The development of microfabricated cantilevers (MC) for atomic force microscopy (AFM) signified an important milestone in establishing approaches to micro-electro-mechanical systems sensors. MC sensors inherited not only the idea of a microfabricated AFM probe, but also the elegant "optical lever" read-out scheme of an AFM instrument, in which the reflection from a laser beam focused on an MC tip is projected onto a position-sensitive photodetector (PSD).1 As the cantilever bends, the projected spot slides along the PSD axis. Although several other readout principles can be used in MC sensors, the optical lever scheme provides an efficient way to convert MC bending into an electronic signal. A more recent and advanced spinoff of this principle implemented by our group involves an array of vertical-cavity surface-emitting lasers (VCSELs) and a charge-coupled device (CCD) or complementary metal oxide semiconductor (CMOS) camera as a read detector sensitive to both position and intensity of the light reflected by a cantilever array (see figure 1).
Figure 1: Multi-cantilever sensor with an advanced version of the optical lever read-out uses an array of VCSELs for the source and a CCD or CMOS camera as a read detector sensor.
MC-based sensors involve measurements of nanoscale deflections, nanomechanical resonances, and, in some cases, Q-factors. The general idea behind the sensors is that physical, chemical, or biological stimuli can affect mechanical characteristics of the micro-mechanical transducer in such a way that the resulting change can be measured using optical, electronic, and other methods.2 Depending on the measured parameter - cantilever deformation or resonance frequency - we can class the mode of sensor operation as static or dynamic. Each of these modes, in turn, can be associated with different transduction scenarios.
MC deformations can be caused by external forces exerted on the cantilever or by intrinsic stresses generated on or within the cantilever. The variety of transduction scenarios also stems from the fact that a stimulus to be detected may affect the mechanical state of the transducer directly, or it may undergo one or several transformations before the measured mechanical parameter of the transducer is affected. MC sensors have practical appeal because they provide very sensitive detection of chemical and biochemical species as a result of changes in intrinsic stresses or mass-loading effects associated with interfacial (adsorption-desorption) processes.The Basics
MC transducers operating in dynamic mode are essentially mechanical oscillators, the resonance characteristics of which depend upon the attached mass and the visco-elastic properties of the medium; for instance, adsorption of analyte molecules increases the suspended mass of the resonator, lowering its resonance frequency.3 We can readily observe resonant behavior of MCs using excitation in alternated electric, electromagnetic, or acoustic fields.
The minute sizes and mass of MCs make them susceptible to thermo-mechanical noise. Because thermal energy provides measurable excitation of any MC at room temperature, the sensors may operate in the resonant mode with or without external excitation. Externally excited MCs have detected adsorbates with masses in the femtogram (10-15 g) and, more recently, attogram (10-18) range.4, 5 Our studies have shown that a low-power (1 to 5 mW) focused beam of a modulatable diode laser provides an excellent excitation source for very small bimaterial MCs with resonances in the radio-frequency range. Dimensional scaling of MCs is associated with respective scaling of their mass, frequency, and energy content, so we can fabricate cantilevers with fundamental frequencies ranging from hundreds of megahertz to several gigahertz. Such intertial-mass detectors theoretically could achieve mass sensitivity limits close to a zeptogram (10-21 g).
Figure 2: Mechanisms of analyte-induced stresses in different types of responsive coatings include surfaces expanding as a result of adsorptive processes (top), coatings swelling upon absorption of extra analyte (middle), and analytes binding to nanostructured surfaces (bottom).
For static-mode chemical sensing, we modify the MC so that one of its sides is relatively passive while the other exhibits affinity to the targeted analyte. Three distinctive models help explain how the bending responses of the sensors are generated in the case of different coatings. The first model focuses on surfaces expanding as a result of adsorptive processes. This model is most appropriate when interactions between the MC and its environment are purely surface phenomena, as in the chemisorption of thiol molecules on the smooth surface of a gold-coated cantilever (see figure 2).6 Such processes typically are accompanied by reduction of the interfacial stress.
In the second model, the response is a result of the swelling of the coating upon the absorption of an extra analyte. This model is applicable when a relatively thick coating, which is permeable for an analyte, is present on one side of an MC. We can quantify the effect of the swelling on cantilever bending by an effective surface stress that scales up in proportion to coating thickness. Although this implies that thicker coatings produce proportionally greater responses, stress-slip conditions can arise on smooth surfaces when the effective surface stress becomes comparable to substrate-coating adhesion energy. This limitation can be overcome in the case of MCs with nanostructured surfaces.
Structured interfaces and coatings represent a more complex, but very promising, class of responsive phases for modification of cantilevers in chemical and biological sensor applications. Approaches include surface immobilization of gold nanospheres and chemical dealloying of co-evaporated gold/silver films.7, 8 When one side of the MC transducer is nanostructured, analyte binding can induce cantilever deflections through several additional mechanisms taken into account in the third model.
Analyte-induced deflections of cantilevers with nanostructured surfaces or coatings combine mechanisms of bulk, surface, and intersurface interactions. A combination of these mechanisms facilitates efficient conversion of the energy of receptor-analyte interactions into the mechanical energy of cantilever bending. In particular, we associate penetration of analyte species into nanoscale gaps with strong steric forces. Our recent studies demonstrated that we can obtain increases in MC responses of up to two orders of magnitude by immobilizing receptor molecules on nanostructured surfaces rather than smooth gold ones. Furthermore, nanostructured responsive phases offer an approach to substantially increasing the amount of binding sites per cantilever without compromising their accessibility for the analyte.
We used focused ion beam (FIB) milling to rapidly prototype MC transducers with customized nanomechanical and structural parameters. Beginning with standard, commercially available MC probes, we found FIB milling was sufficient to create a range of transducers with advanced sensing properties. Standard MCs can be fabricated using a combination of bulk and surface micromachining processes. In particular, we can define geometries by using reactive ion etching to photolithographically pattern a structural material such as silicon nitride (SiN) deposited on the top of a silicon wafer. We then pattern a masking layer on the backside and release the suspended cantilever structures with anisotropic bulk etch of the silicon.
Using similar process flows, we can create single-crystal silicon cantilevers with a doping of silicon or an epitaxy of a doped-silicon layer that substitutes for the deposition of a SiN layer (the p-doped silicon acts as a wet-etch stop layer). Fabrication of MCs becomes more compatible with standard CMOS technology if we exclude bulk micromachining steps and follow fabrication approaches based on the use of a sacrificial layer. A typical process flow involves depositing a sacrificial oxide layer, opening anchoring points, and depositing a structural material like a SiN layer or a polysilicon layer. The suspended structures can be patterned in the structural layer using a dry-etch process and released using a wet etch of the sacrificial layer.Analytical Applications
Although standard metal-coated silicon and SiN cantilevers show notable sensitivity to a few analytes, the modification of cantilever surfaces with appropriately designed responsive phases yields well-controlled specificity to volatile organic compounds (VOCs), ionic species, proteins, and DNA fragments. Even cantilevers coated with thin layers of generic industrial polymers like polymethylmethacrylate, polystyrene, polyurethane, or their blends show great promise as detectors of various VOCs.9, 10 By applying principal-component or artificial neural-network analysis to response patterns from arrays of such polymer-modified cantilevers, researchers were able to demonstrate the concept of an artificial nose capable of notable discriminating power in the case of alcohol mixtures and certain natural flavors. Using thin (about 100 nm) films of synthetic receptor compounds, we observed more distinctive selectivity patterns with regard to different classes of VOCs.11
The use of MC transducers in biosensors represents one of the most intriguing trends in the area of advanced analytical systems. Several research teams have identified the potential of MC sensors for biomedical applications as well as proteomics and genomics.12-14 When antibodies or small DNA fragments were immobilized on one side of a cantilever, the presence of complementary biological species produced cantilever deflections in a rather predictable manner. Based on the deflection behavior of the cantilevers modified with biological receptors, even very small mismatches in receptor-analyte complementarity could be detected. The relatively short analysis time (less than one hour), nanomolar sensitivity, and good compatibility of MC transducers with large 2-D-array formats have obvious practical implications.
Cantilever-based sensors have substantially enriched the portfolio of transduction scenarios that can be used to address future needs in high-performance miniaturized analytical systems. Currently available analytical models of MC bending use adaptations of phenomenological macroscopic models that take into account neither nanoscale morphology of MCs nor the molecular mechanism of the analyte-induced stresses. Clearly, theoretical evaluation of environmentally induced stresses in real responsive phases applicable to MC sensors suffers from the gaps between atomic, molecular-level, and continuum 2-D or 3-D models. We anticipate substantial progress in this area in the near future as a result of further advances in simulations of materials properties.
Another challenge on the way to practical MC sensors is on-chip integration of MC transducers, read-out components, and analyte delivery systems. In principle, this challenge can be addressed using already available technologies. The next few years should see a transition of MC sensors from the laboratory concept to a range of marketable devices. oe
- D. Sarid, Scanning Force Microscopy, Oxford University Press, New York, NY (1991).
- N. Lavrik, M. Sepaniak, and P. Datskos, Review of Scientific Instruments 75, p. 2229 (2004).
- K. Ekinci, Y. Yang, and M. Roukes, Journal of Applied Physics 95, p. 2682 (2004).
- N. Lavrik, P. Datskos, Applied Physics Letters 82, p. 2697 (2003).
- B. Ilic, Journal of Applied Physics 95, p. 3694 (2004).
- E. Berger, Science 276, p.2021 (1997).
- N. Lavrik et al., Biomedical Microdevices 3, p. 33 (2001).
- C. Tipple, Analytical Chemistry 74, p. 3118 (2002).
- N. Abedinov, Journal of Vacuum Science & Technology B 21, p. 2931 (2003).
- H. Lang, Analytica Chimica Acta. 393, p. 59 (1999).
- J. Headrick et al., Ultramicroscopy 97, p. 417 (2003).
- J. Fritz, Science 288, p. 316 (2000).
- P. Dutta, Analytical Chemistry 75, p. 2342 (2003).
- K. Hansen, Analytical Chemistry 73, 1567 (2001).