Sensors and Actuators A 96 (2002) 78±85

Design and fabrication of a high-density metal microelectrode array for neural recording Chenyang Xua,*, William Lemonb, Chang Liua a

Micro Actuator, Sensor and System Research Group (MASS), 319 B Microelectronics Laboratory, University of Illinois at Urbana-Champaign, 208 North Wright Street, Urbana, IL 61801, USA b House Rock Research Inc., 95696 Breakaway Road, Brookings, OR 97415, USA Received 3 January 2001; accepted 23 October 2001

Abstract We report a new fabrication technique for realizing a high-density penetrating metal microelectrode array intended for acute multipleunit neural recordings. The microelectrode array consists of multiple metal shanks projecting from a silicon supporting bulk. Neural recording sites, separated by an average spacing of 50 mm, are located at the tip of each shank. Each shank is comprised of two segments to realize both mechanical strength and sharp tissue penetrating ability. A rear support segment is 6 mm long, 40 mm wide, and 30 mm thick. The front segment consists of a 250 mm long and 6 mm thick tapered tip, with the width at its widest point being 15 mm. Electrical insulation of the microelectrode body is achieved by a conformal coating of a 3 mm thick poly(para-chloroxylylene) (Parylene-C) ®lm. Multiple recording sites with precise opening sizes are de®ned by selectively removing Parylene-C from the electrode tips using photolithography and oxygen plasma etching. # 2002 Elsevier Science B.V. All rights reserved. Keywords: Microelectrode array; Parylene; Extracellular recording

1. Introduction 1.1. Multiunit neuron recording It has become increasingly clear that the nervous system achieves its complex functionalities through the cooperative activities of many neurons. Instruments and techniques for in vivo recording of neural activities from a large number of spatially distributed neurons are critical to further understanding of the nervous systems. Until very recently, neurobiologists have largely relied on sequential analysis of single-unit recordings. Such a method limits the throughput of experiments. The dynamic interaction among neurons is only inferable through indirect statistical analysis. Simultaneous monitoring of neural activities from multiple neurons [1,2] can be achieved using various spike-sorting algorithms. However, spike-sorting algorithms based on single-channel electrodes tend to have high error rates. Spike-sorting algorithms based on geometric triangulation are capable of increasing the accuracy of neuron identi®cation. For such algorithms to work properly, neuron probes having multiple recording sites with precisely de®ned spacing are desired. * Corresponding author. Tel.: ‡1-217-265-0808; fax: ‡1-217-333-4051. E-mail address: [email protected] (C. Xu).

Commonly used single-unit neuron recording electrodes include sharpened tungsten-wire probes [3] and capillary glass probes. However, it is dif®cult to attain uniform geometries of ®nished array probes. It is impractical to manually assemble individual probes into an array with controlled small spacing, although past attempts at realizing a linear array or a two-dimensional matrix have been reported [4,5]. The fabrication process is labor intensive and is associated with low-yields. The stereotrode, and later the tetrode, is one of the most reliable multi-channel recording tools that have been developed in recent years [6,7]. A tetrode is generally produced by manually twisting four insulated single wires into a bundle. It uses inexpensive materials andcanbe produced invirtuallyany laboratories. However, the tetrode suffers from a disadvantage that the manual assembly process is neither precise nor uniform. The geometric locations of tetrode wires cannot be precisely controlled. In addition, the spacing between tetrodes is not ¯exible, being naturally determined by the sizes of the wires used. Speci®cally, the spacing between electrodes roughly corresponds to the diameter of the wires, which ranges from 12 to 30 mm. Consequently, a single tetrode can only record from at most several neurons from a limited region. In order to obtain true simultaneous multi-channel recording, there remains at present a need for multi-channel

0924-4247/02/$ ± see front matter # 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 4 - 4 2 4 7 ( 0 1 ) 0 0 7 6 6 - X

C. Xu et al. / Sensors and Actuators A 96 (2002) 78±85

microelectrodes with the following characteristics: (1) controllably small spacing comparable to neuron sizes, (2) ¯exibility to design inter-electrode spacing, and (3) minimal damage to neuron tissues. In addition, such neuron probes must exhibit excellent mechanical robustness (e.g. resistance to fracture and buckling during insertion and recording). 1.2. Microfabricated neuron microelectrodes In recent decades, micromachining technology has enjoyed rapid growth. It has been used to realize integrated neuron probes, with both miniaturized sizes and precision dimensional control. Using photolithography and other associated microfabrication techniques, researchers have demonstrated methods for creating arrays of insertion microelectrodes. One early example of such probes consists of multiple microelectrode recording sites located along the side of silicon shanks [8,9]. Several groups have developed silicon-based probes with multiple electrical recording sites [10±13]. To date, the majority of published works on micromachined microelectrodes involve silicon as a probe shank material. Due to miniaturization of thickness, silicon probes can be made suf®ciently ¯exible. However, it is of interest to develop probe shanks made of other materials, such as metal, which generally exhibits good fracture strain compared to that of silicon [14]. The location of recording sites on the probe shank is important. In many existing microfabricated neuron probes, the recording sites are distributed along the shanks. Although this method generally allows many recording sites to be made, neurons that form contacts to the recording sites are susceptible to damage during probe penetration. On the other hand, neurons contacting the probes are less likely to be damaged during the probe penetration if the probes have recording sites at the tips. Several probes with tip openings have been made [15,16]. For example, Normann and coworkers developed electrode arrays with tip opening as the silicon stimulating needle arrays [15]. The primary application of such probe arrays is to provide stimulation or recording from large neurons found in the cerebral cortex. Due to the nature of the fabrication process, the spacing between probes is relatively large. The size of electrode openings at the tips is relatively large as well (50±200 mm). Recently, a silicon-based probe array with tip opening for intracellular and extracellular recording has been developed [16]. 1.3. New metal microelectrodes The objective of this work is to develop a microfabricated neuron probe array with the following structural and performance characteristics. 1. Recording sites are placed at the tips of micromachined shanks (needles) such that the neurons that are nearest to

79

the recording sites are likely undamaged during shank insertion. In this respect, the new microelectrode inherits the structural characteristics of a conventional tungstenwire probe. 2. The shank is made of a metal material instead of silicon to increase its robustness. Neuron probes made of metal materials have been demonstrated in the past [14]. In this work, we developed a metal shank with unique twosegmented profiles. 3. Each neuron probe contains multiple recording sites with controllable spacing. The spacing is 50 mm for current prototypes, greater than that used in tetrodes. One major advantage of the newly developed neuron probes compared with tetrode is the precision of fabrication and ¯exibility of spacing controls. 2. Design The top and side views of the new multiunit recording microelectrodes are shown in Fig. 1. A typical microelectrode chip consists of seven metal shanks and recording sites. The probe shank is made of electroplated metal such as nickel and NiFe (Permalloy). The spacing between electrode shanks is 50 mm according to current designs. The design is ¯exible. Both the quantity of the electrodes in an array and the spacing between electrodes can be modi®ed. Each single electrode must satisfy the following two design criteria. Firstly, the tip of the electrode must be sharp and have a small cross-section to minimize damage to neural tissues during the ®nal approach. Secondly, the shank must be suf®ciently long, on the order of 5±10 mm, in order to allow for easy observation during the insertion and for the desired depth of penetration into the neuron tissue. In order to satisfy these two criteria, the shank geometry must be optimized. A long shank is susceptible to buckling when it penetrates into neuron tissues. Buckling is highly undesirable from the performance point of view. If a slender shank is made of a fragile material such as silicon, buckling can cause the beam to fracture. In the case of a metal shank, buckling can cause permanent plastic deformation to the shank. For a given shank, the buckling strength can be estimated as follows. Suppose a long shank with length L, width w, and thickness t is loaded with a longitudinal force. The modulus of elasticity of the shank material is denoted E. The threshold force that needs to be exerted to induce buckling is termed the critical force (Pc). The magnitude of Pc is given by [17]: Pc ˆ

p2 EI L2

(1)

The term I refers to the moment of inertia of a shank. For a shank with a rectangular cross-section (w  t), I ˆ wt3 =12.

80

C. Xu et al. / Sensors and Actuators A 96 (2002) 78±85

Fig. 1. A schematic diagram of the design of the neuron probe: (a) plan view of a microelectrode array with seven recording sites; (b) close-up view of the two-tiered shank design with pertinent dimensions; (c) close-up view of the recording tip with the size of the opening and thickness of the insulation layer.

It is note worthy that the value of Pc decreases with increasing length L. Because the shank of the neuron probe must be suf®ciently long (e.g. 5±10 mm) to provide enough traverse distance, its cross-section must be made relatively large to achieve suf®cient Pc. However, the increased cross-section would generally increase the volume of displaced tissues and, consequently, increase the risk of tissue damages. In order to satisfy the requirements for both sharpness and length of the probe, we designed a shank with a unique two-segment pro®le. These two segments have distinct cross-sections. The front segment has a cross-section of 15 mm  6 mm and a length of 250 mm. The rear segment of the shank was designed to be 5.75 mm long and 30 mm thick. The width of the rear part increases gradually from 30 mm in the frontmost region to 50 mm in the rear-most region, with the average width being 40 mm. The critical buckling force for the shank is approximately 8 mN according to Eq. (1). This buckling force is greater compared with that required to penetrate neural tissues. For example, the measured minimal force for penetrating the cockroach olfactory lobe is approximately 0.1 mN.

3. Fabrication methods 3.1. Overview The neuron microelectrode must satisfy a number of structural and materials characteristics. A unique microfabrication process is developed to satisfy these requirements (Fig. 2). The entire fabrication process can be divided into three major phases. In the ®rst phase, we form metal shanks by using electroplating method. The second phase consists of silicon bulk etching to render the metal needles freestanding. The third and ®nal phase consists of deposition of poly(para-chloroxylylene) (Parylene-C) as an insulation layer and selective removal of Parylene-C from the tips. In this section, we describe the overall fabrication processes according to these three phases. 3.2. Formation of microprobe shank We selected the electroplating process to form the shank. The electroplating rate is high (on the order of 5±10 mm/h) and allows desired metal-layer thickness to be reached.

C. Xu et al. / Sensors and Actuators A 96 (2002) 78±85

81

Fig. 2. Schematic diagram illustrating major microfabrication steps for realizing the microelectrode array.

Two materials have been explored as the shank material [18] in our work. To begin, the double-side polished {1 0 0} silicon wafer is coated with 1 mm of thermally-grown silicon dioxide. The oxide layer not only offers insulation for future electrodes from the semiconductingsiliconbulk,but also acts as a mask for bulk silicon etching. Open windows in the silicon oxide material on the bottom side of the wafer are de®ned by photoresist patterning and buffered hydro¯uoric (HF) acid etching. Two metal-layers, Cr and Cu, are then deposited on the wafer as seed layers for the ensuing electroplating process. Since the Cu thin ®lm does not adhere well to the oxide, an Ê thick Cr layer is deposited to promote adhesion initial 200 A Ê between the oxide and the thermally evaporated Cu (3000 A thick). The two-segmented shank is electroplated in two separate steps.

An 8 mm thick layer of photoresist (PR AZ-P4620) is spun onto the wafer and patterned above the seed layer with selective openings that de®ne the electroplated needle structures. The electroplating process is performed at room temperature at a controlled rate of 6 mm/h [19]. The tip formed using this method is shown in Fig. 3. A second electroplating step is performed to selectively increase the thickness of the rear segment [19]. 3.3. Probe shaping The photoresist mold material is removed after the probe shank has been formed. Lead wires to individual shanks are patterned using the seed layer itself. The shanks are freed by removing part of the underlying silicon substrate using anisotropic wet etching. For this process, standard potassium

82

C. Xu et al. / Sensors and Actuators A 96 (2002) 78±85

Fig. 3. An SEM micrograph of the tip region of the microelectrode before Parylene coating showing the sharp tip formed by electroplating.

hydroxide (KOH, 33%, 65 8C) etching is used with the thermal oxide grown at the start of the ®rst phase acting as the etching mask. Because KOH slowly attacks copper and Permalloy at 65 8C under prolonged etching, the front side of the wafer is protected by ®rst pouring high-quality clear wax, followed by complete wafer encapsulation using Sylgard 184 polydimethylsiloxane (PDMS). After the KOH etching is completed, the wax is dissolved in acetone and the PDMS cover is manually removed. We found this protecting method to be robust and able to withstand the KOH etching for up to 10 h. It should be noted that the metal shank can be uniformly coated with biocompatible metal thin ®lms (e.g. Au or Pt) following the release of the shanks to form an optimal interface with neurons. For example, a commercial process for electroless plating of gold thin ®lm on nickel and NiFe surfaces is available (Transene Company). 3.4. Parylene insulation and tip opening The ®nal step involves coating the metal shank with an insulating thin ®lm and selectively opening exposed metal recording sites at the tips. We choose Parylene-C as the insulating. The material offers several unique characteristics, as summarized in the following. Firstly, the Parylene-C material has high electrical resistivity. Secondly, the polymer ®lm exhibits mechanical ¯exibility comparable to that of inorganic insulation layers (e.g. silicon oxide) and is not susceptible to cracking or other defects. Third, the Parylene material is biocompatible with tissue [20,21]. Last, the material can be formed using conformal, uniform, and pin-holefree vapor-phase deposition at room temperature (Fig. 4). Prior to the application of Parylene-C, the probe was ®rst treated with a dilute solution of a silane primer (g-metha-

cryloxypropyltrimethoxysilane) to enhance the adhesion. A layer of 3 mm thick Parylene-C was then deposited onto the probe using a PDSII Labcoater (special coating systems). To precisely produce a small opening at the tip, we use photolithography and oxygen plasma etching. The oxygen plasma etching is performed under a pressure of 500 mTorr and a power density of 0.42 W/cm2. In order to coat the metal shanks with liquid-phase photoresist, a mold is created to match the shape of the probe. Specially, the height of the mold is identical to the thickness of the silicon bulk. With the mold in contact with the probe, the otherwise freestanding shanks are rested against the mold surfaces. This setup allows spin-coating a 20 mm thick AZ-P4620 photoresist ®lm onto the shank. The desired length of the tip region is selectively exposed in a standard UV mask aligner and the exposed photoresist material is removed. This exposes Parylene at the tips of shanks with a speci®ed exposure length. Oxygen plasma etching is used to remove organic materials including both the Parylene and the photoresist. However, only the Parylene at the tip is removed due to the lack of photoresist coverage. The process is completed by ®nally removing the photoresist in acetone solutions. Scanning electron microscopy (SEM) micrographs of the ®nished probe are shown in Fig. 4. 3.5. Probe packaging The ®nished probe is mounted on a modi®ed circuit board and is electrically connected to standard pin connectors, which offer standard interface to most measurement equipment, such as multi-channel preampli®ers offered by Plexon Inc. A plastic rod was glued to the circuit board so that the probe could be mounted onto standard x±y±z probe positioners.

C. Xu et al. / Sensors and Actuators A 96 (2002) 78±85

83

Fig. 4. SEM micrographs of (a) all seven array microelectrode tip regions, and (b) the Parylene and electrode in a Parylene window on the tip.

4. Performance characterization 4.1. Mechanical strength We have veri®ed that the overall neuron probe structure is mechanically strong and can penetrate neural tissue (cockroach brain with pia sheath removed) without

buckling failure. Using ®ne forceps, individual electrode shanks can be elastically deformed laterally or out of plane without breaking. The Parylene-C ®lm adheres well to the metal shanks. The adhesion remained tight, with no degradation noticeable visually or electrically, for the duration of 7 days soaking inside the saline (0.9% NaCl).

84

C. Xu et al. / Sensors and Actuators A 96 (2002) 78±85

Fig. 5. A recording trace of a neuron's spontaneous activity from a representative electrode in the electrode array.

4.2. Electrical properties The impedance associated with the recording probe is measured experimentally. The longitudinal resistance along the Permalloy beam and the resistance of the lead wires are both less than 1 kO, and so can be neglected compared with the primary impedance, which is derived from the contact at the exposed recording sites. We measure the impedance of the probe by submerging only the recording sites in 0.9% saline solution. The testing signal for impedance measurement was sinusoidal (100 mV, 50 nA, 1000 Hz) with respect to a large reference electrode (platinum wire). We determined the current through each individual recording site by monitoring the voltage drop across a precision 2 MO resistor. The resulting voltage drop between the ground electrode and the test electrode was measured to obtain the value of the impedance. The size of the opening at the recording sites in¯uences the impedance value. For a typical electrode with 50 mm2 tip opening area, the measured impedance is on the order of 2±3 MW at 1 kHz, comparable to the impedance of a commercial tungsten probe with similar opening size (Parylene-C coated tungsten probe, A-M systems). The measured inter-electrode resistance is greater than 300 MO.

Recorded signals (Fig. 5) were ampli®ed and fed into an analog-to-digital signal converter, which produces output that is stored and analyzed using MatLab systems. The neural recordings demonstrate that the probes are capable of recording data from all individual electrodes in a probe. However, more work is needed. Although the probe have recorded action potentials from many locations of cockroach nervous system, some experiments did not yield highquality signals (measured as signal-to-noise ratio, or SNR), compared to those typically obtained using tungsten electrodes. The packaging of the probes will be improved in future generations to incorporated a grounded shield to reduce the extent of in¯uence caused by stray electromagnetic radiation. 5. Future work We are currently conducting a number of studies in design, fabrication, and characterization. The shank materials are not considered to be biocompatible for long-term use. A metal thin ®lm can be uniformly coated over the entire shank before the Parylene material is deposited. This ®lm serves to improve the electrical contact characteristics (e.g. reduce noise).

4.3. Neuron recording

6. Conclusions

Neuron recordings were made from somata of antennal lobe projection neurons of female American cockroaches, Periplaneta americana. The animals were cold-anaesthetized and immobilized, dorsal side up, in a Sylgard-lined dish. A window of cuticle was opened between the compound eyes, exposing the antennal lobe and part of the brain. The air sacs on the anterior face of the brain were removed and the sheath surrounding the antennal lobes was removed. The body of the cockroach was superfused with physiological saline at room temperature. The packaged probe was mounted on a x±y±z micromanipulator positioned over the exposed neural tissue, and slowly inserted. The antennal lobe was immobilized on a small stainless steel platform held beneath the brain.

The fabrication processes described in this paper provide precise and reproducible techniques to mass-produce onedimensional metal microelectrode arrays with ¯exible con®gurations. With its high-density and relatively long traverse distance, the microelectrode array is well suited for recording densely packed neurons in order to look for direct evidence of possible interaction between them. The tip exposure area can be precisely controlled to meet impedance and single-unit isolation requirements. Acknowledgements This project was conducted with the ®nancial support of the DARPA MTO Controlled Biological Systems (CBS)

C. Xu et al. / Sensors and Actuators A 96 (2002) 78±85

program under contract number N66001-98-C-8624. Prof. Fred Delcomyn has contributed to this research by assisting with the neurophysiological recording and characterization. References [1] E.M. Schmidt, Computer separation of multi-unit neuroelectric data: a review, J. Neurosci. Meth. 12 (1984) 95. [2] M. Fee, P. Mitra, D. Kleinfeld, Automatic sorting of multiple unit neuronal signals in the presence of anisotropic and non-Gaussian variability, J. Neurosci. Meth. 69 (1996) 175. [3] G.E. Loeb, R.A. Peck, J. Martyniuk, Toward the ultimate metal microelectrode, J. Neurosci. Meth. 63 (1995) 175. [4] J. Kruger, M. Bach, Simultaneous recording with 30 microelectrodes in monkey visual cortex, Exp. Brain Res. 41 (1981) 191. [5] D. Jaeger, S. Gilman, J.W. Aldridge, A multiwire microelectrode for single-unit recording in deep brain structures, J. Neurosci. Meth. 32 (1990) 143. [6] C.M. Gray, P. Maldonado, M. Wilson, B. McNaughton, Tetrodes markedly improve the reliability and yield of multiple single-unit isolation from multiunit recordings in cat striate cortex, J. Neurosci. Meth. 63 (1995) 43. [7] K.D. Harris, D.A. Henze, J. Csicsvari, H. Hirase, G. Buzsaki, Accuracy of tetrode spike separation as determined by simultaneous intracellular and extracellular measurements, J. Neurophysiol. 84 (2000) 401. [8] K.D. Wise, J.B. Angell, A. Starr, An integrated-circuit approach to extracellular microelectrodes, IEEE Trans. Biomed. Eng. 17 (1970) 238. [9] K.D. Wise, J.B. Angell, A low-capacitance multielectrode probe for use in extracellular neurophysiology, IEEE Trans. Biomed. Eng. 22 (1975) 212.

85

[10] K. Najafi, K.D. Wise, T. Mochizuki, A high-yield IC-compatible multi-channel recording array, IEEE Trans. Electron Devices ED-32 (1985) 1206. [11] R.S. Pickard, T.R. Welberry, Printed circuit microelectrodes and their application to honeybee brain, J. Exp. Biol. 64 (1976) 39. [12] M. Kuperstein, D.A. Whittington, A practical 24-channel microelectrode for neural recording in vivo, IEEE Trans. Biomed. Eng. 28 (1981) 288. [13] D.T. Kewley, M.D. Hills, D.A. Borkholder, I.E. Opris, N.I. Maluf, C.W. Storment, J.M. Bower, G.T.A. Kovacs, Plasma-etched neural probes, Sens. Actuat. A 58 (1997) 27. [14] A.B. Frazier, D.P. O'Brien, M.G. Allen, Two-dimensional metallic microelectrode arrays for extracellular stimulation and recording of neurons, PIEEE Proc. Micro. Electro. Mech. Syst. (1992) 195. [15] K.E. Jones, P.K. Campbell, R.A. Normann, A glass/silicon composite intracortical electrode array, Ann. Biomed. Eng. 20 (1992) 423. [16] Y. Hanein, U. Lang, J. Theobald, R. Wyeth, T. Daniel, A.O.D. Willows, D.D. Denton, K.F. Boringer, Intracellular neuronal recording with high aspect ratio MEMS probes, in: Proceedings of the 11th International Conference on Solid-State Sensors and Actuators, 2001, p. 386. [17] W.F. Riley, L.D. Sturges, D.H Morris, Statics and Mechanics of Materials: An Integrated Approach, Wiley, 1995, p. 578. [18] J.T. Ravnkilde, V. Ziebart, O. Hansen, H. Baltes, Mechanical characterization of electroplated nickel±iron, Sens. Mater. 12 (2000) 99. [19] Y. Yi, C. Liu, Assembly of micro optical devices using magnetic actuation, J. Sens. Actuat. 78 (1999) 205. [20] G.E. Loeb, M.J. Bak, M. Salcman, E.M. Schmidt, Parylene as a chronically stable, reproducible microelectrode insulator, IEEE Trans. Biomed. Eng. 24 (1977) 121. [21] E.M. Schmidt, Parylene as an electrode insulator: a review, J. Electrophysiol. Tech. 10 (1983) 19.