Silicon-LIGA Arrays

Figure 7. (a) Low-density neuronal monolayer culture composed of 76 neurons growing over a matrix of 64 electrodes. The recording craters are spaced 40 um laterally and 200 um between rows. The transparent indium tin oxide conductors are 10 um wide. Tissue is mouse spinal cord; culture age is 27 days in vitro; histology is Loots-modified Bodian stain. (From Ref. 60, their Fig. 2, p. 284.) (b) Cultured hippocampal neurons on patterned self-assembled monolayers. A hybrid substrate pattern of trimethyloxysilyl propyldiethylenetriamine (DETA) and perfluorated alkylsi- lane (13F) showing selective adhesion and excellent retention of the neurites to the DETA regions of the pattern. (From Ref. 6, their Fig. 4, p. 18.)

An alternative, batch-oriented, and larger-scale way to fabri­cate multielectrode needle-shaped devices is to combine sili­con technology with the LIGA technique (Lithographie, Gal – vano Abformung) (29). Briefly, in the silicon/LIGA process nickel needles are grown from a combined seed/interconnec­tion layer through narrow channels in 200 ^m PMMA (poly­methylmethacrylate). After removal of PMMA and etching of the seed layer, the electrode needles stand completely electri­cally separated and are connected individually to the leads in the interconnection layer.

In this way, Bielen succeeded at the IMM (Institute fur Microtechnologie in Mainz, Germany) in fabricating a 2-D multielectrode of 4 X 32 needle electrodes, with square as well as round columns or needles. The electrodes have a thickness as low as 15 ^m and an ultimate height of 220 ^m (11).

Silicon/LIGA technology reduces the number of steps but has as a disadvantage the need for synchrotron radiation facilities. Also, the present limit of the electroplating process to 220 jU, m long nickel needles has to be extended to a needle length of about 500 ^m for useful neuroprosthetic and corti­cal applications.

A review of electrode technology and its perspectives can be found in Mortimer et al. (30). An interesting, nonsilicon approach to contact fibers intrafascicularly is the use of tethDistal nerve stump

Guidance channel

Microelectrode array (die)

Proximal stump

Distal stump


Regenerated axons Proximal nerve stump


Figure 6. (a) Schematic representation of an intelligent neural interface implanted into an inter­sected nerve. (From Ref. 43, their Fig. 1.) (b) Schematic drawing of the silicone chamber model with the inserted silicon chip bridging a 4 mm gap between the proximal and distal stumps of a transected rat sciatic nerve (From Ref. 42, their Fig. 3.) (c) SEM photograph view of a fabricated chip with 100 jum diameter holes. (From Ref. 42, their Fig. 2.) (d) SEM photograph of nerve tissue sections distal to a chip with hole diameters of 100 um after 16 weeks of regeneration. Shown is a minifascicular pattern on the distal surface of the chip. The regenerated nerve struc­ture has a smaller diameter than that of the perforated area of the chip. The circumferential perineurial-like cell layer is clearly visible. (From Ref. 42, their Fig. 5, top.)

Silicon and Silicon-Glass Arrays

Figure 4. (a) Overall diagram of a surface-mounted 3-D recording array. Several multishank 2-D probes are inserted through the plat­form and held in place with micromachined spacer bars. (From Ref. 20, their Fig. 1.) (b) Scanning electron microscope (SEM) photographs of a 3-D 4 X 4-shank microelectrode array. The shanks on the same probe are spaced on 150 ^m centers and are 40 ^m wide. The probes are 120 jum apart in the platform. (From Ref. 20, their Fig. 2, bottom.)

Silicon-based microprobe fabrication has been a major and outstanding activity of the Center for Integrated Sensors and Circuits at the University of Michigan and has led to a large number of single-shaft, multishaft, and 3-D stacked micro­electrode arrays, a number of these being supplied with on­board microelectronics (13-22). Fabrication was supported by design studies (23), strength characterization (24), and devel­opment of interconnection technology (25,26). Groups in Utah and Twente tried to fabricate brush or needle-bed 2-D/3-D multielectrodes in silicon or silicon/glass technology, for corti­cal and nerve applications, with about 100 electrodes. As ani­sotropic silicon etching cannot (yet) perform up to the aspect ratios needed for long, slim needles (a 20 ^m diameter, 500 jU, m long needle has an aspect ratio of 25); the first step to obtain a brush structure from a solid piece of silicon is a saw­ing procedure (12,27,28).

Silicon/glass technology has the advantage of high aspect ratios, sufficient lengths of needles, and different lengths of needles in the same device. The disadvantages are the 3-D nature of many of the process steps, the large number of steps, and the difficulty of their integration (12).


Figure 5. SEM photograph of silicon-nickel-LIGA array. Array with 150 um tall, 20 um diameter nickel needles, realized with aligned X – ray lithography and galvanic growing (LIGA) on silicon substrate with 8 um Cu interconnection wiring. Interdistance between columns is 120 um. (From Ref. 11.)


The forward control of muscle by artificial stimulation might gain importance when this control is supplemented by selec­tive feedback information from nerve fibers attached to sen­sors such as muscle spindles, tendon organs, and cutaneous sensors. This asks for insight into selective recording with multielectrodes.

The same type of calculation previously made for the case of selective stimulation of nerve fibers in rat peroneal nerve (isotropic conductor, local approach) (10) could be applied, by reciprocity, to the case where the device is used to sense natu­ral activity from afferent fibers. These calculations would, for example, lead to a (statistically optimal) electrode interdis­tance of 143 jU, m, for the case that there are 250 type I affer­ent fibers in rat peroneal nerve.

However, while an action potential can be triggered by ac­tivation of one node of Ranvier only (stimulation), propaga­tion of an action potential requires about 20 active nodes (re­cording). So it is not allowed to replace the electrode (stimulation) by one node of Ranvier (recording).

Another difference is that nerve fibers will almost always fire as ensembles. Regarding selectivity, when two (not over­lapping in time) action potentials (or ap trains) are sensed by one electrode, the trains can be detected separately when the selectivity ratio S of their amplitudes V1 and V2 exceeds a certain threshold (i. e., when S > Sth; for example, S > 1.1, or S > 2) (compare this to the signal-to-noise ratio; 1.1 means barely visible, 2 is better).

Quantitative insight in this selectivity ratio S as a function of spatial and conductivity parameters may be obtained by the combined use of an electrode lead field model (using the volume conduction model as outlined previously) and a proba­bility model for the positions of active fibers (12). Figure 3 shows a dramatic decrease in the ability to discriminate two trains when the nerve is insulated from its surrounding tissue (i. e., for zero extraneural conductivity), illustrating the importance of a natural wet surrounding of the nerve.


At low current, an electrode can stimulate one fiber if its posi­tion is close to that fiber, compared to other fibers. Increase of current will expand the stimulation volume, thus including more and more fibers.

The ultimate selectivity would be reached if each fiber would have its own electrode. This would require, however both a blueprint of positions of fibers in the nerve so that electrodes could be positioned close to a node of Ranvier, and enough electrodes. In practice, no blueprint is available, and microfabrication has technological limits. Therefore, with a limited number of electrodes, placed optimally (in a statistical sense), it is important to consider and test how selective stim­ulation can be.

In this respect one has to measure the extent to which each electrode controls as few fibers as possible at low current, be­fore potential fields start to overlap with those of other elec­trodes, with increase of current. Greater overlap means lower selectivity.

From another point of view, one might define the efficiency of a multielectrode device: the number of distinct fibers that can be contacted, divided by the total number of electrodes. Greater overlap means reduced efficiency.

Fiber selectivity has been addressed in Rutten et al. (10), among others. It was concluded, on statistical grounds and by overlap experiments, that an electrode separation of 128 ^m was optimal for a rat peroneal nerve fascicle with 350 alpha motor fibers.

Limited force recruitment experiments with a 2 D 24-elec­trode array (electrode separation 120 ^m) (11) yielded that 10 distinct threshold forces could be evoked (efficiency is 10/24 = 42%).


Lesions in the peripheral nervous system in humans can lead to several disabling effects in sensory and motor functions be­cause the primary information carrier, the propagating action potential, can no longer travel from sensory organs to the brain (afferent information, sensory nerve fibers) or from the brain to muscles (efferent information, motoneurons). In many cases, peripheral nerves may ‘‘repair themselves” (re­generation), provided that the source of the lesion (for exam­ple, pressure on the nerve) is removed soon enough or that adequate surgical measures are taken in due time in order to bring nerve stumps together or to transplant nerve sections to bridge a large gap. During the healing process, nerve fibers will first degenerate and then regenerate all the way, from the spinal cord toward the periphery, reusing the old chan­nels of myelin sheaths and connective tissue. The nerve re­generates with a typical speed of 1 mm per day.

However, this ability to regenerate more or less autono­mously is a property of peripheral nerves only. The central nerve fibers of the spinal cord cannot be induced to regener­ate, although extensive research tries to bring this about by manipulating the biochemical environment of the fibers, offer­ing proteins such as neural growth factors or semaphor pro­teins and other agents that may stimulate nerve growth.

If a person has a central neural lesion but no harm to the peripheral nerves—for example, in paraplegic individuals (with neural interruptions in the spinal cord)—the peripheral nerves may be stimulated artificially by short electric pulses, which evoke propagating action potentials toward the para­lyzed muscles and restore force.