Performance comparison of straight circular and trapezoidal µGC columns for acetone–ethanol separation

This paper presents a comparative design and performance evaluation of two straight micro gas chromatography (µGC) columns with circular and trapezoidal cross-sectional geometries for the chromatographic separation of acetone and ethanol under identical operating conditions. Two geometries were simulated in a computational program, COMSOL Multiphysics, considering helium as a carrier gas and PEG (Polyethylene glycol) as a stationary phase. Parameters such as column length (10–100 cm) and hydraulic diameter (50–600 µm) were considered for both geometries and evaluated in terms of retention time, number of theoretical plates (N), resolution (R s), and height equivalent to a theoretical plat (HETP). The findings show that there is a clear trade-off between efficiency and pressure drop for the two geometrical configurations considered. The straight column of circular cross section configuration results in lower efficiency, which can be noted from the lower theoretical plate number ranging from (0.5–3.9) × 10 3 plates, larger HETP value of 0.027–0.035 cm, and poor resolution R s < 1.25 at lower pressure drop (280–500) Pa and higher flow rate (approximately 0.5 m/s). However, the straight column of trapezoidal cross section shows better separation efficiency due to enhanced mass transfer, band broadening reduction, and a higher surface-to-volume ratio with better flow distribution. Though the performance is improved, this improvement is obtained at the expense of higher pressure drop (∼3.8 × 10 3 Pa) and lower flow rate (∼0.35 m/s). From the above considerations, it can be seen that the trapezoid shape is indeed a better geometry when compared to others in µGCs, especially when used for breath analysis.

Volatile organic compounds (VOCs) represent low-molecular-weight compounds generated by endogenous metabolic mechanisms or exogenous sources such as diet, environment, and microorganisms. Changes in the profile of VOCs have been linked to many physiological and pathophysiological states, leading to the rising popularity of VOC analysis in disease diagnosis without using invasive techniques (Al Bakain et al., 2026). The exhaled breath represents a source of thousands of VOCs generated due to metabolic reactions within the human body and is a convenient, painless, and re-samplable material compared with traditional analyses performed on blood or tissue samples (Ghafourian et al., 2026). Recently published reviews have demonstrated that breathomics represents a promising technique for the early diagnosis and treatment of many diseases, including diabetes mellitus, lung cancer, colorectal cancer, liver diseases, neurodegenerative disorders, respiratory diseases, and infectious diseases (Lagopati et al., 2023; Zhou et al., 2024; Heng et al., 2025). Another systematic review and meta-analysis carried out recently further showed that diagnostics using VOCs can be highly accurate in various cancers, highlighting the high clinical value of the analysis of exhaled air in the future for point-of-care diagnosis (Zhou et al., 2024; Ghawanmeh et al., 2024). However, despite such promising progress, a number of reviews have pointed to the need for high selectivity of analytical methods, standardization of sampling, and proper separation of similarly chemical compounds before detection (Tan et al., 2023; Hara et al., 2025; Tischer et al., 2026; Zhu et al., 2026).

Of the many VOCs found in exhaled breath, acetone and ethanol are among the two most commonly studied oxygenated VOCs due to their diagnostic importance and abundance (Mahnoor et al., 2024; Marfatia et al., 2025). Acetone is one of the primary ketones generated during fatty acid β-oxidation and ketogenesis and has been intensively studied as an indicator of diabetes mellitus, diabetic ketoacidosis, fasting, ketogenic diet, and other metabolic disturbances (Lagopati et al., 2023; Heng et al., 2025). The levels of breath acetone rise dramatically under conditions of metabolic disturbance and therefore serve as one of the most well-established biomarkers for metabolic disturbances (Marzoog and Kopylov, 2026; Palmisani et al., 2025). Conversely, ethanol may be generated either endogenously through microbial fermentation and intestinal metabolism, or exogenously through food consumption or direct exposure and has also been identified as a clinically relevant breath VOC in liver disorders, respiratory conditions, gastroenterological disorders, and certain cancers (Zhou et al., 2024; Heng et al., 2025; Parikh et al., 2026; Rodionova et al., 2026). Due to the presence of acetone and ethanol in exhaled breath and their similar chemical characteristics, poor separation may cause analysis interference. For that reason, separation of acetone and ethanol is vital when using portable breath analyzers or other diagnostic devices based on VOCs (Ghawanmeh et al., 2024; Haripriya et al., 2023).

Gas chromatography-mass spectrometry (GC-MS) continues to be the standard technique in analyzing VOCs due to its high separation ability, sensitivity, and the ability to identify different compounds. However, conventional GC machines in laboratories tend to be bulky, energy consuming, and costly. Miniaturized GC (μGC) has been introduced as a solution to these problems in recent years. Miniaturization involves the use of micro-fabricated columns, small injectors, and mini-detectors to allow quick analysis, low gas flow rate, low energy consumption, and portability (Wang et al., 2026; Chen et al., 2026). Recent studies have shown that μGC technology is emerging as an important tool in environmental analysis, food analysis, process control in industry, and breath analysis since it integrates the separation abilities of traditional GC with the requirement for portability for field work (Rachkidi et al., 2024; Lin et al., 2026). Moreover, developments in artificial intelligence, machine learning, electronic noses, and portable sensors have increased the demand for small-size chromatographic units able to perform high-resolution separation of the VOC mixtures before their detection by the sensors (Lee et al., 2026; Heng et al., 2025).

The separation column has been identified as an important element of the μGC system since it determines the efficiency of the separation process, retention of the analytes, separation time and performance of the whole instrument. The performance of the column can be assessed through the use of chromatographic indices such as number of theoretical plates (N), height equivalent to theoretical plates (HETP), retention time (t r), peak width at half height (PWHH) and the resolution. In chromatography, separation efficiency increases when the number of theoretical plates is increased with minimum plate height and minimum peak broadening (Huber et al., 1975). However, polar stationary phases like polyethylene glycol (PEG) are quite appropriate in the separation of oxygen-containing VOCs like ketones and alcohols, while helium gas is used in most cases due to its good diffusion and good chromatographic performance. In portable μGC systems, it is essential to keep high separation efficiency in centimeters-long columns since this allows for miniaturization without performance loss (Sun et al., 2016).

The recent developments in the fabrication of MEMS have made it possible to produce microcolumns of various geometric shapes, such as rectangular, circular, trapezoidal, serpentine, spiral, and semi-packed, to enhance chromatography performance while keeping the device small (Tian et al., 2018; Rachkidi et al., 2024). Earlier studies have studied the impact of stationary phase coating, pillars design, channels geometry, and fabricating technique on the performance of μGC, resulting in improved efficiency and pressure. Nevertheless, there is only limited research done on the effects of microchannel cross-sectional shape on μGC performance, especially under identical operational conditions and considering clinically relevant VOC pair, such as acetone and ethanol. The relevance of studying the effect of microchannel cross-sectional shape on the performance of μGC with regard to acetone and ethanol is justified by the fact that these VOCs are typical oxygenated VOCs that often coexist in biological fluids and are applicable to breathe testing for diagnosing diseases.

In the current study, a numerical investigation is conducted comparing the performances of two μGC columns that have a circular cross-sectional shape and a trapezoidal one, using the PEG stationary phase and helium carrier gas to separate the acetone-ethanol mixture. The impact of the column geometry is studied by changing the inner diameter from 50 μm to 600 μm at constant length of the column, and changing the column length from 10 cm to 100 cm at constant inner diameter of 250 μm. Evaluation of the performance of μGC columns is made based on the outlet concentration profile, width at half peak, retention time, number of theoretical plates, height equivalent to a theoretical plate, and chromatographic resolution (Lombardi et al., 2024). Through keeping the same operating conditions and material properties for both geometries, this study can isolate the impact of the cross-sectional shape on the performance of the μGC, and it can provide design guidelines for future portable chromatographic systems for VOCs analysis and breath testing.

Governing equations and boundary conditions

The 3D numerical simulations were carried out in COMSOL Multiphysics® software utilizing Laminar Flow and Transport of Diluted Species physics interfaces for modeling of carrier gas flow and analyte transport in µGC columns. As the Reynolds number was below 10, the flow was considered incompressible, laminar and steady. The carrier gas velocity profile was determined by solving the continuity and Navier–Stokes equations, as given in Equations 1 and 2: where u is the velocity vector, ρ is the helium density, μ is the dynamic viscosity, and p is the pressure. Helium was selected as the carrier gas with a density of 0.164 kg m -3 and a dynamic viscosity of 1.96 × 10 −5 Pa s at 298 K and 1 atm. A constant volumetric flow rate of 1 × 10 −8 m 3 s -1 (0.6 SCCM) was prescribed at the inlet, while atmospheric pressure was imposed at the outlet. Fully developed inflow and no-slip wall boundary conditions were applied for both column geometries.

The transport of acetone and ethanol in the helium carrier gas was described using the transient convection–diffusion equation, as given in Equation 3: where c is the analyte concentration, D is the molecular diffusion coefficient, and u is the carrier-gas velocity obtained from the flow solution. The molecular diffusion coefficients were fixed at 2.5 × 10 −10 m 2 s -1 for acetone and 2.0 × 10 −10 m 2 s -1 for ethanol. The inlet concentration was introduced as a transient Gaussian pulse, as defined by Equation 4. where gp1(t) is the Gaussian pulse function with an initial concentration of 4.4 × 10 −5 mol m -3 for both analytes.

The interaction between the mobile phase helium and the PEG stationary phase was modeled using the Robin (mixed) boundary condition, as expressed in Equation 5, which represents reversible analyte partitioning at the wall: where k is the interfacial mass-transfer coefficient, K is the partition coefficient, c is the gas-phase concentration adjacent to the wall, and c s is the stationary-phase concentration. The partition coefficients were maintained constant at 90 for acetone and 190 for ethanol throughout all simulations. This boundary condition enables analyte adsorption and desorption at the PEG-coated wall and thereby reproduces the chromatographic retention mechanism within the µGC column.

For a proper evaluation of both types of columns, it is necessary that all the simulation studies be carried out under the same operating conditions. The conditions include the use of the same values for the hydraulic diameter, length of the column, film thickness, carrier gas, feed rate, Gaussian injection profile, boundary conditions, partition coefficients, and properties of materials.

A physics-controlled ultrafine mesh with boundary layer mesh refinement was applied in order to capture the steep concentration gradients around the interface with the stationary phase. Around 1.2 million degrees of freedom were utilized for each model with mesh independence established through variation of outlet concentrations less than 0.5%. The stationary flow field solution was computed via the use of MUMPS solver with a convergence criterion of 10 –6, while a transient solution was computed through BDF solver with a time step not exceeding 0.01 s. In addition, a parameter study was conducted for K in the range from 50 to 200.

Figure 1 illustrates the numerical model configuration and the boundary conditions employed in the three-dimensional COMSOL Multiphysics simulations. A fully developed laminar flow with a constant volumetric flow rate of 1 × 10 − 8 m 3 s -1 (0.6 SCCM) was specified at the inlet, while a pressure outlet with zero gauge pressure (0 Pa) was imposed at the outlet to establish fully developed outflow conditions. Injection of acetone and ethanol in a temporary manner is described using the Gaussian pulse function (gp 1). This function defines the inlet concentration distribution of the two compounds under investigation. For a fair comparison between the two geometries of the µGC column (circular and trapezoidal), the same inlet and outlet boundary conditions, carrier gas properties, and Gaussian pulse function were used in the simulations. The boundary conditions used in the numerical model are summarized in Table 1.

Model validation

Prior to analyzing the separation efficiency of acetone and ethanol, the validation of the present numerical model was done through replication of the existing finite-element μGC model of Singh et al. (2022). Similar geometry, governing equation, physics interfaces, boundary condition, material property, and operation condition presented in the previous work were used in COMSOL Multiphysics. The simulation results that were obtained were compared against the published velocity field and chromatographic response to confirm the correctness of the current numerical model before its application to acetone-ethanol separation problem.

The above figure (Figure 2) depicts the comparison between the findings of Singh et al. (2022) and the reproduced findings in the current research work. From (Figures 2A,B) below, it can be noted that the reproduced carrier gas velocity distribution is consistent with the published model showing a maximum velocity of 0.45 m s -1, similar to the published result of 0.48 m s -1. The reproduced outlet concentration profile, as depicted from (Figures 2C,D), also shows an excellent agreement with the chromatography of hexane-benzene mixture system with retention times of 21.40 s and 28.89 s for hexane and benzene, respectively.

Further confirmation for the quantitative analysis carried out in Table 2 shows that the reproduced numbers have small errors compared to the numbers provided by the authors in their work. The small discrepancies in the numbers obtained are a result of the numerical discretization process. The accuracy of these numbers obtained indicates that the developed numerical model is accurate enough to capture the underlying transport phenomena taking place in the μGC column. Consequently, the validated numerical model was used to study the separation of acetone and ethanol in straight circular and trapezoidal μGC columns.

In the current research, the case of the acetone and ethanol separation in a capillary column is studied with the help of COMSOL Multiphysics, in which helium is used as the carrier gas, and polyethylene glycol (PEG) serves as the stationary phase. The model Laminar Flow is used to couple with Transport of Diluted Species to allow studying fluid dynamics and mass transfer simultaneously (Qadir et al., 2019). The results of the simulation indicate that the flow within the capillary column is laminar and the velocity profile across the cross-section is parabolic. The velocity is highest at the center and tends towards zero at the wall as a result of no-slip boundary condition (Figures 3B,D). Such behavior is in line with the Poiseuille flow that controls micro-scale and capillary transport systems and forms a vital component in defining residence time and dispersion of analytes (Hinman, 2018). The mesh quality analysis also proves that the numerical model is refined enough, which is the correct resolution of the velocity and concentration gradients. The mesh distributions and skewness quality for the circular and trapezoidal cross-section columns are presented in Figures 3A,C, respectively. In CFD simulations, a high-quality mesh is necessary to reduce the numerical diffusion and enhance the accuracy of the prediction of the species transport. Moreover, the findings show that geometry has a big impact on the distribution of velocity. Whereas circular channels have perfect symmetric parabolic shapes, non-circular shapes cause deviations that influence the dispersion and separation efficiency of solutes. This has a direct effect on the chromatography of a column. Helium carrier gas is a good choice because of its low viscosity and high diffusivity, which increases the rate of mass transfer (Nagy et al., 2015). Meanwhile, PEG stationary phase allows specific interaction with acetone and ethanol, which allows the effective separation on the basis of the variations in the partitioning behavior.

These four subplots present a hydrodynamic response of laminar flow in microchannels, which reveals the dependence of the channel length and characteristic diameter (ID or hydraulic diameter, D h) on velocity and pressure drop. These trends are entirely in line with the Hagen Poiseuille law, which states that the pressure drop rises linearly with channel length and inversely with the fourth power of diameter, whereas the velocity is mostly controlled by flow rate when at steady state (Zareian-Jahromi et al., 2009). These findings affirm that geometric parameters are the most influential in flow resistance in micro-scale systems, and diameter plays a much greater role than length. The behaviour of velocity and pressure versus channel geometry, shown in (Figures 4A–D is a clear indication of the classical hydrodynamic behaviour of the laminar flow in microchannels. In (Figure 4A) in case of a circular microchannel with constant inner diameter (ID = 250 μm) the velocity is nearly constant at about 0.6 m/s at the length of the channel, which implies that the flow is fully developed (Lee et al., 2011). Conversely, the pressure rises between 300 Pa and almost 500 Pa as the length rises between 20 cm and 100 cm and this corresponds to an approximate increase of 65–70 percent. This proves that pressure drop is linearly dependent on channel length, as the Hagen Poiseuille relation, in which pressure drop is directly related to length (Reidy et al., 2006).

The same tendency is observed in (Figure 4C) on the trapezoidal microchannel (D h = 250 μm), where velocity is constant (approximately 0.4 m/s), but pressure rises by almost 8-fold (almost 500 Pa to approximately 4000 Pa). The increase in pressure is greater than in the circular channel and indicates greater resistance to flow as a result of geometrical effects, but remains proportional to length as does the circular channel (Agah et al., 2006; Li et al., 2014). These findings confirm that the channel length has a strong impact on pressure drop and a minimal impact on velocity in the case of incompressible flow with steady flow conditions. Conversely (Figure 4B), and (Figure 4D) emphasize the predominant role of diameter (ID or hydraulic diameter, Dh). In (Figure 4B) with the growth in circular channel diameter (100 μm–600 μm), the velocity decreases rapidly, by about 9 m/s down to less than 1 m/s (95–97 percent decrease) and pressure decreases drastically by about 80,000 Pa to about 2,000 Pa (95–97 Equally, with the trapezoidal channel, (Figure 4D), as Dh increases, the velocity (reduced to 0.5 m/s) and pressure (reduced to less than 5,000 Pa) decrease dramatically (by approximately 9099 fold), as does the pressure. This powerful nonlinear behavior proves the fact that pressure drop is inversely proportional to the fourth power of diameter and therefore diameter is the most sensitive parameter of microchannel flow (Radadia et al., 2010). The quantitative analysis shows that channel length causes a linear change in pressure drop, but the effect of channel diameter (or hydraulic diameter) has a much more extreme nonlinear effect, decreasing pressure drop by as much as two orders of magnitude. The results are in very good accord with the classical microfluidic theory and with the earlier studies which were conducted using the Hagen Poiseuille model (Nazdrajić et al., 2021).

The changing dynamics of the concentration profile over time indicate the influence of the geometry of the channel on convective diffusive transport process (Bruus, 2008). This can be seen from (Figure 5A), where acetone and ethanol have the same Gaussian-shaped inlet profile with the maximum peak of the concentration value of about 3.2 × 10 − 9 mol/m 3 at t ≈ 3 s. The peak in the case of circular microchannel (Figure 5B) is shifted to t ≈ 5s along with broadening of the peak and decreased maximum concentration (∼ 7 × 10 − 10 mol/m 3). It is indicating dispersion and residence time effects. In contrast, the trapezoidal microchannel (Figure 5C) exhibits a more pronounced delay, with peak concentrations appearing at t ≈ 9 – 10 s and further reduction in magnitude (∼ 2.5 × 10 − 10 mol/m 3), along with significant peak spreading. Quantitatively, the residence time increases by nearly 80%–100% and peak concentration decreases by approximately 60%–70% compared to the circular channel. These results indicate that while both geometries exhibit dispersion, the trapezoidal cross-section induces stronger spreading and delay due to increased flow non-uniformity and enhanced diffusive mixing effects (Lan, 2012).

The effects demonstrated in Figure 4 represent the effect of column length and column inner diameter on the retention time and separation efficiency in the circular microchannel. According to (Figure 6A), the retention time for both acetone and ethanol is observed to increase with the increase in column length from 10 cm to 100 cm (Song et al., 2026). In quantitative terms, the retention time of acetone increases from 3.2 s to 7.3 s (approximately 2.3-fold increase), whereas the retention time of ethanol increases from 3.5 s to 8.2 s (∼2.4-fold increase). This trend implies that the retention time depends directly on the length of the column due to the higher residence time of the analyte. In a similar manner, the number of theoretical plates (N) of both acetone and ethanol increases remarkably with increasing length, ranging from 700 to 2900 for acetone and 900 to 4300 for ethanol (nearly 4–5-fold increase in separation efficiency) (Figure 6B) (Zareian-Jahromi et al., 2009). On the other hand, the impact of changing the inner diameter (ID) for a fixed column length of 15 cm can be seen in (Figure 6C) and (Figure 6D). An exponential increase in retention time occurs for IDs less than 300 μm before levelling off. For instance, the retention time of acetone rises from roughly 3.2 s at 50 μm to around 10 s at 300 μm (≈3× increase), while that of ethanol rises from roughly 3.5 s–11 s. Beyond 300 μm, any change will be minimal as it appears to hit its saturation point (Ali et al., 2009). The same effect can be observed with regard to the theoretical number of plates in (Figure 6D) as acetone increases by almost an order of magnitude from roughly 500 plates to around 5500 plates, and ethanol also increases from around 800 plates to 7400 plates. These findings indicate that increasing ID increases dispersion and interaction area until the column reaches hydrodynamic equilibrium (Regmi and Agah, 2018).

(Figure 7) represents the results obtained on the impact of channel length and hydraulic diameter on retention time and resolution for a trapezoidal microchannel. From (Figure 7A), it can be seen that retention time for both acetone and ethanol is directly proportional to column length, with a range of 10 cm–100 cm. Numerically, the retention time for acetone increases from around 3 s–33 s, whereas that of ethanol goes up from ∼4 s to almost 37–38 s. The result is 10–12-fold increase in retention time (Sun et al., 2009). In a similar way, from (Figure 7B), the number of theoretical plates (N) is seen to increase with length non-linearly for both acetone and ethanol. For acetone, it increases from ∼2000 to ∼90,000, while that of ethanol increases from ∼3000 to ∼125,000, showing a multiplication factor of 40–50 (Shakeel, 2015). This indicates that long trapezoidal columns have a high effect on improving separation performance, due to good mass transfer and low band width overlap. However, figures (Figure 7C) and (Figure 7D) show the effects of hydraulic diameter (D h), while fixing the length at 15 cm. Retention time increases exponentially with D h, such that acetone increases from about 3 s at 50 μm to around 10 s at 250–300 μm, while ethanol increases from about 3.5 s to around 11 s, which represents a three-fold increase. After this point, retention time levels off, implying a weak effect at larger diameters. This is also seen in the number of theoretical plates in figure (Figure 7D), which increases rapidly from around 500 to 5,500 for acetone, and 800 to 7,500 for ethanol, before leveling off at around 300 μm (Noye, 2005). The qualitative implication of such behavior is that the increase in D h values leads to more dispersion and interaction until an optimum range where hydrodynamic equilibrium is approached, and any additional increment yields no benefit whatsoever (Du et al., 2018). The trapezoidal column shape shows a high degree of dependency on the length of the column, with significant efficiencies achieved, while its hydraulic diameter shows nonlinearity with saturation (Vial et al., 2011).

What must be highlighted is that the enhanced performance seen in the trapezoidal column is not caused by any difference in operation. The two column models have been simulated by using the same hydraulic diameter, column length, carrier gas flow rate, thickness of stationary phase, inlet concentration, boundary conditions, Gaussian pulse injection, and material properties. Thus, the differences seen in retention factor, theoretical plate number, HETP, and separation efficiency are purely due to the effect of cross-sectional geometry on the flow field and mass transfer in the μGC column. Although the trapezoidal geometry causes a slight increase in pressure drop and reduction in average velocity of carrier gas, these effects are a natural outcome of the modified cross-sectional geometry at the same flow rate conditions.

Comparative assessment of the impact of two geometries of µGC columns straight circular and straight trapezoidal on the efficiency of chromatographic separation of acetone and ethanol was carried out under identical operating conditions. According to the data presented in Table 3, the straight trapezoidal µGC column demonstrated higher efficiency parameters (higher number of theoretical plates up to 125 × 10 3, lower HETP (0.010–0.015 cm), and better R s (up to 1.90) than the straight circular one. In turn, the straight circular shape of the microcolumn had lower pressure drop and higher carrier-gas velocity, yet its effectiveness was lower. The conducted study reveals that the cross-section geometry of the microchannels is a very important parameter in determining the efficiency of µGC and needs to be taken into account during the column design. Thus, the trapezoidal cross-section can be regarded as one of the promising geometries for miniaturized µGC devices, which can find application in fast VOCs detection and breath diagnostics.

The future work will entail the development of the optimized µGC column of trapezoid design on a silicon wafer through MEMS technology and validating the same through experimentation to prove its chromatographic performance. The optimized µGC column will then be incorporated in a µGC system for the rapid detection and diagnosis of VOCs from breath samples.