1 RECENT DEVELOPMENTS AND TRENDS IN MINIATURIZED GAS

2 PRECONCENTRATORS FOR PORTABLE GAS CHROMATOGRAPHY

3 SYSTEMS: A REVIEW
4 Irene Lara-Ibeas1, Alberto Rodríguez Cuevas1,2, Stéphane Le Calvé1*
5 Institut de Chimie et Procédé pour l’Energie, l’Environnement et la Santé (ICPEES, UMR 7515
1

6 CNRS/Unistra), group of atmospheric physical chemistry, 25 rue Becquerel, 67087 Strasbourg Cedex 02,
7 France.

8 2
In’Air Solutions, 25 rue Becquerel, 67000 Strasbourg, France.

10 Abstract

11 In the last years, a growing number of fields such as air quality monitoring, breath analysis or explosives

12 and chemical warfare agents detection requiring fast, on-site, sensitive analysis has led to the

13 development of portable gas chromatography systems. In most cases, these systems integrate a

14 miniaturized gas preconcentrator, which provides a significant enhancement of the sensitivity enabling

15 quantification of analytes present in the sample at trace levels. In this review, the authors have focused

16 on recent developments in these preconcentrators integrated in portable gas chromatography systems.

17 The main materials and fabrication techniques, designs, heating technologies, fluidic connections,

18 adsorbents, and applications are discussed. In addition, an analysis of some factors affecting

19 preconcentration performance is presented. A new figure of merit called Normalized Preconcentration

20 Efficacy (NPE) is proposed to evaluate the performance of these devices in a standardized manner,

21 making possible a more straightforward comparison between different devices.

22 Keywords: Preconcentration, Miniaturization, Microfabrication, Gas chromatography, VOC detection,

23 Normalized Preconcentration Efficacy (NPE)

24

25 1. Introduction

26 Gas chromatography (GC) has been largely used in the detection of volatile organic compounds (VOC)

27 for decades. The versatility of this technique lies in the broad spectrum of detectors and separation

28 columns that can be combined resulting in analytical instruments suitable for many applications such as

1
29 explosives detection [1,2], air quality monitoring [3,4], occupational exposure [5,6] or breath analysis

30 [7,8], among others.

31 Traditional GC benchtop instruments uses capillary columns coated with stationary phases of different

32 polarity to separate the analytes of interest. These columns are coupled to very sensitive detection

33 systems like mass spectrometers (MS) or flame ionization detectors (FID) which enable to identify

34 and/or quantify a great number of chemicals. In these systems, preconcentrators (PC) are usually located

35 upstream the separation column and have two main functions: to narrow the chromatographic band and

36 to improve the detection limit. In benchtop instruments, PCs, also called cold traps, are usually kept

37 sporadically at sub-zero temperatures to promote the adsorption of analytes. After trapping the analytes,

38 the sample is subsequently heated, and a flow of inert gas is employed to transfer the analytes into the

39 separation column. Temperatures of 150–350°C are reached in few seconds in a process called flash

40 desorption which ensures the total desorption of analytes and their transfer to the column in a very sharp

41 band. This technique is very efficient to produce narrow chromatographic peaks and consequently

42 produces a significantly increase of sensitivity, being able to detect chemicals at concentrations in the

43 order of parts per billion (ppb) or parts per trillion (ppt) when coupled to MS or FID. However, these

44 instruments are very bulky and require high power supply, limiting their use for on-site applications.

45 In the last decades, the implementation of stricter environmental laws, especially in Europe [9] and the

46 increasing demand of applications requiring rapid responses to specific events such as chemical warfare

47 agents or explosive detection has led researchers to develop portable gas chromatographs to perform on-

48 site analysis. Although many efforts have been performed since then, the main challenges of portable

49 instruments lie in keeping similar performances to benchtop instruments while the size and the energy

50 consumption are drastically reduced. However, in some cases, these performances are still far from the

51 ones of their benchtop counterparts, especially in terms of sensitivity, since the detection limits of

52 portable GC are usually in the order of 10 – 10,000 ppb whereas the benchtop instruments can very often

53 detect concentrations of less than 1 ppb (typically 0.01 – 1 ppb).

54 Due to their low power consumption and reduced size, compact detectors such as mini photoionization

55 detectors (miniPID) [4,10,11], thermal conductivity detector (TCD) [12], metal oxide (MOX) sensors

2
56 [8,13] and chemiresistors (CR) [6] are typically employed in portable gas chromatography (see Table 1

57 and Table 2). Nevertheless, their sensitivity without any previous preconcentration step varies from ppb

58 (miniPID) to ppm levels in most of them. Therefore, preconcentrators (PC) become a key component in

59 portable GC to reach limits of detection (LOD) at sub ppb levels.

60 During operation, PCs must be rapidly heated to release the trapped analytes in a narrow

61 chromatographic peak as abovementioned which requires high power supply. However, portable

62 instruments have limited power supply, therefore, the thermal mass of PC must be reduced. Other issue

63 arising from the operation of PC in portable devices is the thermal isolation of the PC to avoid heat

64 losses and prevent the damaging of surrounding components. Microfabrication techniques allow to

65 create devices with reduced thermal mass and integrated powerful heating systems [14,15]. However,

66 the problems related to the fluidic interfacing and the need for cleanrooms for their development has

67 brought researchers to adopt new approaches for PC miniaturization. Recently, 3D printing [16] or

68 milling [4] have been employed as alternative fabrication techniques in the fabrication of miniaturized

69 PC. At present, these techniques do not have the features to create microstructures with the same

70 accuracy as microfabrication techniques, but they enable easier fluidic interfacing and may reduce the

71 costs of the prototyping process.

3
72 Table 1. Portable GC coupled to photoionization detectors (PID) with integrated PC and their features.
Preconcentrator
Analysis Sample
Weight Power cons. Target Sensitivity
Ref. Size (cm3) time Volume l (mm) × d (µm) Heating Det. LOD (ppb)
(kg) Material Adsorbent (W) / heating compounds (pg)
(min) (mL) × w (mm) system
rate (°C/s)
Ceramic 0.06 – 0.4
GC-PID [4] 32 × 29 × 14 ∼5 19 20 Cavity (4.6 × 350 × 7.4) Al C-B 5 mg ∼ 44 / 4.8 BTEX PID 3.6 a
heaters (BTEX)

Tube (4 mm i.d., 40 mm CNT sponge CNT 0.13-0.28

GC-PID [17] 35 × 26 × 15 <5 5 90 Glass n. a. / 450 BTEX PID 37.4 a
length) 5 mg sponge (BTEX)

Manifold-shaped cavity Basolite C300 Heating 0.1-1.6

MAP [10] 32 × 29 × 14 ∼5 19 20 Al 210 / 2.5 BTEX PID 6.4 a
4.6 × 350 × 7.4 5.8 mg cartridges (BTEX)
Tube ResSil-B 0.002 - 0.011
GC-PID [18] 31 × 30 × 20 32 ∼35 4000 n. a.g n. a. n. a. 34 VOC PID 140.6 a
0.165 cm i.d. 75 mg (BTEX)
4 Parallel channels SWNTs
GC-PID [19] n. a. n. a. 6 n. a. Si-glass Cr/Pt n. a. 5 VOC PID <1a n. a. a
n.a. × 400 × 0.6 μm 0.15 mg
Frog 4000 Silica gel
25 × 19 × 37 < 2.2 5.5 n. a. n. a. Si-glass n. a. n. a. n. a. PID ~ ppb n. a. a
[11] aerogel

Parallel channels EtQxBox n. a. / 50

GC-PID [20] n. a. n. a. 63 600 Si-glass Pt Benzene PID 1.25 a 2396 a
25 × 1.3 mm × 12 10 mg
Tapered cavity C-B
GC-PID [15] 60 × 50 × 10 <5 16.2 50 Si-glass Pt n. a. / 314 50 VOC µPID n. a. n. a.
8.15 × 250 × 2.9 1.135 mg
C-B 3 mg
Tube Stainless
GC-PID [21] 30 × 17 × 8 <3 15 1000 C-X 2 mg NiCr wire n. a. / 75 10 VOC PID 0.02a -0.36 90a-1051
1.2 mm (i.d.) × 8 cm (l) Steel
C1000 1 mg
73 l: length, d: depth, w: width; Power cons.: power consumption; Det.: Detector; LOD: Limit of Detection, i.d.: inner diameter; n.a.: not available; CNT: carbon nanotubes; PID: photoionization
74 detector; C-B: Carbopack B; C-X: Carbopack X; C1000: Carboxen 1000; a value given for benzene.

75

4
76 Table 2. Portable GC coupled to different detectors other than PID with integrated PC and their features.
Preconcentrator
Analysis Sample
Weight Power cons. Target LOD Sensitivity
Ref. Size (cm3) time Volume l (mm) × d (µm) Heating Det.
(kg) Material Adsorbent (W) / heating compounds (ppb) (pg)
(min) (mL) × w (mm) system
rate (°C/s)
C-B 2.0 mg μCR 16−600
PEMM-2 [6] 20 × 15 × 9 ∼2.1 3 5 2 Cavities (V∼4.7 μL) Si-glass Ti/Pt 1/- 9 VOC 2396 a
C-X 2.3 mg array (150 a)
Ceramic Manifold shaped cavity C-SII
- - ∼50 200 Ceramic Tungsten 18 / 0.25 Ethylene EC 25 5736
PC [22] 61 × 5.5 mm × 23 mm 1.187 g
Cavity with micro-
Zeolite DaY
GC-MOS [8] n. a. n. a. 9 250 pillars Si-glass - - 4 VOC MOS 24 b 22611 b
~13 µm
10 × 400 × 5
Cavity with micro-
GC-CMOS Carbon 15 (1,3,5-
16 × 11 × 11 n. a. n. a. - pillars Si-glass Ni-Cr wire - 7 VOC CMOS -
[23] film TMB)
10 × 250 × 2
iGC3.2 U shape 10 – 2
8 × 10 n. a. 130 24 Glass C-B + C-X Ti/Pt 10.5 / 46 19 VOC 2 CD 766.7 a
[24] n.a. × 300 × 1350 (BTEX)
Cavity with micro- Tenax TA ∼25
Zebra GC [12] 15 × 30 × 10 ∼1.8 < 12 10 Si-glass Cr/Ni 16 / 25 6 VOC TCD 942.1 a
pillars 13 × 240 × 13 ∼200 nm (TEX)
2,4-DNT,
Tapered cavity µCR 0.067-
INTREPID [2] 33 × 29 × 13 5.4 2 1000 Si-glass C−B 2.4 mg Gold - / 75 2,3-DMNB, 500-2200
3.2 mm (l) × 3.5 mm array 0.30
2,6-DNT
Planar Microhotplate
Activated
GC-MOS [25] - - 3 100 1.2 µm thick SiO2/Si3N4 SiO2/Si3N4 Pt 1.02 µW/ºC Benzene MOS < 1000 a -
Carbon
dielectric layer 2 × 4.5
GC-MOX Ten parallel channels
n. a. n. a. ∼67 2750 Si-glass QxCav Pt - / 50 Benzene MOS 0.1 878.5 a
[13] 800 µm depth
77 l: length, d: depth, w: width; Power cons.: power consumption; Det.: Detector; LOD: Limit of Detection, i.d.: inner diameter; n.a.: not available; C-B: Carbopack B; C-X Carbopack X, C1000:
78 Carboxen 1000; CR: chemiresistor; TCD: thermal conductivity detector; MOS: Metal Oxide Semiconductor; CD: Capacitive Detector. CMOS: Complementary Metal Oxide Sensor. MOX: Metal
79 Oxide sensor. QxCav: Quinoxaline bridged Cavitand; EC: electrochemical sensor; n.a.: not available; a value given for benzene; b value given for toluene.

80
81
82

5
 83 Another way to reduce the power consumption is by optimizing the adsorption-desorption process.

84 Therefore, recent efforts of some research groups have been focused on developing foam-based

85 adsorbents that present lower pressure drop and more efficient heat transfer than conventional

86 adsorbents while maintaining a reasonable adsorption capacity. Various types of gas micro

87 preconcentrators and their main parameters and features are presented in Table 3 and Table 4.

88 Researchers working on the development of preconcentrators are often specialists from very different

89 fields, namely microfluidics, electronics, material science, analytical chemistry, or air quality. Given

90 this variety of approaches, different measurement units are used to quantify the same magnitudes: ppb

91 and µg m-3 for gas concentrations, or even in mass unit (ng or pg) of analytes injected into the analytical

92 instrument. This diversity makes comparisons of analytical performances between different systems

93 very difficult, especially when different chemicals are tested. Moreover, a preconcentrator, by definition,

94 concentrates the gas sample for a given time to achieve a certain sensitivity. Thus, the sample volume

95 and the time-resolution are parameters to be considered in the evaluation of portable chromatographs.

96 The present review article covers the last trends in the development of miniaturized preconcentrators for

97 portable gas chromatography. A general overview of gas preconcentration followed by thermal

98 desorption is presented in Section 2; the recent developments in these preconcentrators which includes

99 materials and fabrication techniques, design, heating technologies and fluidic interfacing, adsorbents

100 and applications are reported in Section 3. Additionally, in this section, a comparison of the

101 preconcentration performance of different devices and guidelines to evaluate this performance in a

102 normalized manner are provided. Conclusions are presented in Section 4.

103
104
105

6
106 Table 3. Preconcentrators packed with commercial granular adsorbents.
Target PC Microstructure and dimensions Heating Power cons. (W) / Det. & PF Sample
Ref. Adsorbents Materials volume Application field
compounds l (mm) × d (µm) × w (mm) system Heating rate (°C s-1) LOD (ppb)
Environmental
SVOC and C-X 780µg Two concentric rings of apertures
[26] Si-glass Pt 2 / 150 - 400 n. a. n. a. n. a. monitoring, CWA
VOC C-B 830 μg (8 × 8 mm whole chip)
detection
Tapered cavity
[27] BTEX Tenax TA Si-glass Pt n. a. / 19 MS n. a. n. a. Breath analysis
21 mm (l) × 400 μm (d) × 7.6 mm (w)
Manifold with 4 µchannels Ceramic
[28] Isoprene C-X 20 mg Copper 10.14 / 5.5 FID 352n 18 Breath analysis
12.6 x 2.5 mm x 16 heater
EtQxBox Parallel channels PID Environmental
[20] BTEX Si-glass Pt n. a. / 50 -
10 mg 25 mm (l) × 1.3 mm (d) × 12 mm (w) 1.25 b monitoring
Tenax TA Square cavity FID 5.1-
[29] 12 VOC Glass 2.8 / 17.6 n. a. n. a. -
∼7 mg 2.54 cm (l) × 2.54 cm (d) (chip) 22 10.3a
CBS-II 8 parallel channels Glass-Si- EC
[30] Ethylene Tungsten n. a. 7.7b 220 Food industry
191.0 mg 40.0 mm (l) × 0.9 mm (d) × 2.0 mm (w) Glass 3.8
Spiral microchannel
CBX 1012 GS
120 mm (l) × 0.35 mm (d) × 0.35 mm Si-glass Pt 4.52 / n. a. 2500c Breath analysis
~ 4 mg 800
(w)
CBX 1018 Tube
[7] Acetone
0.95 - 0.2 mm (i.d.) × 0.4 mm (o.d.) × 46.8 mm Glass Ni-Cr wire 8.57 / n. a. GS ∼ 800 c 3000 Breath analysis
11.3 mg (l)
CBX 1018 Tube
0.95 - 0.7 mm (i.d.) × 1.2 mm (o.d.) × 38.2 mm S.S Ni-Cr wire 17.67 / n. a. GS ∼550 c Breath analysis
11.3 mg (l)
HKUST-1 ~ Hotplate Environmental
[31] Benzene Alumina Alumina 0.43 / n. a. MOX 10n
30 mg 3 mm (l) × 3 mm (w) monitoring
Tubular rolled plate Explosives
[32] 4 VOC CB + Tenax Polyimide Gold 1.7 / 40 PID 8-200c 2000
2 mm (l) x 1 mm (i. d.) detection
C-X 1.4 mg 2 tapered cavities Occupational
[5] 17 VOC Si-glass Pt n. a. / n. a. FID 620d 31
C-B 2.0 mg 380 µm deep exposure
Zeolite DaY Rectangular cavity with pillars MOX Explosives
[1] ONT Si-glass Pt n. a. / n. a. 7.2e 500
2.25 mg 5 mm (l) × 400 mm (d) × 10 mm (w) 365 detection
Tapered cavity ECD Environmental
[33] TCE C-X ~2.3 mg Si-glass Pt n. a. / 375 800000d 20000
3.2 mm (l) × 3.5 mm (w) 1.2 monitoring
HayeSep D Tube with micro-pillars Stainless Environmental
[16] Toluene NiCr wire 14.6 / 1.8 MS n. a. 180
119.9 mg 38.68 (l) × 7.69 mm (o.d.) steel monitoring
107 l: length, d: depth, w: width; Power cons.: power consumption; Det.: Detector; LOD: Limit of Detection, i.d.: inner diameter; o.d.: outer diameter; n.a.: not available; C-B: Carbopack
108 B; C-X Carbopack X, CBX: Carboxen; GS: gas sensor; MOX: Metal Oxide sensor; ECD: electrochemical detector; n.a.: not available; Method employed for the calculation of the
109 preconcentration factor: a ratio of the concentration peak area of the PC packed with adsorbent to that of the PC without adsorbent, b ratio between peak area of the detector with and
110 without the presence of a PC, c ratio between the maximum concentration measured at the desorption step and the initial concentration of the injected sample gas, d ratio of the volume
111 of the air sample collected to the volume in which that same mass is contained at the point of detection, n not mentioned.

7
112
113
114 Table 4. Preconcentrators filled with thin film and foam adsorbents.
Target PC Microstructure and dimensions Heating Power cons. (W) / Det. & PF Sample volume Application
Ref. Adsorbents Materials (mL)
compounds l (mm) × d (µm) × w (mm) system Heating rate (°C s-1) LOD (ppb) field
MOF foam 11 × 100 × 8 Environmental
[34] BTEX Si-glass Pt 4.2 / 22 FID 144c 50
85 mm3 Square cavity monitoring
Cavity filled with U-shaped
[35] BA, EB,p-X Tenax-TA Si-glass n. a. n. a. FID n. a. n. a. -
micropillar array
Ethane CNT Rectangular cavity
[36] Si-glass Pt 4.8 / 25 FID 90.2b 150 Breath analysis
Foam 13 mm (l) x 1.5 mm (d) x 11 mm (w)
Benzene,
SWCNT Four parallel channels of 350 μm (d) Environmental
[37] toluene, Si-glass n. a. n. a. PID n. a. n. a.
0.15 mg 10 mm (l) with micropillars monitoring
styrene
Tenax-TA Tapered cavity with cylindrical
Toluene, 7900 //
[38] ~3.7 μm micropillars Si-glass Gold n. a. / 75 FID 150 -
m-xylene 14800a
9 mm (l) × 400 mm (d) × 5 mm (w)
Silicalite 1 Tapered cavity (A) 48.9 ±
8 / n. a.
4.4 mg ~54.0 mm (l) × 10 mm (w) 10.4c
Silicalite 1 Tapered cavity (B) 42.6 ±
8 / n. a. Occupational
[39] Hexane 4.1 mg ~54.0 mm (l) × 10 mm (w) Si-glass MS 15.3 c
Gold 50 exposure
Silicalite 1 Tapered cavity with square pillars (C) 13.7 ±
8 / n. a.
4.6 mg ~54.0 mm (l) × 10 mm (w) 2.5 c
Silicalite 1 Tapered cavity with pillars (D) 30.9 ±
8 / n. a.
4.7 mg ~54.0 mm (l) × 10 mm (w) 1.2 c
SWCNTs 4 microchannels Environmental
[40] Benzene Si-glass Pt n. a. / 3.5 PID n. a. 150
(0.15 mg) 400 μm (d) × 1000 μm (w) monitoring
115 l: length, d: depth, w: width; Power cons.: power consumption; Det.: Detector; LOD: Limit of Detection, n.a.: not available; n.a.: not available; Method employed for the calculation of
116 the preconcentration factor: a ratio of the concentration peak area of the PC packed with adsorbent to that of the PC without adsorbent, b ratio between peak area of the detector with
117 and without the presence of a PC, c ratio between the maximum concentration measured at the desorption step and the initial concentration of the injected sample gas, d ratio of the
118 volume of the air sample collected to the volume in which that same mass is contained at the point of detection, n not mentioned.
119
120
121
122

8
123 2. General overview of gas preconcentration followed by thermal desorption

124 In gas preconcentration devices, a cavity is packed or coated with an adsorbent to capture the molecules

125 of interest. The working principle of preconcentrators (PC) is based on the adsorption of these molecules

126 on the adsorbent followed by a fast thermal desorption (see Figure 1). In the first step, the analytes are

127 collected on an adsorbent either by active flow [41,42] or through passive exposure [43,44]. In active

128 flow sampling, the adsorbent is contained in an enclosed space and a specific sample volume is passed

129 through the adsorbent bed by means of a pump. In this type of sampling, since all the sample volume

130 pass through the adsorbent, a complete trapping of the analytes present in the sample is expected. Due

131 the power requirement of the pumps, active sampling is used in applications where gas concentrations

132 must be measured for short periods of time (typically minutes to hours) [43]. In passive sampling,

133 adsorbents are usually exposed to the sampled medium, i.e., usually indoor or outdoor air, and the

134 theoretical volume of sampling air is determined according to the gaseous diffusion rate of each analyte

135 in the investigated media. Passive sampling is commonly used to monitor gas concentrations during

136 long time periods ranging from some days up to several weeks [43,45].

137 In active as well as passive sampling, the analytes collection is typically conducted at room temperature

138 [41], and sporadically at sub-zero temperatures [40] to enhance the adsorption of analytes. Once the

139 adsorption is completed, the preconcentrator is rapidly heated to relatively high temperatures, i.e., in the

140 wide range 150 – 350 °C and more usually between 250 and 300 °C. This temperature increase results

141 in the desorption of the analytes into a much smaller volume compared to the original sample, thus

142 increasing their concentration prior to the analysis. Furthermore, this rapid increase leads to the injection

143 of analytes in the column in the form of a narrow plug, increasing sensitivity. That is the reason why in

144 some studies, PC are also referred to as preconcentrator-injector (PCI) or preconcetrator-focuser (PCF)

145 [5,33,46,47].

9
146

147 Figure 1. Schematic representation of the adsorption and thermal desorption of VOCs.

148

149 As mentioned above, the preconcentration process consists of two consecutive steps, adsorption and

150 desorption. The efficiency of the whole process is determined by several factors affecting these two

151 stages:

152 • The physico-chemical properties of the adsorbent. Thermal stability, surface area and polarity

153 of the adsorbent dictate the long-term efficiency, maximum adsorption capacity, and strength

154 and nature of adsorbate-adsorbent interactions. Relatively strong interactions are required

155 between the analytes and the adsorbent to capture the analytes at room temperature; however,

156 this process should be reverted when temperature increases. Chemisorption is not desired since

157 this process may involve changes in the chemical structure of the analytes.

158 • The experimental conditions employed. Temperature, sample volume, and analyte

159 concentration during the adsorption step determine the total amount of analyte trapped. During

160 the desorption step, the carrier gas flow rate, desorption temperature and heating rate condition

161 the desorption efficiency and the narrowness of the chromatographic peaks.

162 • Preconcentrator design and configuration of the experimental setup. Geometry of the

163 preconcentrator cavity influences the residence time of the sample and, therefore, the probability

164 of the analytes to be captured. Dead volumes in the fluidic path of the system may contribute to

165 the dilution of the sample prior to the analysis, thus reducing the efficiency of the
10
166 preconcentration process. Similarly, this kind of flow path disruptions tend to generate peak

167 tailing in the chromatograms. In addition, the materials employed in PC fabrication determine

168 the heat transfer and therefore the desorption rate.

169

170 3. Recent developments and trends in miniaturized gas preconcentrators

171
172 Because of the high demand of sensitive and portable analytical instruments, several miniaturized gas

173 preconcentrators have been developed over the years. Miniaturized preconcentrators can be defined as

174 device which serve to increase the analyte concentration in a sample and whose fluidic path have

175 dimensions of microns implying the use of microfabrication techniques. The size of these devices can

176 go from few millimetres up to 10 mm whereas their weight can reach up to few tens of grams. Some

177 examples are displayed in Figure 2. While the working principle of the PC remains the same, new

178 developments including designs, materials, fabrication techniques, or adsorbents have been recently

179 proposed. These developments are reviewed in detail in this section and their main characteristics,

180 performances and future challenges are presented.

181
182 Figure 2. Micro preconcentrators made of diverse materials: (a) silicon-glass (microcavity dimensions:
183 5 mm wide, 10 mm long and 400 µm deep) [8], (b) cooper (device dimensions: 16 mm long, 12.6 mm
184 wide and 2.5 mm deep) [28], (c) stainless steel (tube dimensions: 38.68 mm long and 7.69 mm of
185 outer diameter) [16], (d) ceramics (device dimensions: 61 mm long, 23 mm wide and 5.5 mm deep)
186 [22] and (e) polyimide foil (tube dimensions: 2 mm long and 1 mm of internal diameter) [32].

187 3.1. Materials and fabrication techniques

188 Several materials have been employed for the fabrication of miniaturized gas PCs. PCs are subjected to

189 continuous temperature cycles that conditions the selection of these materials. Ideally, the material

11
190 employed for the fabrication of a PC should be thermally resistant to preserve its structure after several

191 adsorption-desorption cycles which involve rapid heating and cooling. Furthermore, an effective heat

192 transfer from the heaters to the adsorbent is desired to promote the fast desorption of analytes. Therefore,

193 a low heat capacity and a minimal thermal mass are paramount to allow a fast temperature ramp.

194 In this regard, silicon or glass-based PCs machined by MEMS (Micro-Electro-Mechanical Systems)

195 techniques are advantageous due to their reduced thermal mass and, therefore, power consumption.

196 However, the multiple cleanroom-based processes which can be expensive or time-consuming

197 altogether with the high level of research and development required to get a successful device, has led

198 several research groups to develop miniaturized PCs using alternative materials and fabrication

199 techniques.

200 The main advantage of these recently developed devices lies in the possibility to be manufactured in

201 research laboratories without cleanroom requirements making the fabrication process more accessible

202 and, in most cases, reducing time and costs. Furthermore, standard fluidic connections assuring gas

203 tightness are usually easy to integrate in these devices due to their larger dimensions compared to those

204 of MEMS-based devices. However, these dimensions involve higher thermal mass, resulting often in

205 higher power requirements and slower temperature ramps.

206 3.1.1. Silicon

207 In the last decades, MEMS technology has been widely used for the fabrication of PCs [1,5,20,26,27,33–

208 40]; therefore, most of these devices are made of materials related to the electronics industry. So far,

209 silicon has been the most employed material due to its relatively high thermal conductivity and the

210 variety of existing micromachining techniques. A typical preconcentrator made of silicon is illustrated

211 in Figure 2a. This material provides low thermal mass, uniform thermal distribution, and the ability to

212 handle high temperature ramping rates [48], the most relevant features in miniaturized preconcentrators.

213 Silicon micromachining is generally conducted at cleanroom facilities by wet or dry etching. Wet

214 etching techniques involve the use of liquid chemicals such as hydrofluoric acid or potassium hydroxide

215 to remove the exposed surfaces of the substrate. These techniques played an important role in the early

12
216 stages of MEMS development to create microfluidic cavities and channels in silicon at a relatively low

217 cost compared to other etching techniques [29,39]. Wet chemical etching is isotropic and produces

218 rounded side wall microchannels. The depth of the these channels is controlled by the etch rate and etch

219 duration whereas the width of the microchannel can be estimated by the mask opening plus twice the

220 channel depth [49]. Therefore, high aspect ratio features are hard to build using these methods. For that

221 reason, their use in PC fabrication has been limited [39]. On the contrary, dry etching techniques offers

222 the possibility to create high aspect ratio structures. In this process, etching is conducted by ions that

223 collide and react with the exposed surface of the silicon wafer instead of using liquid chemicals [48].

224 The most widespread technique for creating cavities, channels, and pillars with different geometries in

225 silicon-based PC is the deep reactive ion etching (DRIE). DRIE allows to create nearly vertical sidewalls

226 of high surface quality in silicon. For example, using this technique, microchannels 250 µm deep and

227 25 µm wide have been etched to create gas chromatography microcolumns [50]. Once the cavities are

228 etched, the substrate is bonded to another plate usually made of glass resulting in a sealed chip. Anodic

229 bonding is the most common technique for silicon-glass bonding [6–8,19,34,36,37,51–53]. In this

230 process, a DC voltage ranging from 500 to 2000 V is applied between the substrates whereas they are

231 heated to temperatures varying from 300 to 600 °C [48]. The generated electric field leads to the

232 migration of Na+ ions from the silicon/glass interface, leaving oxygen ions (O2−). These ions react to

233 form a layer of siloxane (Si-O-Si) ensuring the irreversible bonding between the substrates [54,55]. As

234 can be seen in Table 3 and Table 4, more than 70% of PCs developed in the last years are silicon-glass

235 hybrid devices. In these devices, the heating system can be easily integrated by patterning a metal layer

236 (typically Pt or Au) on the silicon substrate.

237 3.1.2. Glass

238 Although silicon-glass hybrid PCs are the most common, some devices have been fabricated entirely

239 from glass. McCartney et al. [29] developed a robust micro gas PC made of glass (Borofloat 33 by

240 Schott Co.). In this device, the cavity was created by wet etching using 49% hydrofluoric acid solution.

241 Unlike DRIE, some steps of this wet etching procedure can be conducted outside of a cleanroom. The

242 heating system and temperature sensors were created with a photolithography-free technique enabled

13
243 by laser etching. To seal the chip, the glass die containing the microcavity was bonded to another glass

244 die by thermal fusion bonding.

245 Qin and Gianchadani [24] developed a fully microfabricated gas chromatograph in which all the

246 components were fabricated using glass dies. In this device, the metallization and micromachining were

247 performed on separate glass substrates. The microchannels and ports for fluidic connections were

248 created by sandblasting. This process enables to create high aspect ratio structures and quickly remove

249 large areas of material being suitable for the creation of deep channels as well as through-holes. In the

250 case of the PC, the metallized and the micromachined glass dies were bonded by adhesive bonding using

251 Epotek-377 epoxy, followed by baking at 300 °C for >1 h to minimize outgassing during operation. This

252 technique is a simple, robust, and low-cost bonding procedure that, in addition, do not require a

253 cleanroom to be performed.

254 3.1.3. Metal

255 The well-known drawback of most of silicon and glass micromachining techniques is mainly the need

256 for cleanroom facilities. Furthermore, even though these fabrication processes are well established, PCs

257 made of these materials are very fragile and failures are frequent during the fabrication and the assembly

258 process. Laser etching technology, milling or metal 3D printing techniques have been recently employed

259 as alternative techniques for the fabrication of miniaturized PCs. The late improvements in these

260 techniques provide the possibility to fabricate miniaturized PCs made of more robust materials at a low

261 cost. Han et al. [28] fabricated a micro metal gas preconcentrator (MGP) made of copper (C11000) for

262 the detection of isoprene (see Figure 2b). This material has a thermal conductivity about 2.6 times higher

263 than that of silicon, and a specific heat capacity of about a half of that of silicon promoting temperature

264 uniformity along the chip. Furthermore, copper (C11000) has good chemical stability, can be

265 micromachined by traditional manufacturing methods and is much more robust than silicon or glass.

266 The MGC layout consist of 4 microchannels (each is 1 mm width, 3 mm long and 500 µm depth), and

267 two symmetrical manifold fluidic systems etched in a C11000 substrate by laser etching technology

268 (LET). This substrate was bonded using of vacuum diffusion welding (VDW) to another copper plate

269 in which inlet and outlet ports were etched.

14
270 Huang et al. [16] printed a stainless steel PC using jet binder printing (BJP) technique (see Figure 2c).

271 In this additive technique, the parts are built by spreading the printing powder in a layer-by-layer fashion.

272 After the printing process, the device must be heated in an oven, first at 195°C for 2h for the binder

273 phase to cure and then, at 460°C for 2h for the decomposition of the remaining binder phase. Finally,

274 the device is sintered in a high-vacuum oven at 1250°C for 6 h. To minimize the porosity of the printed

275 parts, sintering additives such as boron nitride (BN) were added to the stainless-steel powder used for

276 the printing. BJP technique allows to manufacture devices with submillimeter internal features that can

277 operate at high temperatures. A standard 10-32 port thread was machined to connect commercial

278 compression fittings resulting in a gas tight connection and facilitating its integration into the GC system.

279 Aluminium has been also used in the manufacture of miniaturized PCs. Lara-Ibeas et al. [10] proposed

280 a PC made of aluminium for the detection of BTEX. In this device, the microfluidic cavity and two

281 symmetrical manifolds were machined on an aluminium substrate by traditional precise milling

282 methods. Two porous metal filters were integrated at both sides of the cavity to retain the adsorbent.

283 The device was covered with another aluminium plate and the sealing was assured by four screws and

284 a Viton gasket. Rodriguez-Cuevas et al. [4] manufactured a micro gas PC made of aluminium in their

285 research lab using also precision milling. The device consisted of an aluminium block with two lateral

286 tubes. Custom-made PEEK reducing unions were inserted in these tubes resulting in reliable gas-tight

287 connection to the GC system.

288 3.1.4. Other materials

289 Other materials have been employed in the fabrication of miniaturized PCs but to a much lesser extent.

290 Zaidi et al. [56] developed a micro preconcentrator made of aluminum nitride ceramics (see Figure 2d)

291 for the detection of ethylene. The chip was manufactured using ceramic technology and the two parts

292 were bonded by co-firing. Unlike other microdevices, this PC can be easily integrated in the

293 chromatographic system due to the two cylindrical hoses glued to the PC for the gas inlet and outlet.

294 Camara et al. [32] reported the use of polyimide foil for the fabrication of a PC (see Figure 2e). The

295 main advantages of using this material are low-cost and easy processing. Additionally, this PC are

15
296 flexible, and the diameter of the inlet/outlet can be adjusted to work at higher flow rates than in their

297 silicon counterparts, reaching up to 1.5 L/min.

298

299 3.2. Design and microstructures

300 The main microstructures reported in the literature can be divided in two types: planar hotplate and three

301 dimensional (3D) structures. Planar hotplates consisted of a planar substrate coated with an adsorbent

302 selective to the target molecules [57]. This design is much easy to fabricate and offers a low power

303 consumption, however, owing to its configuration, the amount of adsorbent that can be hosted in these

304 devices is limited, which reduces the adsorption capacity and, therefore, their preconcentration

305 performance. This major drawback has led to a change of trend in PC design which has shifted

306 progressively from planar to 3D preconcentrators. In the last decade, only few planar preconcentrators

307 have been reported in the literature. Lahlou et al. [25] developed a planar micropreconcentrator-injector

308 for the detection of benzene. This device consisted of a planar rectangular membrane of a 1.2 µm thick

309 SiO2/Si3N4 dielectric layer, which was suspended on a silicon substrate. An activated carbon layer was

310 deposited on the membrane by drop-coating and used as adsorbent. Leidinger et al. [31] used alumina

311 microhotplates to fabricate a planar micropreconcentrator for air quality monitoring. Two metal-organic

312 frameworks (MOF), HKUST-1 and MIL-53, were deposited on the hotplates by drop coating and their

313 preconcentration efficiency was evaluated for benzene and toluene.

314 Currently, most of micro-PCs present a 3D structure. This configuration offers a higher collection

315 capability since the amount of adsorbent packed is greater than in the case of planar PCs. Nonetheless,

316 several aspects must be considered to ensure good performance of these devices:

317 - Uniform flow distribution inside the device to optimize the adsorption through the whole

318 adsorbent bed;

319 - Low pressure drops to limit the risk of gas leaks and decrease operation costs due to high

320 pressure in the microdevices;

321 - Large contact surface and long contact time between the adsorbate and the adsorbent to increase

322 the probability of the molecules to be adsorbed and trapped;

16
323 - Efficient and fast heating but also uniform heat distribution.

324 Taken into account these aspects, several 3D PCs with a variety of geometries have been proposed in

325 the literature such as circular spiral [58], array of parallel channels [28] or single U-shaped channel [24]

326 (see Figure 3).

Figure 3. Preconcentrators with different geometries: circular spiral (a) [58], array of parallel
channels (b) [28] and U-shape cannel (c) [24].

327

328 Due to the narrowness of the microchannels altogether with the flow resistance added by the adsorbent

329 particles, some of the one-single channel designs often shows relatively high pressure drops. Therefore,

330 designs leading to lower pressure drop, such as a wide rectangular cavity connected to inlet and outlet

331 channels [6,10,22,39] are usually preferred. In these microdevices, prior to fabrication, flow simulations

332 are frequently conducted to estimate the pressure drop and optimize the distribution of the gas flow. Li

333 et al. [59] performed computational fluid dynamics (CFD) simulations and demonstrated the influence

334 of the inlet angle on the flow field near the inlet (see Figure 4a). They reported that uniform gas flow

335 velocity distribution can be achieved when the triangular angles in the inlet and outlet are smaller than

336 120°. Hence, as can be inferred from Table 1 and Table 2, it is very common to observe tapered cavities

337 in micro PCs [8,10,23,28,38,56] as illustrated in Figure 4b.

338
339 Figure 4. (a) Flow simulation in a PC with inlet/outlet angle > 120 ° [59]; (b) Preconcentrator containing
340 micropillars and having inlet and outlet angles > 120 ° [23]; (c) Pressure (green) and flow velocity (blue)
341 simulations in a PC with inlet/outlet manifold fluidic system [60].

17
342

343 To create a more uniform flow distribution, some authors have added symmetrical manifolds at both

344 sides of the main cavity where the adsorbent is packed [10,60] (see Figure 4c). These manifolds split

345 the inlet flow in several channels thus promoting uniformity of the gas flow through the adsorbent bed.

346 In a PC loaded with granular adsorbents, the cavity is usually flat although, in some cases,

347 microstructures have been also included to improve the flow distribution without significant reductions

348 in the amount of adsorbent that can be hosted [1]. To reduce the pressure drop associated with the use

349 of granular adsorbents, thin film adsorbents have been employed. In the case of a PC coated with thin

350 film adsorbents, microstructures are very often integrated to enhance their preconcentration performance

351 [61]. It is known that this type of adsorbents has a limited adsorption capacity due to the reduced surface

352 exposed to the analytes. Therefore, the presence of microstructures, themselves usually coated with

353 adsorbent film, inside the microfluidic cavity greatly increases the exposed adsorptive surface, and

354 additionally optimizes the path followed by the gas molecules, thus increasing their contact time with

355 the adsorbent. Larger contact surface and longer residence time result in higher preconcentration

356 performance due to the higher adsorption capacity and higher probability of the gas molecules to be

357 trapped. Therefore, this approach enhances the sampling capacity while reducing pressure drop.

358 Moreover, since most adsorbents have a low thermal conductivity, these structures serve also to promote

359 a more efficient heat transfer through the adsorbent bed [16].

360 Different microstructures such as square [39] (Figure 5a) and cylindrical pillars [38] (Figure 5b), cross

361 shaped micropillars [37] (Figure 5c), U- [35] (Figure 5d) and V-shaped [1] pillars (Figure 5e) can be

362 found in micro PCs. Alfeeli and Agah [61] studied the influence of the shape and the spacing of

363 micropillars in preconcentration performance. Crisscross shaped pillars were found to be more efficient

364 in terms of preconcentration factor (PF) than ordered and staggered pillars. This PC contained more than

365 3,500 crisscross pillars with dimensions of 30 µm ⨯ 120 µm ⨯ 240 µm (see Figure 5f) and it was coated

366 with 2,6-diphenylene oxide polymer (Tenax TA). A PF >10 000 was found for isopropanol. However,

367 a great amount of 3D microstructures embedded within a microcavity can lead to high pressure drops.

18
368 To minimize the pressure drop, Alfeeli et al. [62] was the first to propose a design consisted of an array

369 of U-shaped pillars embedded within a 7 mm ⨯ 7 mm ⨯ 0.38 mm cavity. These pillars were arranged

370 to divide the gas flow in the forward direction and combine the resulting two flows in the reverse

371 direction. The spacing between the reflectors (side and middle spacing) was varied to investigate its

372 effect on the device performance. The pressure drops of different pillars arrangements were investigated

373 at a flow rate of 1.5 mL/min. The configuration with 300 and 150 µm of middle and side spacing,

374 respectively, exhibited a pressure drop as small as 94 Pa. In comparison configuration with 100 and 250

375 µm of middle and side spacing exhibited a pressure drop of 145 Pa. The same microstructure design

376 using reflectors has been recently employed in another miniaturized PC for the quantification of n-butyl

377 acetate, ethylbenzene and p-xylene [35] (see Figure 5b).

378 However, the fabrication of embedded micropillars requires high cost and even the cavities densely

379 filled with micropillars exhibit relatively low contact surface thus reducing the sample capacity of these

380 adsorbents. Besides, some authors observed high desorption temperatures of above 250°C for complete

381 desorption of the analytes when using micropillars coated with thin-film adsorbents [63,64].

382 More recently, a new approach involving foam adsorbents instead of creating microstructures has been

383 adopted to reduce the pressure drop and also enhance the heat transfer due to the electrical resistance of

384 some of these foams [65].

385
386 Figure 5. Preconcentrator containing a) square [39], b) cylindrical [38], c) cross [37], d) U- [35] and e)
387 V-shaped [1] and f) crisscross shaped micropillars [61].

19
388

389 3.3. Preconcentrator heating technologies

390 During the preconcentration step, a very fast temperature ramp is needed for the flash desorption of

391 analytes; therefore, the heating system is considered a key component in preconcentration devices. The

392 faster this rate is, the sharper the desorption peak becomes and, thus, higher sensitivity and better

393 chromatographic resolution can be achieved. However, since miniaturized PCs are integrated in portable

394 instruments and most of them are battery-powered, the energy required for a flash desorption of analytes

395 should be limited to the possible extent. Hence, a compromise between heating rate performance and

396 energy consumption must be adopted. To this purpose, a wide variety of materials has been employed

397 for the fabrication of these systems, as depicted in Figure 6.

398

399 Figure 6. Diverse heating systems found in micro preconcentrators: (a) Ni-Cr wire [81], (b) platinum
400 heater [66], (c) gold layer deposited by electron beam [39] and (d) gold layer deposited by electroless
401 plating technology [38].

402 Originally, the heating system of the first generation of PCs consisted of metal wires coiled around the

403 device [67]. Although some of these systems continue to be used today, the number of portable devices

404 including this type of system is very small [21,66].

405 The material traditionally employed for these metal wires is Ni-Cr because of its extraordinary resistivity

406 and almost constant temperature coefficient of resistance (TCR) with temperature. These features render

407 this alloy suitable for heating and temperature measurement applications [16]. For example, Tzeng et

408 al. [23] developed a micro gas chromatograph (µGC) containing a MEMS PC coiled by a 6 Ω Ni-Cr

20
409 wire for the detection of seven VOCs associated with lung cancer (see Figure 6a). Jian et al. [21]

410 employed also the same heating system for a stainless-steel tubular PC integrated in a µGC. This

411 configuration required 12 V DC to reach 320 °C in less than 4 s.

412 Nowadays, most devices are fabricated by MEMS techniques which enables the integration of heating

413 systems on the substrate. These systems consist of layers of resistive and conductive metals to form the

414 resistive heaters and the contact pads, respectively. Techniques such as e-beam evaporation [38] and

415 sputtering [27,29] are commonly used for metal deposition of a wide range of metals. Among them,

416 platinum is the most used material in micro PC’s heating systems [1,5,24,27,68–70]. A resistive heater

417 made of platinum is presented in Figure 6b. Similarly to Ni-Cr, platinum has a relatively high resistivity

418 and constant TCR, being suitable to act both as a heater and a temperature sensor. Heating rates ranging

419 from 3.5 to 375 °C s-1 and power consumptions between 1 and 10.5 W have been reported in micro

420 preconcentrators with platinum heaters [20,26,27,33,34,36,40]. Although this metal presents better

421 adherence to substrates than other metals, an adhesive layer of titanium [24,27,33,68] or chrome [40] is

422 generally deposited beforehand to improve its adhesion to the substrate.

423 Other less common metal combinations have been employed for the fabrication of heating systems like

424 chromium-tungsten [29], chromium-nickel [12] and chromium-gold [2,39] (see Figure 6c). Despite their

425 widespread use, metal deposition processes by means of e-beam evaporation or sputtering are sometimes

426 tedious involving the use of several photomasks. To tackle this issue, Kuo et al. [38] proposed to use

427 electroless plating technology for the deposition of gold layers (see Figure 6d). This much simpler

428 technique is based on the chemical reduction of gold and allows the deposition of high-surface-area gold

429 layers on the surface of microchannels. Heating rates up to 75 °C s-1 were achieved using these resistive

430 heaters. Inkjet printing has been also employed to create resistive heaters using a gold nanoparticles-

431 based ink [32]. The main advantages of this technique are the possibility to modify the designs without

432 need for masks and the rapidity of the process.

433 Heating systems consisting of thin film resistive heaters are very convenient for MEMS-based devices;

434 however, as other microfabrication techniques, their fabrication requires to be conducted in a cleanroom.

435 Furthermore, sometimes the adhesion to the substrate is not adequate and the metal layers can detach

21
436 from the surface during the fabrication process or after several heating-cooling cycles. As an alternative

437 to integrated thin film resistances, some groups have employed other commercially available heating

438 technologies. Han et al. [28] used a micro ceramic heater located at the bottom of the substrate in their

439 metal PC. This device was heated to 200 °C in 32 s with a power consumption of 10.14 W thus providing

440 reasonable heating power at a very low cost. Commercial ceramic heaters have been employed in other

441 PC [4] reaching 200 °C in 20 s but with a power supply of 61 W. Other commercial solutions such as

442 heating cartridges have been used in PCs with relatively high thermal mass [10]. Homemade flexible

443 membranes have been fabricated by sandwiching a thin Ni-Cr wire with two Kapton tapes [16]. This

444 membrane was wrapped around a cylindrical PC. By applying 14.6 W of power, the PC reached 315°C

445 in 12 s.

446 Apart from the heating rate, power consumption is a vital aspect of any component of a portable

447 instrument as it can be critical for the autonomy of the whole system. Unfortunately, these data are not

448 systematically provided for all the preconcentrators in the published literature (see Tables 1-4) thus

449 making difficult to evaluate the impact of their integration in a real portable instrument. Therefore, it is

450 recommended to provide power consumption in future publications in the field to facilitate this

451 evaluation.

452

453 3.4. Fluidic connections

454
455 Fluidic interfacing between miniaturized components is a critical aspect in µGC. The issues associated

456 with these components such as losses by adsorption, dead volumes or leaks may condition the

457 performance of the GC system. In the particular case of PCs, the fluidic connections limit their maximum

458 operation temperature thereby reducing the efficiency of the analytes’ desorption. Ideally, connectors

459 must be reliable and low-cost, have a low dead volume and they should allow the interfacing in a

460 reversible manner. However, most of the fluidic connections used in miniaturized PCs exhibit only some

461 of these features.

22
462 The most common method is the use of adhesives to glue fused silica capillaries to the miniaturized PCs

463 [34,53]. Deactivated fused silica is the preferred material used for capillaries due to their inertness, cost

464 and variety of diameters, although the use of nickel capillaries have been also reported [27]. Usually,

465 thermal resistant epoxy or glue are applied around the capillaries and these are inserted into the

466 inlet/outlet ports. Then, more epoxy is applied around the connection to assure hermetic sealing. These

467 adhesives can be cured at room temperature but the curing process may also be conducted at higher

468 temperature to speed up the process. Some examples of high temperature adhesives employed are

469 Duraseal 1531 [53] and Duralco 4703 [34]. In some PCs, additional ports are used to introduce the

470 adsorbent into the PC [53]. Other epoxy resins such as Duraseal [6] or Stycast 2850FT [24] are applied

471 to hermetically seal these ports once the adsorbent introduced.

472 In other devices, epoxy was employed to fix Nanoport assemblies [8,34,36]. Capillaries and flexible

473 tubing can be easily connected to these ports by using commercial fittings. Swagelok® stainless steel

474 connections have been also glued using Loctite Hysol 9492 on the glass die of a micro PC [39]. In the

475 ceramic PC developed by Zaidi et al. [22], two hoses of 3.5 mm of outer diameter were glued to the

476 main structure of the device in which flexible tubes were inserted.

477 Some research groups use additional metal parts to accommodate commercial gas flow connectors. The

478 use of these connectors diminishes the risk of leakage and allows the replacement of a single component

479 of the µGC system. Janssen et al. [30] built a housing made of two different aluminum parts using

480 drilling and Computer Numerical Control (CNC) technology. Standard compression fittings were

481 integrated in these parts to connect the inlet and outlet of the PC. In 3D the printed PC proposed by

482 Huang et al. [16], a standard port thread was machined enabling the use of commercial gas fittings. They

483 used PEEK (polyetheretherketone)-based fittings from IDEX Health Science to connect flexible tubing

484 to the port ensuring leak-free connections under a pressure up to 1 MPa (10 bars) and temperature up to

485 343 °C. Rodríguez-Cuevas et al. [4] fabricated PEEK reducing unions equivalent to the standard

486 stainless steel unions provided by Swagelok. These unions were connected at both sides of the PC

487 facilitating its integration into the GC system and thermally insulating the PC.

23
488 Qin and Gianchandani [24] used standard compression fittings between the two modules of their micro

489 GC whereas the different components were connected using micromachined gas flow connectors. The

490 small size of these latter allowed to arrange the components in a compact manner.

491 Other heat resistant materials such as polymide have been used for fluidic interfacing: Han et al. [28]

492 connected deactivated fused silica capillary tubes using customized polyimide adapters which can

493 operate for 2 h at 300°C.

24
494 3.5. Adsorbents

495 A great part of the PC performance relies on the adsorbent material employed. Its textural properties

496 and chemical affinity towards analytes will be determinant for the efficiency of the adsorption and

497 desorption processes. In this regard, the desired features for an adsorbent are:

498 • Selectivity towards the target compounds. Ideally, the adsorbent should capture only the

499 analytes of interest. In practice, the adsorption of a single compound is usually very challenging.

500 However, during the preconcentration, some interfering species can be eliminated by placing

501 some trapping devices in the upstream sampling flow.

502 • High specific surface area. A large number of adsorption sites generally increases adsorption

503 capacity and prevents early breakthrough of the analytes.

504 • Moderate strength interactions with the adsorbate, enabling considerable adsorption capacity at

505 room temperature and desorption at reasonable higher temperatures.

506 • Thermal stability. Adsorbents are subjected to temperature cycles and their performances must

507 last throughout time.

508 • Low pressure drops and good thermal conductivity to minimize energy consumption and

509 promote an efficient thermal desorption of analytes.

510 Traditionally, adsorbents have been distinguished in two categories: granular and thin films.

511 Nonetheless a new category, foam adsorbents, has been in constant development over the last years.

512 These three groups are displayed in Figure 7.

513

514 Figure 7. Schematics of several types of adsorbents: (a) granular, (b) thin film and (c) foam adsorbents

515 [36].

25
516 Granular adsorbents have generally higher adsorption capacity than other types of adsorbents as the

517 grains are compactly packed, thus occupying almost all the volume available inside the PC. However,

518 the complete filling of the microfluidic cavity usually results in high pressure drops. On the contrary,

519 thin-film adsorbents exhibit very low pressure drop [62] as only the surface of the PC is covered with

520 the adsorbent. However, due to the small amount of the material deposited, their adsorption capacity is

521 limited. On the other hand, the new generation of foam adsorbents tend to reduce the pressure drop while

522 keeping a relatively high adsorption capacity [34,36]. Furthermore, some of these foams have high

523 thermal conductivity leading to an improvement in the heat transfer from the adsorbent to the adsorbates

524 facilitating the desorption and decreasing the power consumption of the device [65].

525 3.5.1. Granular adsorbents

526 Granular adsorbents are usually introduced inside the PC by suction [29,32,38,56]. Most of granular

527 adsorbents are carbon-based. Among them, graphitized carbon blacks are traditionally used for VOC

528 preconcentration due to their hydrophobic character, which make them ideal candidates for VOC

529 sampling in humid environments. The most commonly used are Carbopack® B, which is recommended

530 for the capture of C5-C12 molecules such as isoprene [28], and Carbopack® X, which is more suitable

531 to trap C3-C9. Because these adsorbents are appropriate to trap compounds of intermediate volatility,

532 they are considered to have medium strength. Most commonly, a combination of these adsorbents is

533 employed to enlarge the range of analytes trapped [5,6,24]. Qin and Gianchandani [24] use a

534 combination of Carbopack-X and Carbopack-B, to preconcentrate a mixture of 19 VOC including

535 aromatics, alkanes and halogenated compounds. Lee et al. [15] preconcentrate a gas mixture containing

536 50 VOC that was subsequently separated and detected by four independent GC columns and PID

537 detectors. Other type of carbon-based granular adsorbents are the molecular sieves such as the ones of

538 the Carbosieve and Carboxen series. These materials contain micropores in which very volatile

539 compounds (C2-C5) can be trapped. Carbosieve II has been used for the preconcentration of ethylene

540 [22,30]. Rydosz et al. [7] used Carboxen 1012 for the preconcentration of acetone. A combination of

541 molecular sieve with graphitized carbon blacks is usually employed for the preconcentration of

542 molecules of low and intermediate volatility. Jian et al. [21] developed a preconcentrator containing 3

26
543 mg of Carbopack® B, 2 mg of Carbopack® X, and 1 mg of Carboxen® 1000 for the analysis of 10

544 VOC mixture. In these cases, particular attention should be paid to the location of the adsorbents and

545 the flow direction. Analytes of medium volatility are strongly adsorbed in molecular sieves, thus

546 rendering difficult the subsequent desorption. Therefore, for sampling, adsorbents should be placed from

547 the weaker to the stronger while the desorption should be carried out in the opposite direction.

548 The surface of carbon-based adsorbents is generally relatively non-polar preventing the adsorption not

549 only of water, but also other polar VOC. To enhance their adsorption capacity and selectivity towards

550 polar VOC, Wang et al. [71] coated commercial graphitized carbons (Carbopack B and Carbopack X)

551 with room-temperature-ionic-liquids (RTIL). The RTIL-coated adsorbents exhibited ∼2.5 times larger

552 10% breakthrough volume (Vb10) for organophosphorus vapors than the untreated adsorbents.

553 Furthermore, Vb10 values of the four non-polar reference vapors tested were 11–26 times smaller for the

554 RTIL-coated than for the untreated counterpart.

555 Porous polymers as HayeSep® [72] and Tenax® [27,29] are also commonly employed in

556 preconcentration devices. Chappuis et al. [27] employed this latter polymer for the preconcentration of

557 volatile tobacco markers. McCartney et al. [29] reported the preconcentration of a wide variety of VOC

558 with a low-cost PC filled with Tenax. Using this device, Limit Of Detection (LOD) down to 22 ppb

559 were achieved with only 2 min of sampling, the sampling volume being not provided by the authors.

560 Metal organic frameworks (MOF) have been used as preconcentration materials for VOC due to their

561 large specific surface and variety of chemical properties. HKUST-1 (Cu3(BTC)2) is a microporous MOF

562 with a very high specific surface area (1100-1700 m2 g-1) formed by copper nodes with 1,3,5-

563 benzenetricarboxylic acid linkers between them [73,74]. This material is commercially available as

564 Basolite™ C300 and has been used for the preconcentration of aromatic compounds [10,31]. Another

565 well-established MOF, UIO-66 (Zr6O4(OH)4(BDC)6, BDC=1,4-benzenedicarboxylate), having also

566 high surface area (over 1000 m2 g-1) has been used for the capture and preconcentration of acetone

567 [73,75].

27
568 A general issue in gas preconcentration is the lack of selectivity of the employed adsorbents, which leads

569 to the preconcentration of analytes but also of interfering species. To tackle this issue, Zampolli et al.

570 [13] synthetized in 2009 a quinoxaline-based adsorbent named QxCav (see Figure 8a) for the selective

571 preconcentration of aromatic compounds. This adsorbent has cavities of 8.3Å in which BTEX are

572 adsorbed by means of weak CH–π interactions with the walls, whereas other non-aromatics molecules

573 such as aliphatic compounds and water vapour are not retained. The same group recently developed an

574 improved version of the adsorbent called EtQxBox [20]. Its structure is displayed in Figure 8b. In this

575 complex, aromatic VOC are selectively adsorbed, and toluene, ethylbenzene and xylene (TEX) are

576 adsorbed more strongly than benzene. The difference in the strength of interactions renders possible the

577 benzene desorption at lower temperatures compared to TEX. This singularity allows the separation of

578 analytes during the preconcentration step itself by controlling the desorption temperature and thus a

579 shorter GC column can be used for separation reducing the analysis time.

580

581 Figure 8. QxCav (a) and EtQxBox (b) cavitands employed for the selective preconcentration of
582 aromatic compounds [20].

583

584 3.5.2. Thin film adsorbents

585 This type of adsorbents is generally introduced into the microfluidic cavity in a liquid dispersion at a

586 specific concentration. The dispersion is inserted in the microfluidic cavity by depression and the solvent

587 is then evaporated at room temperature, leaving a film on the inner surface of the cavity [1,38,40]. Gregis

588 et al. [8] employed a dispersion of a commercially available dealuminated zeolite (DaY) in ethanol to

589 deposit a uniform 13 µm thick film. This system allowed the detection of toluene, o-xylene, propanol

590 and cyclohexane at the low ppb levels, except for the cyclohexane. Other types of materials such as

28
591 single-walled carbon nanotubes (SWCNT) have been deposited inside a micro PC for the quantification

592 of VOC at trace levels [37,40]. The use of this adsorbent is advantageous due to the large aspect ratio,

593 high effective surface area, chemical and thermal stability, and hydrophobicity. Other films have been

594 synthetized inside the preconcentration device itself. For example, Tzeng et al. [23] and Wong et al.

595 [76] presented in-situ-synthesized carbon adsorbent films for VOCs analysis. To this purpose, cellulose

596 was inserted in a microfluidic cavity and then pyrolyzed at 600 °C under a nitrogen atmosphere to form

597 the porous carbon film. The resulting adsorbent (2 mg) had a specific surface area of 308 m2 g-1 and the

598 whole system exhibited a huge preconcentration factor of 13,637 for toluene (calculated as the ratio

599 between the peak area obtained with and without the preconcentrator). Almazán et al. [39] synthetized

600 silicalite-1 inside of a micro PC with diverse microstructures. These microstructures increased the inner

601 surface of the PC and thus the amount of zeolite deposited compared to the empty PC. All PCs filled

602 with zeolite thin films showed higher preconcentration efficiency than their fixed bed counterpart, due

603 to the uniform zeolite distribution along the inner surface. This homogeneity improved the gas–solid

604 contact and thus adsorption. Furthermore, by introducing pillars in the cavity, the amount of absorbent

605 coated increases, thereby, the adsorption capacity increases. A uniform deposition of the adsorbent is

606 also important for a more homogeneous heating transfer during the desorption because the adsorbent

607 film is directly deposited on the thermal conductor. Therefore, full desorption is more likely to occur,

608 and it will be conducted simultaneously to the same extent across the adsorbent bed. To increase the

609 inner PC surface, Zhao et al. [35] proposed the formation of silicon nanowires as a surface template

610 prior to the deposition of Tenax® inside the cavity. The use of this template increased the

611 preconcentration performance of more than 188 % for n-butyl acetate compared to the PC only filled

612 with Tenax®.

613

614 3.5.3. Foam adsorbents

615 To enhance the adsorption capacity of the preconcentration devices without greatly increasing the

616 pressure drop, foam adsorbents have been recently developed. Lee and Lim [36] employed a carbon

617 nanotube (CNT) foam for the preconcentration of gaseous ethane. For comparison purposes, the same

29
618 PC was filled with Carbosieve™ SIII, a granular absorbent with a surface area of 1000 m2/g and mean

619 pore diameter of 1 nm. Pressure drop was measured in both devices for flow rates from 1 to 5 mL

620 min-1. Pressure drops using the molecular sieve varied from 2061 to 9919 Pa whereas it was more limited

621 with the CNT foam, where pressure drops ranged between 445 to 2078 Pa. Therefore, the use of this

622 adsorbent decreased five times the pressure drops inside the PC.

623 Jang et al. [65] synthetized a carbon nanotube (CNT) sponge for the preconcentration of aromatic VOC.

624 5 mg of the CNT sponge were packed in a ¼” Pyrex glass tube (4 mm i.d. and 40 mm length). The tube

625 was connected using stainless steel fittings in which two porous metal frits were inserted to ensure

626 electrical contact with the CNT sponge. Due to its electrical resistance (15-20 Ω), this material can be

627 rapidly heated at a rate of 400 °C/s by simply applying a voltage of 26 V. This feature leads to an

628 efficient heat transfer from the adsorbent to the adsorbates promoting a quantitative desorption (>96%)

629 with low power consumption. Using 100 mL sample at concentrations of 36–63 ppb (v/v),

630 preconcentration factors (calculated as the ratio of the peak intensities obtained with the preconcentrator

631 and the sampling loop) of 88 (benzene) and 323 (toluene and o-xylene) were observed.

632 Lee et al. [34] developed a micropreconcentrator using a metal organic framework embedded in a metal

633 foam (MOFM) as an adsorbent. To evaluate its performance, commercial adsorbents (RAD145 and

634 Carbopack B) were packed in the same micropreconcentrator. The MOFM micropreconcentrator

635 exhibited pressure drops 4-3 times lower than the one packed with commercial adsorbents. Furthermore,

636 the analysis of a BTEX mixture revealed that the preconcentration factors achieved with MOFM PCs

637 were 2.6 and 4 times higher than the one packed with commercial adsorbents at a desorption temperature

638 of 150°C.

639

640 3.6. Evaluation of preconcentration performance

641 The existence of many factors affecting the performance of gas preconcentrators renders difficult to

642 establish the basis for a proper and easy comparison between different devices. Traditionally, the

643 performance of gas preconcentrators has been characterized by a figure of merit called preconcentration

644 factor (PF). Nevertheless, there is no general consensus among the authors and different definitions have

30
645 been adopted for the PF. In several studies, the PF is defined as the ratio between the maximum

646 concentration measured at the desorption step and the initial concentration of the injected sample

647 [34,77,78]. In some others, PF is calculated as the ratio between either the peak areas [1,29,38] or the

648 peak heights [1] obtained with the PC without the adsorbent and the PC packed with the adsorbent.

649 Other authors determine PF as the ratio between the peak area obtained using the GC system with and

650 without PC [36], which is close to the very first definition above. However, although PF is frequently

651 used to characterize the PC performance, it is not an intrinsic property of the preconcentrator itself [64]

652 as it depends on several external factors including concentration of the analyte, sampling flow rate and

653 dead volumes in the GC system. Among the intrinsic properties of the PC, the ability to obtain a flash

654 desorption of analytes thus producing sharp chromatographic peaks is of great importance. Figure 9

655 represents the preconcentration factor as a function of the heating rate (°C/s) used for the desorption of

656 different VOC. These data have been extracted from the data reported in Table 3 and Table 4 when both

657 parameters were given in the corresponding papers [28,32–34,36,38]. This figure seems to show that

658 the preconcentrator factor increases significantly with the heating rate for a heating rate value higher

659 than 40°C/s although no significant influence of this parameter was highlighted between 5 and 40°C/s.

660 This behaviour seems to be independent of the nature of the adsorbent. Unfortunately, there is little data

661 on the Figure 9 which implies that these interpretations should be taken with caution. For future

662 publications in the field, we recommend that the authors provide both data on the preconcentration factor

663 and the rate of desorption temperature increase. Thereby, whatever the definition of PF considered, this

664 parameter can be very useful when comparing the performances of similar devices under the same

665 experimental conditions, but it should be used with prudence when comparing various preconcentration

666 devices at different experimental conditions. The use of this parameter is not ideal, however, the

667 preconcentration factor is the only parameter provided in most of the publications to evaluate the

668 performance of the preconcentration step.

669

670

31
671 Indeed, the PF results are in general terms linear dependent with the sample volume, unless they are

672 close to the saturation of the adsorbent. Therefore, PF of a single device may vary several orders of

673 magnitude only by increasing the sample volume as illustrated in Figure 10 for the series of experiments

674 performed with granular adsorbents shown on the right part of this figure [1,7,30,32,33]. For example,

675 in the literature, some authors have achieved PF values as high as 800000 whereas some others were as

676 lower as 5.

677

678 Figure 9. Preconcentration factor (as defined by the authors) of different preconcentrators as a function
679 of heating rate: preconcentrators packed with commercial granular adsorbents ( ) [28,32,33] and
680 preconcentrators packed with other type of adsorbents ( ) [34,36,38]. The subscripts indicate the
681 different methods employed for the calculation of the preconcentration factor: “a” ratio of the
682 concentration peak area of the PC packed with adsorbent to that of the PC without adsorbent, “b” ratio
683 between peak area of the detector with and without the presence of a PC, “c” ratio between the maximum
684 concentration measured at the desorption step and the initial concentration of the injected sample gas,
685 “d” ratio of the volume of the air sample collected to the volume in which that same mass is contained
686 at the point of detection, “n” not mentioned.
687

32
688
689 Figure 10. Preconcentration factor of different preconcentrators as a function of sample volume:
690 preconcentrators packed with commercial granular adsorbents ( )[1,5,7,28,30,32,33] and
691 preconcentrators packed with other type of adsorbents ( ) [34,36,38,39]. The subscripts indicate the
692 method employed The subscripts indicate the different methods employed for the calculation of the
693 preconcentration factor: “a” ratio of the concentration peak area of the PC packed with adsorbent to that
694 of the PC without adsorbent, “b” ratio between peak area of the detector with and without the presence
695 of a PC, “c” ratio between the maximum concentration measured at the desorption step and the initial
696 concentration of the injected sample gas, “d” ratio of the volume of the air sample collected to the
697 volume in which that same mass is contained at the point of detection, “n” not mentioned.

698

699 In light of these discrepancies, it would be more convenient to use a different parameter to evaluate the

700 preconcentration performance of a gas preconcentration device which takes into account the sample

701 volume. We propose to use a new figure of merit called Normalized Preconcentration Efficacy (NPE)

702 presented in equation 1. This parameter represents the efficacy of the preconcentration step compared

703 to an ideal scenario, an injection using sampling loop where all the analyte molecules collected are

704 injected into the column. Therefore, NPE will range between 0 and 100%. To calculate this parameter,

705 we recommend using of a sampling loop with a known volume (i.e., 20-200 µL), a commercial capillary

706 column appropriate for the separation of the target compounds and a benchtop GC equipped with an

707 FID. If the use of a sampling loop is not possible, the injection of a known volume of gas can be

708 conducted manually using a gas-tight syringe.

709 To determine NPE, a gas sample with a specific concentration will be first analysed using the sampling

710 loop connected to the benchtop GC-FID. Then, the sampling loop will be replaced by the gas

711 preconcentrator and a gas sample with the same concentration will be analysed again. Obviously, the

33
712 sample volume used in both tests will be different and must be considered. Therefore, the

713 preconcentration efficacy will be calculated as the ratio between the peak areas normalized by the

714 injected volume in each test as shown in equation 1,

𝐴𝑃𝐶 Eq. 1
⁄𝑉 𝐴𝑃𝐶 ∙ 𝑉𝑆𝐿
𝑃𝐶
𝑁𝑃𝐸 (%) = ∙ 100 = ∙ 100
𝐴𝑆𝐿 𝐴𝑆𝐿 ∙ 𝑉𝑃𝐶
⁄𝑉
𝑆𝐿

715

716 where 𝐴𝑃𝐶 is the peak area measured by the GC-FID when using the preconcentrator, 𝑉𝑃𝐶 is the volume

717 of the air sample collected on the preconcentrator, 𝐴𝑆𝐿 is the peak area measured by the GC-FID when

718 using a sampling loop, and 𝑉𝑆𝐿 is the volume of the sampling loop. Note that peak areas must be

719 measured using the same benchtop equipment at the same experimental conditions. Alternatively, the

720 corresponding peak intensities can replace the peak areas considered in equation 1.

721

722 The sampling loop represents the ideal conditions in which all the analyte molecules collected are

723 injected into the column, thus, it can be considered as a reference. As the measurements using the

724 sampling loop and the gas preconcentrator are conducted at the same experimental conditions (length

725 and type of chromatography column, temperature program, detector), the variations between the two

726 peak areas can be only attributed to the features of the gas preconcentrator. Therefore, unlike the

727 preconcentration factor, this parameter can be used to compare the performances of different gas

728 preconcentration devices. Furthermore, NPE considers the ratio between the sample volumes, and thus,

729 it is independent of the gaseous sample volume used for its determination.

730 Sometimes, the preconcentration performance is evaluated using the whole GC system instead of the

731 single PC device. In this case, the sensitivity improvement due to the integration of a preconcentration

732 cannot be assessed and the system is evaluated as a whole. Considering the GC system, it is very

733 common to use the limit of detection (LOD) in concentration units like parts per million (ppm) or parts

734 per billion (pbb) to describe the sensitivity of the instrument. Chromatographic peaks areas increase

735 linearly when the sample volume increases [4,10]. For instance, a relatively small sample volume will

34
736 provide better temporal resolution while a large volume will provide a lower LOD. Therefore, since

737 LOD strongly depends on the sample volume, it is more accurate to the describe the sensitivity of each

738 system in a non-volume dependent magnitude such as the lowest detectable mass of analyte. This mass

739 can be calculated as the lowest LOD achieved multiplied by the sample volume. Only few publications

740 provide sensitivity in mass units, but this parameter has been calculated for several portable GC

741 instruments and presented in Table 1 and Table 2 for comparison. Using this parameter, variation of

742 several orders of magnitude can be observed among different portable GC systems (see Table 1 and

743 Table 2) whereas LOD of these instruments are very similar when they are expressed in ppm or ppb.

744 For instance, the GC developed by Alberto et al. [4] present an LOD of 0.06 for benzene whereas the

745 GC reported by Jiang et al. [21] exhibits an LOD of 0.02 ppb for this chemical; however, this latter has

746 a sensitivity of 90 pg while the sensitivity of the first one is only 3.6 pg.

747 3.7. Applications

748 In the last decades, the quantification of a wide variety of VOC molecules has generated considerable

749 research interest. These molecules are very often found at trace levels; therefore, portable GC integrating

750 PC units have been developed for a broad spectrum of applications. Since there is increasing evidence

751 of the effects of air quality on human health [80,81], the majority of the portable analytical systems

752 developed by research groups are focused on air quality and occupational exposure monitoring.

753 Figure 11 represents the sensitivity (in pg) of different portable GC as a function of air sample volume

754 (mL) used for VOC detection with either photoionization detector (PID) [4,10,11,15,17–21] or other

755 detectors such as thermal conductivity detector (TCD) [12], chemiresistor (CR) [2,6], capacitive detector

756 (CD) [24] and metal oxide sensors [8,13,23,25]. The data presented in Figure 11 have been extracted

757 from the literature when both sensitivity (in pg) and gas sample volume were available. These data are

758 detailed in Table 1 and Table 2. Regarding the own definition of the sensitivity (in pg) already mentioned

759 in this manuscript, the sensitivity should be independent of the sample volume. This is moreover more

760 or less the case for the portable GC coupled to other detectors different from PID with preconcentrators

761 packed with commercial granular adsorbents [2,6,22,24] and other adsorbents [8,12,13] since most of

762 them exhibited sensitivity around 500-2400 pg, except two of them where sensitivity values were 22611

35
763 pg [8] and 5736 pg [22] whatever the nature of the adsorbent used. This obvious trend is not observed

764 for the GC coupled to PID [4,10,17,18,20,21] but this observation could be explained by the variation

765 of other key parameters in these studies such as the heating rate.

766 In Figure 11, the general trend shows that, among the portable GC integrating miniaturized

767 preconcentrators, GC-PID are usually much more sensitive than the portable GC coupled to other

768 detectors. However, if we do not consider Plasma or discharge micro PID (D-µPID) [82,83],

769 photoionization detectors are more specific to unsaturated organic compounds such as aromatic species

770 and carbonyl compounds and they are not able to quantify other VOC with higher ionization energies,

771 whereas other detectors such as capacitive detectors or metal oxide sensors are more universal and are

772 able to detect a large number of VOC families.

773
774 Figure 11. Sensitivity of different portable GC as a function of sample volume: portable GC-PID with
775 preconcentrators packed with commercial granular adsorbents ( ) [4,10,18,21] and other adsorbents (
776 ) [17,20] and portable GC coupled to detectors different from PID with preconcentrators packed with
777 commercial granular adsorbents ( ) [2,6,22,24] and other adsorbents ( ) [8,12,13].

778

779 Qin and Gianchadani [24] developed a micro gas chromatography system in which all the components

780 were microfabricated (see Figure 12a). This device was able to detect a mixture of common indoor air

781 pollutants including alkanes, aromatic hydrocarbons, aldehydes, halogenated hydrocarbons and terpenes

782 at ppb levels. Wang et al. [6] developed a belt-mountable µGC prototype to monitor personal exposure

36
783 able to analyse a sample containing a mixture of 21 VOC in 3.5 min (see Figure 12b). The PC consist

784 of two microfluidic cavities separated by rows of pillars and packed with ∼2 mg of Carbopack B and

785 Carbopack X. This system has a LOD from 16 to 600 ppb depending on the compound. Trzciński et al.

786 [20] reported the fabrication of a sensor prototype for environmental benzene monitoring. This device

787 was packed with a customized adsorbent that selectively trap benzene over the rest of aromatic

788 compounds. Garg et al. [12] developed a µGC for the detection of hazardous air pollutants such as

789 benzene, toluene, tetrachloroethylene, chlorobenzene, ethylbenzene, and p-xylene (see Figure 12c). The

790 instrument integrates a MEMS-based PC with embedded pillars coated with a thin layer of Tenax TA

791 as adsorbent. By sampling for 10 min at a flow rate of 1 mL/min, a limit of detection of ∼1 ng was

792 achieved.

793 More recently, portable analytical systems for the detection of explosives have been also reported. Since

794 these compounds easily degrade into a variety of species, the analysis is often focused on these by-

795 products instead on the explosive molecule itself. Mohsen et al. [1] designed a µGC for the detection of

796 an explosive-related compound, namely ortho-nitrotoluene. The system consisted of a silicon PC and a

797 GC column coupled to a chemical gas sensor and allowed to detect down to 365 ppb of this compound

798 in the presence of interfering compounds such as toluene and water. Collin et al. [2] presented a very

799 compact GC for the quantification of 2,4,6-trinitrotoluene (TNT) derived products (see Figure 12d). The

800 miniaturized PC consisted of a cavity etched in a silicon substrate, sealed with a Pyrex plate and packed

801 with 2.4 mg of Carbopack B. For 1 L sample, LOD of 0.30 ppb (2161 pg), 0.067 ppb (499 pg), and 0.12

802 ppb (894 pg) were achieved for 2,3-dimethyl-2,3-dinitrobutane (DMNB), 2,6-dinitrotoluene (2,6-DNT)

803 and 2,4-dinitrotoluene (2,4-DNT), respectively.

804 Ethylene monitoring is of great interest to predict shelf life of climacteric fruits such as bananas. In this

805 context, Zaidi et al. [22] presented a portable GC system with an integrated miniaturized PC packed

806 with Carbosieve SII for the selective detection of ethylene (see Figure 12e). For a sample volume of 200

807 mL, the instrument has a LOD of 25 ppb (5,736 pg) and no interference was observed from other

808 ripening gases such as CO2, O2 and humidity.

37
809 The low LOD and continuous monitoring of ethylene allow to identify any issues with the goods thus

810 avoiding unnecessary transports and reducing costs and pollution emissions.

811

812 Figure 12. Pictures of portable chromatographs from a) Qin and Gianchadani [24], b) Wang et al. [6],
813 c) Garg et al. [12] d) Collin et al. [2] and e) Zaidi et al. [22].

814

815 Recent progress in medicine showed that some VOC such as acetone, ethanol, isoprene or heptane are

816 biomarkers of certain diseases including asthma, chronic obstructive pulmonary disease and lung cancer

817 [84,85]. These biomarkers are present in human breath at the ppb level; therefore, an accurate and

818 sensitive analysis in near real time may constitute a non-invasive method for early diagnosis of these

819 serious diseases. To this purpose, several µGC integrating PC units have been developed. Gregis et al.

820 [8] reported a µGC to monitor four biomarkers (propanol, toluene, o-xylene and cyclohexane) associated

821 to lung cancer. The PC was composed of a microcavity with embedded micropillars filled with zeolite

822 DaY. This system exhibited LOD of 24 ppb (22,611 pg), 5 ppb (5,428 pg), 21 ppb (12,904 pg) and 112

823 ppb (96,379 pg) for toluene, o-xylene, propanol and cyclohexane, respectively for a sample volume of

824 250 mL. The LOD obtained for the first three compounds was maintained even in samples containing

825 CO2 and water levels similar to those found in real human breath samples. Han et al. [28] reported a

826 low-cost PC for the preconcentration of isoprene, a typical biomarker of chronic liver disease. This

827 device was made of copper and packed with 20 mg of Carbopack X. Isoprene concentrations as low as

38
828 10 ppb (501 pg) can be achieved for a 18 mL sample when this PC is coupled to a GC-FID. However,

829 despite the great advances made in the last years, high concentration of CO2 (~ 40,000 ppm) and the

830 relative humidity of 100 % at 25 °C [8] present in human breath make accurate analysis of human breath

831 still a challenge.

832

833 4. Conclusions

834 Recent trends in the development of miniaturized gas preconcentrators for portable gas chromatography

835 are reported in this paper. Alternative fabrication techniques able to build miniaturized devices, different

836 fluidic layouts promoting flow uniformity, new adsorbent materials with low pressure drops and the

837 integration of standard gas flow fittings for more reliable interfacing are some of the advances made in

838 the last years. These developments can improve the overall features of portable GC like energy

839 consumption, cost, sensitivity and reproducibility.

840 Nowadays, MEMS-based techniques such as photolithography and dry etching continue to be the most

841 popular for the fabrication of micro PC. Even though these techniques are the most convenient

842 technology for micro fabrication of preconcentrators, they are not free of drawbacks. The fragility of the

843 devices produced by this technology and the lack of standards fluidic interfacing are major challenges

844 as it is evidenced by the presence of few MEMS-based PCs in commercial instruments. Furthermore,

845 despite their potential for high-volume production, the development of these devices involves the use of

846 cleanrooms facilities what can be expensive and time-consuming for scientific research.

847 Given this reality, some new approaches are being adopted. Laser etching, micro-milling or additive 3D

848 printing in metal are three big promises to overcome the problems in the preconcentrators derived from

849 the MEMS technology. The opportunity of using alternative materials other than silicon opens a new

850 range of possibilities regarding PC design and integration of standard fluidic connections. Nonetheless,

851 a balance must be pursued since these other techniques do not achieve the performance of MEMS-based

852 PC in terms of power consumption, heating rates and consequently narrowness of chromatographic

853 peaks.

39
854 Traditionally, granular adsorbents have been used in miniaturized PC. Due to the issues associated with

855 high pressure drops and poor heat transfer, thin film adsorbents were lately developed. Currently, a third

856 generation of promising adsorbents, foam-based adsorbents, is being developed. This new type of

857 materials provides a balance between a reasonable adsorption capacity and limited pressure drop.

858 Moreover, foams exhibited better heat transfer efficiency than their granular and thin film counterparts

859 due to its electrical resistance, resulting in a quantitative desorption with lower power consumption.

860 However, commercial granular adsorbents are still frequently used in many devices since their

861 behaviour and characteristics are well-known.

862 Finally, there is a need to establish a standard metric to evaluate the preconcentration performances of

863 the different PCs that allows to compare the results provided by the different authors. In this review, we

864 propose the use of the Normalized Preconcentration Efficacy (NPE) in future publications as a figure of

865 merit to overcome this problem.

866 In addition, in order to evaluate the overall analytical GC performances, LOD should be expressed in

867 both ppb or ppm levels but also in pg or ng to introduce the sample volume considered in the

868 experiments. Indeed, the LOD and PF are linear dependent with the sample volume as long as we are

869 far below the saturation of the adsorbent. This sample volume, the heating rate and the power

870 consumption should be also mentioned in all future publications in the field. This set of information will

871 facilitate the comparison of analytical performances between PCs reported in several studies and LOD

872 between µGC.

873 Acknowledgments
874 This publication is part of a project that has received funding from the European Union’s Horizon 2020

875 research and innovation program under the Marie Skłodowska-Curie grant agreement No 643095, the

876 Clean Sky 2 Joint Undertaking under the European Union’s Horizon 2020 research and innovation

877 program under grant agreement No 687014 and the LIFE program, grant number LIFE17

878 ENV/FR/000330. This project was also supported through the European project ELCOD. The ELCOD

879 project was implemented as a part of the INTERREG V Oberrhein/Rhin Superieur program and was

40
880 supported by the European Regional Development Fund (ERDF) and the co-financed project partners

881 Region Grand Est in France and the countries of Baden-Württemberg and Rhineland-Palatinate.

882 The authors are also grateful to the CNRS, the University of Strasbourg and In’Air Solutions.

883

884 References
885 [1] Y. Mohsen, H. Lahlou, J.-B. Sanchez, F. Berger, I. Bezverkhyy, G. Weber, J.-P. Bellat,
886 Development of a micro-analytical prototype for selective trace detection of orthonitrotoluene,
887 Microchemical Journal. 114 (2014) 48–52. https://doi.org/10.1016/j.microc.2013.12.001.
888 [2] W.R. Collin, G. Serrano, L.K. Wright, H. Chang, N. Nuñovero, E.T. Zellers, Microfabricated
889 Gas Chromatograph for Rapid, Trace-Level Determinations of Gas-Phase Explosive Marker
890 Compounds, Anal. Chem. 86 (2014) 655–663. https://doi.org/10.1021/ac402961t.
891 [3] I. Lara-Ibeas, C. Trocquet, R. Nasreddine, C. Andrikopoulou, V. Person, B. Cormerais, S.
892 Englaro, S. Le Calvé, BTEX near real-time monitoring in two primary schools in La Rochelle,
893 France, Air Qual Atmos Health. 11 (2018) 1091–1107. https://doi.org/10.1007/s11869-018-
894 0611-3.
895 [4] A. Rodríguez-Cuevas, I. Lara-Ibeas, A. Leprince, M. Wolf, S. Le Calvé, Easy-to-manufacture
896 micro gas preconcentrator integrated in a portable GC for enhanced trace detection of BTEX,
897 Sensors and Actuators B: Chemical. 324 (2020) 128690.
898 https://doi.org/10.1016/j.snb.2020.128690.
899 [5] J. Bryant-Genevier, E.T. Zellers, Toward a microfabricated preconcentrator-focuser for a
900 wearable micro-scale gas chromatograph, Journal of Chromatography A. 1422 (2015) 299–309.
901 https://doi.org/10.1016/j.chroma.2015.10.045.
902 [6] J. Wang, N. Nuñovero, R. Nidetz, S.J. Peterson, B.M. Brookover, W.H. Steinecker, E.T. Zellers,
903 Belt-Mounted Micro-Gas-Chromatograph Prototype for Determining Personal Exposures to
904 Volatile-Organic-Compound Mixture Components, Anal. Chem. 91 (2019) 4747–4754.
905 https://doi.org/10.1021/acs.analchem.9b00263.
906 [7] A.M. Rydosz, D. Michon, K. Domanski, W. Maziarz, T. Pisarkiewicz, Various Preconcentrator
907 Structures For Determination of Acetone in a Wide Range of Concentration, Advances in
908 Electrical and Electronic Engineering. 14 (2016) 59-65–65.
909 https://doi.org/10.15598/aeee.v14i1.1525.
910 [8] G. Gregis, J.-B. Sanchez, I. Bezverkhyy, W. Guy, F. Berger, V. Fierro, J.-P. Bellat, A. Celzard,
911 Detection and quantification of lung cancer biomarkers by a micro-analytical device using a
912 single metal oxide-based gas sensor, Sensors and Actuators B: Chemical. 255 (2018) 391–400.
913 https://doi.org/10.1016/j.snb.2017.08.056.
914 [9] Directive 2008/50/EC of the European Parliament and of the Council of 21 May 2008 on ambient
915 air quality and cleaner air for Europe, 2008. http://data.europa.eu/eli/dir/2008/50/oj/eng
916 (accessed October 23, 2019).
917 [10] I. Lara-Ibeas, A. Rodríguez-Cuevas, C. Andrikopoulou, V. Person, L. Baldas, S. Colin, S. Le
918 Calvé, Sub-ppb Level Detection of BTEX Gaseous Mixtures with a Compact Prototype GC
919 Equipped with a Preconcentration Unit, Micromachines. 10 (2019) 187.
920 https://doi.org/10.3390/mi10030187.
921 [11] J.-C. Soo, E.G. Lee, R.F. LeBouf, M.L. Kashon, W. Chisholm, M. Harper, Evaluation of a
922 portable gas chromatograph with photoionization detector under variations of VOC
923 concentration, temperature, and relative humidity, Journal of Occupational and Environmental
924 Hygiene. 15 (2018) 351–360. https://doi.org/10.1080/15459624.2018.1426860.
925 [12] A. Garg, M. Akbar, E. Vejerano, S. Narayanan, L. Nazhandali, L.C. Marr, M. Agah, Zebra GC:
926 A mini gas chromatography system for trace-level determination of hazardous air pollutants,

41
927 Sensors and Actuators B: Chemical. 212 (2015) 145–154.
928 https://doi.org/10.1016/j.snb.2014.12.136.
929 [13] S. Zampolli, I. Elmi, F. Mancarella, P. Betti, E. Dalcanale, G.C. Cardinali, M. Severi, Real-time
930 monitoring of sub-ppb concentrations of aromatic volatiles with a MEMS-enabled miniaturized
931 gas-chromatograph, Sensors and Actuators B: Chemical. 141 (2009) 322–328.
932 https://doi.org/10.1016/j.snb.2009.06.021.
933 [14] J. Yeom, C.R. Field, B. Bae, R.I. Masel, M.A. Shannon, The design, fabrication and
934 characterization of a silicon microheater for an integrated MEMS gas preconcentrator, J.
935 Micromech. Microeng. 18 (2008) 125001. https://doi.org/10.1088/0960-1317/18/12/125001.
936 [15] J. Lee, M. Zhou, H. Zhu, R. Nidetz, K. Kurabayashi, X. Fan, Fully Automated Portable
937 Comprehensive 2-Dimensional Gas Chromatography Device, Anal. Chem. 88 (2016) 10266–
938 10274. https://doi.org/10.1021/acs.analchem.6b03000.
939 [16] X. Huang, T. Bauder, T. Do, H. Suen, C. Boss, P. Kwon, J. Yeom, A Binder Jet Printed, Stainless
940 Steel Preconcentrator as an In-Line Injector of Volatile Organic Compounds, Sensors. 19 (2019)
941 2748. https://doi.org/10.3390/s19122748.
942 [17] D.-W. You, Y.-S. Seon, Y. Jang, J. Bang, J.-S. Oh, K.-W. Jung, A portable gas chromatograph
943 for real-time monitoring of aromatic volatile organic compounds in air samples, Journal of
944 Chromatography A. 1625 (2020) 461267. https://doi.org/10.1016/j.chroma.2020.461267.
945 [18] K.M. Skog, F. Xiong, H. Kawashima, E. Doyle, R. Soto, D.R. Gentner, Compact, Automated,
946 Inexpensive, and Field-Deployable Vacuum-Outlet Gas Chromatograph for Trace-Concentration
947 Gas-Phase Organic Compounds, Anal. Chem. 91 (2019) 1318–1327.
948 https://doi.org/10.1021/acs.analchem.8b03095.
949 [19] J. Sun, N. Xue, W. Wang, H. Wang, C. Liu, T. Ma, T. Li, T. Tan, Compact prototype GC-PID
950 system integrated with micro PC and micro GC column, Journal of Micromechanics and
951 Microengineering. 29 (2018) 29. https://doi.org/10.1088/1361-6439/aaf42c.
952 [20] J.W. Trzciński, R. Pinalli, N. Riboni, A. Pedrini, F. Bianchi, S. Zampolli, I. Elmi, C. Massera, F.
953 Ugozzoli, E. Dalcanale, In Search of the Ultimate Benzene Sensor: The EtQxBox Solution, ACS
954 Sensors. 2 (2017) 590–598. https://doi.org/10.1021/acssensors.7b00110.
955 [21] R.-S. Jian, Y.-S. Huang, S.-L. Lai, L.-Y. Sung, C.-J. Lu, Compact instrumentation of a μ-GC for
956 real time analysis of sub-ppb VOC mixtures, Microchemical Journal. 108 (2013) 161–167.
957 https://doi.org/10.1016/j.microc.2012.10.016.
958 [22] N.A. Zaidi, M.W. Tahir, M.J. Vellekoop, W. Lang, Design of Novel Ceramic Preconcentrator
959 and Integration in Gas Chromatographic System for Detection of Ethylene Gas from Ripening
960 Bananas, Sensors. 18 (2018) 2589. https://doi.org/10.3390/s18082589.
961 [23] T.-H. Tzeng, C.-Y. Kuo, S.-Y. Wang, P.-K. Huang, Y.-M. Huang, Y.-M. Huang, W.-C. Hsieh,
962 Y.-J. Huang, P.-H. Kuo, S.-A. Yu, S.-C. Lee, Y.J. Tseng, W.-C. Tian, S.-S. Lu, A Portable Micro
963 Gas Chromatography System for Lung Cancer Associated Volatile Organic Compound
964 Detection, IEEE Journal of Solid-State Circuits. 51 (2016) 259–272.
965 https://doi.org/10.1109/JSSC.2015.2489839.
966 [24] Y. Qin, Y.B. Gianchandani, A fully electronic microfabricated gas chromatograph with
967 complementary capacitive detectors for indoor pollutants, Microsystems & Nanoengineering. 2
968 (2016) 15049. https://doi.org/10.1038/micronano.2015.49.
969 [25] H. Lahlou, J.-B. Sanchez, Y. Mohsen, X. Vilanova, F. Berger, E. Llobet, X. Correig, V. Fierro,
970 A. Celzard, I. Gràcia, A planar micro-concentrator/injector for low power consumption
971 microchromatographic analysis of benzene and 1, 3 butadiene, Microsystem Technologies. 18
972 (2012) 489–495.
973 [26] C. Zhan, M. Akbar, R. Hower, N. Nuñovero, J. A. Potkay, E. T. Zellers, A micro passive
974 preconcentrator for micro gas chromatography, Analyst. (2020).
975 https://doi.org/10.1039/D0AN01485K.
976 [27] T.H. Chappuis, B.A.P. Ho, M. Ceillier, F. Ricoul, M. Alessio, J.-F. Beche, C. Corne, G. Besson,
977 J. Vial, D. Thiébaut, B. Bourlon, Miniaturization of breath sampling with silicon chip: application
978 to volatile tobacco markers tracking, J. Breath Res. (2018). https://doi.org/10.1088/1752-
979 7163/aad384.

42
 980 [28] B. Han, H. Wang, H. Huang, T. Liu, G. Wu, J. Wang, Micro-fabricated packed metal gas
981 preconcentrator for enhanced monitoring of ultralow concentration of isoprene, Journal of
982 Chromatography A. (2018). https://doi.org/10.1016/j.chroma.2018.08.058.
983 [29] M.M. McCartney, Y. Zrodnikov, A.G. Fung, M.K. LeVasseur, J.M. Pedersen, K.O. Zamuruyev,
984 A.A. Aksenov, N.J. Kenyon, C.E. Davis, An Easy to Manufacture Micro Gas Preconcentrator
985 for Chemical Sensing Applications, ACS Sensors. 2 (2017) 1167–1174.
986 https://doi.org/10.1021/acssensors.7b00289.
987 [30] S. Janssen, T. Tessmann, W. Lang, High sensitive and selective ethylene measurement by using
988 a large-capacity-on-chip preconcentrator device, Sensors and Actuators B: Chemical. 197 (2014)
989 405–413. https://doi.org/10.1016/j.snb.2014.02.001.
990 [31] M. Leidinger, M. Rieger, T. Sauerwald, C. Alépée, A. Schütze, Integrated pre-concentrator gas
991 sensor microsystem for ppb level benzene detection, Sensors and Actuators B: Chemical. 236
992 (2016) 988–996. https://doi.org/10.1016/j.snb.2016.04.064.
993 [32] M. Camara, P. Breuil, C. Pijolat, J.P. Viricelle, N.F. de Rooij, D. Briand, Tubular gas
994 preconcentrators based on inkjet printed micro-hotplates on foil, Sensors and Actuators B:
995 Chemical. 236 (2016) 1111–1117. https://doi.org/10.1016/j.snb.2016.06.121.
996 [33] T. Sukaew, H. Chang, G. Serrano, E.T. Zellers, Multi-stage preconcentrator/focuser module
997 designed to enable trace level determinations of trichloroethylene in indoor air with a
998 microfabricated gas chromatograph, Analyst. 136 (2011) 1664–1674.
999 https://doi.org/10.1039/C0AN00780C.
1000 [34] J. Lee, J. Lee, S.-H. Lim, Micro gas preconcentrator using metal organic framework embedded
1001 metal foam for detection of low-concentration volatile organic compounds, Journal of Hazardous
1002 Materials. 392 (2020) 122145. https://doi.org/10.1016/j.jhazmat.2020.122145.
1003 [35] B. Zhao, F. Feng, X. Yang, F. Luo, H. Zhou, X. Li, Improved Performance of Micro-
1004 Preconcentrator Using Silicon Nanowires as a Surface Template, in: 2019 20th International
1005 Conference on Solid-State Sensors, Actuators and Microsystems Eurosensors XXXIII
1006 (TRANSDUCERS EUROSENSORS XXXIII), 2019: pp. 1281–1284.
1007 https://doi.org/10.1109/TRANSDUCERS.2019.8808319.
1008 [36] J. Lee, S.-H. Lim, CNT Foam-Embedded Micro Gas Preconcentrator for Low-Concentration
1009 Ethane Measurements, Sensors (Basel). 18 (2018). https://doi.org/10.3390/s18051547.
1010 [37] Y. Zhang, J. Sun, X. Zhu, J. Liu, Z. Ning, Mini pretreatment system integrated with micro pre-
1011 concentrator and micro GC column, Sensor Review. 37 (2017) 137–141.
1012 https://doi.org/10.1108/SR-09-2016-0154.
1013 [38] C.-Y. Kuo, P.-S. Chen, H.-T. Chen, C.-J. Lu, W.-C. Tian, Development of micromachined
1014 preconcentrators and gas chromatographic separation columns by an electroless gold plating
1015 technology, J. Micromech. Microeng. 27 (2017) 035012. https://doi.org/10.1088/1361-
1016 6439/aa5aa3.
1017 [39] F. Almazán, I. Pellejero, A. Morales, M.A. Urbiztondo, J. Sesé, M.P. Pina, Jesús Santamaría,
1018 Zeolite based microconcentrators for volatile organic compounds sensing at trace-level:
1019 fabrication and performance, J. Micromech. Microeng. 26 (2016) 084010.
1020 https://doi.org/10.1088/0960-1317/26/8/084010.
1021 [40] X.F. Zhu, Z.W. Ning, J.H. Sun, T.J. Ma, Y.N. Zhang, J.H. Liu, Micro-Fabricated Pre-
1022 Concentrator Filled with Single-Walled Carbon Nanotubes as Adsorbent Material, Key
1023 Engineering Materials. 645–646 (2015) 681–686.
1024 https://doi.org/10.4028/www.scientific.net/KEM.645-646.681.
1025 [41] M. Clément, S. Arzel, B. Le Bot, R. Seux, M. Millet, Adsorption/thermal desorption-GC/MS for
1026 the analysis of pesticides in the atmosphere, Chemosphere. 40 (2000) 49–56.
1027 https://doi.org/10.1016/S0045-6535(99)00230-1.
1028 [42] T. Mokalled, J. Adjizian Gérard, M. Abboud, C. Trocquet, R. Nassreddine, V. Person, S. le
1029 Calvé, VOC tracers from aircraft activities at Beirut Rafic Hariri International Airport,
1030 Atmospheric Pollution Research. 10 (2019) 537–551. https://doi.org/10.1016/j.apr.2018.09.009.
1031 [43] L. Tuduri, M. Millet, O. Briand, M. Montury, Passive air sampling of semi-volatile organic
1032 compounds, TrAC Trends in Analytical Chemistry. 31 (2012) 38–49.
1033 https://doi.org/10.1016/j.trac.2011.08.007.

43
1034 [44] L. Vallecillos, A. Borrull, R.M. Marcé, F. Borrull, Passive sampling to control air quality in
1035 schools: Uptake rate determination and application, Indoor Air. 30 (2020) 1005–1017.
1036 https://doi.org/10.1111/ina.12684.
1037 [45] M. Lévy, J. Al-Alam, C. Ridacker, S. Massemin, M. Millet, Use of XAD®-2 passive air samplers
1038 for monitoring environmental trends of PAHs, PCBs and pesticides in three different sites in
1039 Strasbourg and its vicinity (east of France), Atmospheric Environment. 195 (2018) 12–23.
1040 https://doi.org/10.1016/j.atmosenv.2018.09.052.
1041 [46] G. Serrano, T. Sukaew, E.T. Zellers, Hybrid preconcentrator/focuser module for determinations
1042 of explosive marker compounds with a micro-scale gas chromatograph, Journal of
1043 Chromatography A. 1279 (2013) 76–85. https://doi.org/10.1016/j.chroma.2013.01.009.
1044 [47] Y. Qin, Y.B. Gianchandani, iGC2 : an architecture for micro gas chromatographs utilizing
1045 integrated bi-directional pumps and multi-stage preconcentrators, J. Micromech. Microeng. 24
1046 (2014) 065011. https://doi.org/10.1088/0960-1317/24/6/065011.
1047 [48] A. Ghosh, C.R. Vilorio, A.R. Hawkins, M.L. Lee, Microchip gas chromatography columns,
1048 interfacing and performance, Talanta. 188 (2018) 463–492.
1049 https://doi.org/10.1016/j.talanta.2018.04.088.
1050 [49] W.I. Wu, P. Rezai, H.H. Hsu, P.R. Selvaganapathy, Materials and methods for the
1051 microfabrication of microfluidic biomedical devices, in: X. (James) Li, Y. Zhou (Eds.),
1052 Microfluidic Devices for Biomedical Applications, Woodhead Publishing, 2013: pp. 3–62.
1053 https://doi.org/10.1533/9780857097040.1.3.
1054 [50] M.A. Zareian-Jahromi, M. Ashraf-Khorassani, L.T. Taylor, M. Agah, Design, Modeling, and
1055 Fabrication of MEMS-Based Multicapillary Gas Chromatographic Columns, Journal of
1056 Microelectromechanical Systems. 18 (2009) 28–37.
1057 https://doi.org/10.1109/JMEMS.2008.2007267.
1058 [51] T.H. Chappuis, B.A.P. Ho, M. Ceillier, F. Ricoul, M. Alessio, J.-F. Beche, C. Corne, G. Besson,
1059 J. Vial, D. Thiébaut, B. Bourlon, Miniaturization of breath sampling with silicon chip: application
1060 to volatile tobacco markers tracking, J. Breath Res. (2018). https://doi.org/10.1088/1752-
1061 7163/aad384.
1062 [52] S. Zampolli, I. Elmi, G.C. Cardinali, L. Masini, F. Bonafè, F. Zardi, Compact-GC platform: A
1063 flexible system integration strategy for a completely microsystems-based gas-chromatograph,
1064 Sensors and Actuators B: Chemical. 305 (2020) 127444.
1065 https://doi.org/10.1016/j.snb.2019.127444.
1066 [53] J. Wang, J. Bryant-Genevier, N. Nuñovero, C. Zhang, B. Kraay, C. Zhan, K. Scholten, R. Nidetz,
1067 S. Buggaveeti, E.T. Zellers, Compact prototype microfabricated gas chromatographic analyzer
1068 for autonomous determinations of VOC mixtures at typical workplace concentrations,
1069 Microsystems & Nanoengineering. 4 (2018) 17101.
1070 https://doi.org/10.1038/micronano.2017.101.
1071 [54] C. Iliescu, H. Taylor, M. Avram, J. Miao, S. Franssila, A practical guide for the fabrication of
1072 microfluidic devices using glass and silicon, Biomicrofluidics. 6 (2012) 016505.
1073 https://doi.org/10.1063/1.3689939.
1074 [55] Z. Cui, Wafer Bonding, in: D. Li (Ed.), Encyclopedia of Microfluidics and Nanofluidics,
1075 Springer US, Boston, MA, 2008: pp. 2179–2183. https://doi.org/10.1007/978-0-387-48998-
1076 8_1682.
1077 [56] N.A. Zaidi, M.W. Tahir, M.J. Vellekoop, W. Lang, A Gas Chromatographic System for the
1078 Detection of Ethylene Gas Using Ambient Air as a Carrier Gas, Sensors. 17 (2017) 2283.
1079 https://doi.org/10.3390/s17102283.
1080 [57] H. Lahlou, X. Vilanova, X. Correig, Gas phase micro-preconcentrators for benzene monitoring:
1081 A review, Sensors and Actuators B: Chemical. 176 (2013) 198–210.
1082 https://doi.org/10.1016/j.snb.2012.10.004.
1083 [58] D. Michoń, A. Rydosz, K. Domański, W. Maziarz, T. Pisarkiewicz, Detection of acetone in
1084 exhaled breath with the use of micropreconcentrator and a commercial gas sensor, in: B.
1085 Swatowska, W. Maziarz, T. Pisarkiewicz, W. Kucewicz (Eds.), 2016: p. 1017511.
1086 https://doi.org/10.1117/12.2263268.

44
1087 [59] M. Li, S. Biswas, M.H. Nantz, R.M. Higashi, X.-A. Fu, A microfabricated preconcentration
1088 device for breath analysis, Sensors and Actuators B: Chemical. 180 (2013) 130–136.
1089 https://doi.org/10.1016/j.snb.2012.07.034.
1090 [60] E.H.M. Camara, P. Breuil, D. Briand, N.F. de Rooij, C. Pijolat, A micro gas preconcentrator with
1091 improved performance for pollution monitoring and explosives detection, Analytica Chimica
1092 Acta. 688 (2011) 175–182. https://doi.org/10.1016/j.aca.2010.12.039.
1093 [61] B. Alfeeli, M. Agah, MEMS-Based Selective Preconcentration of Trace Level Breath Analytes,
1094 IEEE Sensors Journal. 9 (2009) 1068–1075. https://doi.org/10.1109/JSEN.2009.2025822.
1095 [62] B. Alfeeli, H. Vereb, A. Dietrich, M. Agah, Low pressure drop micro preconcentrators with
1096 cobweb Tenax-TA film for analysis of human breath, in: 2011 IEEE 24th International
1097 Conference on Micro Electro Mechanical Systems, 2011: pp. 916–919.
1098 https://doi.org/10.1109/MEMSYS.2011.5734575.
1099 [63] M. Akbar, M. Agah, A Microfabricated Propofol Trap for Breath-Based Anesthesia Depth
1100 Monitoring, J. Microelectromech. Syst. 22 (2013) 443–451.
1101 https://doi.org/10.1109/JMEMS.2012.2227949.
1102 [64] B. Alfeeli, Chemical Micro Preconcentrators Development for Micro Gas Chromatography
1103 Systems, (2010). https://vtechworks.lib.vt.edu/handle/10919/29305 (accessed September 30,
1104 2019).
1105 [65] Y. Jang, J. Bang, Y.-S. Seon, D.-W. You, J.-S. Oh, K.-W. Jung, Carbon nanotube sponges as an
1106 enrichment material for aromatic volatile organic compounds, Journal of Chromatography A.
1107 1617 (2020) 460840. https://doi.org/10.1016/j.chroma.2019.460840.
1108 [66] A. Rydosz, W. Maziarz, T. Pisarkiewicz, K. Domański, P. Grabiec, A gas micropreconcentrator
1109 for low level acetone measurements, Microelectronics Reliability. 52 (2012) 2640–2646.
1110 https://doi.org/10.1016/j.microrel.2012.05.012.
1111 [67] C.-J. Lu, E.T. Zellers, A dual-adsorbent preconcentrator for a portable indoor-VOC microsensor
1112 system, Analytical Chemistry. 73 (2001) 3449–3457.
1113 [68] J. Lee, M. Jung, S. Barthwal, S. Lee, S.-H. Lim, MEMS gas preconcentrator filled with CNT
1114 foam for exhaled VOC gas detection, BioChip Journal. 9 (2015) 44–49.
1115 https://doi.org/10.1007/s13206-014-9106-y.
1116 [69] B. Han, G. Wu, H. Wang, J. Wang, Micro-fabricated packed metal gas preconcentrator for low
1117 detection limit exhaled VOC gas measurements, in: IEEE, 2017: pp. 269–273.
1118 https://doi.org/10.1109/3M-NANO.2017.8286278.
1119 [70] J. Wang, N. Nuñovero, Z. Lin, R. Nidetz, S. Buggaveeti, C. Zhan, K. Kurabayashi, W.H.
1120 Steinecker, E.T. Zellers, A Wearable MEMS Gas Chromatograph for Multi-Vapor
1121 Determinations, Procedia Engineering. 168 (2016) 1398–1401.
1122 https://doi.org/10.1016/j.proeng.2016.11.391.
1123 [71] J. Wang, J. Ma, E.T. Zellers, Room-temperature-ionic-liquid coated graphitized carbons for
1124 selective preconcentration of polar vapors, Journal of Chromatography A. 1609 (2020) 460486.
1125 https://doi.org/10.1016/j.chroma.2019.460486.
1126 [72] R. Huang, M. Lu, P. Wang, Y. Chen, J. Wu, M. Fu, L. Chen, D. Ye, Enhancement of the non-
1127 thermal plasma-catalytic system with different zeolites for toluene removal, RSC Adv. 5 (2015)
1128 72113–72120. https://doi.org/10.1039/C5RA13604K.
1129 [73] I. Lara-Ibeas, C. Megías-Sayago, A. Rodríguez-Cuevas, R. Ocampo-Torres, B. Louis, S. Colin,
1130 S. Le Calvé, Adsorbent screening for airborne BTEX analysis and removal, Journal of
1131 Environmental Chemical Engineering. (2019) 103563.
1132 https://doi.org/10.1016/j.jece.2019.103563.
1133 [74] I. Lara-Ibeas, C. Megías-Sayago, B. Louis, S. Le Calvé, Adsorptive removal of gaseous
1134 formaldehyde at realistic concentrations, Journal of Environmental Chemical Engineering. 8
1135 (2020) 103986. https://doi.org/10.1016/j.jece.2020.103986.
1136 [75] X. Huang, Z. Huang, L. Zhang, R. Liu, Y. Lv, Highly efficient cataluminescence gas sensor for
1137 acetone vapor based on UIO-66 metal-organic frameworks as preconcentrator, Sensors and
1138 Actuators B: Chemical. 312 (2020) 127952. https://doi.org/10.1016/j.snb.2020.127952.
1139 [76] M.-Y. Wong, W.-R. Cheng, M.-H. Liu, W.-C. Tian, C.-J. Lu, A preconcentrator chip employing
1140 μ-SPME array coated with in-situ-synthesized carbon adsorbent film for VOCs analysis, Talanta.
1141 101 (2012) 307–313. https://doi.org/10.1016/j.talanta.2012.09.031.

45
1142 [77] E.H.M. Camara, Développement d’un micro-préconcentrateur pour la détection de substances
1143 chimiques à l’état de trace en phase gaz, (2009). https://tel.archives-ouvertes.fr/tel-00448980
1144 (accessed October 11, 2019).
1145 [78] R. Kalidoss, S. Umapathy, An overview on the exponential growth of non-invasive diagnosis of
1146 diabetes mellitus from exhaled breath by nanostructured metal oxide Chemi-resistive gas sensors
1147 and μ-preconcentrator, Biomed Microdevices. 22 (2020) 2. https://doi.org/10.1007/s10544-019-
1148 0448-z.
1149 [79] V. Gold, ed., The IUPAC Compendium of Chemical Terminology: The Gold Book, 4th ed.,
1150 International Union of Pure and Applied Chemistry (IUPAC), Research Triangle Park, NC, 2019.
1151 https://doi.org/10.1351/goldbook.
1152 [80] M.P. Tsakas, A.P. Siskos, P. Siskos, Indoor Air Pollutants and the Impact on Human Health,
1153 Chemistry, Emission Control, Radioactive Pollution and Indoor Air Quality. (2011).
1154 https://doi.org/10.5772/18806.
1155 [81] F. Ahmed, S. Hossain, S. Hossain, A.N.M. Fakhruddin, A.T.M. Abdullah, M.A.Z. Chowdhury,
1156 S.H. Gan, Impact of household air pollution on human health: source identification and
1157 systematic management approach, SN Appl. Sci. 1 (2019) 418. https://doi.org/10.1007/s42452-
1158 019-0405-8.
1159 [82] M. Akbar, M. Restaino, M. Agah, Chip-scale gas chromatography: From injection through
1160 detection, Microsystems & Nanoengineering. 1 (2015) 15039.
1161 https://doi.org/10.1038/micronano.2015.39.
1162 [83] G. Coelho Rezende, S. Le Calvé, J.J. Brandner, D. Newport, Micro photoionization detectors,
1163 Sensors and Actuators B: Chemical. 287 (2019) 86–94.
1164 https://doi.org/10.1016/j.snb.2019.01.072.
1165 [84] M.P. Van der Schee, N. Fens, P. Brinkman, L.D.J. Bos, M.D. Angelo, T.M.E. Nijsen, R. Raabe,
1166 H.H. Knobel, T.J. Vink, P.J. Sterk, Effect of transportation and storage using sorbent tubes of
1167 exhaled breath samples on diagnostic accuracy of electronic nose analysis, Journal of Breath
1168 Research. 7 (2012) 016002.
1169 [85] I. Horváth, P.J. Barnes, S. Loukides, P.J. Sterk, M. Högman, A.-C. Olin, A. Amann, B. Antus,
1170 E. Baraldi, A. Bikov, A European Respiratory Society technical standard: exhaled biomarkers in
1171 lung disease, European Respiratory Journal. 49 (2017) 1600965.
1172

46