Hybrid GC platform: a micro gas chromatography system with a simple configuration for low-concentration VOC analysis

Abstract

A compact hybrid gas chromatography (GC) platform was developed by integrating a previously reported hybrid μ-GC column chip (hybrid chip) and a commercial photoionization detector. The hybrid chip enabled both gas preconcentration and separation in a single device, allowing for a highly compact and simple platform design with a volume of 0.62 L. With a sample volume of 40.8 mL and an analysis time of 20 minutes, it achieved detection limits of 19.3, 22.8, 30.4, and 24.4 ppb for benzene, toluene, ethylbenzene, and ortho-xylene, respectively. The linear ranges were 0.25–1 ppm for benzene and toluene, 0.25–1.5 ppm for ethylbenzene, and 0.25–2 ppm for ortho-xylene. The peak capacity ranged from 5.34 to 8.81, with full width at half height between 0.22 and 0.5 min. Importantly, the detection limit for benzene was below US workplace air concentration limits set by the American Conference on Governmental Industrial Hygienists (ACGIH) and National Institute for Occupational Safety and Health (NIOSH), demonstrating the platform's potential for indoor air monitoring. Furthermore, portability was enhanced through the integration with a battery and carrier gas filter pack. The platform consumed 2.65 W during analysis (20 minutes), and assuming one cycle consists of 20 minutes of analysis and 10 minutes of stand-by operation, the system could theoretically operate for 70 cycles (35 hours) on a single charge. Field testing with classroom and laboratory air samples confirmed the potential applicability of the platform. In addition, partial qualitative separations were achieved for alkanes, alcohols, aldehydes, and ketones, suggesting broader utility in fields beyond indoor air monitoring.

---

Yeongseok Lee received his Ph.D. degree from the Department of Mechanical Systems at Kookmin University, Republic of Korea, where he now serves as a research professor. His work focuses on developing MEMS-based preconcentration, separation, and detection chip components and integrated gas analysis systems for multiple chemical compounds in the fields of health, safety, and environment.

Sangkyun Lee is a master's student in the Department of Mechanical System Engineering at Kookmin University, Republic of Korea. His research interests include development of multi-functional porous materials related to gas sensing.

Woojin Jang received his Ph.D. degree from the Department of Mechanical Systems at Kookmin University, Republic of Korea. He is focusing on gas analysis methods using artificial intelligence and electronic circuits for gas analysis systems.

Junwoo Lee is a master's student in the Department of Mechanical System Engineering at Kookmin University, Republic of Korea. Currently, he is working on developing functional material based chemiresistive sensors for detecting multiple and low-concentration chemical compounds.

Yuntaek Choi is a doctoral student in the Department of Mechanical Systems at Kookmin University, Republic of Korea. His research field is improvement of micro gas chromatography for effective separation of various chemicals known as environmental pollutants.

Si-Hyung Lim received his B.S. and M.S. degrees in Mechanical Design and Production Engineering from Seoul National University, Republic of Korea, in 1994 and 1996, respectively, and his Ph.D. degree in Mechanical Engineering from the University of California, Berkeley, USA, in 2005. He worked as a postdoctoral researcher in the Center of Integrated Nanomechanical Systems (COINS) and Berkeley Nanoscience and Nanoengineering Institute (BNNI), the University of California, Berkeley, USA, from 2005 to 2007. He is currently a professor and the supervisor of Nanomechatronics Laboratory in the School of Mechanical Engineering, Kookmin University, Republic of Korea. His present research focuses on integrated micro gas analysis systems for environmental and medical applications.

---

Introduction

The analysis of various low-concentration volatile organic compounds (VOCs) is becoming increasingly important in safety-, environment-, and health-related fields.^1–3 From a safety and environmental standpoint, the hazards of VOCs present in indoor/outdoor air, soil, drinking water, and food have been demonstrated for a long time through numerous toxicity and risk assessments.^4–6 Even at low concentrations, VOCs have caused respiratory, neurological, and skin-related diseases in both residents and workers.^7–9 Notably, class 1 carcinogenic VOCs, such as benzene, have been linked to serious health conditions, including cancer and congenital disorder, especially after prolonged exposure.¹⁰ Recently, the medical field has placed greater emphasis on early diagnosis and accurate point-of-care (POC) testing rather than follow-up treatment.^11,12 Alongside this, several studies have suggested that the VOCs found in media such as urine, exhaled breath, and sweat may serve as potential biomarkers for intractable diseases such as cancer, indicating that VOC analysis could be useful for early diagnosis or POC healthcare in the future.^13–15 However, the instruments used for VOC analysis, such as thermal desorption-gas chromatography-mass spectrometry (TD-GC-MS), are expensive (>$100 000), bulky (>1000 L), and difficult to operate and maintain, hindering their use in the abovementioned applications.

To address the abovementioned limitations, various researchers have developed microfabricated GC (μ-GC) systems, which have small volumes for portability and low prices, for the on-site and real-time analysis of VOCs.^16–18 As shown in Table 1, these μ-GC systems generally include a micro gas preconcentrator (μ-PC) to improve sensitivity, a μ-GC column to improve selectivity, and a mini- or micro-detector to sense VOCs.^19–23 These three components are the main chip devices of a μ-GC system. The personal exposure monitoring microsystem (PEMM-2), compact GC platform, and μ-GC with a photoionization detector (μ-GC–PID) listed in Table 1 are representative and relatively recent μ-GC systems. These systems demonstrate how various types of sensors can be integrated with μ-GC technology. Specifically, each system utilizes a different type of detector: a micro-chemiresistive sensor array, a micro-thermal conductivity detector (μ-TCD), and a micro photoionization detector (μ-PID), respectively.^19–21

Table 1 Summary and comparison of recently developed μ-GC systems

Year [ref. no.]	μ-GC system	Size	PC (adsorbent; size)	GC (stationary phase; size)	Detector	Analysis performance	Carrier gas	Power	Other mechanical components
2019 (ref. 19)	PEMM-2	20 × 15 × 9 cm^3 (2.7 L)	μ-PC (Carbopack-B/X; 1.4 × 4.1 cm^2)	μ-GC (polydimethylsiloxane; 7.1 × 2.7 cm^2)	μ-Chemi-resistive sensor array	21 VOCs LOD: 33–600 ppb	Helium canister	Battery	• 3-way valves (2 ea)
									• Needle valves (2 ea)
2020 (ref. 20)	Compact GC platform	12 × 8 × 7.5 cm^3 (0.72 L)	μ-PC (graphitized carbon; not described)	Micropacked GC (Carbograph2; not described)	μ-TCD or mini-PID	8 VOCs LOD: not described	Filtered ambient air or helium canister	Wall	• Fluidic mother board
									• Pilot valves (2 ea)
									• Pump
2021 (ref. 21)	μ-GC–PID	27 × 24 × 10 cm^3 (6.48 L)	Tube-type PC (Carbopack B; length: 4.5 cm)	μ-GC (polydimethylsiloxane; 3.2 × 3.2 cm^2)	μ-PID	6 VOCs LOD: 0.18–0.93 pg	Helium canister	Wall	• 3-way valve
									• Valve
									• Pump
2023 (ref. 22)	MPCA system	23 × 10.5 × 20 cm^3 (4.83 L)	MEMS-fabricated multisensing progressive cellular chip composed of 3 PCs, 3 GC columns, 6 capacitive detectors, and 3 PIDs (4 × 5.6 cm^2) (PC: Carboxen 1003, Carbopack-X, Carbograph2) (GC: polydimethylsiloxane)			14 VOCs LOD: 0.16–315 ppb	Filtered ambient air	Battery	• Valves (3 ea)
									• 3-way valve
									• Pumps (3 ea)
2024 (ref. 23)	iCube	10.8 × 9.4 × 8.5 cm^3 (0.86 L)	Not described	Integrated μ-column and μ-PID (2.6 × 3.0 cm^2) (GC: 5% diphenyl, 95% polydimethylsiloxane)		8 VOCs LOD: not described	Filtered ambient air	Battery	• 3-way valves (2 ea)
									• Pump
									• Check valves (2 ea)
2024 (this work)	Hybrid GC platform	9.4 × 7.4 × 8.9 cm^3 (0.62 L)	Hybrid μ-GC column chip (PC/GC: UiO-66-S; 2 × 2 cm^2)		Mini-PID	8 VOCs LOD: 19.3–30.4 ppb (BTEX)	Filtered ambient air	Battery	• 3-way valve
									• Pump

---

Unlike conventional single and non-selective detectors such as μ-TCD or μ-PID, which provide a single output per analyte based on bulk physical or chemical properties, sensor arrays consist of multiple semi-selective sensing elements that generate distinct response patterns for different compounds. This enables pattern-based recognition, making them particularly valuable for differentiating chemically similar analytes and enhancing qualitative analysis.

The integration of sensor arrays with μ-GC was first proposed by Frye-Mason et al. in 1999 and later developed into a robust analytical platform by the Zellers group.²⁴ Their PEMM-2 system remains one of the most comprehensive examples, combining a μ-chemiresistive sensor array with a GC column to enable analyte identification using both retention time and multichannel response patterns.¹⁹ This configuration allowed resolution of coeluting VOCs through principal component analysis (PCA) and was validated using 21-VOC mixtures and mock field tests, demonstrating its practical utility for portable, selective VOC exposure monitoring.

Sensor array-based systems such as PEMM-2, as well as those employing more conventional detectors like the compact GC platform and the μ-GC–PID, incorporate state-of-the-art technologies in device components along with advances in structural and functional design, enabling improvements over commercial portable GC instruments in terms of volume, functionality, cost, and power consumption, while maintaining or even exceeding comparable analytical performance.^25–27 However, despite these achievements, widespread commercialization in various fields remains limited, largely because of issues related to the overall volume and cost, which make these systems less attractive to potential users.

Several strategies have been proposed to reduce the overall volume and cost of portable μ-GC systems. These include minimizing power consumption to reduce battery size, selecting compact carrier gas solutions, and simplifying system architecture through component integration. However, such optimizations must be approached with caution. Factors such as power requirements, battery capacity, and the use of carrier gas canisters affect not only the volume, weight, and cost of the system, but also its analytical performance. These components interact in complex ways, and trade-offs between volume reduction and system performance must be carefully evaluated to ensure reliable and accurate operation. For instance, while helium offers superior chromatographic performance, its use necessitates pressurized canisters and regulators, which add bulk and may incur higher maintenance and replacement costs, particularly in long-term or field applications.

Among the possible strategies, consolidating key components while maintaining adequate performance is considered an effective approach. This integration can reduce fabrication costs by limiting the number of microfabricated components and can help minimize system volume by reducing interconnections between discrete components.

The importance of minimizing interconnections is exemplified by the compact GC platform developed in 2020.²⁰ Although this platform did not incorporate device-level integration of components such as the μ-PC, μ-GC, and detector, it achieved a compact volume of 0.72 L by employing a micromachined fluidic motherboard that consolidated valves and channels to connect components without separate fittings. This approach demonstrated that reducing interconnections alone can contribute meaningfully to system miniaturization. In this sense, it indirectly highlights the potential benefits of device integration, where components like the μ-PC, μ-GC, and detector are integrated into a single chip, thereby eliminating even the internal fluidic connections between them.

The multisensing progressive cellular architecture (MPCA) system reported in 2023 featured a chip that integrated three μ-PCs, three μ-GC columns, and nine detectors, forming a highly parallelized and fully valveless configuration.²² Each of the three cells incorporated a combination of two capacitive sensors and a μ-PID, providing diverse detection modalities based on different physicochemical sensing mechanisms. This design enabled the simultaneous analysis of multiple VOC streams and supported broader chemical coverage. While the system's multisensing capability suggests the potential for analyte identification through both retention time and differential response patterns, a fully developed analytical framework based on pattern recognition has not yet been implemented. Nonetheless, the MPCA system demonstrated strong performance with respect to the number of analyzable VOC species and its limit of detection (LOD), highlighting its potential as a next-generation μ-GC system. However, the integration of 15 major components constrained the degree of volume reduction that could be achieved.

In contrast, the iCube, which was developed in 2024, had an ultracompact system size (0.86 L) due to its use of an iCPID chip, which integrates a μ-column with a μ-PID.²³ While this design-based integration strategy can decrease the system volume, it fundamentally relies on adding additional components to existing ones. As a result, such integration requires extensive and complex microfabrication processes to fabricate multiple functional components within a single chip, which limits the potential for cost reduction.

To overcome these limitations, a hybrid μ-GC column chip (hybrid chip) in which both μ-PC and μ-GC roles were performed by a single component was developed in our previous study.^28,29 The multifunctionality of the hybrid chip was achieved by utilizing metal–organic frameworks (MOFs) with gas adsorption and desorption properties, which served as both the adsorbent and the stationary phase.^30–32 This integration approach is categorized as material-based integration, as it exploits the multifunctionality of a single material within the same physical space of a single chip to achieve multiple functions. In contrast, design-based integration refers to the co-fabrication of multiple distinct micro-components through a series of intricate microfabrication steps, where each component is responsible for a separate function and is located in a different region of the chip (Fig. S1†). Therefore, material-based integration can offer significant advantages in terms of manufacturing simplicity and cost-effectiveness, as it eliminates the need for additional microfabrication processes required for functional integration (Table S1†).

In this paper, we report a hybrid GC platform containing a hybrid chip capable of analyzing representative VOCs, including benzene, toluene, ethylbenzene, and ortho-xylene (BTEX), through retention time-based qualitative analysis and peak area-based quantitative analysis. The advantages of the material-based integration of this hybrid chip include requiring fewer additional components, a simplified configuration, and a decreased system volume. The LOD, quantitative analysis range, and repeatability of the developed platform were verified using standard BTEX mixtures. In addition, the practical utility of the platform was demonstrated by comparing the air quality analysis results with those obtained using a commercial TD-GC-MS system. Furthermore, the platform's capability was demonstrated through retention time-based separations of various homologous series, including C₅–C₉ alkanes, C₂–C₄ alcohols, C₂–C₆ aldehydes, and C₃–C₅ ketones.

Experimental

Materials

Zirconium chloride, terephthalic acid, acetic acid, N,N-dimethylformamide, and methanol, which were used for the synthesis and coating of UiO-66-S, were purchased from Sigma Aldrich (St. Louis, USA). A mixture of benzene, toluene, ethylbenzene, and o-xylene, each at a concentration of 1 ppm, was purchased from Rigas (Daejeon, Republic of Korea). In addition, 10 ppm gaseous butane, a 20 ppm BTEX mixture, 10 ppm styrene, and a mixture of 10 ppm xylene isomers were obtained from DK gas (Hwaseong, Republic of Korea). C₅–C₁₀ alkanes, C₂–C₅ alcohols, C₂–C₇ aldehydes, C₃–C₆ ketones, and molecular sieve 5A for a carrier gas filter pack were obtained from Sigma Aldrich (St. Louis, USA). All components of the hybrid GC platform are listed in Table S2.†

Preparation of a hybrid μ-GC column chip

The hybrid chip was fabricated by using a previously reported microfabrication process, with minor modifications to the inlet and outlet regions to facilitate modular integration.^28,33Fig. 1a shows a schematic diagram of the sequential fabrication of the hybrid chip. A serpentine-type column channel with a width of 150 μm and a depth of 400 μm was fabricated on the front surface of the 6-inch silicon substrate via photoresist (PR) patterning and deep reactive ion etching. On the back side of the silicon substrate, micro-thin film heaters and a resistive temperature detector (RTD) were fabricated through PR patterning, Ti/Pt sputtering (Ti: 200 Å, Pt: 4000 Å), lift-off, plasma-enhanced chemical vapor deposition, and oxide dry etching. Subsequently, the inlet and outlet of the hybrid chip were formed on the glass via dry film resist patterning and sandblasting. Finally, the chip was completed by anodic bonding between the open front side and the glass containing the inlet and outlet.

Fig. 1 Preparation of the hybrid μ-GC column chip. (a) Diagram of the micromachining procedure, (b) front side of the hybrid chip, (c) back side of the hybrid chip, (d) photographs of the hybrid chip before and after coating with UiO-66-S, and (e) SEM images of a cross-section of the hybrid chip.

Fig. 1b and c show the front and back sides of the hybrid chip, respectively. The width and length of the hybrid chip were both 2 cm, and the total length of the column channel was approximately 1.5 m. The resistances of the two micro heaters and the RTD deposited on the backside were 28 Ω and 161 Ω, respectively, at room temperature. UiO-66, a well-known material for the preconcentration and separation of BTEX, was selected as the hybrid material in this study.^29,34–36 In our previous work, we synthesized UiO-66 with controlled particle sizes and observed that the preconcentration performance was inversely related to the particle size. Based on this result, we designated the smaller-particle-size variant as UiO-66-Small (UiO-66-S), which was employed in the present study. UiO-66-S was synthesized via a solvothermal method and deposited inside the channel through dynamic coating, following our previously reported protocol. All conditions for the synthesis and coating were identical to those described in our previous work.²⁹

Fig. 1d presents photographs before and after coating, showing that the front side of the chip became slightly white after UiO-66-S coating. Fig. 1e shows scanning electron microscopy images of a cross-section of the chip, and the enlarged images revealed that UiO-66-S particles were deposited on the wall surface of the column channel.

Assembly and setup of the hybrid GC platform

The hybrid GC platform containing a hybrid chip and a PID as the main components was assembled. Additional components, such as a three-way solenoid valve, a micro pump, a microprocessor, a carrier gas filter pack, and batteries, were also included. In addition, all interconnections were implemented without any adhesives, as shown in Fig. S2.† A key feature of the assembled hybrid GC platform was its usefulness for analyzing low concentrations of BTEX, as it employed fewer mechanical components than other developed μ-GC systems (Table 1).

The battery and the carrier gas filter pack are crucial supplementary components that enhance the platform's portability. In particular, the battery plays a key role in minimizing dependence on external infrastructure by enabling temperature ramping of the hybrid chip, which is the most power-intensive component in the system. Other low-power modules, including the control or operation electronics and sensor interface, operate using a 5 V supply provided via the microprocessor connected to any portable computing devices.

The carrier gas filter pack, which replaced high-purity helium, hydrogen, or nitrogen cylinders or canisters, allowed complete portability of the μ-GC system. The carrier gas filter pack was fabricated from aluminium and contained a winding internal gas passage filled with a 5A molecular sieve. As ambient air flows through this passage, the interfering gases and moisture are removed, ultimately yielding purified air for use as the carrier gas.

The performance of the filter pack is shown in Fig. S3.† The filter pack was connected to a commercial flame ionization detector (FID, ChroZen GC–FID, Young-In Chromass, Republic of Korea), and 0.25 ppm and 20 ppm BTEX samples were continuously injected through a mass flow controller; 0.25 ppm could pass through the system without any BTEX detection for more than 300 min, whereas saturation was reached in 73 min for 20 ppm samples (Fig. S3a†). In addition, the filter pack was attached to the inlet of the hybrid chip, preconcentrated with 1 ppm BTEX for 10 min, and subsequently mounted on the GC–FID to verify the preconcentration effect. The results confirmed that no BTEX vapors were adsorbed onto the hybrid chip, as BTEX removal was achieved by the filter pack (Fig. S3b†). Finally, Fig. S3c† shows that the relative humidity of the ambient air, measured using a temperature and humidity sensor module (DHT11, ASAIR, China), was decreased from 45–46% to 7–9% after passing through the filter pack for 4.85 min. Based on the breakthrough time observed at 20 ppm, and assuming 10-minute carrier gas sampling cycles at 0.25 ppm, the filter pack is recommended to be regenerated after approximately 80–90 cycles. However, due to the co-adsorption of water vapor during each sampling cycle, the effective adsorption capacity may decrease under humid conditions, potentially shortening the usable cycle life in real-world applications. Regeneration was achieved by removing the 5A molecular sieve from the filter pack and placing it in a vacuum oven at 200 °C for approximately 12 hours.

Fig. 2a presents a photograph of the open platform, highlighting the internal configuration, including the carrier gas filter pack, battery, PID module, and hybrid chip module. Fig. 2b shows the fully assembled hybrid GC platform, with dimensions of 9.4 cm in width, 7.4 cm in length, and 8.9 cm in height, corresponding to a total volume of 0.62 L. This compact unit is thus smaller than other reported μ-GC systems. Owing to its small volume and lightweight design (0.5 kg), the platform is suitable for deployment in confined indoor spaces and shows strong potential for wearable applications, such as belt-mounted operation, especially with minor modifications like the addition of Velcro attachments for flexible positioning depending on the user's environment.

Fig. 2 Hybrid GC platform. (a) A photograph showing the internal configuration of the hybrid GC platform, (b) a photograph showing the outside and dimensions of the hybrid GC platform, and (c) a schematic diagram of the analytical process supported by the hybrid GC platform.

Fig. 2c shows a schematic diagram illustrating the analysis process supported by this hybrid GC platform. In most platform configurations, diaphragm pumps are typically placed at the end of the flow path to minimize the risk of contamination. However, in our system, the pump had to be positioned midstream to ensure a stable and adequate flow rate to the hybrid chip. Therefore, additional experiments were conducted to evaluate whether the pump introduced contamination. 1 ppm BTEX was introduced through the pump to the PID, and the results were compared with those obtained using only the mass flow controller (MFC). The experiment aimed to evaluate whether internal components of the pump, such as O-rings or the diaphragm, caused contamination. If contamination was present, BTEX adsorption could lead to lower PID signals compared to the MFC-only condition. Conversely, when pure nitrogen was introduced, the release of previously adsorbed substances could result in higher PID signals than in the MFC-only condition. Despite initial concerns, no significant contamination attributable to the pump was observed under the tested conditions, as shown in Fig. S4.† Nevertheless, to minimize any risk of contamination during extended operation, it is generally preferable to position the pump at the terminal end of the flow path.

The platform was ultimately configured to operate in three distinct modes: stand-by, analysis, and cleaning (Table 2). The stand-by mode is automatically activated upon powering the system. In this mode, filtered air is supplied through the tubing, hybrid chip, and detector, while the hybrid chip is maintained at 40 °C to ensure thermal readiness and to stabilize the PID detector. The analysis mode consists of a total duration of 20 minutes, with the first 10 minutes dedicated to sample gas preconcentration, followed by a 10-minute temperature ramping of the hybrid chip for separation and detection. During the preconcentration step, a three-way valve is connected to the sample gas, whereas during the separation and detection step, it switches to the carrier gas filter pack. Lastly, the cleaning mode replicates the separation and detection sequence of the analysis mode for 10 minutes. This mode is intended to remove any residual gases and adsorbed water vapor remaining in the hybrid chip, particularly after periods of system inactivity. During the initial 4.85 minutes following system restart, water vapor present in the internal flow path may not be fully removed and can be adsorbed by UiO-66-S within the hybrid chip (Fig. S3†). Activating the cleaning mode helps desorb unwanted species, including water vapor and residual gases, and restore the chip's baseline condition. In practice, the cleaning mode was operated whenever the system had been turned off even briefly or when PID signal values were unstable during stand-by mode.

Table 2 Mode, operation, and power consumption of the hybrid GC platform

Mode	Operation		Component					Power consumption
			3-way valve	Pump	Hybrid chip	PID	Processor
Mode 0	Stand-by		Off (3 → 2)	On	On (40 °C)	On	On	Mode 0: 0.89 W
			—	0.3 W	0.32 W	0.17 W	0.1 W
Mode 1	Analysis	Preconcentration step	On (1 → 2)	On	On (40 °C)	On	On	1.59 W (10 min)	Mode 1: 2.65 W (20 min)
			0.7 W	0.3 W	0.32 W	0.17 W	0.1 W
		Separation & detection step	Off (3 → 2)	On	On (40–150 °C)	On	On	3.71 W (10 min)
			—	0.3 W	3.14 W	0.17 W	0.1 W
Mode 2	Cleaning		Off (3 → 2)	On	On (40–150 °C)	On	On	Mode 2: 3.71 W (10 min)
			—	0.3 W	3.14 W	0.17 W	0.1 W

---

During the most important mode (analysis), the platform consumed 2.65 W over a 20-minute period, including a 10-minute preconcentration step (1.59 W) followed by a 10-minute separation/detection step (3.71 W). Excluding the power supplied from a portable computing device to the microprocessor, the batteries (44.4 Wh) were used solely for the temperature ramping of the hybrid chip. In analysis mode, the battery supplies 1.73 W over a 20 minute cycle (0.58 Wh). Under ideal conditions, this corresponds to approximately 76 analysis cycles, equivalent to 25 hours of continuous operation. However, due to the absence of an active cooling system, a 10 minute stand-by period is required between analysis cycles. When a single cycle is defined as 20 minutes of analysis followed by 10 minutes of stand-by (with an average consumption of 1.26 W over 30 minutes, 0.63 Wh), the platform can theoretically operate for up to 70 cycles, corresponding to 35 hours of operation.

During the preconcentration step, the hybrid chip was maintained at a consistent temperature of 40 °C. Because the platform lacked a dedicated flow control unit, the flow rate was altered by temperature changes in the hybrid chip. Consequently, the hybrid GC platform was designed to maintain a preconcentration temperature of 40 °C, accounting for practical application conditions, including heat transfer from other components and warm room environments.

The maintenance and ramping of the temperature during the preconcentration step and the separation/detection step were achieved through proportional–integral–derivative (PiD) feedback control of the micro heater on the basis of the resistance values measured by the RTD, which corresponded to the temperature of the hybrid chip. Fig. 3a shows that the flow rate remained at 4.08 mL min⁻¹ during the preconcentration step at 40 °C and then decreased to 2.40 mL min⁻¹ as the temperature of the hybrid chip increased during the subsequent separation/detection step. The flow rate was measured using a mass flow meter (TSM-D220, MK Precision, Republic of Korea). This reduction in flow rate is attributed to an increase in pressure drops across the microchannel caused by the elevated gas viscosity at higher temperatures, which exceeded the pressure delivery capability of the diaphragm pump. In addition, because the PID is highly sensitive to flow rate fluctuations, the pulsation inherent to diaphragm pumps can generate significant noise signals, potentially degrading the separation resolution. However, as shown in Fig. S5,† the hybrid chip acted as a damper that mitigated the pulsation effect, thereby reducing noise and improving signal stability.

Fig. 3 Temperature control of the hybrid chip using PiD feedback control. (a) Changes in the flow rate as a result of temperature ramping of the hybrid GC chip and (b) robustness test results obtained at various temperatures (0–40 °C).

Additionally, PiD feedback control effectively maintained a constant temperature even under various temperature conditions, as demonstrated using a temperature and humidity chamber (TH-KE-100, JEIO TECH, Republic of Korea). Specifically, the time required to achieve a constant chip temperature of 40 °C when the ambient temperature was 0 °C was 1.38 min (Fig. 3b). During subsequent ramping steps, the temperature was successfully increased under different conditions with no significant variations, confirming the robustness of the platform under varying temperature environments.

As a result, during the preconcentration step, the platform preconcentrated the sample gas at 40 °C for 10 min at a flow rate of 4.08 mL min⁻¹ (sampling volume: 40.8 mL, sampling flow rate: 4.08 mL min⁻¹, sampling time: 10 minutes). In the subsequent separation/detection step, the flow rate decreased to 2.40 mL min⁻¹ as the hybrid chip temperature was ramped from 40 °C to 150 °C in 10 min at a ramping rate of 11 °C min⁻¹. All time and temperature parameters could be adjusted as needed. In addition, most of the samples were collected using Tedlar bags, which were purged five to seven times with nitrogen before use. All analyses were completed within three hours of sample collection, and residual gases were not reused. Lastly, Tedlar bags were additionally cleaned if necessary and were not reused beyond two days.

Results and discussion

Standard BTEX analysis performance

The ability of the hybrid chip to perform both preconcentration and separation was attributed to the presence of UiO-66-S, which facilitated the adsorptive separation of the BTEX vapors.^34–39 UiO-66-S possesses a three-dimensional porous structure composed of zirconium-based metal clusters and terephthalic acid organic linkers, forming cages and windows with sub-nanometer dimensions. These structural features enable the diffusion and adsorption of BTEX molecules via host–guest interactions, including van der Waals forces and π–π stacking between the benzene rings of BTEX and the aromatic linkers in UiO-66-S.⁴⁰ While molecular size (kinetic diameter) contributes to separation selectivity, it should be noted that the retention order of BTEX species in this study generally follows their relative volatilities.^41–43 Specifically, their vapor pressures at 25 °C are as follows: benzene (12.7 kPa), toluene (3.8 kPa), ethylbenzene (1.3 kPa), and ortho-xylene (0.9 kPa), indicating that volatility likely plays a key role in determining elution order. This volatility-driven separation mechanism accounts for most of the observed trends, except for isomeric xylenes, which often require more selective stationary phases due to their similar physicochemical properties.

Fig. 4a shows a chromatogram obtained by preconcentrating 0.25 ppm, 0.5 ppm, and 1 ppm BTEX vapors for 10 min, followed by separation and detection for 10 min (mode 1: analysis). As previously mentioned, detection occurred in the order of decreasing vapor pressure, with variations in peak height and width corresponding to different concentrations. Additional experiments were conducted using 1.5 ppm, 2 ppm, and 3 ppm BTEX vapors to determine the linear range of the quantitative analysis performed with the platform (Fig. S6†). The peak widths at 10% peak height varied from 0.5 to 1.19 minutes, corresponding to relatively low peak capacities of 5.34 to 8.81 (Table S3†). Notably, in the concentration range of 0.25–1 ppm, the peak capacity tended to increase from 7.87 to 8.81, whereas at higher concentrations (1.5 to 3 ppm), a sharp decrease in peak capacity was observed, dropping from 5.34 to 6.30.

Fig. 4 BTEX analysis results obtained by using the hybrid GC platform. (a) Chromatograms of 0.25, 0.5, and 1 ppm BTEX, (b) peak areas for various BTEX concentrations, (c) chromatograms of 5 and 10 ppb BTEX, and (d) RSDs for the peak areas of BTEX obtained during 30 repeatability tests conducted over two consecutive days.

In gas chromatography, quantitative analysis of a specific peak is typically performed by utilizing the correlation between the analyte concentration and peak area. Fig. 4b presents the peak area as a function of concentration within the range of 0.25–3 ppm. The results for benzene and toluene exhibited a linear range from 0.25 ppm to 1 ppm, whereas those for ethylbenzene and ortho-xylene demonstrated different linear ranges, spanning from 0.25 ppm to 1.5 ppm and from 0.25 ppm to 2 ppm, respectively. The linear ranges for ethylbenzene and ortho-xylene were wider because these vapors gave rise to a lower signal intensity (height of the peak) than did benzene and toluene at the same concentration, so the results of these two analytes reached the voltage output limit (official voltage limit: 2.85 V and measured peak voltage: 2.93 V) of the PID more slowly as the concentration increased.

Three factors explained the low signal intensities of ethylbenzene and ortho-xylene. First, the flow rate of the carrier gas decreased as the temperature of the hybrid chip increased during the separation/detection step, resulting in peak broadening and a low intensity. Second, the temperature increase of the hybrid chip, which was controlled via PiD feedback, did not exhibit a first-order profile but instead exhibited a rightward–downward curve. Third, the analytes and carrier gas may have reached the PID detector while still heated, as they were discharged directly from the heated hybrid chip. Since the sensitivity of PID detectors is known to decrease at higher temperatures, this thermal influence may have contributed to the relatively low signal intensities.

As shown by the red line in Fig. 3a, the rate of temperature increase gradually decreased over time. Consequently, the TD of ethylbenzene and ortho-xylene adsorbed on the hybrid chip could not be promoted as quickly as that of benzene and toluene, which eventually led to peak broadening and a decrease in the peak height for ethylbenzene and ortho-xylene. As shown in Fig. S6a,† benzene and toluene gave rise to sharp voltage signals because of the high flow rate and rapid desorption rate. The detection voltage for these compounds at a concentration of 1 ppm nearly reached the limit output of the PID. When the concentration exceeded the output voltage limit, the peaks exhibited blunted vertices, and the peak area eventually lost its linear relationship with concentration (Fig. S6b–d† and 4b). Unlike benzene and toluene, ethylbenzene and ortho-xylene gave rise to less intense PID detection signals, reaching their limits at concentrations of 1.5 ppm and 2 ppm, respectively. As a result, the linear range for the quantitative analysis performed with the platform varied for each analyte and was largely influenced by the detector specifications.

The LOD was calculated from the standard deviation (SD) and slope, as suggested by the United States Food and Drug Administration.⁴⁴ More specifically, a regression curve of peak area versus concentration was generated using peak areas obtained at various concentrations. The LOD was calculated from the slopes of these curves and the SDs of the peak areas determined from repeated experiments at a specific concentration. The equation, the LODs for each BTEX, and the corresponding experimental details are provided in Table S4.† To increase the reliability of the LODs, standard deviations derived from a large number of samples were required. Therefore, 30 replicate experiments with 0.25 ppm BTEX were conducted under identical conditions. The calculated LODs for BTEX were 19.3 ppb, 22.8 ppb, 30.4 ppb, and 24.4 ppb, respectively (Table S4†). Specifically, for benzene, the ACGIH threshold limit value (TLV) is 0.02 ppm and the NIOSH recommended exposure limit (REL) is 0.1 ppm.^45,46 When compared with these regulatory guidelines, the platform's LOD for benzene meets the required sensitivity for detection. However, the limit of quantification (LOQ), generally assumed to be three times the LOD, does not fully meet the regulatory thresholds. This indicates that while reliable detection may be limited under current conditions, the platform still shows potential for detection and warning of benzene exposure.

Visual evaluation of the chromatograms for 10 ppb and 5 ppb BTEX (Fig. 4c) confirmed that BTEX vapors were detected at both concentrations, with LODs that differed from the abovementioned results. Another authoritative approach for LOD determination is the signal-to-noise ratio (SNR) method. LOD values determined using this method are provided in Table S5,† along with relevant details.⁴⁴ The noise was found to be 0.02 V during the preconcentration step and 0.02 V during the separation/detection step via a blank experiment (Fig. S7†). For 5 ppb BTEX, the SNR values ranged from 0.8 to 2.55, which did not satisfy the acceptable LOD criterion of SNR > 3. However, for 10 ppb BTEX, the SNR values were 3.7, 6.05, 2.15, and 1.75, respectively, indicating that 10 ppb benzene and toluene were detectable using the hybrid GC platform.

Repeatability, a crucial factor for evaluating the performance of an analytical instrument, was assessed in terms of the relative standard deviation (RSD), which was calculated from 30 replicate experiments with 0.25 ppm BTEX conducted over two consecutive days (15 measurements per day) to determine the abovementioned SDs. The RSD values for the BTEX vapors were calculated to be 2.07%, 2.07%, 3.49%, and 3.61%, respectively (Fig. 4d). According to NIOSH Method 3900, which outlines analytical procedures for VOCs in workplace air, an RSD of up to 5% is generally considered acceptable for chromatographic methods, indicating that the repeatability of the platform is within a satisfactory range.⁴⁷ Furthermore, given that RSD values of 5% or less are typically acceptable for evaluating performance parameters such as the LOD and LOQ, the LODs calculated for this platform were considered reasonably accurate.⁴⁸

Application test

The analysis of low-concentration VOCs is becoming increasingly important in various fields. BTEX have been particularly prominent focuses in research, as they not only pose health risks as harmful gases that can cause diseases in humans but also serve as potential biomarkers for certain medical conditions.^49–51 In this study, to evaluate the practical applicability of the developed platform, air quality analyses with real samples were performed using the hybrid GC platform with a primary focus on detecting BTEX in air samples. Because the presence of BTEX in real air samples is uncertain, air samples were collected in Tedlar bags under the same time and space conditions from both a general classroom and a laboratory with significant chemical usage, and it was hypothesized that BTEX were more likely to be present in the latter environment.

To evaluate the reliability of the analysis results, samples collected under the same conditions were also analyzed using TD-GC-MS, and the results were compared with those obtained from the hybrid GC platform. The classroom air samples were collected in a nearby classroom and transferred to the laboratory in the same building for analysis within 10 minutes, whereas the laboratory air samples were collected and analyzed immediately at the same laboratory location. Additionally, samples collected at the same time and under identical conditions were sent to the Korea Polymer Testing & Research Institute (KOPTRI) for TD-GC-MS analysis and analyzed within one day of collection. For these indoor air samples, because VOC concentrations are typically very low, the preconcentration time was extended to 60 min, which is six times the standard setting, while the separation/detection step settings remained unchanged. The setup for the TD-GC-MS analysis is described in Table S6.†

Fig. 5a and b show chromatograms of these air samples analyzed via TD-GC-MS, and Fig. 5c and d show chromatograms of the same samples analyzed via the hybrid GC platform. As shown in Fig. 5a and b, aromatic hydrocarbons, including benzene, toluene, ethylbenzene, para-xylene (X_P), styrene, and ortho-xylene (X_O), were detected in the classroom and laboratory air samples (all peaks from the blank experiment were excluded from the discussion). These VOCs can originate from various indoor sources, including computers, screens, desks, chairs, wall paint, flooring materials, and even outdoor air. In the laboratory air sample, hexamethyldisiloxane (HMDS) and trichloroethylene (TCE) were also detected (Fig. 5b). The HMDS likely originated from volatilization caused by exposure to ultraviolet light or heat during surface-related research, while TCE was thought to originate from the use of a gas cylinder containing standard 10 ppm TCE for another study.

Fig. 5 Analysis results for indoor air samples collected in a classroom and a laboratory. (a and b) Chromatograms obtained via TD-GC-MS and (c and d) chromatograms obtained via the hybrid GC platform.

The analysis of single samples using TD-GC-MS did not allow quantitative calibration, but examining the differences in the peak intensities or areas allowed simple quantitative comparisons of the VOCs detected in the classroom and laboratory air samples. Notably, with the exception of toluene and X_O, the common VOCs detected in the classroom sample showed larger peak areas, suggesting higher concentrations. Although the cause of these high concentrations cannot be definitively determined without detailed follow-up investigations, they were likely caused by the powerful ventilation systems installed in the laboratory.

Since only X_O was included in the standard BTEX analysis described in the previous section, additional experiments involving styrene and different xylene isomers were performed to assess the feasibility of qualitative analysis based on retention time using the developed platform. Fig. S8† presents chromatograms for xylene isomers (Fig. S8a†), styrene (Fig. S8b†), and a mixed sample containing benzene, toluene, ethylbenzene, xylene isomers, and styrene (Fig. S8c†). Although complete baseline separation was not achieved for all analytes in the mixed sample, the differences in retention times were sufficient to distinguish individual compounds, thereby demonstrating the feasibility of qualitative analysis using the developed platform.

The separation of xylene isomers remains a challenge in chromatography, especially for differentiating between meta-xylene (X_M) and X_P. The boiling points of these isomers are very close, with X_P at 138 °C, X_M at 139 °C, and X_O at 144 °C. Because of this similarity, separation between X_M and X_P is generally difficult on conventional GC stationary phases. However, in this study, UiO-66-S was confirmed to provide sufficient retention force to achieve such separation. This outcome is attributed to the combined effects of vapor pressure differences and geometric effects within UiO-66-S. The vapor pressures of the xylene isomers at 25 °C are 1.17 kPa for X_P and approximately 0.88 kPa for both X_M and X_O. In addition, a geometric interaction mechanism contributed significantly, wherein stronger van der Waals forces arise as bulkier molecules interact more closely with the pore walls. The structural features of UiO-66-S, including the diameters of the octahedral and tetrahedral cages and the window (1.1 nm, 0.8 nm, and 0.5–0.7 nm, respectively), closely match the kinetic diameters of the vapor molecules, with X_P at 0.67 nm, X_M at 0.71 nm, and X_O at 0.74 nm. This allows a larger difference in retention time between xylene isomers.^52,53 As previously mentioned, retention time increases as vapor pressure decreases. However, styrene (7.32 min) eluted earlier than X_O (7.67 min) despite having a slightly lower vapor pressure (0.7 vs. 0.88 kPa). This behavior is attributed to styrene's smaller kinetic diameter (0.61 nm vs. 0.74 nm), which likely facilitated faster diffusion through the pores of UiO-66-S due to reduced steric hindrance. Similar size-dependent transport behavior has been reported in zeolite systems, where molecules with smaller kinetic diameters exhibited faster migration despite lower volatility.⁵⁴

The results in Fig. 5c and d, which were obtained via the developed platform, exhibited two notable trends when compared with those obtained using TD-GC-MS. First, in terms of qualitative analysis, most of the gas peaks identified in the TD-GC-MS chromatograms were also observed in the chromatograms obtained via the hybrid GC platform. For the classroom air sample, the same major VOC peaks were detected by both systems. In the laboratory air sample, all peaks detected by TD-GC-MS were also captured by the platform, with the exception of the styrene peak. Second, a similar trend in the relative concentration differences of VOCs was observed across the two systems (Fig. S9†). For example, the peak areas for toluene and X_O were smaller in the classroom air sample than in the laboratory air sample in both the TD-GC-MS and platform analyses. In contrast, the peak areas for other VOCs such as benzene, ethylbenzene, and X_P were larger in the classroom air sample, and this trend was also consistently observed in the results obtained from both analytical systems. Although compounds such as HMDS and TCE were detected in the TD-GC-MS results and could also be detected by the PID used in the hybrid GC platform, co-elution may still occur even for VOCs with defined retention times. Nevertheless, the results suggest that the platform has potential for both qualitative and quantitative analysis of BTEX and other target VOCs in indoor air environments.

However, to verify the applicability of the platform, including indoor air quality analysis, and to explore ways to expand its applicability, it is essential to discuss the analytes that the hybrid GC platform can detect from multiple perspectives in addition to quantitative metrics such as the LOD and linear range. The analytes that can be analyzed on this platform can be controlled or limited in three ways: the temperature during the preconcentration step, the maximum TD temperature, and the ionization energy of the PID. To determine the analyzable range of the platform, experiments and analyses were conducted using various classes of compounds, including alkanes, alcohols, aldehydes, and ketones. With the exception of butane (C₄ alkane), all the tested analytes were vaporized from liquid samples via bubbling, collected in a Tedlar bag, and subsequently diluted with pure nitrogen to decrease their concentrations prior to the experiments. In addition, analytes with ionization energies exceeding the detection capability of the PID (10.6 eV), such as C₁–C₃ alkanes, C₁ alcohol, and C₁ aldehyde, were excluded from the experiments.

Fig. 6a–d present chromatograms of the alkanes, alcohols, aldehydes, and ketones, respectively, analyzed on the platform after a preconcentration time of 3 min. Fig. 6a shows that butane underwent TD during the preconcentration step at 40 °C, so it was detected prior to the separation/detection step. In contrast, C₅–C₉ alkanes were at least partially separated, although baseline separation was not achieved. In a similar way, experiments on the other compounds were conducted, and the results demonstrated that analysis was feasible for C₂–C₄ alcohols, C₂–C₆ aldehydes, and C₃–C₅ ketones (Fig. 6b–d). Although the vapor pressure differences among alkane mixtures are larger than those of alcohols, aldehydes, and ketones, which could favor separation, alkanes are nonpolar molecules that lack functional groups capable of interacting with UiO-66-S. As a result, meaningful separation between them was likely limited.

Fig. 6 Experimental results of the hybrid GC platform for the analysis of alkanes, alcohols, aldehydes, and ketones. (a) Chromatograms of C₄–C₁₀ alkanes, (b) chromatograms of C₂–C₅ alcohols, (c) chromatograms of C₂–C₇ aldehydes, and (d) chromatograms of C₃–C₆ ketones.

In addition, C₁₀ alkane, C₅ alcohol, C₇ aldehyde, and C₆ ketone, which had potential to be preconcentrated and detected by the PID, were not observed in the chromatogram (Fig. 6). This was due to the insufficient TD temperature of 150 °C, and additional experiments confirmed that the analytes preconcentrated in the hybrid chip were thermally desorbed when ramped up to 200 °C (Fig. S10†). The peak widths and peak capacities determined from these experiments are summarized in Table S3,† and the analytical range of the platform is presented in Table 3. Based on these retention time data, the developed hybrid GC platform successfully resolved eight VOCs from different chemical classes, including alcohol, alkane, ketone, and aromatics, with distinguishable retention times even for structurally similar compounds such as ethylbenzene, X_M, and X_O (Fig. 7). This demonstrates the platform's capability for qualitative discrimination of VOCs with different functional groups and molecular structures.

Table 3 Analyzable and non-analyzable analytes detected via the hybrid GC platform

Analyte		Formula	Vapor pressure (kPa at 25 °C)	Boiling point (°C)	Preconcentration^a (40 °C)	Separation^b (40–150 °C)	Detection^c (ionization energy: 10.6 eV)	Analysis	Retention time (min)
a Temperature of the hybrid chip during the preconcentration step. b Temperature range of the hybrid chip during the separation/detection step. c Ionization energies for both the PID sensor (10.6 eV) and the VOCs. If the VOC's ionization energy exceeds that of the PID sensor, the VOC cannot be detected.
Aromatics	Benzene	C6H6	12.7	80	Yes	Yes	Yes (9.24)	Analyzable (VP: 0.9–12.7 kPa, BP: 80–144 °C)	3.12
	Toluene	C7H8	3.8	111			Yes (8.82)		4.88
	Styrene	C8H8	0.7	145			Yes (8.4)		7.32
	Ethylbenzene	C8H10	1.3	136			Yes (8.76)		6.20
	p-Xylene	C8H10	1.2	137			Yes (8.44)		6.40
	m-Xylene	C8H10	0.9	139			Yes (8.56)		7.02
	o-Xylene	C8H10	0.9	144			Yes (8.56)		7.67
Alkanes	Methane	CH4	4550	−162	Not tested	Not tested	No (12.96)	Not analyzable
	Ethane	C2H6	3240	−89			No (11.56)
	Propane	C3H8	940	−42			No (11.07)
	Butane	C4H10	240	−1	No	No	Yes (10.63)
	Pentane	C5H12	53.3	36	Yes	Yes	Yes (10.35	Analyzable (VP: 0.42–53.3 kPa, BP: 36–151 °C)	2.63
	Hexane	C6H14	17.6	69			Yes (10.13		3.37
	Heptane	C7H16	6.0	98			Yes (9.92)		3.70
	Octane	C8H18	1.5	126			Yes (9.8)		4.45
	Nonane	C9H20	0.42	151			Yes (9.72)		6.05
	Decane	C10H22	0.13	174	Yes	No	Yes (9.65)	Not analyzable
Alcohols	Methanol	CH3OH	16.9	65	Not tested	Not tested	No (10.85)	Not analyzable
	Ethanol	C2H5OH	7.9	78	Yes	Yes	Yes (10.43)	Analyzable (VP: 1.2–7.9 kPa, BP: 78–118 °C)	2.17
	Propanol	C3H7OH	2.9	97			Yes (10.17)		4.83
	Butanol	C4H9OH	1.2	118			Yes (10.04)		8.53
	Pentanol	C5H11OH	0.5	138	Yes	No	Yes (9.78)	Not analyzable
Aldehydes	Methanal	HCHO	518	−19	Not tested	Not tested	No (10.87)	Not analyzable
	Ethanal	CH3CHO	101	20	Yes	Yes	Yes (10.23)	Analyzable (VP: 1.1–101 kPa, BP: 20–129 °C)	3.08
	Propanal	C2H5CHO	27	49			Yes (9.95)		4.02
	Butanal	C3H7CHO	10.6	75			Yes (9.86)		5.32
	Pentanal	C4H9CHO	3.5	103			Yes (9.74)		6.67
	Hexanal	C5H11CHO	1.1	129			Yes (9.72)		8.37
	Heptanal	C6H13CHO	0.36	153	Yes	No	Yes (9.65)	Not analyzable
Ketones	Propanone	CH3COCH3	30.8	56	Yes	Yes	Yes (9.69)	Analyzable (VP: 4.3–30.8 kPa, BP: 59–128 °C)	4.35
	Butanone	C2H5COCH3	11.6	80			Yes (9.51)		5.98
	Pentanone	C3H7COCH3	4.3	101			Yes (9.31)		8.72
	Hexanone	C4H9COCH3	1.5	128	Yes	No	No (∼9)	Not analyzable

---

Fig. 7 Chromatogram obtained using the hybrid GC platform, showing the separation of eight VOCs belonging to four chemical classes: alkane, alcohol, ketone, and aromatics.

One of the key advantages of the hybrid GC platform is that, with the exception of forced constraints such as the PID ionization energy, parameters such as the temperature during the preconcentration step, the final desorption temperature, and the temperature ramping rate can be precisely controlled to target specific gases or adjust the analytical range. Moreover, by changing the MOF, the pore size or window size can be adjusted to induce molecular sieving effects, enabling more refined chemical targeting within a narrower range.

Notably, C₆–C₉ alkanes, C₄ alcohol, C₅–C₆ aldehydes, and C₃–C₅ ketones analyzed in Fig. 6 have a high probability of being present in exhaled breath, suggesting the potential application of the platform for noninvasive diagnosis of specific diseases via exhaled breath analysis.^55,56 Because the developed platform lacks a built-in humidity removal unit, several issues should be considered when analyzing highly humid samples such as exhaled breath, which typically exceeds 90% relative humidity. The first concern is the potential degradation of UiO-66-S due to moisture exposure. However, UiO-66-S exhibits high hydrolytic stability because its zirconium clusters form strong metal–oxygen bonds, making the framework more resistant to water-induced structural collapse than many other MOFs.^57–59 The second concern is that water vapor may adsorb into the pores of UiO-66-S, reducing the available adsorption sites for VOCs, lowering the effective preconcentration factor, and ultimately suppressing the PID response.

To evaluate this, 1 ppm BTEX samples were analyzed under both dry (0% RH) and humid (70–72% RH) conditions. As shown in Fig. S11,† no significant difference in signal intensity was observed between the two cases. This result is attributed to two main factors. First, the absolute volume of water vapor present in the 40.8 mL sampling volume at 70% RH is relatively small and unlikely to significantly interfere with either UiO-66-S adsorption or PID detection. Second, even if water adsorption slightly reduced UiO-66-S's available adsorption capacity, the platform's PID detector output reached saturation within the tested concentration range. Therefore, the limiting factor for the measured signal was more dominantly governed by the PID's output voltage limit than by any potential reduction in adsorption capacity due to water vapor. While this finding suggests that moisture interference may be negligible within the current platform's linear range, further investigation would be essential if the detector or system configuration were modified.

Conclusions

In this paper, we reported a hybrid GC platform for the quantitative and qualitative analysis of BTEX; this platform had a compact volume of 0.62 L and a simple configuration consisting of only one hybrid chip, PID sensor module, three-way valve, pump, filter pack, batteries, fluidic connectors, and electronics for control and power supply. This level of miniaturization was achieved by employing a hybrid chip, developed to enable both gas preconcentration and separation, through a material-based integration strategy leveraging the multifunctional properties of MOFs. The platform was designed to operate for up to 70 battery-powered cycles (35 hours), assuming that one analytical cycle consists of 20 minutes of analysis and 10 minutes of stand-by operation.

The platform achieved LODs of 19.3 ppb, 22.8 ppb, 30.4 ppb, and 24.4 ppb and linear ranges of 0.25–1 ppm, 0.25–1 ppm, 0.25–1.5 ppm, and 0.25–2 ppm for the BTEX vapors, respectively. These results were obtained using a total sample volume of 40.8 mL, with 20-minute analysis mode consisting of a 10-minute preconcentration step followed by a 10-minute separation and detection step. The LOD and linear range for benzene indicated that the platform had the potential to detect harmful levels of this compound, as defined by the regulations established by the ACGIH and NIOSH. Repeatability was confirmed with RSD values of 2.07–3.61% obtained from 30 replicates.

Comparative analyses with TD-GC-MS showed notable consistency in both qualitative and quantitative results, demonstrating the platform's practicality. Moreover, the platform was able to differentiate a wide range of analytes, including alkanes, alcohols, aldehydes, and ketones, based on their distinct retention times. Notably, some of these analyzable VOCs are commonly found in exhaled breath, highlighting the potential applicability for noninvasive medical diagnostics. However, given that breath samples typically contain high humidity levels, future studies should examine the impact of water vapor on MOF performance and PID sensitivity. This could be addressed by incorporating humidity control components or implementing signal correction strategies. With these capabilities, the hybrid GC platform could be further developed into a more compact and cost-effective instrument for broader use in safety-, environment-, and health-related applications, particularly if expanded to include quantitative analysis of a wider range of VOCs, including BTEX.

However, certain limitations such as a relatively low peak capacity due to broad peak widths and a narrow linear range resulting from limited detector specification need to be addressed through several strategies. First, for the hybrid chip, further material-related optimization such as improvements in the coating method, mass, thickness, and uniformity is needed to enhance overall performance. Second, the use of low-flow-compatible detectors (e.g., μ-PID or μ-TCD), which have small dead volumes, may help improve response and recovery times, which in turn can contribute to increased peak capacity by reducing peak broadening. Finally, employing a sensor array capable of pattern-based analysis, in addition to retention time, could significantly enhance the platform's qualitative performance by enabling more robust discrimination of complex VOC mixtures.

Data availability

Data for this article are available at the OSF Storage at [https://osf.io/jtk53/?view_only=0f3fbaf1ae374349a6941639b0fbcbc5]. TD-GC-MS data analyzed by Korea Polymer Testing & Research Institute (KOPTRI), described in Fig. 5a and b, are not available, due to the institution's policy prohibiting the provision of modifiable raw data. Instead, official TD-GC-MS analysis reports by the KOPTRI are uploaded at the same link.

Author contributions

Yeongseok Lee: conceptualization, resources, data curation, formal analysis, investigation, visualization, validation, writing – original draft. Sangkyun Lee: formal analysis, data curation, visualization, resources. Woojin Jang: resources, software. Junwoo Lee: resources. Yuntaek Choi: resources. Si-Hyung Lim: supervision, conceptualization, methodology, project administration, writing – review & editing, funding acquisition.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This study was supported by a National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) [grant number: RS-2025-00518573 & RS-2024-00412249], Republic of Korea.

References

1. J. Lee, J. Lee and S. H. Lim, J. Hazard. Mater., 2020, 392, 122145.
2. X. Zhou, X. Zhou, C. Wang and H. Zhou, Chemosphere, 2023, 313, 137489.
3. K. Schmidt and I. Podmore, J. Biomarkers, 2015, 2015, 981458.
4. H. Lin, H. Jiang, S. Y.-S. S. Adade, W. Kang, Z. Xue, M. Zareef and Q. Chen, Crit. Rev. Food Sci. Nutr., 2022, 63, 8226–8248.
5. C. Sonne, C. Xia, P. Dadvand, A. C. Targino and S. S. Lam, J. Build. Eng., 2022, 62, 105344.
6. R. Yadav and P. Pandey, Int. J. Plant Environ., 2018, 4, 14–26.
7. J. T. Brinke, S. Selvin, A. T. Hodgson, W. J. Fisk, M. J. Mendell, C. P. Koshland and J. M. Daisey, Indoor Air, 1998, 8, 140–152.
8. A. Lamplugh, M. Harries, F. Xiang, J. Trinh, A. Hecobian and L. D. Montoya, Environ. Pollut., 2019, 249, 518–526.
9. J. Liu, J. H. Seo, Y. Li, D. Chen, K. Kurabayashi and X. Fan, Lab Chip, 2013, 13, 818–825.
10. D. Loomis, K. Z. Guyton, Y. Grosse, F. El Ghissassi, V. Bouvard, L. Benbrahim-Tallaa, N. Guha, N. Vilahur, H. Mattock and K. Straif, Lancet Oncol., 2017, 18, 1574–1575.
11. J. D. Emery, K. Shaw, B. Williams, D. Mazza, J. Fallon-Ferguson, M. Varlow and L. J. Trevena, Nat. Rev. Clin. Oncol., 2014, 11, 38–48.
12. A. M. Zeid, I. M. Mostafa, B. Lou and G. Xu, Lab Chip, 2023, 23, 4160–4172.
13. Q. Wen, P. Boshier, A. Myridakis, I. Belluomo and G. B. Hanna, Metabolites, 2020, 11, 17.
14. F. Monedeiro, R. B. dos Reis, F. M. Peria, C. T. G. Sares and B. S. De Martinis, J. Breath Res., 2020, 14, 026009.
15. J.-W. Yoon and J.-H. Lee, Lab Chip, 2017, 17, 3537–3557.
16. Y. Lee, H. Son, J. Lee and S.-H. Lim, Micro Nano Syst. Lett., 2024, 12, 11.
17. B. P. Regmi and M. Agah, Anal. Chem., 2018, 90, 13133–13150.
18. E. Lussac, R. Barattin, P. Cardinael and V. Agasse, Crit. Rev. Anal. Chem., 2016, 46, 455–468.
19. J. Wang, N. Nunovero, R. Nidetz, S. J. Peterson, B. M. Brookover, W. H. Steinecker and E. T. Zellers, Anal. Chem., 2019, 91, 4747–4754.
20. S. Zampolli, I. Elmi, G. C. Cardinali, L. Masini, F. Bonafè and F. Zardi, Sens. Actuators, B., 2020, 305, 127444.
21. M. Wei-Hao Li, A. Ghosh, A. Venkatasubramanian, R. Sharma, X. Huang and X. Fan, ACS Sens., 2021, 6, 2348–2355.
22. W. Liao, D. Winship, I. Lara-Ibeas, X. Zhao, Q. Xu, H.-T. Lu, T. Qian, R. Gordenker, Y. Qin and Y. B. Gianchandani, Anal. Chem., 2023, 95, 2157–2167.
23. X. Huang, S. Yang, W. Li, R. Nidetz, R. Sharma, A. D. Sivakumar, C. Thota, H. Zhu, W. Wu, S.-Y. Jeong and X. Fan, Sens. Actuators, B, 2024, 410, 135717.
24. G. C. Frye-Mason, presented in part at International Microprocesses and Nanotechnology Conference, Yokohama, Japan, July, 1999.
25. Scientific Equipment Source, FROG-5000 (Defiant Technologies, Inc.), https://www.defiant-tech.com/pdfs/FROG-5000-Manual-2017-Rev3.pdf.
26. Scientific Equipment Source, VOCAM (Defiant Technologies, Inc.), https://www.defiant-tech.com/pdfs/VOCAM-Heated-PID-GC-System-for-Continuous-Monitoring.pdf.
27. Scientific Equipment Source, X-PID 9500 (Drägerwerk AG & Co.), https://www.draeger-mo.com/media/files/x-pid-9000-9500-pi-9104798-en-master.pdf.
28. Y. Lee, Y. Choi, J. Sim, J. Kim and S. H. Lim, Lab Chip, 2024, 24, 658–667.
29. Y. Lee, J. Lee, S. Hwang, H. Choi, J. Kim and S. H. Lim, Chem. Eng. J., 2024, 157075.
30. N. A. Khan, Z. Hasan and S. H. Jhung, J. Hazard. Mater., 2013, 244–245, 444–456.
31. M. Alhamami, H. Doan and C. H. Cheng, Materials, 2014, 7, 3198–3250.
32. J. R. Li, J. Sculley and H. C. Zhou, Chem. Rev., 2012, 112, 869–932.
33. J. Lee and S. H. Lim, Sensors, 2019, 19, 3706.
34. N. Chang and X. P. Yan, J. Chromatogr. A, 2012, 1257, 116–124.
35. R. Navarro Amador, L. Cirre, M. Carboni and D. Meyer, J. Environ. Manage., 2018, 214, 17–22.
36. R. Ou, W. Zhu, L. Li, X. Wang, Q. Wang, Q. Gao, A. Yuan, J. Pan and F. Yang, Sep. Purif. Technol., 2021, 266, 118087.
37. G. Gumilar, J. Henzie, B. Yuliarto, A. Patah, N. Nugraha, M. Iqbal, M. A. Amin, M. S. A. Hossain, Y. Yamauchi and Y. V. Kaneti, J. Mater. Chem. A, 2022, 10, 6662–6678.
38. Y. Wang, Z. Wang, Y. Gao, Y. Yuan, J. Liu, J. Yan and Y. Chen, RSC Adv., 2022, 12, 30262–30269.
39. X. Zhang, X. Lv, X. Shi, Y. Yang and Y. Yang, J. Colloid Interface Sci., 2019, 539, 152–160.
40. E. Ploetz, H. Engelke, U. Lächelt and S. Wuttke, Adv. Funct. Mater., 2020, 30, 1909062.
41. D. Wu and A. Navrotsky, J. Solid State Chem., 2015, 223, 53–58.
42. L. Hu, W. Wu, L. Gong, H. Zhu, L. Jiang, M. Hu, D. Lin and K. Yang, Angew. Chem., Int. Ed., 2023, 62, e202215296.
43. W. W. Zhao, C. Y. Zhang, Z. G. Yan, L. P. Bai, X. Wang, H. Huang, Y. Y. Zhou, Y. Xie, F. S. Li and J. R. Li, J. Chromatogr. A, 2014, 1370, 121–128.
44. R. Suresh, S. V. J. Anarthanan, R. Manavalan and K. Valliappan, Int. J. Pharm. Sci. Res., 2010, 1, 123–132.
45. American Conference of Governmental Industrial Hygienists, ACGIH Data Hub - 2025, https://www.acgih.org/data-hub.
46. The National Institute for Occupational Safety and Health, NIOSH Pocket Guide to Chemical Hazards, https://www.cdc.gov/niosh/idlh/intridl4.html.
47. NIOSH Manual of Analytical Methods (NMAM), 5th edn, https://www.cdc.gov/niosh/nmam/pdf/NMAM_5thEd_EBook-508-final.pdf.
48. N. A. Epshtein, Pharm. Chem. J., 2021, 55, 91–96.
49. Y. Saalberg and M. Wolff, Clin. Chim. Acta, 2016, 459, 5–9.
50. P. C. Moura, M. Raposo and V. Vassilenko, Biomed. J., 2023, 46, 100623.
51. J. Zhou, Z. A. Huang, U. Kumar and D. D. Y. Chen, Anal. Chim. Acta, 2017, 996, 1–9.
52. R. D. Arrua, A. Peristyy, P. N. Nesterenko, A. Das, D. M. D'Alessandro and E. F. Hilder, Analyst, 2017, 142, 517–524.
53. B. Gao, M. Huang, Z. Zhang, Q. Yang, B. Su, Y. Yang, Q. Ren and Z. Bao, Chin. J. Chem. Eng., 2019, 27, 818–826.
54. H. Zheng, L. Zhao, J. Ji, J. Gao, C. Xu and F. Luck, ACS Appl. Mater. Interfaces, 2015, 7, 10190–10200.
55. A. Z. Temerdashev, E. M. Gashimova, V. A. Porkhanov, I. S. Polyakov, D. V. Perunov and E. V. Dmitrieva, Metabolites, 2023, 13, 203.
56. E. Gashimova, A. Osipova, A. Temerdashev, V. Porkhanov, I. Polyakov, D. Perunov and E. Dmitrieva, Anal. Methods, 2021, 13, 4793–4804.
57. F. Ahmadijokani, H. Molavi, M. Rezakazemi, S. Tajahmadi, A. Bahi, F. Ko, T. M. Aminabhavi, J. Li and M. Arjamand, Prog. Mater. Sci., 2021, 125, 100904.
58. D. Bůžek, S. Adamec, K. Lang and J. Demel, Inorg. Chem. Front., 2020, 8, 720–734.
59. E. Acuna-Yeomans, P. J. Goosen, J. J. Gutiérrez-Sevillano, D. Dubbeldam and S. Calero, J. Mater. Chem. A, 2024, 12, 25233–25243.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5lc00268k

---