From a systems perspective, a microscale probe for recording is the front-end sensor component in a comprehensive neural probe system that also includes packaging, an electronics interface, signal computation and storage (Kipke 2004). A stimulating probe is the front-end actuation component in a similar system. While various types of microelectrode arrays are functionally similar in terms of basic neural recording (or stimulation) requirements, on a deeper level the devices are differentiated on the basis of (i) primary design parameters (e.g., the number, spatial arrangement, and size of recording sites), (ii) materials, (iii) fabrication and assembly techniques, (iv) packaging, and (v) extensibility for integration of advanced components (e.g., bioactive coatings and/or electronics). In our work in the CNCT, we have found that in many cases the issue at hand is not to improve the sensor function of the probe per se, but rather to develop new packaging or designs that better meet application requirements.

Microwire arrays were the first type of multi-site probes developed for neural recording and stimulation and remain in use today. There have been many different types of microwire arrays developed for many different types of experiments (Salcman 1976; McNaughton 1983; Legendy 1984; Liu 1999; Williams, J. C. 1999; Kralik 2001; Ulbert 2001). In general, the various approaches have the common attribute of forming an array of sensors by assembling the microwires into patterned bundles. The precision and consistency of the bundle varies with the intricacy of the assembly process. The recording sites are the exposed tips of the insulated microwires. The microwire diameter ranges from approximately 12 µm up to 100 µm depending on application requirements. Some microwire arrays allow for moveable electrodes using micropositioners in order to localize or track neurons (McNaughton 1983). Tetrode microwire arrays consist of four microwires tightly twisted such that the wire tips are very closely spaced in order to improve spike discrimination (McNaughton 1983; Gray, C.M. 1995). 

Although arrays of bundled metal microelectrodes have proven to be useful for studying neural circuits, they do have significant limitations. Perhaps foremost is that the limited design space does not lend itself to the type of reconfigurable engineered probe system that is needed to effectively support diverse experimental requirements. For example, the exact geometrical configuration of hand-built arrays is not reproducible. Additionally, there are significant limitations on the size and layout of recording/stimulation sites. An interest in increasing the number of recording sites without increasing the volume of the array spurred the onset of a number of attempts to create a microelectrode array using high-precision photolithographic techniques employed in the microelectronics industry. Using these methods, a single recording site can be made as small as the tip of the smallest wire electrode. The microelectrode shank, the portion that supports the recording sites and displaces the tissue, can also carry multiple recording sites and can be at least as small as a single-wire electrode. In general, these methods are the same as those used to create integrated circuits and therefore utilize similar substrate, conductor, and insulating materials. Fabrication typically starts on a wafer substrate and the electrode features are added using a number of photolithographically patterned thin-film layers that are defined by etching. These methods are attractive because they result in highly reproducible, batch-processed devices that have features defined to ±1 µm. Many of the techniques are compatible with the inclusion of on-chip circuitry, which is an important characteristic because, as the density of sites increases, so does the number of interconnect leads and packaging complexity. Multiplexing becomes imperative as the lead count approaches 32 or more, especially if the electrode is to be chronically implanted. In addition, on-chip buffering and amplification can reduce crosstalk, noise coupling and signal attenuation, improving the overall signal-to-noise ratio (Csicsvari 2003).

The origin of photoengraved probes extends back to 1969 when Prof. Kensall Wise, then at Stanford, reported the development of the first silicon neural probe (Wise 1969; Wise 1970). It seemed like a natural fit of technology to applications at the time, an idea that has largely been realized through sustained and focused efforts over the last 35 years. Many materials and approaches have been considered and reduced to practice at Michigan(BeMent 1986; Najafi 1986; Ji 1991; Hetke 1994; Hoogerwerf 1994; Bai 2000; Bai 2001; Vetter 2003) (Figure B.2) and elsewhere (Prohaska 1977; May 1979; Prohaska 1979; Kuperstein 1981; Reitbock 1983; Takahashi 1984; Barth 1985; Campbell 1991; Jones 1992; Kovacs 1994; Nordhausen 1996; Kewley 1997; Williams, J. C. 1999; Bragin 2000; Burmeister 2000; Rousche, P. J. 2001). Silicon recording micoelectrodes were recently reviewed by Hetke et al (Hetke 2002). 

Several groups have made notable progress developing silicon probes for recording and stimulation. Prof. Richard Normann and his colleagues at the University of Utah have developed a microelectrode array with a high density of penetrating shafts, referred to as the Utah Electrode Array, or UEA (Campbell 1991; Jones 1992; Nordhausen 1996). Each shaft is 1 to 1.5 mm long and projects down from a 0.2-mm-thick glass/silicon composite base. The device is formed from a monocrystalline block of silicon using a diamond dicing saw and chemical sharpening. The resulting silicon shafts are electrically isolated from one another with a glass frit and from the surrounding tissue with deposited polyimide, silicon nitride, or parylene. The tip-most 50 to 100 µm of each shaft is coated with platinum to form the electrode site. Arrays have been fabricated with up to 100 (10x10) penetrating shafts that are typically spaced on 400 µm centers. Interconnection to the electrode sites is accomplished by bonding either individual, insulated 25-µm wires or a multilead polyimide ribbon cable to bond pads on the top of the array. While the shafts are sharp, they are dense and can cause significant tissue dimpling during normal insertion. A high-velocity insertion technique using a pneumatic device was developed to alleviate this problem (Rousche, P.J. 1992). Although the UEA was originally designed and has been successfully used for acute and chronic recording in cat cortex(Maynard 1997; Rousche, P.J. 1998), modified designs have now been developed for cat peripheral nerve (Branner 2000; Branner 2001) and other structures. A variety of acute and chronic versions of the device, along with instrumentation associated with their use, are available commercially from Bionic Technologies (www.bionictech.com).

Another relatively new set of research efforts is directed toward development of planar neural electrodes fabricated using SOI wafers (Ensell 2000; Hofmann 2002; Cheung 2003). SOI wafers are manufactured with an oxide layer buried a specified distance below the top silicon surface (6 to 25 µm for these neural devices). This oxide layer provides an etch-stop that accurately defines the final thickness of the electrodes. Metal conductors and electrode sites are defined photolithographically and are insulated by deposited layers of silicon nitride and silicon dioxide. Deep reactive ion etching (DRIE) is used to etch through the entire thickness of the device from the front, stopping on the buried oxide layer. The backside of the wafer is then patterned and etched through using DRIE to the other side of the buried oxide layer. The electrodes are finally released by etching the buried oxide in buffered HF. The final thickness of the shanks of these devices is determined by the depth of the buried oxide layer. Shanks from 6 to 25 µm have been fabricated. Bond pad regions are typically left at full wafer thickness (>500 µm). Single and multiple shank acute designs have been fabricated with up to 32 sites and have been used successfully for acute recordings. In anticipation of chronic use, one group has added a process step to include surface topology on the devices in an effort to improve mechanical anchoring in the tissue (Cheung 2003). In addition to being compatible with the inclusion of on-chip circuitry, this type of fabrication offers the advantage that all of the process steps are compatible with current foundry techniques.

While there have been interesting developments in microelectrode technologies, the Michigan probe technology stands out because of its established design rules, well-characterized fabrication process, and available packaging (recently reviewed by Wise et al. (Wise 2004)). Michigan probes are the standard by which most other scientific neural probes are judged. Michigan probes now constitute a platform technology that enables the systematic development of engineered probe systems for diverse neuroscience applications. This is substantiated by the impact that Michigan probes have made in many areas of neuroscience (discussed in Section C-Progress Report). 

This design flexibility becomes very important in the transition from acute experimental preparations (anesthetized animals for hours) to chronic preparations (typically awake and behaving animals for days to several weeks and beyond).   Each type of experiment has its particular set of requirements for probe systems, which may include moveable probes, closely spaced sites, or fluid delivery. In many cases, the transduction function of the probe itself is not an issue; the challenge lies in packaging or handling or interfacing with the device. An engineered probe system such as the Michigan platform is indicated in order to adequately and efficiently address these diverse application needs.   This extension of existing technologies to new applications is the first objective of the proposal.