Microelectronic Engineering 85 (2008) 22902298

Contents lists available at ScienceDirect

Microelectronic Engineering
journal homepage: www.elsevier.com/locate/mee

Suspended channel MEMS fabrication process for advanced device design and bio-analytical functionalization
Alexander Vladimirov Grigorov
MIC Department of Micro and Nanotechnology, Technical University of Denmark, DTU-Bldg 345E, DK-2800 Kgs. Lyngby, Denmark

a r t i c l e

i n f o

a b s t r a c t
This paper reports a simple and novel process that allows the fabrication of suspended channel MEMS devices. Two basic types of suspended structures were fabricated using the new process, and a possible functionalization mechanism for bio-analytical detection and sensing using spotting and capillary action was successfully tested. Possible uses of the fabricated structures are described, as specic existing types of biochemical MEMS sensors. 2008 Elsevier B.V. All rights reserved.

Article history: Received 16 April 2008 Received in revised form 2 July 2008 Accepted 30 July 2008 Available online 27 August 2008 Keywords: MEMS Suspended channels Angled deposition Surface stress detection Resonant cantilever Mass detection

1. Introduction Suspended channel MEMS devices are not a new concept [15]. The novelty of our proposed process is that it allows the fabrication of suspended channels out of metals, such as aluminum. Previous suspended channel devices are largely fabricated out of silicon, silicon oxide or silicon nitride. The advantage of our process is that the suspended channels structures are thin-walled, which makes the fabricated devices more responsive to piezo-resistive detection of stress induced on the devices surface by biochemical processes and interactions. Thin-walled channel structures are also lighter and thus more suitable for mass measurements using resonant actuation and detection. 2. Process description A structure dened by a line on a positive lithography mask, translates into a grove developed into a polymer resist layer on a substrate. This grove already does have one of the desired shapes it is U-shaped in cross-section, i.e. it is essentially an open channel. This shape similarity was used to fabricate suspended metal channel devices. Fig. 1 illustrates the processing steps in detail. Positive resist is spun on a silicon wafer, baked, the mask pattern is dened using lithography, and the polymer resist layer is
E-mail address: cacxo@abv.bg 0167-9317/$ - see front matter 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.mee.2008.07.015

developed, leaving a pattern designed to consist of wide grooves with narrow trenches on their sides. A metal layer is deposited using EBD (electron beam deposition) at 45 angle to the substrate, by using multiple deposition steps from multiple sides. The narrow trenches parallel to the groove dening the channel are more narrow than they are tall, meaning that 45 angle depositions do not cover their side walls; the deposited metal is shadowed by one of the sides of the narrow trench and covers the opposite sidewall only partially. The bottom of such trenches remains uncovered by metal. An easy way to visualize this effect would be to imagine standing next to the leeward side of a building in heavy rain if the building is tall and the wind strong enough, one will not get wet. This geometrical effect separates the deposited metal layer along the lines of the narrow trenches. The resist layer is then dry etched using oxygen plasma to facilitate the next processing step. A exible polymer adhesive layer is then applied to the top side of the chip. The tape is then peeled-off taking the top metal layer with it; it was found that most metals adhere much better to the tape than to the resist underneath. The metal remaining on the substrate is then under-etched using reactive ion etching to release the suspended metal structures. A holder was made to facilitate subsequent angled depositions. It consists of a aluminum metal plate bent at 45 to the horizontal, to which a chip with the drawn lithography pattern is attached. The pattern was angled at 45 to the chip sides. Rotating the chip four times by 90 between depositions was used to change the deposition angle, and completely cover all groove side walls, thus making solid open channels.

A.V. Grigorov / Microelectronic Engineering 85 (2008) 22902298

2291

Fig. 1. Suspended channel structures fabrication process and holder diagram.

Fabrication of rectangular cross-section structures (closed channel) was also attempted by the introduction of an extra resist spinning, plasma-ashing and metal deposition steps. Secondary resist was spun on top of the chips after metal deposition. This layer was then plasma-ashed to reduce its thickness below the level of

the channels formed by the U-shaped cross-sectional structures. Controlling the uniformity and the thickness of the top resist layer proved to be difcult, and the yield extremely low; the channels were lled with resist, as intended, but there was also a residual thin resist layer left between the primary and secondary metal lay-

2292

A.V. Grigorov / Microelectronic Engineering 85 (2008) 22902298

ers in some places on the chip which prevented uniform peel-off. Some devices and parts of complete systems were successfully fabricated, with a yield of about 5%. With that in mind, all subsequent development into metal channel devices and systems was done using the open channel process, which is also a completely dry process; wet processing steps lower yield and reproducibility for most processes and they also destroy suspended fragile structures, especially if ultrasound is used for lift-off. It must be pointed out that shadow masking and self-shadowing are well known concepts in the MEMS community, but they have never been used in conjunction with angled metal deposition, to fabricate suspended channels. The process was standardized as follows: 1 lm wide beams and 250 nm wide separation trenches were used throughout the subsequent research. On a mask level this translates to using just two line widths 1 lm and 250 nm. The separation between the two linewidths was standardized to 250 nm. 500 nm thick layer of ZEP-520A resist were used as a primary resist layer. It was spun on 2 inch wafers at 2000 rpm, 100 rpm/s for 30 s. Baking was done at 160 C for 5 min. Electron beam writing was done at 6 nA current, 280 lc/cm2 dosage, 100 keV acceleration voltage and 10 nm steps. ZED-N50 was used to develop the written patterns for 50 s. The four subsequent metal depositions were plain aluminum layers each 50 nm thick in the direction of deposition. Up to 120 nm thick metal layers were found to produce equally good devices using the same masks, which allows for ner control of gaps/parallel beams spacing. The metal layers were deposited using electron beam deposition (EBD) at a rate of 3 nm/s. Timed plasma ashing was done using oxygen plasma at 300 scm ow rate and 400 W power for 15 min. Plasma ashing removal of the remaining resist after peel-off was performed using the same recipe for 60 min. Blue lab tape (Scotch TM) that was manually applied to the top of the chips was used as polymer adhesive layer. A standard silicon anisotropic RIE recipe was used to under-etch the suspended structures. Etching times were varied between 15 and 45 min depending on what was needed to completely under-etch and release the specic devices shape that was being fabricated. 3. Mask design specics Fig. 2 illustrates a simple four devices (whitstone bridge) system lithography mask, and shows important aspects of the mask design. 3.1. Anchoring

The anchor edges were found to be the main cause of structural bending, after under-etching both suspended structures and the edge of the structures (anchors) connecting the suspended structures to the substrate. A solution was found for this problem; the edge of the anchors was cut at regular intervals. Narrow slits were introduced at regular distance along the anchor edge. Anisotropic reactive ion etching of silicon was found to under-etch the anchors inwards from the edge, but very little through the narrow slits, so the slits effectively relieved the strain mismatch between the released anchor edge and non-released anchor middle after under-etching, but the suspended structures attachment points were not underetched through them. This allowed the fabrication of suspended beam structures that are perfectly parallel in the horizontal plane. 3.2. Yield Yield for the new peel-off fabrication method was in the order of 2050%. Peeling-off a top metal layer is an isotropic process; it has directionality with respect to the pattern. This leads to ripping of the top metal layer at stress-points such as corners, the rest of the layer remaining attached to the resist layer after the peel-off. To solve this, stress areas were stripped with an alternating pattern of narrow 250 nm lines 250 nm away from each other at the mask level. This served to under-etch those areas during the plasma-ashing step, and release them prior to peel-off. This solution raised the yield for the overall, completely fabricated micro-uidic systems to 100%. Writing solid patterns on the anchor plates was by far the greatest cause of high e-beam writing time. In order to drive the cost down a dot pattern was introduced on the anchoring plates. This reduced the cost and writing time ten fold, while still producing anchoring plates large enough for macro-sized probes electrical contact. This approach also eliminated the need for introducing a UV lithography step, in order to fabricate large at areas (anchors/electrical contact pads). 3.3. Rapid prototyping The described channel MEMS process lends itself well to rapid prototyping. Large number of devices can be fabricated on complete wafers, as long as the angle of deposition is kept constant over the whole wafer, i.e. if the deposition equipments source is sufciently far away from the deposited surface. This allows rapid prototyping where arrays of devices are fabricated on a single chip, or wafer, so that many parametric permutations and combinations are fabricated in one go. This approach was used in subsequent research into straight cantilever bridge structures. 4. Functionalization

Most metals have a thermal expansion coefcient that is higher than that of silicon; this thermal coefcient mismatch and the elevated temperatures of deposition lead to the build up of internal stress gradient across the thickness of a metal layer. After release this stress manifests itself by expanding the structures lengthwise and curving them towards the substrate. Our experience with conventional/at metal structures fabricated by lift-off processing shows that metal deposition rates affect the internal stress of the resulting metal layer; low deposition rates lead to high stresses and vice versa. Consequently, as high as possible deposition rates were used with the available equipment in our angled deposition peel-off process. This still left us with a metal layer that is somewhat compressively stressed prior to release.

Liquid ow along the channels formed by U-shaped cross-sectional structures was successfully performed using capillary action. This was done by connecting the channels formed by the suspended structures to channels on the anchors. These were extended across and outside the anchors to a reservoir where a drop of liquid was spotted using a syringe. Acetone, Isopropanol, DI water and Glycerin were all experimented with; all were found to ow along the channels after spotting; Acetone, Isopropanol and DI water ow extremely fast along those channels and took approx. 3060 s to ll all the channels and circumnavigate the whole system, glycerin ow speed was very slow in comparison, approx. 10 lm/s. Fig. 3 shows the actual micro-fabricated system, using the lithography mask in Fig. 2.

A.V. Grigorov / Microelectronic Engineering 85 (2008) 22902298

2293

Fig. 2. Sample system mask design, details and denitions.

5. Suspended channel structures 5.1. Flat spring structures Suspended structure are subjected to the same expansion (due to deposition internal stresses) after release as the expansion affecting plain uncut anchor edges. This means that it is relatively easy to fabricate meandering (Fig. 3), coiled (Fig. 4a) or mixed coiled-meandering structures (Fig. 4b) as they are effectively at springs that are capable of accommodating expansion after release, which aids shape preservation. 5.2. Straight structures Fabricating a conventional double-clamped cantilever beam proved difcult initially; the structural expansion after release makes such a device buckle in or out of plane, which means that it is impossible to reliably position such a structure next to an actuating or sensing electrode (Fig. 5).

The buckling of bridge structures suggested the solution to the problem itself. If the center point of a beam buckles after release, than this displacement can be used to stretch a beam suspended between the center points of two beams buckling outwards, i.e. a structure that has a H shape observed from above. The direction of buckling can be controlled by using beams that are angled at their center, in the direction of intended buckling [6]. Fig. 6 illustrates this novel stress-inversion mechanism, and denes its researched aspects. When compressive stress inside the structures is translated into elongation after release, a beam that is suspended between the center points of two outwards moving buckle-beams, elongates less sideways than the center point displacements of the buckle-beams. This stretches the center beam sideways, and converts compressive stress into tensile stress. This mechanism was investigated parametrically by fabricating variations of such outwards stretching buckle-beam structures. Different buckle angles, buckle-beam lengths, number of beams and beam grouping/spacing were tried.

2294

A.V. Grigorov / Microelectronic Engineering 85 (2008) 22902298

Fig. 3. Sample micro-fabricated system with four meandering suspended channel devices.

Fig. 4. Other suspended channel devices (a) coiled suspended channel device and (b) mixed suspended channel device.

Three horizontal offsets (h) were tried 2, 4 and 6 lm, and segments lengths were 20, 30 and 40 lm, i.e. nine permutations of each suspension mechanism type (four-beam, eight-beam and double four-beam) were fabricated. The parametric investigation showed that small angle (a) buckle-beam groups, that are tightly packed together were not successful. One structure type was found to perform its intended task best (double four-beam). In consists of two groups of four parallel buckle-beams that have a large buckle-beam angle and are connected by a straight beam at their central axis. This structure is different than the rest as it is more resilient to twisting in the vertical plane, having four attachment points to the anchors that are spaced apart. The beam suspended between two such devices was found to be perfectly straight and aligned with an electrode parallel to it. The parameters for this structure are 20 lm segment length, 6 lm horizontal offset, double four-beam type. Buckle-beam angle (a) is 17. This structure was used in all subsequent research into bridge cantilevers beams, either as is, or by scaling it. Scaling was done by maintaining a and beam widths, and proportionately increasing beam lengths (bridge, buckle-beam segments and beam connecting buckle-beam groups). This is not necessarily the only possible mechanism; any grouping or number of buckle-beams along a central connecting beam

that is long enough would probably work; the buckle-beams were grouped at both ends of the connecting beam for ease of fabrication and high yield this arrangement results in just two top metal layer rectangles in between the device segments, that need to be peeled-off. Fig. 7 shows the mask design implementation and the actual device.

6. Channel structures as MEMS sensing elements 6.1. Analogue sensing elements The suspended open channel structures types that have been fabricated (coiled, meandering, mixed and straight bridges suspended by stress-inversion buckle-beam mechanisms) could be used for piezo-resistive measurements, where a change in the overall structure stress is induced by a bio-molecular layer, or multiple layers, on the channel walls [710]. One can envision a whitstone bridge micro-system comprising four suspended channel devices, which are functionalized (or not) separately by using more than one reservoirs for spotting and capillary action, connected to individual devices. An individual device, in this context, does not need to be a single suspended structure. It could consist of several suspended

A.V. Grigorov / Microelectronic Engineering 85 (2008) 22902298

2295

Fig. 5. Conventional straight cantilever bridge buckling out-of-plane.

Fig. 6. Buckle-beam suspension mechanism principle and parametric investigation device specics.

structures with small anchors in between them, arranged in sequence with a single channel running across the anchors and structures. This may be needed in order to optimize an individual devices electrical resistance for best piezo-resistive measurement sensitivity; a single suspended metal structure has extremely low resistance. 6.2. Digital sensing elements This research has demonstrated that a device consisting of parallel buckle-beams (Fig. 6, double four-beam mechanism) fabricated with the new process displaces along its central axis, under the inuence of in-plane stress. The same device could possibly be used for a bio-molecular layer induced stress detection by positioning an electrode contact on the displacement axis so that the devices tip (the foremost buckle-beam central point) and electrode make electrical contact after stress induced in the channel walls has resulted in displacement along that axis; compressive stress moves such a device in the direction of buckling, tensile stress moves such a device opposite the direction of buckling. Such a device could also be used in a capillary action micro-system array, like the one described above, and would probably offer better sensitivity, as surface stress would result in a digital signal (on/ off), i.e. a biochemical interaction on the channel wall resulting in stress (and overall buckling) would trigger an electrical contact, rather than be the cause for an analogue signal. Precise digital measurements of analyte concentrations could be done using an array of those devices where the electrode-tip gap increases across

the array. In this fashion, high concentrations would result in most of the devices in the array making electrical contact (most devices would displace by a large amount), and low concentrations would result in just the smaller gap devices making electrical contact. Similarly, one could vary the devices segment lengths in an array, which would also result in different displacements for a constant analyte concentration. One and the same single channel could connect all the devices on such an array, rather than having separate reservoirs per device. It must be pointed out that this is not the only buckle-beam arrangement that can be used for motion/displacement amplication; cascaded buckle-beam devices [6] are also a useful approach to contact detection using hollow beams and stress induced on the channels inner surface. Most static electro-thermal actuator MEMS [11,12] device congurations would probably work for this purpose, as the two MEMS device families (static actuation electrothermal MEMS and the hollow suspended channel stress detection by electrical contact MEMS, that is proposed here) are essentially one and the same from an engineering stand point of view both types of devices seek to translate a beam elongation into a deection of a point of a suspended device. 6.3. Mixed digital/analogue sensing elements A mixed solution between contact detection and piezo-resistive measurements is also possible with square coil/spiral devices (Fig. 4a), where surface channel stress results in internal coil rotation in the plane of the substrate, until the coil makes electrical

2296

A.V. Grigorov / Microelectronic Engineering 85 (2008) 22902298

Fig. 7. Buckle-beam suspension mechanism mask design and actual micro-fabricated device.

contact with itself. This would lower its electrical resistance in a step fashion, i.e. such a device could also have a digital response to surface stress and be used in a parametrically varying array, as for the buckle-beam devices. 6.4. Resonant cantilever mass detection The fabrication of straight suspended channels opens the possibility for resonant cantilever mass sensing [1315], where a resonant bridge device is used to measure the density of the liquid inside it, or the mass of a solid residue after liquid evaporation from an open channel. There are two ways to actuate a double-clamped bridge device to resonance, and our device allows both. An AC signal can be applied between the electrode and cantilever and measure the current while sweeping the input frequency

actuation and detection [17]. In that case, applying an AC current through the cantilever would both actuate the cantilever and also generate an EMF current proportional to the amplitude of movement. Measuring the EMF current with respect to the AC signal frequency would also reveal the resonant peaks. The mass sensitivity of a cantilever beam sensor is proportional to its mass and inversely proportional to its resonant frequency, i.e. a cantilever beam sensor that is light and has high resonant frequency is more sensitive than one that is heavy and has low resonant frequency [18]. The resonant frequency of a cantilever is proportional to the root of its stiffness and the stiffness [19] of a cantilever beam is proportional to its second moment of area:where k is the cantilevers stiffness, m is the cantilevers mass, E is Youngs modulus of the cantilevers material, I is the second moment of area, L is the cantilevers length, q is the cantilevers material density and A is the cantilevers cross-sectional area.

range, most commonly known as electrostatic actuation and detection [16]. A DC current can be passed through the electrode which would serve to generate a radial electromagnetic eld for electromotive

The second moment of area [20] of a beam is a measure of how far away the material that makes it is, from it central axis. In other words, a hollow tube used as a mass sensor is better than a solid rod of the same mass and length.

A.V. Grigorov / Microelectronic Engineering 85 (2008) 22902298

2297

Fig. 8. Suspended open channel cross-section and second moment of area dimensions and denitions.

The second moment of area of the aluminum U-cross-section shaped double-clamped cantilevers (Fig. 8) is and their theoretical resonant frequency fres is 750 kHz, using the above for I.

One could also fabricate composite U-shaped cross-section structures using the same process, e.g. a structure that is made by four consecutive metal depositions, which is then oxidized for

By comparison, a conventional double-clamped beam of the same mass, cross-sectional area and length (300 nm thick and 300 nm wide rectangular cross-section) has a second moment of area I = Ab2/12 = 0.68 1027 m4 and a resonant frequency of fres = 80 kHz, i.e. an order of magnitude improvement in theoretical mass sensitivity has been achieved. As an added bonus, aluminum has a high electrical conductivity and can support high current densities compared to silicon and poly silicon, which aids resonant actuation and detection by external magnetic elds. The true resonant frequency of our fabricated beam bridge structures is probably raised further due to the tensile stress applied to the beams by their buckle-beam suspension mechanism, which improves mass sensitivity by an unknown amount.

electrical insulation between channel and uid, and then covered with gold by another four consecutive depositions. This would allow easy functionalization of the insides of the channel, as in an SPR chip [21], while separating the uid and the channel electrically. One could also directly functionalize the metal oxide with a bio-molecular layer, although this is probably more difcult than with a gold layer. 7.2. Functionalization outlook More than one solutions could be delivered to individual devices by consecutive spotting, i.e. a solution is spotted in a reservoir, the uid is delivered to the device via capillary action, the solution evaporates from the open channel and reservoir, then the process is repeated with a different solution. One could have an array of such micro-systems on a chip, and perform many functionalizations and measurements in parallel using an automated spotting system, as in PCR array technology. 7.3. Other MEMS elements outlook The angled deposition process also supports full integration of static channel devices such as chromatography columns, and could be used to create a fully-integrated lab-on-chip total analysis system for bio-applications, with various suspended and static devices operating in sequence or in parallel. Those could be calorimetric MEMS, chromatography columns, static suspended stress detection MEMS, dynamic/mass-sensing MEMS, etc.; a MEMS device design requires a suspended, electrically conductive structure, allowing for movement and electricity, and biological sensing requires a way to deliver bio-agents and molecules (a solution) to it this can be achieved using U-shaped channels and capillary action. 7.4. System outlook Capillary action only works while the liquid channels are dry; once full, the ow would stop. One way to alleviate this problem

7. Outlook 7.1. Process outlook The angled deposition peel-off fabrication process is material independent as it uses a geometrical self-shadowing effect. Hence, this process could be used to fabricate complex structures out of any material that can be deposited directionally, or any materials combination in a laminar composite, by using multiple angled depositions. The internal stresses arising at the fabrication stage can be dealt with using careful engineering design devices composed of high second moment of area beams, buckle-beam suspension, self-relaxing devices, stripped anchors, etc. The process is also substrate independent, as long as the substrate and resist layer can be etched separately. So, one could envision the same structures fabricated on a glass substrate, which is cheaper than silicon. E-beam lithography is not necessary for implementing this process. Any process capable of producing narrow trenches in a resist layer (compared to the linewidth of intended channels) would be sufcient, i.e. one could also use laser lithography, UV lithography, X-ray lithography, nanoimprint, etc.

2298

A.V. Grigorov / Microelectronic Engineering 85 (2008) 22902298 [4] D. Sparks, S. Massoud-Ansari, M. Straayer, R. Smith, R. Schneider, J. Cripe, Y. Zhang, in: Proceedings of Sensors Expo Fall 2002, 2002, p. 171. [5] Y. Zhang, S. Tadigadapa, N. Naja, Transducers01 (2001) 1460. [6] L. Que, J.-S. Park, Y.B. Gianchandani, Bent-beam electro-thermal actuators for high force applications, in: Micro Electro Mechanical Systems, 1999. MEMS99. Twelfth IEEE International Conference, 1999, pp. 3136. [7] G.Y. Chen, T. Thundat, E.A. Wachter, R.J. Warmack, Source, J. Appl. Phys. 77 (8) (1995) 36183623. [8] Y. Arntz, J.D. Seelig, H.P. Lang, J. Zhang, P. Hunziker, J.P. Ramseyer, E. Meyer, M. Hegner, Ch. Gerber, Nanotechnology 14 (2003) 8690. [9] F. Huber, M. Hegner, C. Gerber, H.J. Guntherodt, H.P. Lang, Biosens. Bioelectron. 21 (8) (2006) 15991605. [10] A. Boisen, J. Thaysen, H. Jensenius, O. Hansen, Ultramicroscopy 82 (14) (2000) 1116. [11] E.S. Kolesar, P.B. Allen, J.T. Howard, J.M. Wilken, N. Boydston, Thin Solid Films 355356 (1999) 295302. [12] A.V. Grigorov, Z.J. Davis, P. Rasmussen, A. Boisen, Microelectron. Eng. 7374 (2004) 881886. [13] K.S. Hwang, J.H. Lee, J. Park, J.H. Park, T.S. Kim, Miniaturized chemical and biochemical analysis systems, in: 7th International Conference, 2003, pp. 12671269. [14] B. Ilic, H.G. Craighead, S. Krylov, W. Senaratne, C. Ober, P. Neuzil, J. Appl. Phys. 95 (7) (2004) 36943703. [15] G. Abadal, Z.J. Davis, B. Helbo, X. Borrise, R. Ruiz, A. Boisen, F. Campabadal, J. Esteve, E. Figueras, F. Perez-Murano, N. Barniol, Nanotechnology 12 (2) (2001) 100104. [16] W.C. Tang, T.C.H. Nguyen, M.W. Judy, R.T. Howe, Sensor. Actuat. A 2123 (1990) 328331. [17] A.N. Cleland, M.L. Roukes, Appl. Phys. Lett. 69 (18) (2002) 22532255. [18] N.V. Lavrik, Michael J. Sepaniak, Panos G. Datskos, Rev. Sci. Instruments 75 (7) (2004) 22292253. [19] T. Lamminmaki, K. Ruokonen, I. Tittonen, T. Manila, O. Jaakola, A. Oja, H. Seppa, P. Seppala, J. Kiihamaki, Electromechanical analysis of micromechanical SOIfabricated RF resonators, in: 2000 International Conference on Modeling and Simulation of Microsystems MSM 2000, 2000, pp. 217220. [20] A.M. Howatson, P.G. Lund, J.D. Todd, Engineering Tables and Data, second ed., Chapman & Hall 1991, ISBN 0-412-38970-3. [21] J.M. McDonnell, Curr. Opin. Chem. Biol. (2001) 572577.

would be to use an enclosed channels system (rectangular crosssection channels) with a uid reservoir at one end and an open area evaporator at the other end of the system. Capillary action may not be sufcient to drive a liquid through such a system quickly enough for fast assays, and capillary action depends on the channels top layer contact angle with respect to the uid. To solve this problem one could also integrate one or more suspended MEMS pumping devices, that consist of one or more channels, and have a reciprocating motion in the axis of the channel(s), but whose channels have isotropic resistance to uid ow (connected arrowheads channel form as seen from above). One could use electrostatic actuator to move such a device along its axis. Electro-osmotically driven ow is also a possibility. One could also use a reciprocating motion MEMS heat-engine to drive the pumping element(s). Such an engine could use an enzymatic reaction, or incorporate a secondary uidic system for a high energy density uid such as hydrazine which is catalyzed within a platinum covered channel. Such a system need not have specic applications. A MEMS device that switches the uid ow between separate devices could possibly make it universally applicable to a wide range of biomolecular assays. The switching action could also be provided by specic bio-molecular agents (such as DNA/RNA in the liquid solution under study), rather than externally. References
[1] T.P. Burg, S.R. Manalis, Appl. Phys. Lett. 83 (13) (2003) 26982700. [2] M.B. Stern, M.W. Geis, J.E. Curtin, J. Vac. Sci. Technol. B (Microelectron Nanometer Struct.) 15 (1997) 28872891. [3] P. Enoksson, G. Stemme, E. Stemme, Sensor. Actuat. A 46 (1995) 327.