŜT from data matrix X and finding new residual matrix Xnew.

Page 4 of 9

Xnew = X - (Ĉ .ŜT) (2) 6- If the number of sensitivity profiles (i) in Ŝ is less than of the number of pure components (nc) steps 2 to 5 are repeated, else algorithm is terminated. In chemometric resolution techniques, alternating least squares (ALS) is a simple, effective and widely used procedure that applies an iterative method to calculate C and S 49. The first step in GAFA-ALS is to find the rank of a data matrix, which is the number of pure chemical components. The second step is to set an initial value of Ĉ or Ŝ. As mentioned before, GAFA gives a unique solution by taking advantage of local rank information. The estimated elution profile (Ĉ) or spectral profile (Ŝ) by GAFA was used as an initial estimate of iterative ALS for further refinement under additional constraints such as non-negativity and/or unimodality. ALS is carried out through the applying of constraints and forces the iterative optimization to model the profiles. Even though good estimates can often be obtained with GAFA alone, constraints are applied after the initial estimation by GAFA to correct errors in the initial estimation and improve the results. Convergence parameters can be defined by the specific number of iterations or the relative difference between sums of the squares of residuals (RDSSR). RDSSR is equal to: (3) ((SSRold - SSRnew) / SSRold)×100 Where SSRold is sum of squares of residuals (SSR) at one iteration before new iteration and SSRnew is SSR at new iteration. If RDSSR < convergence criterion value (CCV) (0.1% as default); then the optimization stops. Hybrid algorithm that combines GAFA and ALS, is proposed for this study for multivariate curve resolution of two-dimensional chromatography data. The flowchart of this hybrid algorithm is illustrated in Figure 3. In this study, data pretreatments and GAFA algorithm are carried out in MATLAB environment and the well-known MCRALS GUI 2.0 was used for ALS procedure 50.

RESULTS AND DISCUSSIONS For resolving the pure sensitivity and elution profiles of MCC-MOS data during the measurement cycle, its normalized conductance response was studied by GAFA-ALS. As stated before, ALS algorithm at first needs initial estimation profiles of the sensitivity (S), or elution (concentration, C) profiles. To generate these initial estimation profiles, GAFA algorithm, which can adequately deal with the overlapping and embedded areas even with a narrow selective window size was used. As the conductance and elution profiles must take positive values, initial estimation profiles were generated in GAFA algorithm by applying the non-negativity constraint to sensitivity and elution profiles.

Figure 3. Flow chart of the GAFA-ALS algorithm used to resolution of MCC- temperature modulated MOS array datasets.


Page 5 of 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 4. (a) 3D plot of simulated ternary data of MCC-MOS for one modulated sensor; (b) plot of simulated elution profile; (c) plot of simulated response of one modulated sensor.

Simulated data. To get insight into the resolution method and performance of the algorithm in MCC-temperature modulated MOS array data, an overlapping and embedded MCCMOS array data with dimensions of 1000 × 40 (elution points × spectral points) was simulated as shown in Figure 4. The 3D plot of this modulated sensor’s response is shown in Figure 4a. The elution profiles were constructed by Clifford–Tuma model 51–53 with three peaks centered in 35, 100 and 200 (Figure 4b). The sensitivity profiles were simulated with three Gaussian peaks centered in 10, 20 and 35, and have the sigma of 7, 3 and 2.3, respectively in 40 thermally modulated point (Figure 4c). Initial estimates of the sensitivity and elution profiles of the species are determined based on the detection of selective region or purest variables by GAFA. Its SVD indicated the presence of three major components in it. At the first step, simulated MCC-MOS data is explored for local rank analysis by GAFA to find first pure or purest region (selectivity window 1), that is shown in Figure S4a. This algorithm proceeds by elimination of the pure component from data matrix and performing GAFA on residual matrix to find the next pure component region (next selectivity window), that is shown in Figure S4b. Pure components are found one by one until all the components are determined. The results of GAFA analysis at each step are illustrated in Figure S4 for simulated MCC-MOS data. By elimination of each pure component, the rank of the residual matrix decreases. For components where a rank-one window (selectivity window) is present, the shape of the unit profiles correctly recovered 54.

For species where a rank-one window is not present, purest variables are used and constraints are applied after the initial estimation by GAFA to correct errors in the initial estimation and refine the results. For this purpose, constraints such as non-negativity, selectivity matrix, have been very efficient. The zero concentration windows and selectivity matrix also can obtain by GAFA. The initial estimations of elution profiles data by GAFA are shown in Figure 5 for simulated MCC-MOS data. As it can be seen, Figure 5 demonstrates proper initial estimation of elution profiles by GAFA. As previously mentioned, using the selectivity region in each step of GAFA algorithm results in a unique solution for initial estimation. Estimated elution profiles of simulated data by GAFA were considered as the initial values for iterative calculation by ALS. A non-negativity constraint was adopted in both the elution and sensitivity profiles. Elution and sensitivity profiles of simulated data was finally obtained by repeating iterations until convergence.

Figure 5. Initial estimation for three-component simulated MCCMOS data, by GAFA, (a) initial estimation of sensitivity profile, (b) initial estimation of elution profile

Figure 6. Block diagram of quantitative estimation of the VOCs concentration by GAFA-ALS, (a) row wise agumentation, (b) column and row wise agumentation, (c) resolved sensitivity profiles, (d) resolved elution profiles, (e) a typical obtained calibration curve.



Page 6 of 9

Figure 7. a) 3D plot of row wise augmented experimental data of standard 1 obtained from the MCC-MOS array detector; b) initial estimation of elution profiles by GAFA algorithm for row wise augmented experimental data of standard 1, c) resolved elution profiles of row wise augmented experimental data of standard 1 by GAFA-ALS.

Experimental data. Five standard mixtures of VOCs in the air sample were prepared with different concentration levels as it was described in Table 1. The experimental data from the CASI-MCC-MOS array detector of these five standards, with dimensions of 1200×80 (elution points× sensitivity points) were studied. Table 1. Concentration of experimental standard mixtures Ethanol (ppm)

Acetone (ppm)

n-Hexane (ppm)

Benzene (ppm)

Toluene (ppm)

Standard 1

50

450

350

250

150

Standard 2

150

50

450

350

250

Standard 3

250

150

50

450

350

Standard 4

350

250

150

50

450

Standard 5

450

350

250

150

50

At first to do quantitative estimations and better resolution, experimental data of the CASI-MCC-MOS array detector was augmented in column and row wise modes 29. Temperature modulation data matrix of each sensor [1200x10] was augmented row wise so each row is a combination of temperature modulated signal of all eight sensors [1200 x (10x8)]. Elution profile of each standard sample was augmented column wise and each column is combination of elution profiles for five samples (Figures 6a and 6b). This process leads to a [(5×1200) × (10×8)] augmented matrix. Row wise augmented experimental data of standard 1 with dimensions of 1200×80 (elution points× sensitivity points) as a typical multi sensors case is shown in Figure 7a. The number of factors was estimated to be five by SVD, and local rank analysis was obtained by GAFA. GAFA with nonnegativity constraint was used to generate initial estimations of pure elution profiles. Pure elution profiles showed a good agreement with the expected retention times with the experimental ones, and allowed the detection of five compounds. The initial estimation of elution profiles achieved by GAFA for row wise augmented experimental data from standard 1 by MCC-MOS are shown in Figure 7b. Estimated elution profiles of row wise augmented experimental data of standard 1 by GAFA were considered as the initial values for iterative resolution by ALS.

Figure 8. a) 3D Plot of column and row wise augmented experimental data of analysis of five standards by MCC-MOS array detector, b) resolved elution profiles of column and row wise augmented experimental data of five standards by GAFA-ALS.

A non-negativity constraint was done in both the elution and sensitivity profiles. Elution and sensitivity profiles of row wise augmented experimental data of standard 1 were finally obtained by repeating iterations until convergence. The resolved elution profiles of row wise augmented experimental data of Standard 1 by GAFA-ALS are shown in Figure 7c.


Page 7 of 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry Table 2. Quantitative results, relative errors of calibration and coefficients of determination for five VOCs

Ethanol

Acetone

n_Hexane

Benzene

Reference (ppm)

50

150

250

350

450

Calculated (ppm)

41.3

152.8

259.7

357.0

439.2

Relative error (%)

-17.4%

1.9%

3.9%

2.0%

-2.4%

Calculated (ppm)

41.9

145.6

265.6

364.7

432.3

Relative error (%)

-16.3%

-2.9%

6.2%

4.2%

-3.9%

Calculated (ppm)

43.7

146.0

273.4

340.4

446.5

Relative error (%)

-12.6%

-2.7%

9.4%

-2.7%

-0.8%

Calculated (ppm)

41.7

155.3

254.3

358.8

439.9

Relative error (%)

-16.6%

3.5%

1.7%

2.5%

-2.2%

Calculated (ppm)

45.7

142.2

257.3

376.3

428.6

Relative error (%)

-8.7%

-5.2%

2.9%

7.5%

-4.8%

R2 0.996

0.991

0.993

0.997

0.987

Toluene

To study the possibility of quantitative analysis and better resolution, five specified standard mixtures in Table 1 were analyzed by the MCC-MOS array detector with the above described method. Experimental data obtained in these analyses were augmented in column and row wise modes, and eight sensors / five sets augmented data that were resolved by GAFAALS as previously described (Figure 6c and Figure 6d), initial estimations for this eight sensors / five set augmented data were done by GAFA by applying non-negativity constraint to the elution and sensitivity profiles. Column and row wise augmented experimental data from the analysis of these five standards with dimensions of 6000×80 (elution points× sensitivity points) were studied as an eight sensors / five sets case that is shown in Figure 8a. The number of factors was estimated to be five by SVD and local rank analysis was obtained by GAFA. GAFA with non-negativity constraints was used to generate initial estimations of pure elution profiles. Resolved pure elution profiles by GAFA-ALS (Figure 8b and Figure 6d) showed an agreement with the expected retention times and concentration for standard samples, and allowed the quantitative determination of these five compounds by using the area under pure elution profiles of each VOCs peak (Figure 6e). Four measurements have been performed for each standard sample to obtain four 6000×80 matrices and by described GAFA-ALS procedure, average of four replications in each concentration level for each compound was resulted in calibration curves. Linear calibration models were used to determine the concentrations of these compounds in all samples. Quantitative results of five VOCs, their relative errors of calibration and coefficients of determination (R2) are shown in Table 2 and Figure S5. As shown in Table 2 and Figure S5, the acceptable calibration errors with high coefficients of determination (R > 0.98) were obtained.

MOS array detector involves of eight customized designed MOS sensors. GAFA-ALS has been used as a good multivariate curve resolution algorithm to extract valuable information from temperature-modulated MOS sensors array detector dataset. On these systems, because this type of dataset, in particular has overlapping and embedded areas and changes with time, GAFA-ALS algorithm is proper to estimate this type of data. Results indicate excellent linearity over concentration ranges and repeatability that make it proper for quantitative analysis as well. The future goal is to make this device appropriate for air quality monitoring in space applications and utilize it as a powerful and sensitive method in laboratories and space missions.

ASSOCIATED CONTENT Supporting Information Additional figures and a table and their descriptions exist for the paper, including deactivation procedure, function of air sampler pump and six port valve during one cycle of analysis, selection of gas sensors for the MOS array detector, raw data acquired from CASI-MCC-MOS array detector, waveforms of the pulse voltage applied to the MOS heaters; temperature modulated MOS array data of standard 1 for elution point number 220, rank map of each step of three-component simulated MCC-MOS data that is obtained by GAFA and calibration curves for five VOCs. This material is available free of charge on the ACS Publications website.

AUTHOR INFORMATION Corresponding Author *Phone: +98 31 55912338; Fax: +98 31 55912397. E-mail address: [email protected]

Notes The authors declare no competing financial interest.

CONCLUSIONS

ACKNOWLEDGMENT

In this paper, it has been demonstrated a continues air sampling and injecting system to introduce air samples into multicapillary columns and unique design for coupling the homemade temperature modulated MOS array detector with MCC to facilitate detection of volatiles in air samples. It has been shown that MCC columns can be operated at room temperature and using of pure air as a carrier gas, thus the CASI-MCC-MOS array detector is operative, simple and low cost for determination of VOCs in complex air samples. On the other hand, the MOS sensors array detector is considered as a detector for MCC gas chromatographs. This homemade temperature modulated

The authors are grateful to the University of Kashan for supporting this work by Grant NO. 159181/8.

REFERENCES (1)

Chutjian, A.; Bornstein, B. J.; Conroy, D. G.; Croonquist, A. P.; Darrach, M. R.; Edgu-Fry, E.; Francis, G. R.; Fry, D. J.; Garkanian, V.; Girard, M. A.; Haemmerle, V. R.; Heinrichs, W. M.; Kidd, R. D.; MacAskill, J. A.; Rust, T. M.; Schaefer, R. T.; Thomas, J. L.; Toomarian, N.; Walch, M. J.; Christensen, M.; Demonbrun, D.; Vanholden, R.; Holland, P. M.; Shortt, B. J. In International Conference On Environmental Systems; SAE International, 2007; p 7.


Analytical Chemistry (2)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(3) (4)

(5)

(6) (7) (8) (9) (10) (11) (12)

(13) (14) (15) (16) (17) (18) (19)

(20) (21) (22) (23)

(24) (25) (26)

(27)

Limero, T.; Reese, E.; Cheng, P.; Trowbridge, J. Int. J. Ion Mobil. Spectrom. 2011, 14 (2–3), 81–91. James, J. T. In International Conference On Environmental Systems; SAE International, 1998; p 8. Cincinelli, A.; Martellini, T.; Amore, A.; Dei, L.; Marrazza, G.; Carretti, E.; Belosi, F.; Ravegnani, F.; Leva, P. Sci. Total Environ. 2016, 572 (Supplement C), 333–339. Kim, K.-H.; Szulejko, J. E.; Jo, H.-J.; Lee, M.-H.; Kim, Y.-H.; Kwon, E.; Ma, C.-J.; Kumar, P. Environ. Pollut. 2016, 215 (Supplement C), 340–346. Rosenberger, W.; Beckmann, B.; Wrbitzky, R. J. Chromatogr. B 2016, 1019, 117–127. Wilson, A. D. Metabolites. 2015. Righettoni, M.; Amann, A.; Pratsinis, S. E. Mater. Today 2015, 18 (3), 163–171. Güntner, A. T.; Koren, V.; Chikkadi, K.; Righettoni, M.; Pratsinis, S. E. ACS Sensors 2016, 1 (5), 528–535. Graf, M.; Frey, U.; Taschini, S.; Hierlemann, A. Anal. Chem. 2006, 78 (19), 6801–6808. Jie, D. F.; Wei, X.; Zhou, H. L.; Pan, J. M.; Ying, Y. B. Appl. Spectrosc. Rev. 2017, 52 (2), 101–122. Lupan, O.; Postica, V.; Gröttrup, J.; Mishra, A. K.; de Leeuw, N. H.; Carreira, J. F. C.; Rodrigues, J.; Ben Sedrine, N.; Correia, M. R.; Monteiro, T.; Cretu, V.; Tiginyanu, I.; Smazna, D.; Mishra, Y. K.; Adelung, R. ACS Appl. Mater. Interfaces 2017, 9 (4), 4084–4099. Eranna, G.; Joshi, B. C.; Runthala, D. P.; Gupta, R. P. Crit. Rev. Solid State Mater. Sci. 2004, 29 (3–4), 111–188. Hernández, P. T.; Hailes, S.; Parkin, I. Sens. Actuators. B. Chem. 2017, 242, 1281–1295. Korotcenkov, G. Sens. Actuators. B. Chem. 2005, 107 (1), 209– 232. Korotcenkov, G. Mater. Sci. Eng. R Reports 2008, 61 (1), 1–39. In Advanced Technology for Human Support in Space; National Academies Press: Washington, D.C., 1997; pp 61–81. Hunter, G.; Xu, J.; Evans, L.; Vander Wal, R.; Berger, G.; Liu, C. C. In ECS Transactions; ECS, 2006; Vol. 3, pp 199–209. Jang, B.-H.; Landau, O.; Choi, S.-J.; Shin, J.; Rothschild, A.; Kim, I.-D. Sens. Actuators. B. Chem. 2013, 188 (Supplement C), 156–168. Benkstein, K. D.; Raman, B.; Lahr, D. L.; Bonevich, J. E.; Semancik, S. Sensors Actuators, B Chem. 2009, 137 (1), 48–55. Gosangi, R.; Gutierrez-Osuna, R. Sens. Actuators. B. Chem. 2013, 185, 201–210. Herrero-Carrón, F.; Yáñez, D. J.; Rodríguez, F. de B.; Varona, P. Sens. Actuators. B. Chem. 2015, 206, 555–563. Vergara, A.; Llobet, E.; Brezmes, J.; Ivanov, P.; Cané, C.; Gràcia, I.; Vilanova, X.; Correig, X. Sens. Actuators. B. Chem. 2007, 123 (2), 1002–1016. Matindoust, S.; Baghaei-Nejad, M.; Shahrokh Abadi, M. H.; Zou, Z.; Zheng, L.-R. Sens. Rev. 2016, 36 (2), 169–183. Ngo, K. A.; Lauque, P.; Aguir, K. Sens. Actuators. B. Chem. 2007, 124 (1), 209–216. Aggio, R. B. M.; de Lacy Costello, B.; White, P.; Khalid, T.; Ratcliffe, N. M.; Persad, R.; Probert, C. S. J. J. Breath Res. 2016, 10 (1), 17106. McGuire, N. D.; Ewen, R. J.; de Lacy Costello, B.; Garner, C. E.;

(28)

(29) (30) (31) (32) (33) (34)

(35) (36) (37) (38) (39) (40)

(41) (42)

(43) (44) (45) (46) (47) (48) (49) (50) (51) (52) (53) (54)

Page 8 of 9 Probert, C. S. J.; Vaughan, K.; Ratcliffe, N. M. Meas. Sci. Technol. 2014, 25 (6), 65108. Shepherd, S. F.; McGuire, N. D.; de Lacy Costello, B. P. J.; Ewen, R. J.; Jayasena, D. H.; Vaughan, K.; Ahmed, I.; Probert, C. S.; Ratcliffe, N. M. J. Breath Res. 2014, 8 (2), 26001. Montoliu, I.; Tauler, R.; Padilla, M.; Pardo, A.; Marco, S. Sens. Actuators. B. Chem. 2010, 145 (1), 464–473. Zhang, L.; Tian, F.; Pei, G. Measurement 2014, 54 (Supplement C), 31–39. Tohidi, M.; Ghasemi-Varnamkhasti, M.; Ghafarinia, V.; Bonyadian, M.; Mohtasebi, S. S. Int. Dairy J. 2017. Zhang, L.; Tian, F.-C. Anal. Chim. Acta 2014, 816, 8–17. Rogers, P. H.; Benkstein, K. D.; Semancik, S. Anal. Chem. 2012, 84 (22), 9774–9781. Güntner, A. T.; Pineau, N. J.; Mochalski, P.; Wiesenhofer, H.; Agapiou, A.; Mayhew, C. A.; Pratsinis, S. E. Anal. Chem. 2018, 90 (8), 4940–4945. Righettoni, M.; Tricoli, A.; Pratsinis, S. E. Anal. Chem. 2010, 82 (9), 3581–3587. Vergara, A.; Benkstein, K. D.; Montgomery, C. B.; Semancik, S. Anal. Chem. 2014, 86 (14), 6753–6757. Lee, A. P.; Reedy, B. J. Sens. Actuators. B. Chem. 1999, 60 (1), 35–42. Tauler, R. Chemom. Intell. Lab. Syst. 1995, 30 (1), 133–146. García-González, D. L.; Aparicio, R. Food Chem. 2010, 120 (2), 572–579. Taylor, C.; Lough, F.; Stanforth, S. P.; Schwalbe, E. C.; Fowlis, I. A.; Dean, J. R. Anal. Bioanal. Chem. 2017, 409 (17), 4247– 4256. Spinelle, L.; Gerboles, M.; Kok, G.; Persijn, S.; Sauerwald, T. Sensors 2017, 17 (7), 1520. Zampolli, S.; Elmi, I.; Stürmann, J.; Nicoletti, S.; Dori, L.; Cardinali, G. C. Sens. Actuators. B. Chem. 2005, 105 (2), 400– 406. Denawaka, C. J.; Fowlis, I. A.; Dean, J. R. J. Chromatogr. A 2014, 1338. Hoh, E.; Mastovska, K. Journal of Chromatography A. 2008, pp 2–15. Lars Blomberg. J. Chromatogr. A 1975, 115 (2), 365–372. Golay, M. J. E. J. High Resolut. Chromatogr. 1988, 11 (1), 6–8. Ruzsanyi, V.; Fischer, L.; Herbig, J.; Ager, C.; Amann, A. J. Chromatogr. A 2013, 1316 (Supplement C), 112–118. Manne, R. Chemom. Intell. Lab. Syst. 1995, 27 (1), 89–94. Tauler, R.; Kowalski, B.; Fleming, S. Anal. Chem. 1993, 65 (15), 2040–2047. Jaumot, J.; de Juan, A.; Tauler, R. Chemom. Intell. Lab. Syst. 2015, 140, 1–12. Eklöv, T.; Mårtensson, P.; Lundström, I. Anal. Chim. Acta 1997, 353 (2), 291–300. Clifford, P. K.; Tuma, D. T. Sensors and Actuators 1982, 3 (0), 255–281. Clifford, P. K.; Tuma, D. T. Sensors and Actuators 1982, 3 (C), 233–254. Liang, Y. Z.; Kvalheim, O. M.; Keller, H. R.; Luc Massart, D.; Kiechle, P.; Erni, F. Anal. Chem. 1992, 64 (8), 946–953.


Page 9 of 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


For TOC only

ACS Paragon Plus Environment

9

Multicapillary gas chromatography - ACS Publications - American

Recommend Documents