A HYBRID BACTERIA AND MICROPARTICLE DETECTION PLATFORM ON A CMOS CHIP: DESIGN, SIMULATION AND TESTING CONSIDERATIONS Zhao Lu1, Jaouad El-Fouladi 1, Sylvain Marte1 1 and Yvon Savaria 2 1
NanoRobotics Laboratory, Department of Computer and Software Engineering, 2 Department of Electric Engineering, École Polytechnique de Montréal, Campus of the University of Montréal, Montréal, QC, CANAD
Abstract- This paper presents a hybrid bacteria and microparticles detection platform based on a CMOS technology. Vertical face to face microelectrode arrays are implemented onto CMOS chips by connecting the metal and via layers together. A CMOS post-processing procedure based on Deep Reactive Ion Etching (DRIE) is used to release the microelectrodes and to construct microchannels in between. With medium flow of the fluid, Bacteria and microparticles are allowed to pass through the microchannels, where impedance variations are measured using a microelectrode pair on the wall, and then detected by electronic circuits embedded on the same chip. This microelectronic/microfluidic hybrid system targets screening individual bacterium or microparticle with high throughput and accuracy. The system architecture is presented first, followed by the detailed model, design, simulation and parameters of the prototype. The CMOS post-processing, specific packaging and testing procedures are also introduced in this paper. Finite element analysis method (FEM) and circuit simulations confirm that a single microparticle, 5 ȝm in diameter, can be detected by the proposed microsystem. Based on preliminary etching results, pairs of released electrodes 10 ȝm *2 ȝm *8 ȝm (L x W x H), also contribute to validate the feasibility of the CMOS postprocessing procedure.
Detecting bacteria or microparticles is becoming more and more important in the field of biology, pharmaceutical industry, bio-medicine and anti-bio-terrorism. In food- or air-borne disease control, early detection of single bacterium rather than later detection of larger concentrations is critical for disease control. The current most widely adopted technology for this type of task, named flow cytometry , is based on fluorescent reactions when the targeted particles pass through the detecting area. Due to the required light source and the complexity of its detector, fluorescent based flow cytometry with parallel detection is very difficult to achieve. Furthermore, the targeted particles or cells have to be prepared, generally by coating them with a fluorescent label before detection. It limits the application of this technology, and it also increases the overall detection time. Thus the throughput is low and the screening speed is relatively slow. Another conventional bacteria detection technology is based on electrochemical sensors [2~5], also referred to as amperometric or impedimetric sensors. These sensors detect changes in the electrical characteristics of the medium containing the bacterial cells. Although the electrochemical sensors offer advantages such as high sensitivity, low cost and ease of integration onto a
MEMS/NEMS device, long detection times (usually a 12 hours to 7 days process) are expected due to the long preamplification period if the initial concentration of bacteria is very low. Meanwhile, most bacteria and particles are not motile, and the diffusion rate of the bacteria and particles is very slow; especially under low Reynolds number laminar conditions. It takes a long migration time for target bacteria approaching the detection or sensing area where the electrodes are implemented. The required signal processing is another challenge for impedimetric sensors. Due to the very small impedance signal, very precise and complex signal processing circuits are needed. With the recent development of MEMS and microfluidic technology, especially bioMEMS or Lab-on-Chip, combined with conventional microelectronic technology, traditional biosensors can be integrated onto CMOS chips. Although most of these systems need some post-processing procedure or special package, their integration with a CMOS chip not only significantly reduces the fabrication cost and alleviates the requirements of signal processing, but also greatly increases the sensitivity, throughput, robustness, and reliability of this kind of system. Among the benefits expected from the miniaturization of CMOS based biosensors: we can cite a reduction in the required bio-reagent and samples volume. Another benefit is that fast detection and analysis can be performed. Moreover, high density of biosensors with multiple functions can be implemented onto a same substrate, which may greatly increase the screening speed. Indeed, on-chip microelectronic circuits make parallel signal processing and production of real-time analysis reports possible. In the last decade, different kinds of CMOS based biosensors have been presented for bacteria detection, bio-analysis and neuron activity monitoring [6-8]. For most of these biosensors, the top metal layer or metal deposited with postprocessing is used to construct coplanar electrodes arrays for electrical impedance spectroscopy. Highly integrated microelectrode arrays and on-chip detection have also been reported. However, confined by the size of the microelectrode and the planar orientation, the sensitivity of these devices is not sufficient to detect bacteria with very low concentration. Meanwhile, the microelectrodes in these biosensors have to be coated with some noble and biocompatible material such as Gold and Platinum, which not only increases cost, but also requires complicated microfabrication procedures. In this paper, we present a CMOS based microfluidic/microelectronic system for faster single
bacterium or microparticle detection. Metal and Via layers of CMOS chip are used to construct vertical microelectrode arrays. The microchannel between a pair of microelectrodes allows a liquid medium containing the bacteria or microparticles to flow from a microchamber on the surface of a CMOS chip through the substrate down to the other side of the chip. On-chip microelectronic circuits monitor the impedance variations in the channel, when the measured impedance is beyond some pre-defined threshold, the existence of the target bacteria will be indicated and counted. The real-time results can be collected and output to a computer through I/O ports on the same chip. This system aims at parallel, fast single bacterium detection with high accuracy and screening speed. By re-configuring the sensing circuits, this microsystem can also be used for microparticle detection, bio-analysis, cell-culture and so on. In this paper, the architecture and working principle of this system is first presented. Then, its detailed design and simulation results are introduced. The CMOS-postprocessing and related testing and packaging issues are also discussed. Finally, the optimization, design consideration, and potential applications of this system are also introduced. II.
A. High Level Overview The proposed microelectrical/microfluidic hybrid microsystem  consists of a microelectrode array, microelectronic circuits, a top microchamber, an array of microchannels, and outlets. As shown in Fig 1, the microelectronic circuits and microelectrodes array are implemented directly on a CMOS chip using standard 0.18 ȝm CMOS technology. After the microchannels are created using a post-processing procedure based on DRIE technology, the microchamber and outlets are constructed with epoxy as explained in the sequel.
passing between a pair of microelectrodes, the change of impedance will be measured by sensing circuits integrated on the CMOS chip. The presence of a bacterium will be identified when the measured impedance is beyond some predetermined threshold. Because the microchannels go through the substrate, a small pressure difference between the top microchamber and the bottom outlet will force the fluid medium to pass through the device. The waste sample is collected on the back side of the device. The flow rate has to be calibrated according to the dynamic response (frequency) of the sensing circuits. The sample can be injected by either top outlet shown in Fig 1. When sample injection is performed through one outlet, the other outlet is blocked. Consequently, the liquid sample can only flow through the device by the microchannels. Both top outlets are functional only when a microchamber cleaning procedure is required. A very thin layer of silicon oxide is left during the post-processing procedure, to protect the microelectrodes from corrosion/erosion. Such a layer is biocompatible. B. Specific parameters of a first implementation A cut of the proposed electrode structure that we are implementing is illustrated in Fig 2. The vertical microelectrode structure is created by stacking the Metal and Via layers of a 0.18 ȝm standard CMOS chip.
b. Fig. 2. a) Cross-sectional view. b) 3-D view of the microchannel and microelectrode based on a typical 0.18ȝm CMOS technology.
Fig. 1. Schematic of the proposed hybrid microsystem.
The proposed system works as follows. First, a liquid medium containing the bacteria is injected into the microchamber above the microelectrode array. Then, the medium enters into the microchannels. Controlled electrical currents are injected through the microelectrodes embedded on the wall of the microchannels. If there is a bacterium
The height of the microelectrodes is around 8.05 ȝm after connecting 6 Metal layers and 5 Via layers. The microelectrodes are defined by stacking metal rectangles. For instance, in a typical design, 10 ȝm x 2 ȝm (L x W) stacked metal rectangles are separated by 10 ȝm. In order to avoid the potential cross-talk among pairs of microelectrodes, the distance between the microelectrode pairs is kept at 80 ȝm. A total of 80 microelectrode pairs are implemented on the first prototype. The die has a total area
of around 2700 ȝm x 1300 ȝm (W x H). Metal wires connect the microelectrodes to adjacent sensing circuits. The silicon oxide between each pair of microelectrodes is removed during the post-processing procedure using DRIEICP dry etching technology. A through-wafer microchannel is created between each electrode pair by etching through the silicon substrate. After the post-processing procedure, the microfluidic component is covered with transparent, dielectric polymer, to protect bonding pads and wires on the top of the CMOS chip and to avoid electrical shorts while providing outlets for test sample injection. The planned height of the microchamber is 1 mm, which provides a volume of 15 ȝL for each test. microchamber. The bottom microchamber used to collect the waste sample is fabricated separately and bonded to the bottom of the chip. The microchamber is constructed with Polydimethylsiloxane (PDMS) using conventional soft lithography technology. III. DESIGN AND SIMULATION In this section we present some design constraints of the sensing microelectronic circuit. It is designed to be robust and flexible to meet various potential applications related to the proposed microsystem. As this experimental system is designed to prove concepts in spite of a great deal of uncertainty, in our first prototype, different sizes of microelectrodes are implemented. This will be invaluable for evaluating the performance of the circuit and to guide us toward the selection of the best parameters that will allow to meet requirements of different applications. Specifically, 5 µm x 5 µm, 6 µm x 6 µm, 8 µm x 8 µm, 10 µm x 10 µm, 12 µm x 12 µm, 15 µm x 15 µm and 20 µm x 20 µm (microelectrode length x microchannel width) are chosen for this design. In a first considered design of the stimulation and sensing circuits, the fluid medium containing the bacteria/microparticles to be detected is assumed to be an electrolyte having a conductivity ranging from 0.5 to 5 S/m. We also consider that the bacteria/microparticles to be detected are essentially non-conducting, with a conductivity much lower than that of the electrolyte. Based on available information, the expected conductivity can be around 0.1 pS/m. Based on various studies with COMSOLTM finite element models of the physical structure, combined with detailed circuit simulations, the preferred solution that was retained for our experimental system isolates the microelectrodes from the fluid media with thin dielectric layers. To form these layers, the electrodes are coated with a thin layer of a dielectric material (either silicon oxide or Parylene) which results in the formation of a relatively high value capacitance at the interface between each microelectrode and the electrolyte. Based on material properties, microelectrode dimensions and target dielectric thickness, the value of this capacitance ranges from 2pF to 15pF. The resulting electrical model for each microelectrode pair is shown in Fig 3.
The resistance Rsol reflects the finite conductivity of the electrolyte. The value of Rsol is greater when a nonconductive bacterium/microparticle is passing through the microchannel between the electrodes. The C_ox capacitors are due to the thin dielectric layers on the surface of the electrodes. For a given electrode pair, as their geometry is the same and the means of producing the dielectric is equivalent, both capacitance values should be almost equal. However, some slight variations on these values should be considered. They can be caused by variations of the process of the microelectrodes that creates the structure. With our previous observations, we can conclude that it is very important for the circuit we design to be robust to many uncertainties related to the design parameters of this experimental device. For instance, we expect variations of the conductivity of the electrolyte due to biological activities. As mentioned earlier, the electrolyte conductivity is ranging from 0.5 to 5 S/m. The nature and dimensions of the bacteria/microparticles to detect can also change. For instance, one may want to detect polythene beads of 3 µm, 5 µm or 6 µm diameter or bacterium with diameters that can range from 1 µm to 3 µm. At last, variations of the parameters controlling microelectrodes fabrication and coating could have a significant incidence. These considerations are capital issues in order to get a working circuit. To design the detection circuit, we took advantage of the presence of the relatively large capacitance C_ox that can integrate a DC current. In that case, the resulting circuit can reduce to Fig 4, for which Equations (1), (2) and (3) apply.
Fig. 4. The equivalent circuit when injecting a DC current into a microelectrode pair.
Thus if we inject a reference current “I”, the voltage across a microelectrode pair is given by equation (3). It is a linear relationship with a slope inversely proportional to C_ox, and a value at the origin is directly proportional to Rsol. If, for instance, C_ox is a constant, a change in Rsol results in a vertical translation of the voltage created across an electrode pair.
(1) (2) (3)
Fig. 3. Electrical model for each microelectrode pair.
Therefore the conceptual diagram of Fig 5 is proposed to model the sensing mechanism associated with each electrode pair.
Fig. 5. Conceptual diagram of the sensing mechanism.
When the Charge switch is on, the Discharge switch is forced off (mutually exclusive control) and the reference current is injected in the upper microelectrode, generating a voltage modeled by equation (3). This voltage is fed to the input of a buffer with a threshold voltage set to Vdd/2. After some time, the Charge switch is turned off and The Discharge switch is turned on. The input of the buffer is then grounded, and both capacitances are gradually discharged. After a suitable time, we can restart this process. As a result a pulse train is created at the output of the circuit and the width of the pulses composing this train is related to the value of Rsol. The greater is Rsol, the wider are the pulses. Hence by analyzing this wave, the system can automatically determine when a bacterium or microparticle passed by. The actual circuit that implements this detection concept is shown in Fig 6. The reference current is provided from outside the microchip and it is injected into the detection circuit using a simple current mirror. This is necessary as the design of the circuit proceeds in parallel with the development of the post-processing microfabrication steps, and different applications may require very different stimulation currents. Ultimately, in a fixed application, with well characterized post-processing steps targeting some low-cost system, the current source would be on-chip. At this stage, we expect that fully external control of the injected current is essential for characterization tests.
The Charge switch is implemented using a PMOS transistor and the Discharge switch is implemented using a NMOS transistor. Thus, as mutually exclusive conduction is desired, only one control signal is needed for both switches. The buffer is easily implemented with two simple CMOS inverters. In order to simulate the behaviour of this circuit, we initially carried out simulations using COMSOL software in order to find the expected range of Rsol. For these simulations, we used non-conducting spherical beads of 5µm in diameter. The lengths of the electrodes were 6µm, 10µm and 20µm and the widths of the channels were of 6µm, 10µm and 20µm respectively (square channel section assumed).The microbead is located at the centroid of the volume defined by the microelectrodes pair. The conductivity of the electrolyte is 1 S/m. Table 1 presents the values of resistances found. It is obvious that smaller microelectrodes with narrower channels provide better discrimination. When the channel is much larger than the microbead, more reference current flows around the microbead, thus reducing the sensitivity of the circuits. TABLE I
FEM SIMULATION RESULTS FOR Rsol WITH VARIOUS MICROELECTRODES, AND CHANNEL SIZES. Length of Microelectrode (µ m)
Rsol with a bead (kȍ)
Rsol without a bead (kȍ)
Discrimination ratio (%)
Fig. 7. Simulation results of delay time according to the impedance variations between the microelectrodes.
Based on those expected resistance values, it is possible to simulate the behaviour of the designed circuit. Fig 7 shows a typical simulation result for a C_ox value of 5pF, a charging time of 1.5µs, a discharging of 50 ns, and a reference current of 1µA. The figure shows two simulations Fig. 6. CMOS stimulus generation and detection circuit.
results. One is for a value of Rsol equal to 100k and the other is for a value of 140k. This simulation results confirm the expected linear relationship described by equation (3) as well as the pulse wave generated. It also shows that a large Rsol produces a larger pulse width. Table 2 shows widths of the pulses in the train for a value of C_ox of 8pF, a charging time of 3.5µ s, a discharging time 0.5µs, a reference current of 1µA and Rsol values ranging from70k to 150k. This table confirms the linear relation between the value of Rsol and the pulse width.
photoresist on top of the CMOS chip to provide an additional protection layer. A typical photolithographic process is used for this purpose. Then, the DRIE technology is adopted to remove the silicon oxide between a pair of microelectrodes and etch through the silicon substrate. In Fig.8, an optical microscopy image illustrates part of the microelectrode array and the scanning electron microscope (SEM) image that depicts a pair of released microelectrodes. Recall that the microchannel is to be etched through the dies in between these electrodes.
TABLE II WIDTH OF PULSE FOR VARIOUS Rsol VALUES Rsol (k)
Pulses width (ns)
To characterize the impact of possible process variations, we performed a set of simulations with a C_ox value halved. The variation of the pulse width with Rsol decreased. In general, as the value of C_ox increases, the slope of the voltage generated at the input of the inverters is smaller. A larger time is then needed to reach the threshold of the inverters in the proposed microelectronic circuit. The circuit proposed was designed in a manner which makes it flexible in multiple ways. Firstly, one can control the injected current depending on the values of the impedance to measure. It is then possible to vary the charging and discharging time for the capacitors C_ox. Note that the same circuit can be used for a wide range of C_ox and Rsol, making the circuit robust against inaccurate knowledge of the device parameters and variations of these values. IV. FABRICATION A. CMOS Post-processing The CMOS post-processing procedure includes two steps, passivation layer patterning and DRIE etching. On the die received from the foundry (TSMC through CMC Microsystems), the passivation layer on the microchannel area is removed. So the silicon oxide between a pair of microelectrodes is exposed and ready for the following etching process. However, the thickness of the original passivation layer is not enough to protect the surrounding circuits and microelectrodes during the DRIE dry etching procedure. Thus, the first step is to pattern a layer of
Fig. 8. Optical and SEM images of the released microelectrode and microchannel after the silicon oxide is etched off.
B. Packaging To avoid any electrical shorts caused by the liquid samples on the chip and to provide a microfluidic interface for sample injection, a specific package is developed. As illustrated in Fig 9., the raw die is first mounted on a chip carrier for wire bonding. Then a special Epoxy is patterned above the area of the microelectrode array by direct writing technology . After that, liquid polymer is poured on the surface of the CMOS chip to cover it completely, including pads, wires and previously patterned Epoxy. After the liquid polymer is fully cured, the whole device is set on a hot plate, when the temperature reaches about 70 oC, the Epoxy inside is melted and can be extracted through the outlets by applying a vacuum. In order to remove the melted Epoxy entering into microchannels during the heating procedure,
hot water is injected for cleaning the debris of Epoxy in the microchamber and microchannels. Finally, the device is installed on a dedicated PCB for testing.
Fig. 9. a) Wire bonding, b) Epoxy is patterned on the area of microelectrode array by direct writing technology. c) Polymer is poured on the surface of the CMOS chip to protect the pads and bonding wires. d) After the epoxy is melted and removed, the microchamber and outlets are formed by the surrounding transparent epoxy.
The ultimate goal of this project is to provide a portable hand-held microsystem for on-site usage. By adopting a standard CMOS technology, the task of fabricating high density array of face to face microelectrodes can be achieved easily and the on-chip detection circuit also greatly increases the signal to noise level. As it is a first prototype for validating the ideas and evaluating the performance of the circuit and feasibility of the microfabrication process, the parameters are chosen for better understanding of challenges and leaving enough space for further adjustment to meet requirements of different applications. Based on our simulations and the preliminary microfabrication results, single bacterium or microparticle can be identified by the proposed microsystem. However, there is always a balance between the performance and cost of a design. For example, the circuit simulation results suggest a narrow microchannel compatible with the size of targeting bacteria or microparticles for better sensitivity. However, producing smaller microchannels, for example, less than 10 ȝm, will significantly increase the complexity of the post-processing procedure, thus increasing cost accordingly. Meanwhile, in order to achieve higher screening speed and throughput, higher density of the microelectrodes are needed, which could raise the background noise and cross-talk of circuits and reduce the performance of the system. With the concept of design for test, this first prototype aims to develop a uniform platform for validating the performance of the circuits and microelectrodes under different conditions and providing data for further optimization. The proposed
testing scheme comprises the following steps. Firstly, testing will be performed with different kinds of medium such as electrolyte with different conductivity, deionised water, PBS, and bacteria medium. The collected data can be used not only to evaluate the function of the circuit, but more importantly, to establish a reference database. The anti-corrosion/erosion capability of microelectrodes and the long-time permeability of the package will also be checked at this stage. Then, microparticles with varied size will be used to determine the sensitivity of the system and to calibrate the circuits as well. Meanwhile, controlled medium injection will used to investigate the relationship between the flow rate of the sample injection and dynamic response of the circuit. Finally, a diluted bacteria medium will be used to demonstrate the function of the system and then a concentrated bacteria medium will be used to determine the maximum frequency/minimum response time of the system in practical operating conditions. Moreover, by collecting results corresponding to different species of bacteria, microparticles or other bio-entities, a more precise and wide database can be established for future on-site detection. By adjusting the circuits’ parameter, this microsystem can be reconfigured to fulfill detection tasks. It should be stressed that the proposed microelectronic circuit is designed to produce a high yield microchip. Indeed, its simplicity and small size make it possible to have an independent circuit for each microelectrodes pair. Thus, a bug that appears in one cell doesn’t propagate to the other cells. Moreover, the system is designed to be totally self-referenced, making it very robust against variations of any kind. This self-reference comes from the choice made to create a pulse train. By comparing the widths of the pulses independently for each cell, they become selfreferenced. The information about a problem in a cell is then taken into account in the pulse train generated. Finally, detection of a microparticle between two microelectrodes only relies on comparisons of results generated by the same cell. VI. CONCLUSION AND FINAL REMARKS This paper proposed a microfluidic/microelectronic hybrid system based on CMOS technology. On this Labon-Chip microsystem, face to face microelectrodes are constructed by stacking Metal and Via layers of a CMOS chip. CMOS post-processing with the DRIE technology is used to release the microelectrode array and form the microchannels in between. A dedicated package is developed to integrate microchambers and outlets on top of the CMOS chip and to isolate the circuits from the microfluidic component as well. The liquid medium containing bacteria or particles can be injected into the mirochamber and microchannel, where the microelectrode pairs are implemented on the wall. finite element modeling and circuit simulations were used to confirm that single bacterium or particles can be identified by the presented detection circuits. The reported preliminary etching results also partly validate the proposed post-processing procedure. This microsystem is designed to detect pathogenic bacteria or other micro-size bio-entities present in a diluted liquid media in minutes.
VII. ACKNOWLEDGEMENTS This work was initially supported by a grant from the Canadian Institute for Robotics and Intelligent Systems (IRIS). It was recently supported in part by a Canada Research Chair (CRC) in Development, Fabrication, and Validation of Micro/Nanosystems, the Canada Foundation for Innovation (CFI), the National Sciences and Engineering Council of Canada (NSERC), and the Government of Québec. The authors also thank CMC Microsystems for access to various design tools and CMOS fabrication technologies. REFERENCES  D. Huh, W. Gu, Y. Kamotani1, J. B. Grotberg and S. Takayama, “Microfluidics for flow cytometric analysis of cells and particles, “ Physiol. Meas., vol. 26, pp. 73-98, 2005.  M. R. Stephen, C. A. Evangelyn, “Design and Fabrication of a Microimpedance Biosensor for Bacterial Detection”, IEEE Sensor Journal. vol.4, no.4, pp. 434-440, 2004.  K.Cheung, S. Gawad, and P.Renaud, “Impedance spectroscopy flow cytometry: On-chip label-free cell differentiation,” Cytometry Part A, vol. 65A, pp. 124-132, 2005.  R. Gomez-Sjoberg, D.T. Morisette, and R. Bashir, “Impedance microbiology-on-a-chip: Microfluidic bioprocessor for rapid detection of bacterial metabolism,” Journal of Microelectromechanical Systems, vol. 14, no. 4, Aug. 2005  S.M Radke and E.C. Alocilja, “Design and fabrication of a microimpedance biosensor for bacterial detection,” IEEE Sensors Journal, vol. 4, no. 4, Aug. 2004  B. Eversmann, et al, “A 128 x128 CMOS Biosensor Array for Extracellular Recording of Neural Activity”, IEEE Journal of Solidstate Circuits, vol. 38, no. 12, pp.2306-2317, 2003.  Y, L, E. Smela, N.M.Nelson, and P.Abshire, “Cell-lab on a chip: a CMOS-based Microsystem for Culturing and Monitoring Cells”, Proceedings of the 26th Annual International Conference of the IEEE EMBS San Francisco, CA, USA, September 1-5, 2004.  F. Heer, W. Franks, A. Blau, S. Taschini, C. Ziegler, A. Hierlemann, and H. Baltes, “CMOS microelectrode array for the monitoring of electrogenic cells,” Biosens. Bioelectron, vol. 20, pp.358–366, 2004.  Z. Lu, R. Denomme, and S. Martel, “Micro/Nanoparticle Detection: An Impedimetric Microsensor Based on CMOS Technology”, The 7th IEEE International Conference on Nanotechnology (IEEE-NANO), Hong Kong, China, Aug. 2-5, 2007  D. Therriault, S. White and J. Lewis, “Chaotic Mixing in ThreeDimensional Microvascular Networks Fabricated by Direct-Write Assembly,” Nature Materials, vol. 2, pp. 265-271, 2003.