WO2025235006A1

WO2025235006A1 - Monolithic system for gas sampling and analysis with integrated gas pumps

Info

Publication number: WO2025235006A1
Application number: PCT/US2024/028878
Authority: WO
Inventors: Xiangyu Zhao; Yutao QIN; Yogesh Gianchandani
Original assignee: University of Michigan System
Current assignee: University of Michigan System
Priority date: 2024-05-10
Filing date: 2024-05-10
Publication date: 2025-11-13
Anticipated expiration: 2026-11-10

Abstract

A monolithic gas chromatography system is presented for gas sampling and analysis with integrated gas pumps. Components required for sampling and analysis of chemical compounds, including multiple Knudsen pumps, separation column, and detector, are integrated into a single microfabricated chip. The on-chip Knudsen pumps work together to control all the required gas flows for operation without the need for valves or external flow controls.

Description

MONOLITHIC SYSTEM FOR GAS SAMPLING AND ANALYSIS WITH INTEGRATED GAS PUMPS

FIELD

[0001] The present disclosure relates to a monolithic system for gas sampling and analysis with integrated gas pumps.

BACKGROUND

[0002] Analysis of vapor mixtures is important for many applications, such as process monitoring and personal exposure monitoring. Many monitoring applications require rapid, in-situ, continuous, and quantitative results. Conventional gas chromatography instruments are typically large, benchtop, expensive, and high-power instruments which are not suited for these monitoring applications. Microscale gas chromatography ( n GC) systems, with their small size, low cost, and rapid response times have been developed to fit the requirements for these applications better.

[0003] A microscale gas chromatography system typically incorporates a column to separate the analyte molecules, a detector for detection, and pumps and valves to control gas flow. For many applications, the system may additionally incorporate a preconcentrator for analyte molecule collection, allowing the system to perform both gas sampling and analysis (GSA). Most microscale gas chromatography systems incorporate commercial pumps and valves, whereas a few incorporate micropumps and microvalves, or use modified architectures to avoid valves. Additionally, existing microscale gas chromatography systems often contain components that are not microfabricated or are fabricated separately and subsequently connected together using different methods of fluidic connections in a hybrid integration.

[0004] The assembly required to make connections in hybrid integrated systems is mainly accomplished manually, which is time-consuming, prone to errors, and incompatible with future mass production. Hybrid integration allows more flexibility in terms of the individual components and allows the integration of commercial off the shelf components within the microsystem but also increases the system size, increases the fabrication and assembly cost, and introduces cold spots and dead volume, which can affect system performance. The disadvantages of hybrid integration can be addressed by monolithic integration. Additionally, monolithic integration of the microscale gas chromatography is a pathway towards further system miniaturization and decreasing the cost of fabrication and assembly. There has been no report on the monolithic integration of pumps or valves or the monolithic integration of a full gas analysis system.

[0005] The challenges with monolithic integration of pumps and valves with other microscale gas chromatography components are the differences in the structures and fabrication. Most pumps and valves require flexible diaphragms and mechanical actuators. These elements are distinct from preconcentrators and columns, which are typically fluidic chambers or channels incorporating sorptive materials. Different types of detectors may pose other requirements, such as specialized structures, materials, or coatings. Creating a common fabrication process able to fabricate all structures required is a challenge to monolithic integration. Additionally, monolithic integration introduces challenges with thermal isolation between the components because of the proximity of the components and the high thermal conductivity in the commonly used substrates. Yield is another consideration important to monolithic integration, as any nonfunctional component will cause the monolithically integrated chip to be nonfunctional.

[0006] By overcoming these challenges and developing a monolithic system for gas sampling and analysis, a single chip microscale gas chromatography system microfabricated with a common fabrication flow is developed to serve as a low-cost, mass-produced, and ultra-small gas analyzer.

[0007] This section provides background information related to the present disclosure which is not necessarily prior art.

SUMMARY

[0008] This section provides a general summary of the disclosure, and is not a comprehensive disclosure of its full scope or all of its features.

[0009] In one aspect, a gas chromatography system is presented with a separation column, a detector, and a separation pump monolithically integrated onto a substrate. The separation column has an inlet and an outlet and is disposed in a separation flow path. The separation column incorporates a stationary phase material that provides chromatographic separation of molecules and the separation flow path includes an inlet port and an outlet port, such that the inlet port of the separation flow path is fluidly couped to a sample source. The detector is disposed in the separation flow path at the outlet of the separation column. The separation pump is fluidly coupled between the detector and an outlet port of the separation flow path. [0010] The gas chromatography system may further comprise a preconcentrator having an inlet, a chamber, and an exhaust, where the inlet of the preconcentrator is configured to receive a sample and pass the sample through the chamber to the exhaust of the preconcentrator, and the chamber contains a sorbent material therein to trap analytes. In some embodiments, the preconcentrator resides off the substrate; whereas, in other embodiments, the preconcentrator resides on the substrate.

[0011 ] In one embodiment, the detector is further defined as a capacitive detector and the separation pump is further defined as Knudsen pump. The gas chromatography system may further include an inlet pump fluidly coupled between a sampling inlet port and the inlet of the preconcentrator; and a sampling pump fluidly coupled between the exhaust of the preconcentrator and a sampling outlet port; thereby defining a sampling flow path between the sampling inlet port and the sampling outlet port.

[0012] During operation, the sampling pump operates to draw the sample into the preconcentrator during a sampling phase, and the separation pump operates to pull gas through the separation flow path towards the detector during a separation phase. More specifically, a controller is operably coupled to each of the inlet pump, the sampling pump, and the separation pump. The controller activates the sampling pump and deactivates the inlet pump during the sampling phase, and the controller activates the separation pump and deactivates the sampling pump during the separation phase. The controller also controls the separation pump to prevent gas from entering the separation flow path during the sampling phase, and the controller controls the inlet pump to prevent gas from entering the sampling flow path during the separation phase.

[0013] The gas chromatography system may also include a flow meter disposed in the separation flow path, where the flow meter includes a heater element and a temperature sensor is positioned downstream in the separation flow path from the heater element.

[0014] In another aspect, the gas chromatography system include a preconcentrator, an inlet pump, a sampling pump, a separation column, a detector, a separation pump, and a controller.

[0015] In an example embodiment, the preconcentrator has an inlet configured to receive a sample from a sampling inlet port and has an exhaust fluidly coupled to a sampling outlet port, thereby defining a sampling flow path between the sampling inlet port and the sampling outlet port. The preconcentrator includes a chamber and sorbent material within the chamber. The inlet pump is fluidly coupled between the sampling inlet port and the inlet of the preconcentrator. The sampling pump is fluidly coupled between the outlet of the preconcentrator and the sampling outlet port. The separation column is disposed in a separation flow path and has an inlet fluidly coupled to the sampling flow path between the preconcentrator and the inlet pump. The detector is disposed in the separation flow path at an outlet of the separation column. The separation pump is fluidly coupled between the detector and a separation outlet port. The controller is operably coupled to each of the inlet pump, the sampling pump, and the separation pump.

[0016] Further areas of applicability will become apparent from the description provided herein. The description and specific examples in this summary are intended for purposes of illustration only and are not intended to limit the scope of the present disclosure.

DRAWINGS

[0017] The drawings described herein are for illustrative purposes only of selected embodiments and not all possible implementations, and are not intended to limit the scope of the present disclosure.

[0018] Figure 1 A is a diagram depicting an example architecture for a monolithic gas chromatography system.

[0019] Figure 1 B is a diagram depicting an example embodiment having only the separation column, detector and pump monolithically integrated onto a substrate.

[0020] Figure 2 is a graph illustrating the operation of the micropumps during the sampling phase and the separation phase.

[0021] Figures 3A and 3B are top and bottom views of a 3D model of a first design for the gas chromatography system, respectively.

[0022] Figure 4 is a cross-sectional view of the first design for the gas chromatography system.

[0023] Figures 5A and 5B are perspective sectional views of a Knudsen pump.

[0024] Figure 6 illustrates the temperature distribution across a pumping unit and along a pumping channel side wall of a Knudsen pump design.

[0025] Figures 7A and 7B are graphs showing pump performance and load lines during sampling and separation, respectively, for the first design of the gas chromatography system. [0026] Figures 7C and 7D are graphs showing pump performance and load lines during sampling and separation, respectively, for the second design of the gas chromatography system.

[0027] Figure 8 is a graph showing the thermal simulation result for the preconcentrator and the separation column.

[0028] Figure 9 is a cross-sectional view of an example capacitive detector.

[0029] Figures 10A and 10B are cross-sectional views of a flow rate sensor without and with airflow, respectively.

[0030] Figures 11 A-11 H are diagrams depicting the fabrication process for a gas chromatography system with a suspended Knudsen pump.

[0031 ] Figures 12A-12M are diagrams depicting an alternative fabrication process for a gas chromatography system with a suspended Knudsen pump.

[0032] Figure 13 is a graph showing the measured preconcentrator and column temperature during preconcentrator desorption.

[0033] Figure 14 is a diagram of an injection test setup to characterize the separation column, where the column is connected between the inlet and flame ionization detector of a benchtop gas chromatography.

[0034] Figures 15A-15C are graphs showing injection test results with hexane and octane.

[0035] Figure 16 is a chromatogram of 50 ppm 2-pentanone, chloropentane, chlorobenzene and octanol.

[0036] Figure 17A is a diagram depicting how three different Knudsen pumps operate together to control the gas flow in the gas chromatography system.

[0037] Figure 17B is a cross-sectional view illustrating the gas chromatography system.

[0038] Figure 17C is a 3D illustration of the gas chromatography system.

[0039] Figure 17D is an example layout of the components of the gas chromatography system.

[0040] Figures 18A and 18B are cross-sectional views of the Knudsen pumps comparing the pumps that uses an array of pumping channels supported by a suspended membrane and the more robust design that uses a single pumping channel supported by the device silicon.

[0041] Figure 19 is a graph showing the flow resistance of the sampling and separation flow paths along with Knudsen pump performance. [0042] Figure 20 is a graph showing the simulated temperature distribution during preconcentrator desorption with a heat sink and heat pipe to act as cooling components.

[0043] Figure 21 is a graph showing estimated flow rate and pressure required for different column dimensions in order to achieve a target separation resolution of 2 between propylene glycol methyl ether (PGME) and propylene glycol monomethyl ether acetate (PGMEA).

[0044] Figures 22A-22I are diagrams depicting the fabrication process for a third design of the gas chromatography system.

[0045] Figure 23 is a side view showing the monoGSA chip connections to the daughterboard, the motherboard and the microcontroller.

[0046] Figure 24 is a graph showing temperature readings from the monoGSA chip during separation with external cooling components to reduce thermal crosstalk.

[0047] Figures 25A-25C are graphs showing injection test results with PGME and PGMEA mixture.

[0048] Figure 26 is a chromatogram of a 1 :1 :1 mixture of ethyl acetate, propyl acetate, butyl acetate with and without the inlet pump operation.

[0049] Figure 27 is a chromatogram of benzene, toluene, chlorobenzene, and dichlorobenzene.

[0050] Figures 28A and 28B are chromatograms of the sampling and separation of ethyl acetate, propyl acetate, and butyl acetate and 2-methoxyethanol, o-xylene, benzaldehyde, respectively.

[0051] Figures 29A and 29B are a layout and cross-sectional view, respectively, of the third design of the gas chromatography system with additional thermal isolation.

[0052] Figure 30A shows the geometry of the monoGSA chip used for thermal simulation.

[0053] Figure 30B is a cross-sectional view of the monoGSA chip.

[0054] Figures 31 A-31 L are diagrams depicting the fabrication process for a fourth design of the gas chromatography system.

[0055] Figures 32A-32K are diagrams depicting the fabrication process for the fourth design of the gas chromatography system using anodic bonding.

[0056] Figure 33 is a graph showing expected separation flow rate when suspended membrane Knudsen pumps are used for the separation pump. [0057] Figures 34A-34K are diagrams depicting the fabrication process for a fifth design of the gas chromatography system.

[0058] Figure 35 is a diagram depicting an example architecture for a monolithic gas chromatography system with an optional off-chip preconcentrator.

[0059] Corresponding reference numerals indicate corresponding parts throughout the several views of the drawings.

DETAILED DESCRIPTION

[0060] Example embodiments will now be described more fully with reference to the accompanying drawings.

[0061] Figure 1 A depicts an example architecture for a monolithic gas chromatography system 10. The gas chromatography system 10 is generally comprised of a preconcentrator 12, a separation column 14, and a detector 15. Additional components of the gas chromatography system 10 include an inlet pump 11 , a sampling pump 13, a separation pump 16, and a controller 18. Each of these components are further described below. It is to be understood that only the relevant components of the system are discussed in relation to Figure 1 A, but that other components may be needed to control and manage the overall operation of the system including but not limited to flow rate sensors, temperature sensors, and calibration sources which improve the performance and reliability of the system. In the figures, the sampling pump is labeled as Knudsen Pump 1 , inlet pump is labeled as Knudsen Pump 2, and the separation pump is labeled as Knudsen Pump 3.

[0062] The preconcentrator 12 has an inlet configured to receive a sample from a sampling inlet port 6 and an exhaust fluidly coupled to a sampling outlet port 8, thereby defining a sampling flow path between the inlet port 6 and the sampling outlet port 8. The preconcentrator contains a chamber with sorbent particles, polymer coating, or other types of sorbent material to collect analyte molecules. The sampling flow path further includes the inlet pump 4 fluidly coupled between the inlet port 6 and the inlet of the preconcentrator 12 as well as a sampling pump 13 fluidly coupled between the outlet of the preconcentrator 12 and the sampling outlet port 8.

[0063] The separation column 14 is a flow channel with stationary phase used to separate analyte molecules so that they can be differentiated. In this embodiment, the separation column 14 is disposed in a separation flow path. More specifically, the separation column 14 has an inlet fluidly coupled to the sampling flow path between the preconcentrator 12 and the inlet pump 4. The detector 15 is disposed in the separation flow path at an outlet of the separation column 14. The detector is used to detect and quantify the analyte molecules as they elute from the separation column 14. Lastly, the separation pump 16 is fluidly coupled between the detector 15 and a separation outlet port 9.

[0064] The three micropumps 1 1 , 13, 16 operate together to control gas flow throughout the system without the need for valves or external flow controls. In the example embodiment, the three micropumps are implemented as Knudsen pumps. A controller 19 is operably coupled to each of the three micropumps 1 1 , 13, 16. In this embodiment, the inlet pump 11 , the preconcentrator 12, the sampling pump 13, the separation column 14, the detector 15, and the separation pump 16 are monolithically integrated onto the substrate.

[0065] In the exemplary embodiment, the controller 19 is implemented as a microcontroller. It should be understood that the logic for the control of system 10 by controller 19 can be implemented in hardware logic, software logic, or a combination of hardware and software logic. In this regard, controller 19 can be or can include any of a digital signal processor (DSP), microprocessor, microcontroller, or other programmable device which are programmed with software implementing the above described methods. It should be understood that alternatively the controller is or includes other logic devices, such as a Field Programmable Gate Array (FPGA), a complex programmable logic device (CPLD), or application specific integrated circuit (ASIC). When it is stated that controller 19 performs a function or is configured to perform a function, it should be understood that controller 19 is configured to do so with appropriate logic (such as in software, logic devices, or a combination thereof).

[0066] With reference to Figure 2, operation of the gas chromatography system 10 is split into two phases: a sampling phase and a separation phase. During the sampling phase, the sampling pump 13 operates to draw the sample into the preconcentrator 12. More specifically, the controller 19 activates the sampling pump 13 and deactivates the inlet pump 11. In this way, the inlet pump provides a lower flow resistance path (as compared to the separation flow path) for the sample to flow into the preconcentrator 12. Additionally, the controller 19 controls the separation pump 16 to prevent gas from entering the separation flow path during the sampling phase by actuating the separation pump 16 at a low power level so it acts as a valve and prevents unwanted gas flow from entering the system. [0067] During the separation phase, the separation pump operates to pull gas through the separation flow path towards the detector. More specifically, the controller 19 activates the separation pump 16 and deactivates the sampling pump 13. Analyte molecules are released from the preconcentrator 12 using a thermal pulse to inject the collected analyte molecules into the separation column 14. The separation column 14 in turn separates the injected analyte molecules, and the detector 15 detects the molecules as they elute from the separation column 14. Additionally, the controller 19 controls the inlet pump to prevent gas from entering the sampling flow path during the separation phase.

[0068] In another aspect of this disclosure, the separation column 14, the detector 15 and at least one micropump (i.e., the separation pump 16) are monolithically integrated onto a substrate while the remainder of the components reside off the substrate as seen in Figure 1 B. In this example embodiment, the separation column 14, the detector 15 and the separation pump 16 form a separation flow path. The separation flow path has an inlet port and an outlet port, such that the inlet port of the separation flow path is fluidly couped to a sample source (e.g., a preconcentrator located off the substrate). Except with respect to the differences discussed herein, the remainder of the system is substantially the same as the gas chromatography system 10 described above in relation to Figure 1 A.

[0069] Five particular implementations for the gas chromatography system 10 are set forth below.

[0070] MonoGSA 1 and monoGSA 2 are monolithic implementations of systems performing GSA functions with Knudsen pumps that have pump walls formed with ultrathin dielectrics and supported by a suspended membrane. The fluidic chip incorporates a preconcentrator, separation column, detector, calibration source, two flow rate sensors, and three Knudsen pumps with a footprint of 17 x 20 cm². The Knudsen pumps are designed to provide high performance and are designed to provide sampling and separation flow rates above 0.5 seem. Two separate Knudsen pump designs are used, one using a large upper membrane to support pumping channels and another that splits the large membrane into smaller membranes. Implementations of the gas chromatography system 10 that use these Knudsen pumps are referred to as monoGSA 1 and monoGSA 2, respectively. The system successfully sampled and separated chemicals within a mass range of 86-1 12 g/mol using an external pump providing the flow. [0071] Referring to Figures 3A and 3B, the gas chromatography system is formed from a stack of three dies with a silicon-on-insulator (SOI) die sandwiched between two insulating covers. The SOI die contains the through die structures for Knudsen pumps, through-holes to route fluidic flow and provide thermal isolation and metal traces for heating and sensing. The top insulating cover has fluidic channels, which form the fluidic chambers for pGC components. It is smaller than the other two dies to expose wire pads on two sides of the chip in order to form the electrical connections. The bottom insulating cover has channels to make fluidic connections between different pGC components. A cross-sectional view of all components can be seen in Figure 4.

[0072] The different components of the gas chromatography system must be codesigned in order to obtain a functional system. Components must be designed considering the limited chip size and additional component specific requirements. In particular, the following considerations should be in mind during the design process: Knudsen pumps and detector must be placed in the center of the lithography reticle as they have features with the smallest sizes; the preconcentrator requires high temperatures for operation and should be placed away from other components; thermal isolation should be implemented for the preconcentrator while also considering of the chip size limitations; the length of the column should be maximized within the available chip footprint and pump performance; location of electrical connections must be considered to decrease the parasitic resistance, to decrease the difficulty of assembly and making electrical connection, and to fit within the limited exposure region of the lithography tool; placement of components should minimize dead volume and flow resistance from fluidic connections. Additionally, special consideration has to be given to the Knudsen pump designs as changing the configuration of a pump not only changes the pump performance but also affects flow resistance. In particular, the sampling flow is determined by a combination of sampling pump performance and the sampling path flow resistance, the latter of which is mainly attributed to the intrinsic flow resistance of the inlet pump; the separation flow is determined by a combination of separation pump performance and separation flow resistance, the latter of which is mainly attributable to the intrinsic flow resistance of the column and sampling pump.

[0073] Knudsen pumps generate gas flow by thermal transpiration through narrow channels. When the hydraulic diameter of the channels is comparable to the mean free path, the gas molecules move against a temperature gradient along the channel. With motionless operation, Knudsen pumps provide reliability and long lifetime. Unidirectional Knudsen pumps have a more facile fabrication process and electronic control scheme compared to bidirectional Knudsen pumps. As the Knudsen pumps are placed in the flow paths, they not only provide the pumping flow but also add to the flow resistance. Therefore, careful design of the pumps is required to ensure sufficient flow is available.

[0074] The Knudsen pumps used in the gas chromatography system are Knudsen pumps with microfabricated pumping channels. These pumps contain densely packed arrays of narrow pumping channels made of AI2O3 supported by a membrane of oxide and nitride as seen in Figures 5A and 5B. Each pumping unit of the Knudsen pump consists of three main parts: 1 ) the active pumping area, 2) the outer substrate area, and 3) one thermal isolation zone between the other two parts. The active pumping area consists of dense arrays of rectangular vertical pumping channels with 1 .2-pm wide pumping channels with 10 nm thick sidewalls. Each pumping channel is of the same channel width 1.2 pm, depth 18 pm, and length 25 pm. The pumping channels are supported by the upper oxide-nitride-oxide (ONO) composite membrane from above. This membrane is designed to have a slight tensile composite stress to prevent buckling. Multiple rectangular openings are located periodically in the upper ONO layer; the openings serve as gas outlet windows for pumping channels. A thin-film metal is patterned on top of the upper ONO layer to serve as the resistive heater. The outer substrate area supports the membrane and is connected to the other pumping units or pGC components. A thermal isolation zone is located between the active pumping area and the outer substrate rim to confine the heat generated by the heaters within the active pumping area. The thermal isolation zone is provided by the arrays of gas vertical channels, which are fully covered by the upper ONO layer. The heat sinks are made in the bottom silicon layer. The heat sink dissipates Joule heat either directly to the ambient or laterally through the SOI substrate. The grid-patterned openings in the heat sink serve as the gas inlet windows for the pumping channels. Multiple pumping units are connected in different configurations to ensure that the required pressure and flow rates are achieved for each Knudsen pump in the gas chromatography system.

[0075] Two separate designs are created for the Knudsen pump. In the Knudsen pump used for monoGSA 1 , each pump uses a larger composite membrane of oxide and nitride to support a greater number of pumping channels. The oxide-nitride-oxide (ONO) membrane supporting the pumping channels has a size of 0.4 x1 .2 mm² to support 3354 pumping channels per pumping unit. Each pump is designed to operate at a maximum of 0.6 W to provide a maximum flow rate of 1 .5 seem and blocking pressure of 770 Pa based on analytical calculations using FEA simulations of the temperature (Fig. 6). The sampling pump contains two of these pumping units in parallel, the inlet pump contains one of these pumping units, and the separation pump contains six pumping units in series. The flow resistances for different monoGSA components are estimated based on their geometry (Table 1 ). The load lines and pump performance lines for both sampling and separation are calculated to obtain an estimated separation flow rate of 0.7 seem and sampling flow rate 0.6 seem as seen in Figures 7A and 7B.

Table 1 : Estimated flow resistance of different monoGSA components

[0076] For the Knudsen pump design for monoGSA 2, each pumping unit uses two smaller composite membranes to support the pumping channels to increase the fabrication window of the gas chromatography system. Each smaller composite membrane has a size of 0.45 x 0.45 mm² to support 1245 pumping channels. Compared to the monoGSA 1 design, the Knudsen pump design for monoGSA 2 incorporates additional C shaped perforations in the composite membrane to relieve stress and prevent fracturing. Each pumping unit is designed to operate at a maximum of 0.4 W to provide a maximum flow rate of 0.53 seem and blocking pressure of 650 Pa based on simulations. The sampling pump contains four pumping units in parallel, the inlet pump contains two pumping units in parallel, whereas the separation pump contains six pumping stages in series of two pumping units in parallel (total of twelve pumping units). The load lines and pump performance for both sampling and separation are calculated to obtain an estimated separation flow rate of 0.55 seem and sampling flow rate 0.55 seem (Figs 7C and 7D).

[0077] The preconcentrator 12 is a chamber packed with sorbent particles to adsorb analyte molecules at room temperature (or a lower temperature) during sampling and desorbs the analyte molecules with a thermal pulse during separation. The chamber has a footprint of 7 x 1 .5 mm² with a channel depth of 500 pm for a total volume of 5.3 pL.

[0078] Packed sorbent particles (e.g., Carbograph™ 2) are selected as they are proven materials commonly used in gas chromatography systems. Possible alternatives to sorbent particles is to use thin polymer films of materials such as Tenax TA. The sorbent particles are packed into the chamber through a loading port located on the edge of the chip and are confined by pillar structures with gaps that are smaller than the sorbent particle size. The packing process is performed by providing a vacuum to Port 3 to draw sorbent particles into the preconcentrator chamber through the loading port. The flow resistance of the packed preconcentrator is estimated to be 0.082 kPa/sccm based on the Ergun equations assuming 80 mesh particle size.

[0079] Heating is provided in a closed-loop, controlled manner using a thin film metal heater with an estimated resistance of 37 Q and a thin film metal thermistor with estimated resistance of 2.3 kQ. The preconcentrator is heated up to ~150°C to desorb the collected analyte molecules. The temperature uniformity of the preconcentrator during desorption benefits from the silicon in the SOI substrate, which is thermally conductive and spreads the heat throughout the preconcentrator area.

[0080] The temperature that the preconcentrator requires for desorption is higher than the operating temperatures for the other pGC components. The separation column is typically operated isothermally or is temperature programed with a temperature ramp that increases as separation progresses. As the preconcentrator desorption occurs at the start of separation, the consequent temperature rise that desorption causes through thermal crosstalk is detrimental to separation column performance. Similarly, the capacitive detector performance is negatively impacted by temperature crosstalk, as a change in temperature causes baseline drift in the detector reading. A higher average temperature for the Knudsen pump decreases the performance of the Knudsen pump. In order to prevent the aforementioned impact of preconcentrator heating, a thermal isolation region is placed between the preconcentrator and the Knudsen pumps to thermally isolate the preconcentrator from the rest of the pGC components.

[0081] The thermal isolation is meant to increase the thermal resistance between the preconcentrator and the other pGC components. Silicon has a much higher thermal conductivity of «130 Wrrr¹K“¹ compared to 1 .38 Wnr¹K“¹ for glass and 0.025 Wrrr¹K“¹ for air. The high thermal conductivity of silicon causes the entire chip to heat up along with the preconcentrator, increasing the power requirements of the chip and impacting the performance of the column, detector, and Knudsen pumps.

[0082] In the thermal isolation area, a grid of through holes are made through the handle silicon, buried oxide, and device silicon through deep reactive ion etching (DRIE) and XeF2 etch of silicon. The DRIE holes have a size of 24 x 367 pm² and are separated by 24 pm. Over these holes, there remains a layer of oxide and nitride, which supports the metal traces connecting pGC components to the wire pads. The thermal isolation through-holes covers an area of 3.6 x 16 mm² between the preconcentrator and Knudsen pump. By having these DRIE holes, the effective thermal conductivity in the thermal isolation area drops from 130 Wrrr¹K“¹ to around 15 Wm^-1K“¹. In the thermal isolation area, there are two types of locations that do not have these DRIE holes. The outer edge of chip does not have DRIE holes to maintain the structural integrity of the chip. The area underneath fluidic channels do not have DRIE holes to reduce the chance of leaks that can occur if there are breaks in the membrane.

Transient thermal response of the preconcentrator desorption is simulated using finite element analysis using COMSOL Multiphysics® software (COMSOL Inc., Stockholm, Sweden). The main goal of the simulation is to evaluate the effectiveness of an area with thermal cutouts in the silicon during transient heating. The simulation coupled the electrical and thermal responses to model both the effects of joule heating on the preconcentrator heater and the temperature spread when heating for a short period of time.

[0083] The modeled monoGSA 1 chip is placed on top of a PCB and a thin layer of copper over the PCB to simulate the copper traces. The thermal conductivity for each material is assumed as follows: 1.4 Wnr¹K“¹ for fused silica, 130 Wrrr¹K“¹ for silicon, 14.8 Wrrr¹K“¹ for the thermal isolation, 0.3 Wm^-1K“¹ for the PCB made from FR4, 71.6 Wm^-1K“¹ for the Ti/Pt film, and 0.03 Wm^-1K“¹ for air. The thermal conductance of the thermal isolation region is determined using a thermal simulation of heat spread through silicon with through-holes based on the thermal isolation design. Experimentally measured electrical resistivity at room temperature and temperature coefficient of resistivity of the Ti/Pt thin film, which are 3.8x10^-7 Q m and 0.003 ^QC ^-1 respectively are also incorporated within the model for the joule heater. Heating is provided through metal traces using 17 V from 5 to 15 seconds. The initial temperature is set at 20^QC and external surfaces are assumed to have natural air convection of 20 ^QC. [0084] The results of the thermal simulation are shown in Figure 8. The preconcentrator reached an average of 153.8^QC from a starting temperature of 20^QC by 10 seconds while the column area only reached an average temperature of 45.9^QC. Based on the temperature distributions, the column area remained below 50^QC and 40^QC based on the contour plot and the Knudsen pump area next to the thermal isolation zone is between 50-60^QC. The temperature drop across the thermal isolation is 107.9^QC, which indicates the effectiveness of the thermal isolation cutout.

[0085] Additional thermal isolation can be achieved by increasing the size of the thermal isolation region or by increasing the size of the through-holes inside of the thermal solation region. These methods are not favored in order to maintain structural integrity of the chip, to use the same size of DRIE through-holes in the handle silicon, and to maximize the area for the other pGC components.

[0086] The column is a long fluidic channel with a stationary phase coating for separating analytes based on their partitioning coefficients. The column has dimensions of 25.3 cm x 350 pm x 130 pm. The column takes up a large part of the monoGSA 1 and monoGSA 2 chips with it being designed in a serpentine manner and having a footprint of 10.4 x 14.2 mm². The length of the channel is limited by the size of the chip, the width of the channel is limited by the stationary phase coating method, and the height of the channel is limited by the flow resistance that the separation Knudsen pump can drive. The efficiency of the column varies with the narrower dimension and is independent of the wider dimension, which is why the height of the channel is minimized as much as possible to provide better separation efficiency. It incorporates a 0.4 pm thick polydimethylsiloxane (PDMS) coating on the channels sandblasted in the insulating cover die. A heater and thermistor are included for closed loop temperature control to either have temperature programming for better separation or to maintain an elevated operation temperature to reduce the impact of ambient temperature on separation. The carrier gas for the column is ambient air, which reduces the consumables needed for the system.

[0087] The detector 15 aims to detect analyte molecules as they elute from the separation column to generate a signal that can then be used to identify and quantify the analyte molecules. It is ready understood that there are many different types of detectors that can be incorporated in the gas chromatography system 10. An example implementation for the detector 15 is a capacitive detector where analyte molecules that flow over the detector cause the vapor-sensitive polymer to change its capacitance. [0088] Capacitive detectors are formed using closely spaced interdigitated electrodes coated by a polymer sensitive to analyte molecules as seen in Figure 9. Analyte molecules are absorbed into the coating of the detector as the analyte molecules pass over, which changes the thickness and permittivity of the polymer and, in turn, changes the capacitance. The interdigitated electrodes have a gap and width of 1 .5 pm and a PDMS polymer thickness of 0.4 pm. The sensing area of the detector is 1 .6 mm², which provides an estimated 17 pF initial capacitance based on the simulated value of 1 mm² sensing area providing 11 pF. The polymer layer thickness is less than the gap of the electrodes, which means that the capacitance change is dominated by changes in polymer thickness. The thinner polymer layer provides a faster response, as the detector response time is mainly constrained by the speed of vapor absorption and desorption in the polymer layer. The capacitive detector is placed in the middle of the monoGSA chip because of the fine features of the interdigitated electrodes and because the center of reticles has the highest lithographic resolution for stepper lithography tools. The interdigitated electrodes are on an ONO layer over a conductive device silicon layer, creating the possibility of parasitic capacitance between the metal traces of the capacitive detector and the device silicon. The impact of the parasitic capacitance can be removed through the selection of a proper capacitance to digital converter chip for capacitance reading. The device and handle silicon need to be grounded, which is performed by using a diamond scribe to scratch the device and silicon layers on the side of the chip and using indium solder to connect a wire between the scratch and an analog ground connection.

[0089] A flow rate sensor can provide valuable information about the gas flow within the gas chromatography system 10. The structure of the flow rate sensor is illustrated in Figures 10A and 10B. The flow rate sensor has three thermally isolated bridge-like structures (referred to as bridges), which are suspended over a cavity beneath. Thin-film metals are placed on the bridges to form a heater on the center bridge and thermistors on the left bridge and the right bridge, respectively. The three bridges are isolated from each other such that the heat conduction is minor to heat convection from the heater to the thermistors. The sensing principle is based on the measurements of temperature distribution generated by the heater in the center. At zero-flow condition, the temperature profile is symmetrical about the heater in the center. Thus, the temperatures at the two thermistors are equal. When gas flow is present, the temperature profile is shifted towards the downstream of the flow direction and the asymmetrical temperature profile results in resistance change in the thermistor. The total change of resistances can be measured and converted to a corresponding flow rate.

[0090] The incorporation of a calibration source allows for post deployment calibration of the separation column and detector over the chip’s deployment lifetime, increasing the reliability of the system. The calibration source aims to generate known analyte molecules that can be sampled and subsequently separated. The calibration source provides valuable results verifying the reliability of the results generated by the system and allows for the correction of results if drift is observed. A possible implementation of the calibration source consists of a solid that can be thermally decomposed in ambient air to provide known volatile organic chemicals, which the system can collect and detect for calibration. A chemical such as polypropylene undergoes thermal decomposition at 200^QC to 300^QC in air and can decompose into several molecules that can be separated and detected by the system.

[0091] With reference to Figures 1 1 A-1 1 H, the monoGSA 1 and 2 chip is fabricated lithographically with 5 masks on the SOI die and the two masks for the insulating cover. The SOI has thicknesses of 21 pm, 1 pm, and 400 pm for the device silicon, buried oxide, and handle layer, respectively. Rectangular trenches are etched through the device silicon using deep reactive ion etching (DRIE) to form pumping channels for the Knudsen pump (Fig. 1 1 A). Atomic layer deposition (ALD) of AI2O3 is used to create a conformal coating over the trenches to for pumping channel sidewalls for Knudsen pumps. The trenches are then refilled with polysilicon deposited using low- pressure chemical vapor deposition (LPCVD). An upper dielectric membrane of oxide and nitride is deposited by plasma-enhanced chemical vapor deposition (PECVD) over the device silicon (Fig. 1 1 B). The membrane consisted of 0.6/0.55/0.7/0.55 /0.6 pm thick silicon oxide/ nitride/ oxide/ nitride/ oxide (ONONO) to tune the stress of the membrane and reduce the chance of fracturing during fabrication and operation. Metal is then sputtered (Fig 1 1 C) and evaporated (Fig. 11 D) for metallization with the first metallization layer forming most of the metal features while the second metallization layer forming the fine features capacitive detector and to reduce parasitic resistances. Backside DRIE is used to create through-holes in the handle layer to form heat sinks for pumps and through-holes to route fluidic flow and a layer of ALD AI2O3 is deposited to protect the exposed silicon (Fig. 11 E). A XeF2 vapor etch of silicon is used to release the pumping channel sidewalls and to etch away the remaining silicon in the thermal isolation area (Fig. 1 1 F). Sandblasting is performed (by IKONICS Corp., Duluth, Ml, USA) on insulating cover wafers to create the top and bottom insulating covers of the monoGSA. The fluidic channels have a depth of 180 pm and the ports for tube insertion and the preconcentrator area has a depth of 500 pm. The top and bottom insulating cover are attached to the SOI die using epoxy (#377, Epoxy Technology, MA, USA) and capillary columns are attached. The Sorbent particles are packed into the preconcentrator through a sorbent loading port and the loading port is sealed afterwards (Fig. 1 1 G). A labeled cross-section view of the resultant monoGSA chip can be seen in Figure 11 H. The monoGSA chip has a size of 17 x 20 mm.

[0092] Referring to Figure 12A-12M, an alternative method of fabrication is proposed to reduce the die level post-processing and assembly. The current chip stack has a sandblasted bottom insulating cover to form fluidic connections. The alternative fabrication method will move the fluidic channels from the bottom insulating cover into the handle silicon layer with an additional DRIE step to form DRIE channels of a shallower depth. This alternative fabrication method will increase the difficulty of the SOI fabrication but reduce the overall cost of the system by greatly decreasing the labor and cost of assembly.

[0093] The monoGSA chip is fabricated and assembled and individual components of the monoGSA are separately evaluated to validate the design. The components of significant interest are the preconcentrator heating and thermal isolation performance, the column performance, the detector performance, the pump performance. The system is also evaluated by sampling and separating a chemical mixture to validate operation of the gas chromatography system.

[0094] A constant voltage is applied to the preconcentrator to evaluate the effectiveness of the thermal isolation between the preconcentrator and the rest of the chip. The preconcentrator desorption temperature is limited to 150°C as low temperature solder that melts at 138°C is used to form connections between wire pads and header pins. With 12 V applied for 10 seconds from 5 to 15 seconds, the preconcentrator reached 150°C while the column is elevated to 55°C (Fig. 13). The measured results matched the thermal simulations in the design section. The temperature difference between the preconcentrator and the columns matched the simulation and showed the effectiveness of the thermal isolation in preventing the preconcentrator from heating up the entire chip during thermal desorption.

[0095] The column performance is evaluated with injection tests using the benchtop gas chromatograph (Fig. 14). The separation column is connected between the inlet and flame ionization detector (FID) of a benchtop GC (#Agilent 7890, Agilent, CA, USA). The port next to the inlet pump is connected to the inlet and the port close to the separation pump is connected to the FID. The port close to the separation pump is blocked so that analytes bypass the preconcentrator during injection. The injection used N2 as the carrier gas and occurs at 40°C. The injection results of a hexane and octane is shown in Fig.15A. Injection tests are performed at a range of flow rates from 0.01 seem to 1 seem. The height equivalent to a theoretical plate (HETP) and the theoretical plate number (N), which are measures of separation column efficiency, are extracted from the chromatogram using the following equations: where L is the length of the column, PWHH is the peak width at half height, tn is the retention time of the chemical, to is the retention time of an unretained analyte. Higher N and lower HETP indicate higher column efficiency. The HETP reached a minimum of 2.5 mm for hexane and 0.5 mm for octane in the range 0.05 and 0.20 seem as seen in Figures 15A-15C. [0096] Sampling flow of 0.05 seem (Table 2) and separation flow of 0.04 seem

(Table 3) are obtained.

Table 2: Sampling flow obtained with the on-chip Knudsen pumps of monoGSA 1 Table 3: Separation flow obtained with on-chip Knudsen pumps of monoGSA 1

[0097] Evaluating the system operation is performed with an external pump (#MP6-gas, Servoflo, MA, USA) providing the sampling and separation flow. Analytes are pre-mixed in liquid form and injected into a 4 L Tedlar bag to form vapors of known concentration. During sampling, the vapor is connected to the sample inlet of the gas chromatography system 10 and an external pump is connected to Port 1 to provide a pulling flow. During separation, the sample inlet is blocked with a septum and the external pump is connected to Port 3 to pull air through the system. The preconcentrator is desorbed with a 10 second thermal pulse where the temperature reached 150°C and a separation flow of 0.08 seem is provided by the external pump. There is no column heating during the test. A 50 ppm mixture of 2-pentanone, chloropentane, chlorobenzene, and octanol are used as the test mixture and is sampled for 2 min. The peaks for 2-pentanone, chloropentane, and chlorobenzene are seen at 60 sec, 105 sec, and 231 sec respectively and the chemicals are baseline separated (Fig. 16).

[0098] Next, the monoGSA 3 implementation uses a more robust Knudsen pump design. This implementation of the gas chromatography system 10 monolithically integrates three Knudsen pumps with a preconcentrator, a separation column, and a detector into a 15x15 mm² chip. The Knudsen pumps in this implementation bypass the manufacturability challenges of Knudsen pumps with pumping channels supported by suspended membranes by forming the pumping channels using a thick and unsuspended oxide layer. Although the unsuspended design reduces the temperature gradient for pumping, its consequence is mitigated by the co-design of other components in the gas chromatography system, which incorporates narrower and shorter fluidic channels than monoGSA 1 and monoGSA 2. Unconstrained by power consumption, thermal crosstalk is mitigated by heat dissipation from heat sinks and heat pipes and forced convection from fans. [0099] Compared to the monoGSA 1 and monoGSA 2 designs, the following innovations are made: Knudsen pump design changed to increase its robustness and manufacturability; improved overall manufacturability and yield of the chip by removing suspended membranes from the design and changing design to maximize the manufacturing window; instead of a large thermal isolation region, external thermal management options are used to reduce the impact of preconcentrator heating on the rest of the monoGSA chip; thermistors are used to estimate the Knudsen pump temperatures in order to estimate the pumping performance and provide closed loop flow control; and column design with short column length and small cross-section to match the pressure and flow rates generated by the Knudsen pump design.

[0100] Within the monoGSA chip, the sampling flow through the preconcentrator and the separation flow through the preconcentrator, column, and detector are controlled by the three Knudsen pumps without the need for any valves as seen in Figure 17A. For sampling, Knudsen pump 1 (sampling pump) provides a pulling flow to draw chemicals into the preconcentrator, during which Knudsen pump 2 (inlet pump) is idle within the sampling path, whereas Knudsen pump 3 (separation pump) provides a gentle pressure head that resists a gas flow through the column and the detector. For separation, separation pump pulls the collected chemicals in the preconcentrator through the column and capacitive detector, during which sampling pump is idle, whereas inlet pump provides a gentle pressure head that prevents additional chemicals from entering the system.

[0101] Referring to Figures 17B and 17C, the monoGSA 3 chip is composed of a silicon-on-insulator (SOI) die between two fused silica dies. The SOI die incorporates a 12 pm thick device silicon layer, a 0.4 pm thick buried oxide layer, and a 525 pm thick handle layer. The device silicon layer is covered by a 2 pm thick upper oxide layer, which is covered further covered by a Ti/Pt Metal 1 layer of 0.03/0.1 nm thickness and a Ti/Ni Metal 2 layer of 0.03/0.2 nm thickness. The SOI die incorporates pumping channels of the Knudsen pumps and through-holes for flow routing, all formed through the silicon and oxide layers. Metal 1 provides all the on-chip heaters, thermistors, and detector electrodes, whereas Metal 2 provides the wire pads to form electrical connections and reduces the parasitic resistance for the metal connecting the wire pads to the heater and thermistors. The fused silica dies are 675 pm thick and incorporate 40 pm deep fluidic channels and 500 pm deep gas ports for capillary tube attachment. The top fused silica die forms fluidic channels for the separation column, preconcentrator, and detector. Both the top and bottom fused silica dies contain fluidic interconnect channels. Fused silica is selected as the material to form the pGC components and fluidic interconnect channels due to its high chemical inertness which allows for better separation performance. The locations of the pGC components are designed to minimize the length of fluidic connections, parasitic electrical resistances, and thermal crosstalk.

[0102] The three Knudsen pumps in the monoGSA 3 generate the gas flow and prevent unwanted gas flow from entering the chip. The yield of each component is critical to a monolithic system. Therefore, instead of using the prior approach of building pumping channels with ultra-thin dielectric sidewalls supported by a suspended membrane, a more conservative and manufacturable approach is used for the Knudsen pumps in the monoGSA 3. In these Knudsen pumps, 1 .2x200 pm² openings are etched through the upper oxide layer and device silicon to form narrow channels required for pumping channels and 30x230 pm² channels are etched through the handle silicon to allow through-wafer gas flow. Metal traces surrounding the narrow channel in the upper oxide layer act as Joule heaters to provide the hot side of the temperature gradient. With its lower thermal conductivity than silicon, the upper oxide layer provides the majority of the temperature difference and provides the main pumping effect. The device silicon and the handle silicon dissipate heat, creating the cool side of the temperature gradient. Using the upper oxide layer and device silicon to form the channel walls provides support for the pumping channel and improves the robustness and yield of the pump. Pumping channels are connected in parallel to increase the output flow rate and are placed «1 .5 mm apart to ensure sufficient heat sinking. The differences between these Knudsen pump and the other Knudsen pump designs with suspended membranes are shown in Figures 18A and 18B. The more robust design has a higher efficiency in terms of pumping performance per unit area but has much fewer pumping channels, creating a tradeoff in the performance, specifically the flow rate, for much increased manufacturability and yield.

[0103] In the monoGSA 3 system, the sampling pump provides the sampling flow, the separation pump provides the separation flow, whereas the inlet pump is used to prevent additional chemicals from entering the system during separation. Based on the gas flow needs, sampling pump and separation pump each incorporates 6 parallel pumping channels, whereas the inlet pump incorporates 4 parallel pumping channels.

[0104] Thermistors are placed close (5 pm away) to and far away (1 mm away) from the Knudsen pump heaters to measure the hot and cold side temperatures. These thermistors are formed using thin film metal traces similar to the heaters but have a much larger resistance ranging from. With the appropriate correction factors, i.e., the ratio of the experimental and calculated performance, thermistors could provide estimation and feedback control of the pumping performance.

[0105] The temperature distribution of the Knudsen pump designs are analyzed by electrically and thermally coupled solid mechanics models and finite element analysis (FEA) in COMSOL Multiphysics®. The chip is seated on a thermal interface pad, which bonded it to a printed circuit board (PCB) with a metal heat sink on the bottom side, mimicking the actual usage. The material properties and thicknesses assumed in the models are listed in Table 4. The electrically and thermally coupled models simulated Joule heating by a voltage input applied to the heater. The temperature coefficients of resistivity (TCR) of the metal layers are incorporated into the model to reflect the electrical resistances more accurately during heating.

Table 4: Material properties and thicknesses assumed in the Knudsen pump FEA

[0106] For the simulated cross-sectional temperature distribution of the pumping channel in a typical 4-channel pump, an applied power (l/Vp) of 0.5 W/channel (2 W total) generated an average temperature at the hot ends of the pumping channels (Th.p) of 235^QC, an average temperature at the cold ends of pumping channels temperature (T_c.p) of 133^QC and an average temperature difference AT (=Th._p - T_c.p) of 102^QC. Owing to its low thermal conductivity of 1 .4 W/mK, the upper oxide layer localized the heat near the pumping channel openings. Therefore, the bulk silicon temperature measurement by a thermistor located 1 mm away from the active metal heater is used to represent T_c.p at the bottom of the thermally conductive (130 W/mK) device silicon.

[0107] By sweeping the input heater voltage, the W_p is swept through 0-0.5 W/channel to investigate the temperature distribution and its impact on the pumping performance. The average T values of the 4- and 6-channel pumps are within 4% of each other at a given power level per channel. However, T_avg, defined as the average of Th._p and T_c.p, has a more significant variation of 15%. The T_avg can represent the overall heating of the chip and the effectiveness of heat sinking. The PCB temperature underneath the 6-channel pump is elevated up to 95^QC compared to that of 76^QC underneath the 4-channel pump. At a W_p of 0.5 W/channel, T_avg reached 212^QC for the 6-channel pump and 184^QC for the 4-channel pump, impacting the pumping performance.

[0108] The mass flow rate M through a rectangular pumping channel can be estimated based on Sharipov’s equation as: where AP is the pressure difference applied between the hot (outlet) and cold (inlet) ends (/.e., applied pressure head); P_avg is the average pressure; m is the mass of the gas molecule; B is the Boltzmann constant; and grand qp are the thermal creep and viscous flow coefficients, respectively, which are dependent on T_avg and have been reported in literature. The blocking pressure AP_eq can be calculated by equating M to zero in Eq (3.1 ):

[0109] The flow rate of a Knudsen pump reaches its maximum (Q_ma%) when AP is equal to zero in Eq (3.1 ): where N is the number of parallel pumping channels and p is the mass density of the gas at T_avg and P_avg. It is worth noting the impact of T_avg on AP_ec? and Qmax. Based on the tabulated gr and qp values, gr increased with T_avg whereas qp decreased with T_avg. For AP_eq, the qr and qp largely canceled the impact of T_avg, causing AP_ec? to be minimally dependent on T_avg. In contrast, the Qmax has a higher dependency on T_avg. Therefore, the calculated performance of the Knudsen pumps showed a nonlinear relation to W_p. This effect is more noticeable in the Qmax of 6-channel pump at a high total W_p. For instance, at 0.5 W/channel, the 15% higher T_avg of the 6-channel pump results in 7% lower Q_max/channel compared to the 4-channel pump.

[0110] The pumping characteristics of the sampling pump and the separation pump are estimated from experimental testing of a separate, standalone six-channel Knudsen pump with the same pump design and co-fabricated with the monoGSA 3 on the same SOI wafer. The standalone six-channel Knudsen pump provided a maximum flow rate of 0.037 seem and a blocking pressure of 340 Pa when powered at 2.10 W (0.35 W per pumping channel). Based on the measured pump performance at 0.35 W per pumping channel of a standalone pump and the measured flow resistances (Table 5), the flow vs. pressure of the Knudsen pump and flow resistances of the sampling and separation paths are plotted, in which the intercept between the Knudsen pump performance line at 2.10 W and the two flow resistance lines indicated the expected flow rates to be 0.01 1 seem for sampling and 0.008 seem for separation as seen in Figure 19.

[0111]

Table 5: Measured flow resistance of monoGSA 3 components

[0112] Microfabricated preconcentrators typically use packed sorbent particles or sorptive films to collect chemicals during sampling and thermally desorb the chemicals in the subsequent analysis. In this monoGSA 3 implementation, the preconcentrator is a 0.48 pL fluidic chamber with a heater and a thermistor. The heater is coated with a 4 pm thick polydimethylsiloxane (PDMS) as the sorptive film. PDMS is selected because it has been previously used for preconcentration, and because the same material can also be used in the separation column and the detector to decreases the fabrication complexity. Other polymers, such as Tenax TA, may also be used to enhance the preconcentration.

[0113] In this implementation, the preconcentrator is heated to 130^QC to desorb chemicals during separation. This temperature is much higher than the operating temperature of the other pGC components. Therefore, the aforementioned thermal crosstalk is a critical issue to be addressed. Proven methods of improving thermal dissipation and mitigating thermal crosstalk, such as adding thermal isolation cutouts, adding heatsinks and heat pipes, and adding a fan, are implemented to reduce thermal crosstalk during preconcentrator desorption. An aluminum heat sink is placed above and a copper-water heat pipe is placed below the monoGSA chip for thermal dissipation. Screws are used to hold the heat sink, monoGSA chip, and heat pipe in place and apply pressure on thermal interface pads to provide a conformal contact between the different components. A fan is placed matching the orientation of the heat sink fins to provide airflow and increase the amount of dissipation provided by the heatsink.

The transient thermal response of the preconcentrator desorption is simulated by FEA using COMSOL Multiphysics®. The model incorporates the designed dimensions of the monoGSA 3 chip, incorporates the surrounding heat dissipation components, and applies Joule heating at the preconcentrator heater. Critical parameter values used in the model are listed in Table 6. In the simulation, the heat sink fins and the far end of the heat pipe are set at 35°C and 30°C, respectively, which are the experimentally measured values. A 26.5 V voltage pulse is applied to the heater during 2-20 seconds. The resulting temperatures of the preconcentrator and the column are averaged over their designed areas. As shown in the simulation results of Figure 20, the preconcentrator is heated rapidly to 1 1 1 °C during 2-5 s and further to 124°C during 5-20 s, which are sufficient for desorption. The column temperature remains below 60°C during the desorption and falls quickly below 40°C within 5 s afterward. The other pGC components underwent similar temperature profiles. As evident from the simulation results, the thermal dissipation from the heat sink and heat pipe allowed sufficiently high temperatures for thermal desorption while maintaining low levels of thermal crosstalk to the rest of the chip.

Table 6: Material properties and thicknesses assumed in the FEA simulation of preconcentrator desorption

[0114] The separation column is a serpentine flow channel with 200 pm width, 40 pm height, 4.5 cm length, and coated with 1 pm thick PDMS stationary phase. The relatively small column cross-section dimensions are selected to mitigate the impact of the relatively small flow rate from the Knudsen pump (Fig. 21 ). At a given flow rate, a smaller cross-section increases the flow velocity toward the optimal separation condition of the column. Although the smaller column cross-section causes a higher flow resistance and hence requires a higher pressure head from the Knudsen pump, this requirement is suited to the Knudsen pump design in this implementation. Additionally, the relatively small column height improves separation, which mitigates the separation performance loss from the small column length, which is limited by the chip footprint.

[0115] A capacitive detector is implemented in the monoGSA 3 because of its advantages in simplicity and structural compatibility with the preconcentrator and column. The capacitive detector is a fluidic chamber that incorporates interdigitated electrodes coated with a sorptive polymer layer. In this implementation, the interdigitated electrodes have 2 pm tine width and gap over a sensing area of 1 .28 mm² and are coated with 0.4 pm thick PDMS. This capacitive detector configuration provides a nominal capacitance (Co) reading of around 7 pF. The electrodes are on top of a 2 pm thick oxide layer, which allows a portion of the electric field lines to traverse through and penetrate into the device silicon, forming parasitic capacitance. The impact of parasitic capacitance can be addressed by the readout circuit, which includes a capacitive-to-digital converter (#AD7746, Analog Devices, Norwood, MA, USA) that provides a high tolerance to parasitic capacitances between the sensing electrodes and the ground. Therefore, the device silicon layer must be electrically grounded rather than floating; this grounding is provided via a pad of Ti silicide that forms a low resistance connection to the device silicon layer. The capacitive-to-digital converter, when configured properly and operated in the differential sensing mode, has been confirmed in our separate internal tests to achieve an RMS noise as low as 0.013 fF for the practical detector capacitance range (on the order of 10 pF).

[0116] Referring to Figure 22A-22I, the monoGSA 3 chip is a stack of two fused silica dies that sandwich a silicon-on-insulator (SOI) die. Eight lithography masks are used to form all the pGC components with the SOI die processed using six and fused silica die processed using two masks. The SOI die has a 12 pm device silicon, 0.38 pm buried oxide, and 525 pm handle layer, while fused silica dies have a thickness of 675 pm. The process starts with plasma-enhanced chemical vapor deposition (PECVD) of a 2 pm thick oxide layer on top of the device silicon (Fig. 22A). The deposited oxide insulates the metal traces formed in the subsequent steps from the device silicon and provides the temperature gradient for Knudsen pumps. A section of this oxide is patterned and wet etched to allow access to the device silicon for silicide formation (Fig. 22B). Titanium is deposited onto the exposed device silicon area using the lift-off process and evaporation. The deposited titanium is then annealed using rapid thermal processing to 775^QC to form titanium silicide (Fig. 22C). The first metal layer of 30/100 nm Ti/Pt (Fig. 22D) and a second metal layer of 30/200 nm Ti/Ni (Fig. 22E) are patterned using the third and fourth masks and deposited through evaporation. Front side deep reactive ion etching (DRIE) is performed to form the pumping channels of the Knudsen pumps and through holes to route the gas flow based on the fifth mask (Fig. 22F). Backside DRIE is then performed to etch 30x200 pm² through-holes through the handle layer to form channels for gas flow (Fig. 22G). PDMS for the preconcentrator, separation column, and capacitive detector are then deposited and crosslinked (Fig. 22H). Two fused silica dies have 40 pm and 500 pm deep channels patterned by sandblasting using the seventh and eighth masks (performed by IKONICS® Corporation, MN, USA). The 40 pm deep channels are used in all parts of the chip except for the gas ports after the Knudsen pumps, which are 500 pm deep to allow capillary tube attachment. The fused silica dies are bonded to the SOI chips (Fig. 22I) using a low outgassing epoxy (Epotek-377, Epoxy Technology Inc., MA, USA). The monoGSA 3 chip footprint is 15x15 mm². [0117] To facilitate testing and evaluation, fused silica capillary tubes of 250 pm inner diameter are attached at the gas ports of each Knudsen pump using an epoxy (#Stycast2850FT, Henkel, Germany) for facile connections to a sample source and test setup. Depending on actual application scenarios, these capillary tubes are not always necessary.

[0118] In addition to the monoGSA chip, electronic interface and software controls are required to interface with and control the gas chromatography system. This involves routing the electrical connections to control the heating and reading thermistor and capacitive detector values. The monoGSA chip is placed onto a dedicated daughterboard PCB, which also contains the thermal management measures, and connected to a motherboard, which contains the control, power supply, and readout electronics. A microcontroller contains software is used to control the system operation.

[0119] In Figure 23, the electrical connection between the PCB and the monoGSA chip is made using spring-loaded pins. Alternative connection methods such as soldering, wire bonding, or conductive epoxy are possible. For soldering, the solder material and temperature profiles must be carefully selected to prevent shorting between wire pads and the device silicon and to prevent delamination of the wire pads from the chip. Wire bonding creates a fragile connection. Conductive epoxy can cause shorts if applied incorrectly. Compared to these options, spring-loaded pins bypasses the use of conductive silver epoxy or soldering, which reduces assembly cost and ensures a consistent contact with tolerance of mechanical shocks.

[0120] A 3D printed housing is used to align the monoGSA chip to the spring- loaded pins. A heat sink is used to press the monoGSA chip down towards the PCB. This arrangement ensures electrical contact between the metal pads of the monoGSA chip and the spring-loaded pins. This arrangement also ensures a good thermal contact between the monoGSA chip and the cooling elements.

[0121] The daughterboard is electrically interfaced with a microcontroller ^Raspberry Pi 3 B+, Raspberry Pi Foundation, Cambridge, UK) through a dedicated motherboard PCB, which included analog-to-digital converters (ADCs), capacitance-to- digital converters (CDCs), relays, and other power electronics. The microcontroller communicates with ADCs and CDCs via inter-integrated circuit (l²C) protocols for temperature sensing and uses general-purpose input and output (GPIO) pins and l²C to control the preconcentrator heating and Knudsen pump actuation through either relays or buck converters. Relays controlled using pulse width modulation (PWM) may cause large voltage ripples depending on the relay and PWM frequency. Because Knudsen pumps require very low voltage fluctuations, a well-designed buck converter circuit is preferred for Knudsen pump heating.

[0122] An implementation of the software package that incorporates two separate parts is developed to control the operation of the gas chromatography system. The first part includes hardware control and data readout implemented on the microcontroller. The data readout includes the readout of temperatures and detector signal and the control voltage applied for heating different components. The Raspberry Pi provides 26 general-purpose input/output (GPIO) pins, supports numerous open-source libraries in Python (in which the hardware control and data readout are developed), and includes high computation power for future data processing such as signal processing the chromatograms generated by the pGC for chemical recognition and quantification. A second part of the software is a user interface (Ul) implemented in C# on a laptop computer. The Ul receives the user-defined run method for hardware control (which includes the timing and voltages that should be applied and when to read detectors) and plots the readout data in real time. The microcontroller used a standard transmission control protocol (TCP) to communicate with a laptop-based Ul, allowing remote control of the system.

[0123] The monoGSA 3 design with robust Knudsen pumps is fabricated, components are tested, and a full system, with the monoGSA chip, electronics, and software control, is assembled. Extensive testing is performed to verify operation and investigate the humidity resistance, sampling and separation capabilities, and the repeatability. Fabrication at a commercial microfabrication foundry is also performed to demonstrate the chip’s manufacturability.

[0124] The preconcentrator is powered with 26.5 V to heat the preconcentrator area to 130^QC for 18 s and the separation flow started 2 s after the start of preconcentrator desorption as seen in Figure 24. The column temperature remained under 70^QC during the preconcentrator desorption and falls quickly to below 35^QC for the rest of the separation time period. With the cooling components, a 60^QC difference between the preconcentrator and column temperatures is observed during preconcentrator desorption. The measured temperature profile tracked with the simulated temperature profile for preconcentrator desorption, and the measured temperatures are 2% and 1 1 % higher than the simulated temperatures for the preconcentrator and column, respectively. [0125] The microfabricated column is evaluated with a benchtop GC (#Agilent 7890, Agilent, CA, USA). For this test, Port 1 is blocked with a septum, inlet pump is connected to the benchtop GC inlet, and separation pump is connected to the flame ionization detector (FID). A mixture of propylene glycol methyl ether (PGME) and propylene glycol monomethyl ether acetate (PGMEA) is injected into the benchtop GC inlet and carried by 0.020 seem N2 flow through the monoGSA chip, which is maintained at 40^QC inside the benchtop GC oven. The resultant chromatogram (Fig. 25A) shows that the short microfabricated column can separate the two injected chemicals even at the low flow rate.

[0126] The separation performance of the column can be represented by the number of theoretical plates (A/) and the height equivalent to a theoretical plate (HETP), which can be calculated from the chromatogram of an isothermal separation experiment: where t is the retention time of the chemical, to is the retention time of an unretained analyte, and PWHH is the peak width at half height, and L is the length of the column. The t and PWHH are obtained from the chromatogram while to is estimated from the flow rate and dimensions of the separation and guard column. A higher column efficiency is indicated by a lower HETP, which is affected by the average flow velocity of the carrier gas.

[0127] To characterize the HETP dependence on the flow velocity, the aforementioned test of the separation column is repeated with the carrier gas flow rate swept from 0.01 seem to 0.4 seem. As shown from the results, the optimal flow rate for PGME is 0.090-0.1 10 seem, at which the HETP is 0.33 mm as calculated from a measured fo ot 50.3 s, a measured PWHH of 1 .0 s, and calculated to of 29.1 s (Fig. 25B). The optimal flow rate for PGMEA is 0.090-0.110 seem, at which the HETP is 0.16 mm as calculated from a measured fo ot 72.3s, measured PWHH of 6.1 s, and a calculated to of 29.1 s (Fig. 25C). At the optimal flow rate, the number of plates for the 4.5 cm column is 136.4 plates for PGME and 281 .3 plates for PGMEA. Note that this optimal flow rate of the column is higher than the actual flow rate of 0.009 seem that separation pump is able to provide for separation. At the operational flow rate of the system, the number of plates is 24.6 plates for PGME and 38.25 plates for PGMEA. [0128] The measured flow rates at different power levels for the sampling pump, the inlet pump, and the separation pump are shown in Table 7, Table 8, and Table 9 respectively. Experimentally measured flow rates of the gas chromatography system matched the expected flow rates. When the sampling pump is powered by 2.10 W to provide the sampling flow and the separation pump is powered by 0.84 W to resist a flow through Port 3, the sampling flow rate is measured at 0.01 1 seem and the flow through Port 3 is measured at 0 seem. When the separation pump is powered by 2.10 W to provide the separation flow and the inlet pump is powered by 0.61 W to resist a flow through Port 2, the separation flow rate is measured at 0.009 seem and the flow through Port 2 is measured at 0 seem. The power required for the pump to resist unwanted gas flow is dependent on the flow rate in the main path and is obtained by adjusting the power applied to the pump (separation pump during sampling and inlet pump during separation) that resists flow until a flow rate is no longer measured.

Table 7: Flow rates measured for sampling pump at different power levels

Table 8: Flow rates measured for inlet pump at different power levels

Table 9: Flow rates measured for separation pump at different power levels

[0129] The use of the inlet pump during separation to prevent additional chemicals from entering the system is demonstrated using a mixture of three acetates (Fig 26). Without inlet pump operation, the baseline is elevated and chemicals peaks appeared without the sampling step, which showed that additional chemicals are sampled into the system during separation. With inlet pump operation, the baseline is free of chemical peaks, which means that additional chemicals are prevented from entering the system.

[0130] The monoGSA chip is designed to analyze reagents, products, or significant byproducts of known catalysis reactions to monitor reaction progress and catalyst health. An analysis run consisted of two steps, sampling and separation. The sampling step lasted two minutes, during which the headspace of mixtures placed in a 2 ml sample vial is sampled through Port 2. The sample vial simulated the samples that can be extracted throughout a chemical reaction for in situ reaction monitoring. The impact of humidity, the ability of the system to monitor changes in chemical composition, and ability of the system to sample and separate a wide range of chemicals is demonstrated.

[0131] The impact of humidity on the monoGSA system is evaluated using three different relative humidity levels. The level of humidity in the ambient air of the laboratory is 15-30% relative humidity at 20^QC measured using a commercial humidity and temperature sensor (#00215CA, AcuRite, Wl, USA). A higher level of humidity (65-70% relative humidity) is obtained by placing the sample vial and capillary tubes connected to the monoGSA chip ports into an enclosed container with a small amount of deionized water, allowing the humidity inside the container to equilibrate, and measuring the humidity with the same commercial humidity sensor. A small amount of water is injected into the sample vial to achieve 100% relative humidity in the sampled chemical mixture. The capacitive detector response of the tested chemicals matched for all three relative humidity levels and water peaks are not observed.

[0132] The ability of the system to monitor the change in mixture composition is characterized using mixtures of hexene and heptanal and mixtures of PGME and PGME. During a chemical reaction, the percentage of the reagent decreases, whereas the percentage of the product increases. Such a change in concentration is emulated using sample vials containing different concentrations of the reagent and product at 20^QC and five analysis runs are performed at each concentration to assess the repeatability of the system. The changes in peak height in the chromatograms tracked with the changes in the percentage of the reagent (hexene and PGME) and the product (heptanal and PGMEA) in the mixtures. The variation in retention time at the same concentration is ± 7.6% for hexene, ± 6.25% for heptanal, ± 6.10% for PGME, and ± 6.32% for PGMEA. The variation in retention time is below 13.5% for heptanal, PGME, and PGMEA across the different concentrations tested. For hexene, the variation in retention time at the same concentration of hexene in the mixture is below 6%, but there is a shift in retention time from mixtures with less than 50% hexene compared to neat hexene. The shift in retention time is caused by the column overload, which changed the peak shape and retention time of the hexene peak. The monoGSA 3 is also reliable in terms of the peak heights across the tests for each concentration. The peak height for tests with neat chemicals is 35.6 ± 1 .0 fF for PGME, 31 .3 ± 0.4 fF for PGMEA, 79.2 ± 2.8 fF for hexene, and 29.0 ± 1 .3 fF for heptanal which showed a variation of less than 5% in peak height. Similar levels of variation are observed for the other concentrations of chemicals.

[0133] The results are repeatable in terms of retention time and peak heights, which means the chromatograms can be used with a calibration curve to estimate the concentration of a chemical in the mixture and provide concentration information needed to monitor chemical reactions. The calibration curve plots the ratio of normalized peak heights against the concentration of a chemical in the liquid mixture. The peak height has previously been shown to be highly proportional to the concentration of the chemical for capacitive detectors and this trend continues for the monoGSA 3 system. The normalized peak height is defined as: where p is the normalized peak height, p is the peak height at the sampled concentration, and pNeat is the peak height for the neat chemical. The peak height is normalized to linearize the relationship between the peak height and the concentration of the chemical and to reduce the change uncertainty of concentration estimates across the full scale. A ratio of the peak heights is used because the concentration in the headspace sampled by the system is dependent on the vapor pressures of all chemicals in the mixture and is calculated using where Pi is the ratio of normalized peak heights, pi and >2 are the normalized peak heights for the chemicals in the mixture. The confidence interval (Cl), which provides the estimated range of the average concentration value, and the prediction interval (F7), which provides estimated range for concentration are used to create the calibration curve. The calibration curves can be used to estimate the percentage of a chemical or peak height ratio. Using the prediction intervals, the concentration of the hexene and heptanal mixture can be estimated to a range of ±6.5% and the concentration of PGME and PGMEA can be estimated to a range of ±8.5%.

[0134] The ability of the monoGSA 3 to separate a range of Kovats retention indices is evaluated using a mixture of benzene, toluene, chlorobenzene, and dichlorobenzene. These are typical chemicals involved in the hydrodealkylation reaction followed by a chlorination reaction used to form dichlorobenzene. In Figure 27, the four chemicals in mixture are observed in the chromatogram. Results of neat chemical tests corroborated the identity of each peak, and the retention times of each chemical are at their expected locations based on the chemical’s Kovats retention index. Based on this result, the monoGSA 3 is capable of separating chemicals with retention indices from as low as 600 (benzene) to as high as 1000 (dichlorobenzene). Additional chemical results showing sampling and separation are obtained for mixtures of ethyl acetate, propyl acetate, and butyl acetate (common chemicals in food monitoring applications) and mixtures of 2-methoxyethanol, o-xylene, benzaldehyde (common chemicals in infectious disease biomarker detection) are obtained, which shows the ability of the monoGSA chip for other applications such as food monitoring and infectious disease marker detection as shown in Figures 28A and 28B.

[0135] This monoGSA 3 design with robust Knudsen pumps is also fabricated in a commercial microfabrication foundry. There are minimal changes in the layout and fabrication process used by the commercial facility to fabricate the monoGSA chip in order to demonstrate the manufacturability of the system. The changes include using different thicknesses for the deposited heater metal traces, using a SOI wafer with different thicknesses, a split of the front DRIE step into two separate steps in order to fabricate features with different sizes. These changes are made to match the fabrication capabilities of the foundry and are superficial. The commercial foundry is able to successfully fabricate the monoGSA chips with a high yield. The chips have also demonstrated to have functional Knudsen pumps with comparable pumping performance and a comparable separation performance.

[0136] Based on the performance of the fabricated chips, additional improvements can be made to monoGSA designs in order to increase the microsystem performance. Improvements to thermal isolation and individual components can be implemented to increase the separation performance. Additionally, the use of different types of Knudsen pumps

[0137] A monoGSA 4 design builds upon the monoGSA 3 design with robust Knudsen pumps. It maintains the robustness and manufacturability while also improving the system performance by improving thermal isolation between components. Compared to the other monoGSA implementations, the following innovations are made: improved thermal isolation by adding additional thermal cutouts between components; increased pump performance by increasing the number of Knudsen pump channels to increase the operational flow rate closer to the optimal flow rate; increased separation column length and increased capacitive detector sensing area to improve system performance; and process flow changed to allow for wafer level bonding of the bottom insulating cover to decrease the assembly cost and maintain structural integrity of the chip during fabrication.

[0138] With reference to Figures 29A and 29B, the monoGSA 4 chip is a single chip composed of a silicon-on-insulator (SOI) die between two glass dies that contains the Knudsen pumps, preconcentrator, column, and detector. System performance is improved by increasing thermal isolation between components, increasing the separation and detection performance of the column and detector, increasing the preconcentration factor of the preconcentrator, and increasing the performance of the Knudsen pump.

[0139] Thermal isolation is added to thermally isolate different components to increase performance and improve usage of the chip area. The thermal isolation between components reduces the impact of thermal crosstalk of heating the preconcentrator, column, and Knudsen pumps and decreases the power required for heating. The improved isolation is especially important for the column area as it allows the column to be independently heated for temperature programming. Additionally, adding thermal isolation allows the length of the column to be increased as less space is required between different components.

[0140] The preconcentration factor of the preconcentrator is related to the surface area covered by the polymer coating. By increasing the surface area of the preconcentrator with the introduction of micropillars, the preconcentration factor of the system can be increased, thereby improving the limit of detection for the system.

[0141] The robust Knudsen pump designs are used to maintain the yield and manufacturability of the chip. The thermal isolation and thermal management measures allow for a higher number of pumping channels to be connected in parallel for the Knudsen pumps to increase the achievable flow rate, increasing the separation flow rate towards the optimal flow rate for the column geometry.

[0142] The fabrication process is improved by changing the fabrication process to allow for wafer level bonding of the bottom insulating cover. There are several different wafer bonding technologies and two specific technologies, solid liquid interdiffusion (SLID) and anodic bonding are selected for attachment of the bottom insulating cover. The support of the back insulating cover enables the use of thermal isolation as it provides additional support for the chip. Additionally, SOI wafers with thinner handle silicon layers can also be used with the additional support, further increasing the thermal isolation between components.

[0143] Thermal isolation takes the form of a grid of cutouts that goes entirely through the SOI wafer. Silicon material in the SOI wafer is the main route for thermal transfer due to the high thermal conductivity of silicon and the thickness of the wafer. To reduce the thermal crosstalk between components, the amount of silicon in the area between components needs to be reduced.

[0144] The inclusion of thermal isolation benefits the performance of all monoGSA components. The preconcentrator requires thermal isolation to reduce the thermal mass that needs to be heated up during thermal desorption. The separation column requires thermal isolation to reduce the impact of temperature programming on other components. The detector requires thermal isolation as its capacitance measurements are impacted by temperature, and a stable temperature is required to reduce baseline drifts and other artifacts. Knudsen pump performance is improved as the impact on cold side temperature from the heating of other components is reduced.

[0145] Etching away silicon to form thermal isolation creates challenges for the structural integrity of the chip during fabrication, dicing, and assembly. Additionally, the fabrication of thermal isolation needs to be compatible with the existing fabrication process and features. Therefore, a grid of long thermal cutouts is used for thermal isolation to balance isolation performance and structural integrity of the chip and wafer level bonding of the bottom insulating cover is performed after the creation of the thermal isolation cutouts to provide structural support during subsequent fabrication, dicing, and assembly steps.

[0146] The transient thermal response of the preconcentrator desorption with the thermal isolation is simulated using finite element analysis using COMSOL Multiphysics® software (COMSOL Inc., Stockholm, Sweden). The main goal of the simulation is to evaluate the effectiveness of an area with thermal cutouts. The simulation couples the electrical and thermal responses to model both the effects of joule heating on the preconcentrator heater and the temperature spread when heating for a short period of time.

[0147] The model is based on the preconcentrator heating simulation presented above. A preconcentrator with the same materials and thermal management measures of thermal interface pad, heat sink, and heat pipe in shown in Figure 30A. In Figure 30B, the thermal isolation cutouts are modeled as a grid of cutouts with the important dimensions being the length, width, and gap of the thermal cutout. The thermal simulations investigated different dimensions for the thermal cutout trenches, widths for the thermal isolation area, and the addition of additional backside DRIE trenches underneath the component. Heating is provided in the simulations as a total power level of 10 W applied over the preconcentrator area for 20 seconds and the temperature difference between the preconcentrator area and the other components is used to measure the effectiveness of the thermal isolation.

[0148] Without any thermal isolation, the preconcentrator heats up to 93.7^QC, whereas the column and capacitive detector heated up to 47.3 ^QC and 41.0^QC, respectively. Adding a 1.5 mm thermal isolation region increased the temperature difference between the preconcentrator and the column. Results show that thermal isolation is improved by: increasing length of the thermal cutout and reducing the gap between cutouts, increasing the thermal isolation area width, and adding additional DRIE cutouts underneath the components. Additional methods, such as reducing the handle layer silicon thickness, may also improve the thermal isolation but may be incompatible with the fabrication process flow. The most effective thermal isolation configuration has a thermal cutout width of 30 pm, length of 400 pm, and gap of 30 pm and has additional DRIE cutouts underneath the preconcentrator. The thermal isolation increases the temperature that the preconcentrator reaches when the same power is applied for heating and also decreases the impact of preconcentrator heating on other components. The preconcentrator reached 134.8^QC while the column and capacitive detector heated up to 42.2^QC and 39.1 ^QC respectively, a 100% increase in the temperature difference between the preconcentrator and column compared to the case without thermal isolation.

[0149] The performance of the column and the detector can be improved for monoGSA 4 by increasing the length of the column and the sensing area of the capacitive detector, respectively.

[0150] The column length in monoGSA 3 is 4.5 cm with a cross-sectional area of 40 pm by 200 pm and is limited by a combination of different factors including, the flow provided by the pumps, footprint of the chip, fabrication limitations, and thermal crosstalk. The separation performance of the column can be represented by the number of theoretical plates (A/), which is directly related to the length of the separation column. The column can be lengthened to improve separation. The performance of the Knudsen pumps must still be considered when increasing the column length, but with some pump configuration changes listed in the later section, the column length can be increased to 6 cm with further increases possible with additional improvements to the Knudsen pump. Additionally, with the inclusion of thermal isolation cutouts in this implementation of the monoGSA 4, the column can be temperature programmed to optimize the separation to reduce the separation time.

[0151] The sensitivity of the capacitive detector to different chemicals is related to the sensing area. The monoGSA 3 capacitive detector has a conservative design of 2 pm tine width and gap over a sensing area of 1 .28 mm² to ensure manufacturability. These dimensions resulted in the capacitance detector having a capacitance value of roughly 7 pF. Given that the CDC chip used to perform capacitance readings (#AD7746, Analog Devices, MA, USA) is capable of measuring capacitances of 17±4 pF, the capacitive detector sensing area can be increased by a factor of more than 2.4 to achieve a detector sensing area of 3.1 mm², which provides roughly 17 pF. [0152] The preconcentrator is a chamber coated with polymer coating. Increasing the surface area of the preconcentrator with the use of micropillars increases the amount of analyte collected during sampling, which subsequently increases the concentration of chemicals injected into the column and improves improve the detection limit of the system. The impact of different possible micropillar dimensions are listed in Table 10. Assuming the same preconcentrator footprint of 12 mm², a 60% increase in the available surface area is possible with a preconcentrator depth of 100 pm and a micropillar width of 30 pm, length of 240pm, and gap of 100 pm. The actual dimensions of the pillars need to be determined by the micromachining capabilities used to machine the top insulating cover and care in the preconcentrator design has to be taken to ensure that airflow containing the analytes to be sampled will reach all coated surfaces.

Table 10: Surface area available for sorbent polymer coating of the preconcentrator with different micropillar configurations assuming a preconcentrator footprint of 12 mm².

[0153] The three Knudsen pumps in the monoGSA 4 design generate the gas flow and prevent unwanted gas flow from entering the chip. This implementation of the gas chromatography system uses the more robust Knudsen pump design. The column has an optimal separation flow rate of 0.09-0.10 seem based on injection tests and increasing the performance of Knudsen pump to achieve a higher separation flow rate closer to the optimal flow rate will increase the separation performance of the monoGSA system. [0154] In order to increase the separation flow rate towards the optimal flow rate, the number of pumping channels in parallel are increased from 6 to 9 for both sampling pump and separation pump. The increase in pumping channels for separation pump increases the maximum flow rate. The increase in pumping channels for sampling pump decreases the flow resistance contributed by the sampling pump to the separation flow path.

[0155] The temperature distribution of the Knudsen pump with increased pumping channels is simulated with finite element analysis (FEA) in COMSOL Multiphysics®. The chip is seated on a thermal interface pad, which provided an interface to thermal management measures of a heat pipe and heat sink that have been added to the simulation. The material properties and thicknesses assumed in this FEA is same as for the monoGSA 3 implementation. The results of the 9 channel Knudsen pump are shown in Table 1 1 and demonstrated that there is an increase of 21 % in the blocking pressure and 52% in the maximum flow rate with the increase in the number of pumping channels.

Table 1 1 : The simulated results of the Knudsen pumps with different number of pumping channels in parallel. W_p Is the power per channel, Th._pand T_c.p are the hot and cold end temperatures of the pumping channel, P_eq is the blocking pressure, and Qmax is the maximum flow rate.

[0156] Based on the estimated pump performance and flow resistances, the flow vs. pressure of the Knudsen pump and flow resistances of the sampling and separation paths are plotted, in which the intercept between the Knudsen pump performance line and the two flow resistance lines indicated the expected flow rates to be 0.024 seem for sampling and 0.013 seem for separation. The increase in separation flow rate from 0.008 seem to 0.013 seem improve the HETP from 1.82 mm to 1.15 mm, a decrease of 37%. Combined with the increase in separation performance from the increased column length, the separation performance is expected to have a total increase of 106% in the number of theoretical plates (based on Eq. 3.5). Further changes to the Knudsen pump design, such as by increasing the thickness of the top oxide layer or changing the pumping channel dimensions, can also be implemented to improve the separation performance.

[0157] With reference to Figures 31 A-31 L, the monoGSA 4 chip remains a stack of three dies, two insulating covers that sandwich a silicon-on-insulator (SOI) die. The top insulating cover is made from fused silica to reduce the number of chemical active sites in the separation column, whereas the bottom insulating cover is made from a borosilicate glass to create a better match between the coefficient of thermal expansion of silicon and glass during fabrication. Two different wafer bonding methods are proposed for bonding the bottom insulating cover to the SOI layer, solid liquid interdiffusion (SLID) and anodic bonding. SLID involves depositing thin layers of different metals onto wafers and using the deposited metal layers to create a robust hermetic seal. Anodic bonding has typically involved bonding of a silicon waver with a glass wafer with the application of a voltage, but it has also been shown to bond for SOI and glass wafers and glass. No matter the wafer bonding process, the use of wafer level bonding decreases the assembly complexity of the monoGSA 4 design.

[0158] The fabrication process starts with plasma-enhanced chemical vapor deposition (PECVD) of a 2 pm thick oxide layer on top of the device silicon (Fig. 31 A). The deposited oxide insulates the metal traces formed in the subsequent steps from the device silicon and provides the temperature gradient for Knudsen pumps. A section of this oxide is patterned and etched to allow access to the device silicon for silicide formation (Fig. 31 B). Titanium is deposited onto the exposed device silicon area using the lift-off process and evaporation. The deposited titanium is then annealed using rapid thermal processing to 775^QC to form titanium silicide (Fig. 31 C). The first metal layer of 30/100 nm Ti/Pt (Fig. 31 D) and a second metal layer of 30/200 nm Ti/Ni (Fig. 31 E) are patterned using the third and fourth masks and deposited through evaporation. Front side deep reactive ion etching (DRIE) is performed to form the narrow pumping channels of the Knudsen pumps based on the fifth mask (Fig. 31 F) and a subsequent frontside DRIE is performed to form the wider trenches for through-holes to route flow and to form thermal isolation based on the sixth mask (Fig. 31 G). It may be advisable to temporarily attach a carrier wafer on to the top side of the wafer to ensure the structural integrity during the subsequent processes.

[0159] At this part of the process, different steps are performed for the two different methods (SLID and anodic wafer bonding) of bonding the bottom insulating cover. For SLID wafer bonding, a metal layer of Ti/Ni/Au is deposited on the handle silicon (Fig. 31 H). Backside DRIE and reactive ion etch (RIE) is then performed to etch through-holes through the handle layer to form channels for gas flow and thermal isolation (Fig. 31 1). On a borosilicate glass wafer, flow channels are etched using either sandblasting or buffered hydrogen fluoride and a coating of Ti/Ni/Sn/Au is deposited onto the borosilicate glass wafer (Fig. 31 J). SLID wafer bonding is performed to bond the SOI wafer and the glass wafer and then dicing is performed to singulate each device (Fig. 31 K). A top insulating cover is fabricated using fused silica flow channels patterned by micromachining. The fused silica top cover is bonded to the SOI chips using a low outgassing epoxy (Fig. 31 L).

[0160] Figures 32A-32K show the fabrication process of the monoGSA 4 design using anodic wafer bonding. Backside DRIE and reactive ion etch (RIE) is performed to etch through-holes through the handle layer to form channels for gas flow as seen in Figure 32H. On a borosilicate glass wafer, flow channels are etched using either sandblasting or buffered hydrogen fluoride (Fig. 32I). Anodic wafer bonding is performed to bond the wafer and dicing is performed to singulate each device (Fig. 32J). A top insulating cover is fabricated using fused silica flow channels patterned by micromachining. The fused silica top cover is bonded to the SOI chips using a low outgassing epoxy (Fig. 32K).

[0161] The monoGSA 5 design improves upon the monoGSA 4 design by using a combination of suspended membrane Knudsen pumps and robust Knudsen pump designs to further improve the separation performance while maintaining chip yield.

[0162] The suspended membrane Knudsen pump design provides a higher flow rate compared to the robust Knudsen pump design due to its thermal isolation and greater number of pumping channels. On the other hand, the robust Knudsen pump design has a higher level of robustness and manufacturability. A monoGSA design that incorporates both Knudsen pump designs can achieve a higher performance while maintaining a high level of manufacturability.

[0163] The gas flow in the gas chromatography system is controlled by three Knudsen pumps. In order to maintain system manufacturability while increasing the separation flow rate, only the separation pump will use a suspended membrane design. The sampling pump and the inlet pump will use the robust Knudsen pump design. The separation pump is selected to use the Knudsen pump as it is responsible for providing the separation flow. The separation flow rate in monoGSA 3 and 4 did not reach the optimal separation flow rate for the column and increases to the separation flow rate towards the optimal flow rate will provide the greatest improvement to the separation performance.

[0164] For flow rate improvement, the 0.4 x1.2 mm² suspended membrane Knudsen pump design (Knudsen pump used in monoGSA 1 ) is selected for the separation pump to increase the separation flow rate. Each pumping unit of the monoGSA 1 design provides a maximum flow rate of 1 .5 seem and blocking pressure of 770 Pa when operated at 0.6 W. As the flow rate is much higher than that of the robust KP designs for the monoGSA 3 design, the design places the pumps in series to improve the blocking pressure.

[0165] The estimated flow resistance for the separation flow is 36.3 kPa/sccm (the sampling pump using the robust pump design with 9 channels, preconcentrator, 6 cm long column, and detector). As the number of pump stages connected in series increases, the separation flow rate increases as shown in Figure 33. To reach the optimal flow rate of around 0.100 seem of the designed column geometry, five pumping stages in series is needed for separation pump. An 828% increase in separation efficiency (based on the number of theoretical plates) compared to the monoGSA 3 design is expected with the increase in separation flow rate.

[0166] Referring to Figures 34A-34K, the monoGSA 5 chip is formed from a stack of three dies, two insulating covers that sandwich a silicon-on-insulator (SOI) die. The top insulating cover is made from fused silica to reduce the number of chemical active sites in the separation column, while the bottom insulating cover is made from a borosilicate glass to create a better match between the coefficient of thermal expansion of silicon and glass during fabrication. The fabrication starts with an SOI wafer (Fig. 34A). Front side DRIE is performed to form pumping channels for the Knudsen pump (Fig. 34B). Atomic layer deposition (ALD) of AI2O3 is used to create a conformal coating over the trenches to for pumping channel sidewalls for Knudsen pumps. The trenches are then refilled with polysilicon deposited using low-pressure chemical vapor deposition (LPCVD). An upper dielectric membrane of oxide and nitride is deposited by plasma- enhanced chemical vapor deposition (PECVD) over the device silicon and patterned (Fig. 34C). The membrane should be designed to be slightly tensile to reduce the chance of fracturing of suspended membranes during fabrication and operation and also provide enough thermal isolation. Titanium is deposited onto an area with exposed device silicon using the lift-off process and evaporation. The deposited titanium is then annealed using rapid thermal processing to 775^QC to form titanium silicide (Fig. 34D). Two metallization steps occur, depositing layers of 30/100 nm Ti/Pt and 30/200 nm Ti/Ni through evaporation on the top surface of the SOI wafer (Fig. 34E). For SLID wafer bonding, a metal layer of Ti/Ni/Au is deposited on the handle silicon (Fig. 34F). Backside DRIE is then performed to etch through-holes through the handle layer to form channels for gas flow (Fig. 34G). ALD of AI2O3 is performed to protect the backside trenches and reactive ion etching of the buried oxide is performed to complete the channels. On a borosilicate glass wafer, flow channels are etched using either sandblasting or buffered hydrogen fluoride and a coating of Ti/Ni/Sn/Au is deposited onto the borosilicate glass wafer (Fig. 34H). SLID wafer bonding is performed to bond the SOI wafer and the glass wafer and then dicing is performed to singulate each device (Fig. 34I). A XeF2 vapor etch of silicon is used to release the pumping channel sidewalls and to etch away the remaining silicon in the thermal isolation area (Fig. 34J). Polymer is deposited in the preconcentrator, column, and detector areas and a top insulating cover is fabricated using fused silica flow channels patterned by micromachining. The fused silica top cover is bonded to the SOI chips using a low outgassing epoxy (Fig. 34K) to finalize the chip.

[0167] Several design considerations need to be taken into account to allow the preconcentrator to be monolithically integrated into the monoGSA. These include the thermal considerations, fabrication and assembly considerations, and the limited size.

[0168] The thermal considerations refer to the fact that monolithic integration and the high thermal conductivity of silicon used in the fabrication process limits the desorption temperature that the preconcentrator temperature can reach. A table of commonly used sorbent materials can be seen below.

Table 12: Common sorbent materials used in preconcentrators and their suitability for on-chip and off-chip preconcentrators.

Different sorbent materials require different desorption temperatures required to fully desorb collected chemicals, with more retentive sorbents requiring higher desorption temperatures. The on-chip preconcentrator is limited in the range of desorption temperatures it can support without negatively impacting the other monolithically integrated components and therefore cannot incorporate certain sorbents that are highly retentive (i.e. highly retentive carbon adsorbents such as Carboxen 1003).

[0169] Certain types of sorbent materials can be more easily incorporated in the fabrication process. Sorbent materials can come in different forms, such as coatings, particles, or other non-conventional forms such as foams. Sorbent particles are very commonly used in preconcentrators and while they can be incorporated into the on-chip preconcentrator, they do require an additional manual assembly step after the monoGSA chip is assembled where particles are packed into the preconcentrator. Sorbent particles, specifically Carbograph™ 2, has been used in monoGSA 1 and 2. Sorbent materials in the form of coatings are the most easily incorporated into the fabrication process of the monoGSA. Two examples of coatings are PDMS and Tenax TA and the use of PDMS coating for preconcentration is demonstrated in monoGSA 3. Other, more unconventional forms of sorbent materials, such as carbon nanotube foams, have been demonstrated in literature but these materials are oftentimes incompatible with the monoGSA fabrication process.

[0170] The limited size of monoGSA chip also limits the size of the on-chip preconcentrator and the amount of sorbent material that can be incorporated into the on- chip preconcentrator. The overall adsorption capacity a preconcentrator is dependent on the amount of sorbent material the preconcentrator contains. As such, the limited size of the on-chip preconcentrator also puts a limit on the preconcentrator factor that can be achieved, which ultimately impacts the limit of detection for the monoGSA system.

[0171] Due to the listed design challenges, there are some use cases where analytes of interest cannot be sampled with the monoGSA's on-chip preconcentrator. To improve the preconcentration of chemicals, an off-chip preconcentrator that is not limited by the previously mentioned design challenges can be attached to the monoGSA system at Port 1 next to sampling pump (Fig. 35). In this context, off-chip means that the preconcentrator is not monolithically integrated with the monoGSA chip, but the off-chip preconcentrator can still be a microfabricated chip. This off-chip preconcentrator provides additional adsorption during sampling and desorbs the sampling into the existing on-chip preconcentrator for a subsequent sharper injection during separation, allowing for an increase in the overall limit of detection for the system. As the off-chip preconcentrator is not monolithically integrated, the off-chip preconcentrator can be designed to use different types of stronger sorbent materials to target a wide range of analytes and a larger internal volume to increase its adsorption capacity without impacting the performance of the monoGSA system. However, the off-chip preconcentrator should have a low flow resistance to minimize the impact on sampling and separation flow rates for the whole system.

[0172] The foregoing description of the embodiments has been provided for purposes of illustration and description. It is not intended to be exhaustive or to limit the disclosure. Individual elements or features of a particular embodiment are generally not limited to that particular embodiment, but, where applicable, are interchangeable and can be used in a selected embodiment, even if not specifically shown or described. The same may also be varied in many ways. Such variations are not to be regarded as a departure from the disclosure, and all such modifications are intended to be included within the scope of the disclosure.

[0173] The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting. As used herein, the singular forms "a,” "an," and "the" may be intended to include the plural forms as well, unless the context clearly indicates otherwise. The terms "comprises," "comprising," “including,” and “having,” are inclusive and therefore specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. The method steps, processes, and operations described herein are not to be construed as necessarily requiring their performance in the particular order discussed or illustrated, unless specifically identified as an order of performance. It is also to be understood that additional or alternative steps may be employed.

[0174] When an element or layer is referred to as being "on," “engaged to,” "connected to," or "coupled to" another element or layer, it may be directly on, engaged, connected or coupled to the other element or layer, or intervening elements or layers may be present. In contrast, when an element is referred to as being "directly on," “directly engaged to,” "directly connected to," or "directly coupled to" another element or layer, there may be no intervening elements or layers present. Other words used to describe the relationship between elements should be interpreted in a like fashion (e.g., “between” versus “directly between,” “adjacent” versus “directly adjacent,” etc.). As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.

[0175] Although the terms first, second, third, etc. may be used herein to describe various elements, components, regions, layers and/or sections, these elements, components, regions, layers and/or sections should not be limited by these terms. These terms may be only used to distinguish one element, component, region, layer or section from another region, layer or section. Terms such as “first,” “second,” and other numerical terms when used herein do not imply a sequence or order unless clearly indicated by the context. Thus, a first element, component, region, layer or section discussed below could be termed a second element, component, region, layer or section without departing from the teachings of the example embodiments.

[0176] Spatially relative terms, such as “inner,” “outer,” "beneath," "below," "lower," "above," "upper," and the like, may be used herein for ease of description to describe one element or feature's relationship to another element(s) or feature(s) as illustrated in the figures. Spatially relative terms may be intended to encompass different orientations of the device in use or operation in addition to the orientation depicted in the figures. For example, if the device in the figures is turned over, elements described as "below" or "beneath" other elements or features would then be oriented "above" the other elements or features. Thus, the example term "below" can encompass both an orientation of above and below. The device may be otherwise oriented (rotated 90 degrees or at other orientations) and the spatially relative descriptors used herein interpreted accordingly.

Claims

CLAIMS What is claimed is:

1 . A gas chromatography system, comprising: a separation column having an inlet and an outlet and disposed in a separation flow path, wherein the separation column incorporates a stationary phase material that provides chromatographic separation of molecules and the separation flow path includes an inlet port and an outlet port, such that the inlet port of the separation flow path is fluidly couped to a sample source; a detector disposed in the separation flow path at the outlet of the separation column; and a separation pump fluidly coupled between the detector and an outlet port of the separation flow path, wherein the separation column, the detector, and the separation pump are monolithically integrated onto a substrate.

2. The gas chromatography system of claim 1 further comprises a preconcentrator having an inlet, a chamber, and an exhaust, where the inlet of the preconcentrator is configured to receive a sample and pass the sample through the chamber to the exhaust of the preconcentrator, and the chamber contains a sorbent material therein to trap analytes.

3. The gas chromatography system of claim 2 wherein the preconcentrator resides off the substrate.

4. The gas chromatography system of claim 1 wherein the detector is further defined as a capacitive detector and the separation pump is further defined as Knudsen pump.

5. The gas chromatography system of claim 1 further comprises an inlet pump fluidly coupled between a sampling inlet port and the inlet of the preconcentrator; and a sampling pump fluidly coupled between the exhaust of the preconcentrator and a sampling outlet port; thereby defining a sampling flow path between the sampling inlet port and the sampling outlet port.

6. The gas chromatography system of claim 5 wherein the sampling pump operates to draw the sample into the preconcentrator during a sampling phase, and the separation pump operates to pull gas through the separation flow path towards the detector during a separation phase.

7. The gas chromatography system of claim 5 further comprises a controller operably coupled to each of the inlet pump, the sampling pump, and the separation pump, wherein the controller activates the sampling pump and deactivates the inlet pump during the sampling phase, and the controller activates the separation pump and deactivates the sampling pump during the separation phase.

8. The gas chromatography system of claim 7 wherein the controller controls the separation pump to prevent gas from entering the separation flow path during the sampling phase, and the controller controls the inlet pump to prevent gas from entering the sampling flow path during the separation phase.

9. The gas chromatography system of claim 5 wherein the inlet pump, the sampling pump and the separation pump are further defined as Knudsen pumps.

10. The gas chromatography system of claim 5 wherein the inlet pump, the preconcentrator, the sampling pump, the separation column, the detector, and the separation pump are monolithically integrated onto the substrate.

11 . The gas chromatography system of claim 5 further comprises an off-chip preconcentrator having an inlet fluidly coupled to the sampling outlet port and residing off the substrate.

12. The gas chromatography system of claim 5 further comprises a flow meter disposed in the separation flow path, where the flow meter includes a heater element and a temperature sensor positioned downstream in the separation flow path from the heater element.

13. A gas chromatography system, comprising: a preconcentrator having an inlet configured to receive a sample from a sampling inlet port and having an exhaust fluidly coupled to a sampling outlet port, thereby defining a sampling flow path between the sampling inlet port and the sampling outlet port, wherein the preconcentrator includes a chamber and sorbent material within the chamber; an inlet pump fluidly coupled between the sampling inlet port and the inlet of the preconcentrator; a sampling pump fluidly coupled between the outlet of the preconcentrator and the sampling outlet port; a separation column disposed in a separation flow path and having an inlet fluidly coupled to the sampling flow path between the preconcentrator and the inlet pump; a detector disposed in the separation flow path at an outlet of the separation column; a separation pump fluidly coupled between the detector and a separation outlet port; and a controller operably coupled to each of the inlet pump, the sampling pump, and the separation pump.

14. The gas chromatography system of claim 13 wherein the sampling pump operates to draw the sample into the preconcentrator during a sampling phase, and the separation pump operates to pull gas through the separation flow path towards the detector during a separation phase.

15. The gas chromatography system of claim 13 wherein the controller activates the sampling pump and deactivates the inlet pump during the sampling phase, and the controller activates the separation pump and deactivates the sampling pump during the separation phase.

16. The gas chromatography system of claim 14 wherein the controller controls the separation pump to prevent gas from entering the separation flow path during the sampling phase, and the controller controls the inlet pump to prevent gas from entering the sampling flow path during the separation phase.

17. The gas chromatography system of claim 12 wherein the detector is further defined as a capacitive detector.

18. The gas chromatography system of claim 12 wherein the inlet pump, the sampling pump and the separation pump are further defined as Knudsen pumps.

19. The gas chromatography system of claim 12 wherein the separation column, the detector, and the separation pump are monolithically integrated onto a substrate.