Neuron-Microdevice Connections - Caltec

Neuron-Microdevice Connections - CaltechTHESIS
CaltechTHESIS
A Caltech Library Service
About
Browse
Deposit an Item
Instructions for Students
Neuron-Microdevice Connections
Citation
Regehr, Wade Gordon
(1988)
Neuron-Microdevice Connections.
Dissertation (Ph.D.), California Institute of Technology.
doi:10.7907/nshf-ww49.
Abstract
A new method for long-term recording and stimulation applicable to cultured neurons has been developed. Silicon-based microelectrodes have been fabricated using integrated-circuit technology and micromachining. The chronic connection is made by positioning the electrode tip into contact with the cell body, and gluing the device to the bottom of the culture dish. These "diving-board electrodes" consist of an insulated lead exposed only at the tip sealed to the cell body of a cultured neuron. A two-way electrical connection to
Helisoma
B19 neurons has been established for up to four days. Preliminary experiments with cultured superior cervical ganglion neurons indicate diving-board electrodes can be used with cultured neurons larger than 20 µm in diameter.
In a related technique
Helisoma
neurons grown on special dish containing a multielectrode array were found to seal to the dish electrodes, establishing similar long-term connections. This capability will make it possible to conduct experiments with either diving-board electrodes or dishes that cannot be performed using conventional techniques.
Item Type:
Thesis (Dissertation (Ph.D.))
Subject Keywords:
Applied Physics
Degree Grantor:
California Institute of Technology
Division:
Engineering and Applied Science
Major Option:
Applied Physics
Thesis Availability:
Public (worldwide access)
Research Advisor(s):
Rutledge, David B.
Thesis Committee:
Rutledge, David B. (chair)
Bellan, Paul Murray
Lester, Henry A.
Bower, James M.
Pine, Jerome
Defense Date:
25 March 1988
Record Number:
CaltechETD:etd-11092007-084226
Persistent URL:
DOI:
10.7907/nshf-ww49
Default Usage Policy:
No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:
4478
Collection:
CaltechTHESIS
Deposited By:
Imported from ETD-db
Deposited On:
05 Dec 2007
Last Modified:
16 Apr 2021 23:16
Thesis Files
Preview
PDF
- Final Version
See Usage Policy.
12MB
Repository Staff Only:
item control page
CaltechTHESIS is powered by
EPrints 3.3
which is developed by the
School of Electronics and Computer Science
at the University of Southampton.
More information and software credits
Neuron-Microdevice Connections

Thesis byWade G. Regehr

In Partial Fullfillment of the Requirements
for the Degree of
Doctor of Philosophy

California Institute of Technology,
Pasadena, California
1988
(Submitted March 25, 1988.)

TO MY PARENTS AND GRANDPARENTS

Ill

Acknowledgements

I am grateful to have had the opportunity to work on a very interesting

project, and in the course of my research to have worked with many excellent
people. I am fortunate to have had Jerry Pine and Dave Rutledge for advisors.

They provided the environment necessary for the success of such an interdisci
plinary project. In addition to providing scientific guidance, they were under

standing and encouraging. This was never more apparent than their realization
that a trip around the world would be an excellent opportunity for me. I would

like to thank the members of Dave and Jerry’s groups, both past and present.
Not only did they teach me a great deal, we had some good times along the

way. In particular I would like to thank Dean Neikirk who taught me integrated
circuit fabrication. Lastly, I would like to thank my family and friends for their
support. Michelle... thankyou.

Abstract

A new method for long-term recording and stimulation applicable to cul
tured neurons has been developed. Silicon-based microelectrodes have been fab
ricated using integrated-circuit technology and micromachining. The chronic

connection is made by positioning the electrode tip into contact with the cell
body, and gluing the device to the bottom of the culture dish. These “diving-

board electrodes” consist of an insulated lead exposed only at the tip sealed to the
cell body of a cultured neuron. A two-way electrical connection to Helisoma B19

neurons has been established for up to four days. Preliminary experiments with
cultured superior cervical ganglion neurons indicate diving-board electrodes can

be used with cultured neurons larger than 20 μm in diameter.

In a related technique Helisoma neurons grown on special dish containing a
multielectrode array were found to seal to the dish electrodes, establishing similar

long-term connections. This capability will make it possible to conduct exper
iments with either diving-board electrodes or dishes that cannot be performed

using conventional techniques.

Table of Contents

Acknowledgement ......................................................................................... iii
Abstract .......................................................

1. Introduction ...................................

2. Stimulating andRecording From Cultured Neurons With Glass

Loose- Patch Electrodes

2.1 Background .................................................................................................. 9
2.2 Cultured identified Helieoma neurons ....................................................11

2.3 Cultured rat superior cervical ganglionneurons (SCG’s) ................... 13
2.4 Patch electrode tests ................................................................................. 16
3. Diving-Board Electrode Fabrication

3.1 Diving-board electrode: basic idea ......................................................... 26
3.2 Micromachining of silicon........................................................................ 27

3.3 Insulators for saline environments.......................................................... 31
3.4 Fabrication procedure .............................................................................. 38
3.5 Mechanical properties of diving-board electrode ................................. 44

4. Stimulating and Recording with Current-Clamped Loose-Patch

Electrodes

4.1 Equivalent Circuit ................................................................................... 53
4.2 Recording ................................................................................................... 54
4.3 Stimulation .......................

5. Diving-Board Electrode Tests

5.1 Establishing the chronic connection ..................................................... 63
5.2 Recording from identified Helisoma neurons........................................ 67
5.3 Stimulating identified Helisoma neurons .............................................. 70

5.4 Tests on Rat SCG’s .................................................................................. 72
6. Multielectrode Arrays

6.1 Recording and Stimulating Using a Multielectrode Array................ 75
6.2 Dish Fabrication ....................................................................... .............. 76
6.3 Dish Electronics ........................................................................................78

6.4 Experimental Procedure .......................................................................... 78

6.5 Recording ................................................................................................... 81
6.6 Stimulation ..............................................................
7. Future Work.........................

84
88

7.1 Future Experiments with Diving-Board Electrodes ........... . . . . . . . . . 88
7.2 Future Dish Experiments .........................................................................89
7.3 Future In Vitro Applications of Microdevices ..................................... 90
7.4 Impact of Integrated-Circuit Based Micrdevices on Neurobiology .. 93
Appendix

Chapter 1
Introduction
A primary goal of neurobiology is to learn how the nervous system works

by determining how neurons communicate with each other, how they are put
together to form a functioning nervous system, and the means by which a system

of neurons learns based on experience. Many different approaches are required
to begin to understand the nervous system. While it is possible to study neurons
in intact animals, such systems are complex and experiments can be difficult to

interpret. Another approach is to simplify the nervous system and study single
cells and then small networks of neurons. Technological advances will allow many

interesting experiments to be performed on such systems.
Cell-culture techniques, in which neurons are grown in the artificial environ
ment of a tissue culture dish, provide a powerful tool for answering fundamental

neurobiological questions. A great variety of neuronal cell types can be grown in

cell culture providing: (1) the neurons are removed from the animal at the ap
propriate time, (2) culture dishes are prepared to provide a place for the neurons

to attach, and (3) the culture medium provides a correct balance of biologically

necessary molecules [1]. In addition to the many types of embryonic neurons
that have been cultured, recently it has been possible to culture adult identi

fied neurons from various invertebrates, including the leech [2], ffelisoma [3],
and Aply3ia [4]. While most of these large neurons have been characterized

in vivo, the in vitro preparations make it possible to study neurite outgrowth
and synapse formation under experimentally controlled conditions. Stimulating
and recording from neurons is greatly simplified because the individual cells are

visible and accessible, the biochemical environment is easily manipulated, and

the nervous system can be greatly simplified. However there are a number of

interesting questions that remain unanswered because experiments are difficult
or impossible to perform since a long-term, two-way electrical contact is not
available.

There are several ways to make electrical contact to a cell. A method that
provides an excellent signal-to-noise ratio is to insert a glass electrode with a

submicron tip into a cell. Another way to make electrical contact is to whole
cell patch to the cell [5]. A glass pipette several microns in diameter is brought

up against a cell and a seal of several gigohms is obtained by applying suction.
It is possible to break down the membrane beneath the patch and get inside

the cell. However, these techniques have several disadvantages: it is difficult to

maintain such connections for more than a few hours, they are invasive methods

usually damaging the cell, and it is not generally practical to use more than two

electrodes simultaneously.
Another approach is to use extracellular electrodes to stimulate and record
from cultured neurons. Cells have been grown on specially prepared dishes with
a multielectrode array pattern [6]-[10]. Such methods are noninvasive, and have
been used successfully to stimulate neurons [7]-[9], and to record action poten

tials from cultured neurons for periods of weeks [6]. However, this technique is

not sensitive to subthreshold signals, it is often difficult to interpret recordings
since it is possible to record from several cells simultaneously, and a current pulse

passed through an electrode may stimulate more than one cell.
It is also possible to measure the intracellular voltage of cultured neurons

using voltage-sensitive dyes. When these dyes are bound to a cell membrane
they respond linearly to changes in the intracellular potential either by changing

their fluorescence or their absorption [11],[12]. This change follows the potential
change recorded by an intracellular electrode. By having an array of detectors it

is possible to record from many cells simultaneously. However, there are draw

backs in the use of such dyes: illumination breaks down the dye molecules into

toxic byproducts limiting the total exposure time and the number of measure
ments; signals are small, and shot noise results in poor signal-to-noise ratios;

an appropriate dye must be found for each preparation; a complicated array
of detectors and amplifiers, and a data acquisition system is required; and some

means of stimulating cells are required [13], [14]. While voltage-sensitive dyes are
promising for long-term recording, limitations of this technique warrant search
ing for other solutions to the problem of long-term recording and stimulating.
It is also possible to stimulate and record by a loose-patch method [15] that

does not break down the membrane (Figure l.la). Chapter 2 deals with experi
ments using glass pipettes and cultured Heli30ma snail neurons and cultured rat

superior cervical ganglion neurons (SCG’s), demonstrating the feasibility of a
loose-patch type connection for stimulating and recording. Two types of exper

iments were performed: one-electrode experiments with 10 μm-12 μm diameter
patch pipettes; and two electrode experiments with an intracellular microelec
trode used as well, to stimulate the neuron and compare the intracellular poten
tial with the loose-patch electrode potential. However, it is difficult to maintain

the connection for long periods and to communicate with more than two cells
simultaneously due to the large manipulators needed to position the electrodes.
The diving-board electrode is a silicon microdevice used in a manner similar

to a loose-patch electrode (Figure l.lb), but without the need to be held in place
by manipulators. Each electrode is positioned and is then glued to the bottom of

the tissue culture dish. A seal is made between the lower surface of the electrode

and the cell in the same way that a seal is made to the bottom of a glass pipette,
although suction cannot be applied when using this device. This establishes

Figure 1.1. (a) Schematic view of a conventional glass micropipette patch electrode
being used to record from a cell, (b) Neuron sealed to an electrode in the bottom of a
dish, (c) Diving-board electrode in contact with a cell.

a one-to-one connection between a device and a cell for both stimulating and

recording. An insulated gold wire extends vertically from the device out of the
solution to connect to the electronics.
Integrated circuit-technology and micromachining of silicon provide pow
erful tools to fabricate such microelectrodes [16].

Chapter 3 begins with an

introduction to silicon micromachining techniques, illustrating the capabilities
and limitations of such techniques, and motivating the diving-board electrode

fabrication procedure. Also, insulating layers available for use in saline envi
ronments are reviewed to explain the choice of passivation layers used in the

diving-board electrode. Next the electrode fabrication procedure is outlined,
with a more detailed treatment given in Appendix A.

Glass loose-patch electrodes, diving-board electrodes, and sealed dish elec
trodes improve recording and stimulation capabilities by sealing to a neuron.

Chapter 4 discusses the principles of operation of such devices by starting with

an appropriate equivalent circuit, and then considering stimulating and record

ing.
Diving-board electrodes were tested on identified Helisoma neurons, and on

SCG,s. First, a procedure for manipulating the electrodes into place and gluing
them down under water was developed. Then it was necessary to determine

the biocompatibility of the glue, the gluing process, and the electrodes. Once
the biocompatibility was established electrophysiological tests of the device were
performed, including simultaneous recording with an intracellular electrode and

long-term studies. These results are presented in Chapter 5.
Another approach to obtain a chronic two-way connection is to use mul

tielectrode dishes in a manner similar to the loose-patch method by growing
a neuron over an electrode to form a stable seal between the cell and the elec
trode [9]. Such a technique is practical only with large invertebrate neurons such

as Helisoma neurons, but is not effective for smaller vertebrate neurons such as
SCG’s. Chapter 6 deals with the use of such multielectrode arrays, beginning
with their fabrication and moving on to electrode tests using Helisoma neurons.
The results obtained with diving-board electrodes and neuron-capped elec

trode dishes as presented in this thesis are promising, but can be thought of as

a first step. While the capability of chronic stimulation and long term recording

of action potentials has been demonstrated, the techniques must be improved
if subthreshold signals are to be chronically recorded. Integrated circuit tech

niques and micromachining of silicon have yet to significantly impact the field of

neurobiology. However, the potential exists for innovative solutions to difficult
biological problems using such technologies. Chapter 7 is a discussion of ways
in which the operation of multielectrode dishes and multielectrode arrays might

be improved, and applications of such devices. Other devices with biological

applications that could be fabricated using such techniques are discussed.

References

[1] P. G. Nelson, and M. Lieberman ed., Excitable Cells in Tissue Culture,
Plenum Press, New York, 1981.
[2] D. F. Ready, and J. F. Nicholls, “Identified neurons isolated from the leech
CNS make selective connections in cell culture,”Nature, vol. 182, pp. 67-69.
[3] R. G. Wong, R. D. Hadley, S. B. Kater and G. Hauser, “Neunte outgrowth in
molluscan organ and cell cultures: The role of conditioning factor(s)”, J. Neu
roscience, vol. 1, p. 1008, 1981.
[4] D. Dagan, and I. B. Levitan, “Isolated identified Aplysiα neurons in cell
culture,” J. Neurosci., vol. 1, pp. 736-740, 1981.
[5] A. Marty, and E. Neher, “Tight-seal whole-cell recording,” in Single Channel
Recording, ed. B. Sakmann, and E. Neher, Plenum Press, New York, pp. 107122, 1983.
[6] G. W. Gross, A. N. Williams, and J. H. Lucas, “Recording of spontaneous
activity with photoetched microelectrode surfaces from spinal cord neurons
in culture,” J. Neurose. Methods, vol. 5, pp. 13-22, 1982.
[7] J. Pine, “Recording action potentials from cultured neurons with extracellular
microcircuit electrodes,” J. Neurose. Methods, vol. 2, pp. 19-31, 1980.
[8] C. A. Thomas, Jr., P. A. Springer, G. E. Loeb, Y. Berwald-Netter, and
L. M. 0kun, “A miniature microelectrode array to monitor the bioelectric
activity of cultured cells,” Exp. Cell Res., vol. 74, pp. 61-66, 1972.
[9] D. W. Tank, C. S. Cohan, and S. B. Kater, “Cell body capping of array elec
trodes improves measurements of extracellular voltages in micro-cultures of
invertebrate neurons,” IEEE Conference on Synthetic Microstructures, Airlie,
1986.
[10] D. A. Israel, W. H. Barry, D. J. Edell, and R. G. Mark, “An array of
microelectrodes to stimulate and record from cardiac cells in culture,” Am. J.
Physio., vol. 247, Am. J. Physio., vol. 247, pp. H669-H674, 1984.
[11] A. Grinvald, “Real-time optical imaging of neuronal activity,” TINS, pp. 143150, 1984.
[12] B. M. Salzberg, “Optical recording of electrical activity in neurons using
molecular probes, in Cellular Mechanisms of Neurobiology, vol. 3, 1982.
[13] H. Rayburn, J. Gilbert, C-B .Chien, and J. Pine, Noninvasive techniques for
long term monitoring of synaptic connectivity in cultures of superior ganglion
cells, Soc. of Neurosci. 14th Annual Meeting, Abstract 171.1, 1984.
[14] C-B. Chien, W. D. Crank, J. Pine, “Noninvasive techniques for measurement
and long-term monitoring of synaptic connectivity in microcultures of sym
pathetic neurons,” Soc. of Neurosci. 17th Annual Meeting, Abstract 393.14,
1987.
[15] W. Stuhmer, W. M. Roberts, and W. Aimers, “The loose patch clamp,”
in Single Channel Recording, ed. B. Sakmann, and E. Neher, Plenum Press,
New York, pp. 123-132, 1983.

[16] K. E. Peterson, “Silicon as a mechanical material,” Proc. IEEE, vol. 70,
no. 5, pp. 420-457, 1982.
[17] W. G. Regehr, S. B. Kater, J. Pine, “A chronic in vitro microdevice-neuron
connection,” Soc. of Neurosci., 17th Annual Meeting, Abstract 393.15, 1987.

Chapter 2
Stimulating and Recording from Cultured Neurons

with Glass Loose-Patch Electrodes

2.1 Background

Preliminary experiments were performed with glass patch-pipettes to de
termine what seals and electrode diameters would be required for (1) recording

action potentials with good signal-to-noise ratios, (2) stimulating, (3) verify

ing the stimulus with the same electrode, and (4) recording subthreshold post

synaptic potentials. Patch electrodes were used in a method that combines fea
tures from the loose-patch clamp [1], and extracellular stimulation and recording
techniques that have been used both in vivo and in vitro.

Figure 2.1. Loose-patch electrode in (a) voltage-clamp (b) current-clamp

The loose-patch clamp method was developed to map the ,channel distribu
tions in cells without requiring large seal resistances. Glass pipettes are brought

i contact with the cell membrane until a seal of 1-3 MΩ is obtained. The pipette
is operated in voltage-clamp mode as in Figure 2.1 (a). Voltajge steps are applied
to the pipette, the potential of the patch membrane is changed, and current

flows through channels in the patch. Using large diameter glass electrodes to
record from a 50-200 μm2 patch containing thousands of channels, the resulting

membrane currents are large enough to record with good signal-to-noise, without
requiring gigaohm seals. A combination of analog and digital techniques is used
to correct for current lost through the seal resistance. The loose patch clamp

has been used primarily to map channel distributions on muscles [2]-[4].

We used patch electrodes of comparable diameters and had seal resistances
in a similar range to those of loose patch clamping. However, our primary goal

is to record the intracellular potential and to stimulate cells. Also, the final
electrode design utilizes metal electrodes, precluding voltage clamping the patch

membrane (see Chapter 4). Consequently current clamping as in Figure 2.1(b)
was used rather than voltage clamping which is used in conventional loose-patch
recording.

Tests with glass electrodes were performed on both Helisoτna neurons and
SCG’s: to measure typical seal resistances that could be attained without ap

plying suction; to determine breakdown voltage of the cell membrane and the

maximum amplitude of the stimulus current pulse that could be passed through
the electrode, without damaging the cell; to compare the intracellular potential
to the signal recorded using the patch electrode; to determine appropriate stim
ulus paradigms; and to investigate the seals required for whole-cell recording,

breaking down the cell membrane with a current pulse, and reversibly breaking
down the cell membrane as an approach to obtaining improved signal-to-noise
ratios for subthreshold events.

2.2 Cultured Identified Helisoma Neurons

Identified Heli30ma snail neurons were used for initial electrode tests since
they are large (cell bodies are about 50 μm in diameter), robust, and have been

studied extensively both in vivo and in vitro [5]. Neurons were taken from the
buccal ganglia which innervates the muscles of the buccal mass and controls

feeding activity. Based on their size and location in the ganglia it is possible
to identify individual neurons in the ganglia. Electrode tests were performed

on two types of neurons: neuron B5 whose axons branch extensively over the
foregut, and neuron B19 which provides chemical excitatory input to a large

muscle in the buccal mass. Each neuron type has its own distinct physiology,
and morphology, and forms unique connections with other neurons.

2.21 Culture Techniques
Animals from inbred stocks of Heli30ma trivolvis were dissected under sterile

conditions and their buccal ganglia were removed [10].

Buccal ganglia were

treated with 0.2% trypsin (Sigma III) for 30 minutes followed by a 15 minute

rinse in trypsin inhibitor. Trypsin softened the connective tissue within the
ganglion and in the outer ganglionic sheath. Ganglia were pinned to a Sylgard

dish and identified neurons B5 and B19 were exposed by making an incision in
the ganglionic sheath with a tungsten microknife. Then, while observing through
a dissecting scope at a magnification of about 90X, neurons were removed using a
micrometer-drive syringe and a micropipette controlled by a micromanipulator.

Identified neurons were placed in polylysine-coated 35 mm Falcon dishes

containing conditioned medium [7]-[9], which contains factor(s) necessary for
process outgrowth in isolated neurons. It is produced by incubating two adult

Heli30ma brains/ml of Heli30ma defined medium for 72 hours prior to cell plating.

Helisoma defined medium consists of 50% Leibowitz L-15 medium without inor-

Figure 2.2. Helisoma B19 neuron after 3 days in culture.

ganic salts, and 50% salt, buffer, and antibiotic solution in distilled water with

an osmolarity of 130mOsm. The final concentrations are 40 mM NaCl, 1.7 mM
KC1, 4.1 mM CaCl2, 1.5mM MgCl2, 5mM Hepes, with 50μg∕ml gentamycin

sulfate. Helisoma saline consists of 51.3 mM NaCl, 1.7mM KC1, 4.1 mM CaCl2,
1.5mM MgCl2, and 5mM Hepes (pH 7.3) distilled water. Antibiotic saline was
made by autoclaving Helisoma saline, and adding gentamycin sulfate to a final
concentration of 150 μmg∕ml.
Normal cell growth is shown in Figure 2.2. Initially the cell is simply a

sphere that has had all of its processes removed. After several hours in culture
cells “veil” and grow processes. Growth continues for several days and then
stops and the cell is said to be stable state, as indicated by “phase bright”

growth cones.

2.22 Electrophysiology of Cultured B5 and B19 Neurons

Morphologically and electrophysiologically Helisoma B5 and B19 neurons

differ both in vivo and in vitro [6],[10]. B5 neurons have a larger total neu-

rite outgrowth, higher growth rate, and more filopodia per growth cone than

B19 neurons. Serotonin applied to the growth cones of B19 neurons causes them
to retract, while having no effect on the growth cones of B5 neurons [11]. Levels

of serotonin as low as lμM cause B19 neurons to fire and cease growing, while
having no effect on B5 neurons. The action potential half-width for cultured

neurons is 15.6 ± 5.8 msec for B5,s and 6.7 ± 1.2msec for B19,s [6]. Figure 2.3
shows action potentials typical of B5 and B19 neurons. The half-width of action
potentials of cultured neurons increases with their time in culture.

2.3 Cultured Rat Superior Cervical Ganglion Neurons (SCG’s)
Cultured SCG’s provide an excellent preparation for initial studies [12]—[13] :
(1) the well defined trophic factor necessary for long-term growth, nerve growth

factor (NGF), is readily available, (2) ganglia are discrete, easily obtainable

containers filled with a large number of relatively homogeneous neurons, (3) they
are relatively immature at birth, facilitating dissociation and culturing, (4) they

have been extensively studied and characterized both in vivo and in vitro, and
(5) the study of such mammalian neurons compliments the study of Helisoma

invertebrate neurons.
2.31 Culture Techniques
Neonatal rats less than one day old were decapitated just above the arms and

the superior cervical ganglia were removed, cleaned of surrounding tissue, and

placed in a dish containing 4 ml of sterile Hanks solution. 200 μl of 2.5% trypsin
was added, and it was allowed to incubate at 37oC in 5% CO2 for 30 minutes.

Then the ganglia were removed from the trypsin medium, placed in a dish con
taining serum, and each ganglia was pulled apart with tweezers. The contents

of the dish were transferred to a 15 ml centrifuge, triturated 30-50 times, and

Figure 2.3. Electrophysiology of cultured Helisoma (a) B5 and (b) B19 neurons.

filtered through a 44/zm nylon filter. After spinning at 1400r.p.m. the super
natant was removed and the cells suspended in plating medium at a density of

3 × 104 cells/ml. Cells were then added dropwise inside glass rings in culture
dishes containing complete growth medium.
The dishes used to culture the neurons were 35 mm Falcon dishes. Two

different surfaces were used: polylysine, or extracellular matrix from bovine en-

15
dothelial cells [14]. The culture medium used has been described elsewhere [15]
Figure 2.4 is an example of a mass culture of SCG’s.

Figure 2.4. Normal growth of a mass culture of SCG neurons.

20 mV

θ∙4 ∩A

2.32 Electrophysiology of Cultured SCG’s Neurons

Figure 2.5. Intracellular electrophysiology of cultured SCG’s.

16
SCG’s has been studied extensively in cell culture. Two weeks after plating

the cells are large enough to reliably penetrate with an intracellular electrode as
shown in Figure 2.5. The whole-cell patch can be used with younger cells [16],
but with time in culture in a serum containing medium the cells become more

difficult to seal to as the surfaces become coated with material from the medium.
Membrane currents have been studied using a two-electrode voltage clamp [17].

While SCG’s rarely interconnect in intact animals, they form extensive connec
tions in dissociated cell culture [18].

2.4 Patch Electrode Tests
Patch electrodes were made from soda-lime glass (Kimble 73811). Electrode
diameters were varied by changing the heat on the second pull. Pipettes were

filled with extracellular salts that had been filtered to 0.2 μm . These solutions
were used to closely replicate the solution between the electrode and the cell for

diving-board and dish electrodes. Intracellular electrodes were pulled on a Brown

and Flaming P77 pipette puller using 1.0 mm o.d. omega dot borosilicate glass

(WPI 1B100F-4). Electrode and seal resistances were determined by injecting a
current pulse of known amplitude, and measuring the resulting voltage drop.

As expected the seal resistance varied roughly as the inverse of the electrode
diameter, and resistances were in the range 1-3 MΩ for a 10 μm diameter pipette
tip, for SCG’s and 2-6 MΩ for Helisomα neurons. ‘Due to higher resistivity extra

cellular solution the electrode and seal impedances were larger for the Helisoma
studies. The resistivity of media used for SCG’s is approximately 100ohm-cm,

and for Helisoma neurons approximately 300ohm-cm.
Next the breakdown voltage of the patch beneath the membrane was deter
mined. First the electrode impedance and then the seal resistance was measured.

Then current pulses of progressively larger amplitude were applied. An intracel

17
lular electrode was used to determine when the cell membrane was broken down.

For positive going 2 msec pulses the breakdown voltage was determined to be
400 mV ±50 mV (n=8) for SCG’s, with similar results obtained for Helisoma
neurons. There was not a large dependence of breakdown voltage on polarity or
on pulse length for pulse lengths of 0.2 msec-50 msec.
Next, the intracellular potential was compared to the patch electrode sig

nal. This necessitated an intracellular electrode in addition to the loose-patch
electrode. The patch pipette was pressed against the cell body to obtain a stable

seal, then an intracellular electrode was used to penetrate the cell. The intracel
lular electrode was used both to stimulate the cell, and to record the resulting

action potential. This intracellular response was compared to that recorded with

the loose-patch pipette. Signals recorded on the patch electrode were essentially
derivatives of the action potentials (see Chapter 4). For each cell type, Heli-

soma B19, Heli30ma B15, or SCG, the magnitude and duration of the signal

recorded on the patch pipette was very different. This can be attributed to the

fact that both the channels contained within the patch and the rate of rise of

an action potential are very cell dependent. As seen in Figure 2.6, patch signals
recorded from Helisoma B19 neurons are larger than for Helisoma B5 neurons,

due to the faster rate of rise of a B19 action potential.

In the next experiment a cell was stimulated with a patch pipette and an in
tracellular electrode was used to verify the stimulus. Using patch electrodes with

a diameter of 10 μm facilitated stimulating neurons without breaking down the
cell membrane. If significantly smaller pipettes were used stimulation through

the patch was not possible without breaking down the cell membrane. Many

different types of stimuli were used to fire neurons. Although stimulation with
long pulses could be accomplished easily, primarily short stimuli were used since

18
(a) Helisoma B19

Figure 2.6. Neurons stimulated with a current pulse passed through an intracellu
lar electrode (40-50 MΩ), and the response of the cell recorded with the intracellular
electrode, and the loose-patch electrode (diameter ≈ 10μm). (a) Helisoma B19 neu
ron (Rseα∕ = 3.5 MΩ, bandwidth 10Hz-lkHz). (b)A Helisoma B5 neuron (JRseal =
4.0 MΩ, bandwidth 10Hz-lkHz). (c)SCG neuron (R3eaι = 3.0 MΩ, bandwidth 10Hz1kHz).

19
when using a metal electrode, such as the diving-board or dish electrodes, long
voltage pulses will charge up the electrode to potentials large enough to produce

gas. Figure 2.7 shows an example of some of the many stimulus experiments that
were done using SCG’s, and Helisoma B19 and B5 neurons. In 7 (a) a 1 msec long

pulse is used to stimulate a Helisoma B19 neuron, with four traces superimposed.

In 2.7 (b) a 2 msec long pulse is used to stimulate a Helisoma B5 neuron. This
is an example of near-threshold stimulation. If the amplitude of the stimulus
were increased slightly a more reliable stimulus as in 2.7 (a) would be obtained.

However, if the amplitude were further increased by more than about 30% then

the potential would be sufficient to break down the cell membrane. Figure 2.7 (c)

is an experiment in which an SCG is stimulated. Four successively larger stim
uli were used: the first two resulted in no discernible intracellular response, the
third resulted in a 5 msec depolarization, and the fourth stimulated the neuron.

This demonstrates that the response is all or none, as an action potential should
be. Based on these experiments it is possible to reliably stimulate neurons, even

using fairly short stimulus pulses.

In addition to the stimulus experiments of Figure 2.7 negative stimulus
pulses were used. Such negative pulses could be used to reliably stimulate SCG
neurons, but not Helisoma neurons. Figure 2.8 is an example of a patch elec

trode stimulating an SCG, recording the resulting action potential with the same

electrode, and using an intracellular electrode to verify the response. This is
promising because it suggests that using the same electrode for both stimulation
and verification of stimulation is possible. However, the diving-board electrode
and dish electrode are metal, with a complex impedance that has a time constant

extending the stimulus artifact for many milliseconds after the stimulus current
pulse. If the same electrode is to be used for both stimulating and recording

20
the resulting action potential a combination of digital and analog, techniques
similar to those employed in the loose patch clamp, must be used to remove the

stimulus artifact.
In order to improve the signal-to-noise ratio and obtain essentially an in

tracellular connection using a patch electrode it is possible to break down the

membrane and get inside the cell as in whole cell patching [19]. But there is a

problem: if seals are only of the order of several megohms and a hole is put into
the cell beneath the patch electrode then the cell becomes very leaky, the resting

potential of the cell deteriorates, and the cell dies. However, if very short current
pulses of appropriate amplitude are used, the membrane can be reversibly broken

down, leaving a 50 μm2 patch with a resistive leak of about 200 MΩ [20]-[21].
An illustration of the impressive signal-to-noise ratio that can be obtained with
a seal resistance of 3MΩ is given in Figure 2.9. At time t = 0 a 0.5 msec current
pulse of 225 nA was passed through the patch electrode, resulting in a 400 mV

drop across the patch and breaking the membrane down. Five seconds later

the intracellular signal is essentially an intracellular signal that has been atten

uated by a factor of 50, with Raeaι and Rpatch forming a voltage divider so that
V0ut ∞ Rseal∕(Rpatch ÷ Rseal) (see Chapter 4 for a detailed explanation). Based

on this, at t = 5 sec Raeaι ≈ 150 MΩ. At time t = 2 min. the signal recorded by
the patch electrode is much smaller, indicating that the hole in the patch has
begun to seal (Raeaι ≈ 750MΩ). By t = 4 min. the patch has essentially sealed
and only a small differentiated signal is recorded by the patch electrode.
The results of such experiments suggest a technique allowing even electrodes

with seals as low as several megohms to record essentially intracellular signals;
it would be possible to peak inside a cell and record from it for a short time,

let the patch seal, and then come back at a later time and repeat the process

Stimulation of neurons using a current pulse passed through a loosepatch electrode (diameter ≈ 10μm). to stimulate and recording using an intracellular
electrode, (a) Helisoma B19 neuron, ∆i3tiτn = 1 msec, Istiτn = 100 nA, Raeaι — 3MΩ.
(b) Helisoma B5 neuron, ∆f3iim = 2 msec, Iatim = 100 nA, Raeaι = 3MΩ, stimulated
at 0.1 Hz. (c) SCG neuron, ∆f3i,∙m = 0.5 msec, Iatim = 100 nA, Raeαl = 2MΩ.
Figure 2.7.

Figure 2.8. Stimulation of neurons using a current pulse passed through a loose
patch electrode to stimulate and record, and using an intracellular electrode to record,
∆tsijτn = 2 msec, Iatim = -200 nA, Raeαl - 1.5 MΩ, Vpatcfl = -300 mV.

when desired. However, for large diameter pipettes even a hole put in a cell by
this reversible breakdown technique is large enough to seriously damage the cell;

calibrating the signals recorded through a patch membrane with variable seal
resistance is difficult; and the range of membrane voltages resulting in reversible
breakdown is small and it is easy to exceed this range and damage the cell.
For these reasons the approach adopted for the diving-board electrode was to

try to get large seals to facilitate whole-cell recording. This is a reasonable
approach, since gigaohm seals have been obtained in some preparations using

glass patch pipettes without providing suction. Failing this it would still be
possible to noninvasively record action potentials and stimulate neurons with

the diving-board electrode, with the use of reversible-breakdown existing as a
backup technique.

Figure 2.9. Rat SCG neuron stimulated with a current pulse passed through an
intracellular electrode, and the response of the cell recorded with the intracellular
electrode, and the loose-patch electrode at times t = 5 sec, t = 2 min, t = 4 min.
(72seα∕ — 3.0 MΩ, dpatch = 7μm, at time t — 0 patch membrane broken down with
a +130 nA, 0.5 msec, current pulse, bandwidth 10Hz-lkHz, resistance of intracellular
electrode 140 MΩ.

References

[1] W. Stuhmer, W. M. Roberts, and W. Aimers, “The loose patch clamp,” in
Single Channel Recording, ed. B. Sakmann, and E. Neher, Plenum Press, New
York, pp. 123-132, 1983.
[2] J. H. Caldwell, D. T. Campbell, and K. G. Beam, “Na channel distribution
in vertebrate skeletal muscle,” J. Gen. Physiology, vol. 87, pp. 907-932, June
1986.
[3] W. Aimers, W. M. Roberts, and R. L. Ruff, “Voltage clamp of rat and hu
man skeletal muscle: measurements with an improved loose-patch technique,”
J. Physiol., vol. 307, pp. 751-768, 1984.
[4] R. E. Weiss, W. M. Roberts, W. Stuhmer, and W. Aimers, “Mobility of
voltage-dependent ion channels and lectin receptors in the sarcolemma of frog
skeletal muscle,” J. Gen. Physiol, vol. 87, pp. 955-983, June 1986,.
[5] S. B. Kater, “Dynamic regulators of neuronal form and connectivity in the
adult snail Helisoma,” in Model Neural Networks and Behavior, Plenum pub
lishing Corp., 1985.
[6] C. S. Cohan, P. G. Hay don, and S. B. Kater, “Single channel activity differs
in growing and nongrowing growth cones of isolated identified neurons of
Helisoma,” J. Neurosci. Res., vol. 13, 1985.
[7] D. L. Barker, R. G. Wong, and S. B. Kater, “Separate factors produced
by the CNS of the snail Heli30ma stimulate neurite outgrowth and choline
metabolism in cultured neurons, J. Neurosci. Res., vol. 8, pp.419-432, 1982.
[8] R. G. Wong, R. D. Hadley, S. B. Kater, and G. Hauser, “Neurite outgrowth in
molluscan organ and cell cultures: the role of conditioning factor(s),” J. Neu
rosci., vol. 1, pp. 1008-1021, 1981.
[9] R. G. Wong, D. L. Barker, and D. A. Bodnar, “Nerve growth-promoting
factor produced in cultured media conditioned by specific CNS tissues of snail
Heli30ma, Brain Res., vol. 292, pp. 81-91, 1984.
[10] P. G. Haydon, C. S. Cohan, D. P. McCobb, H. R. Miller, and S. B. Kater,
“Neuron-specific growth cone properties as seen in identified neurons of Helisoma,” J. Neurosci. Res., vol. 13, pp. 135-147, 1985.
[11] P. G. Haydon, D. P. McCobb, and S. B. Kater, “ Serotonin selectivity in
hibits growth cone motility and synaptogenisis of specific identified neurons,”
Science, vol. 226, pp. 561-564, 1984.
[12] S. C. Landis, “Environmental influences on the development of sympathetic
neurons,” in Cell Culture in the Neurosciences, eds. J. E. Bottenstein and
G. Sato, Plenum Press, New York, pp. 169-192, 1985.
[13] H. Burton, and R. P. Bunge, “The expression of cholinergic and adrenergic
properties by autonomic neurons in tissue culture,” in Excitable Cells in Tis
sue Culture, eds. P. G. Nelson, and M. Lieberman, Plenum Press, New York,
pp. 1-29, 1981.
[14] D. K. MacCallum, J. H. Lillie, L. J. Scaletta, J. C. Occhino, W. G. Frederick,
and S. R. Ledbetter, “Bovine corneal endothelium in vitro: elaboration and

organization of a basement membrane,” Exp. Cell Res., vol. 139, pp. 1-13,
1982.
[15] R. E. Mains, and P. H. Patterson, “Primary cultures of dissociated sympa
thetic neurons. I. Establishment of long-term growth in culture and studies of
differentiated properties, J. Cell Biol., vol. 59, no. 329, 1973.
[16] J. M. Nerbonne, A. M. Gurney, and H. B. Rayburn, “Development of the
fast, transient outward K+ current in embryonic sympathetic neurons,” Brain
Research, vol. 378, pp. 197-202, 1986.
[17] J. E. Freschi, “Membrane currents of cultured rat sympathetic neurons under
voltage clamp,” Journal of Neurophysiology, vol. 50, no. 6, Dec. 1983.
[18] C.-P. Ko, H. Burton, M. I. Johnson, and R. P. Bunge, “Synaptic transmission
between rat superior cervical ganglion neurons in dissociated cell cultures,”
Brain Research, vol. 117, pp. 461-485, 1976.
[19] A. Marty, and E. Neher, “Tight-seal whole-cell recording,” in Single Channel
Recording, ed. B. Sakmann, and E. Neher, Plenum Press, New York, pp. 107122, 1983.
[20] U. Zimmerman, “Electric field-mediated fusion and related electrical phe
nomena,” Biochimica et Biophysica Acta, vol 694, pp. 227-277, 1982.
[21] U. Zimmerman, and J. Vieken, “Electric field-induced cell-to-cell fusion,”
J. Membrane Biol., vol. 67, pp. 165-182, 1982.

Chapter 3

Diving-Board Electrode Fabrication

3.1 The Idea
The diving-board electrode as shown in Figure 2.1 is a silicon microdevice

used in a manner similar to a loose-patch electrode. Each electrode is manipu
lated into place and is then glued to the bottom of the tissue culture dish.. A

seal is made between the lower surface of the electrode and the cell in the same

way that a seal is made to the bottom of a glass pipette, although suction can
not be applied when using this device. This establishes a one-to-one connection

between a device and a cell for both stimulating and recording.

Figure 3.1. Schematic of a diving-board electrode in contact with a cell.

A long flexible arm makes gentle contact with a cell, while its silicon support

pedestal is permanently mounted to the bottom of the culture dish. A gold

strip sandwiched between two insulating layers leads to a cup-shaped structure
at the end of the diving board. Here the diving board makes a seal to the

neuron, and electrical contact is made by a platinized gold electrode. An insulated
gold wire extends vertically from the device out of the solution to connect to

the electronics. In principal diving-board electrodes can be placed on many
neurons simultaneously, and after doing so the micromanipulators are free for

conventional physiology.
3.2 Micromachining of Silicon

Integrated circuit-technology and micromachining of silicon provide power
ful tools to fabricate very small three-dimensional structures with well defined

mechanical properties [l]-[4], such as the diving-board electrode. These meth
ods have the advantage of being well established for use in the semiconductor
industry, and allow batch fabrication of many devices on one wafer with feature

sizes down to a micron.

In order to build such three-dimensional structures wet-etching techniques
are indispensable. Figure 3.2 shows the characteristic etching of several such
techniques. Figure 3.2(a) and 3.2(b) show a wafer first patterned by etching

a SiO2 masking layer, and then etched by an isotropic etchant.

HF∕HNO3

∕CH3COOH [5] is an example of such an isotropic etch: one with an etch rate

unaffected by crystal orientation.

There are also anisotropic etches that etch along different crystal planes at
very different etch rates. This orientation-dependent etching can be explained
by considering the lower order crystal planes of silicon, which is crystallized in

the diamond structure. Ethylene-diamine/pyrocatechol/water (EDP) etch [6],[7]
has several properties that make it indispensable in the fabrication of many

micromechanical devices: the etch rates in the (100):(110):(lll) directions are
about 50:30:1 μm∕hour∙, and the relative etch rate of gold, S1O2, Si3N4, and heav
ily doped boron regions is extremely small. Figures 3.2 (c) and 3.2 (d) show EDP

ISOTROPIC ETCHING

b) side view

ANISOTROPIC ETCHING (100) ORIENTATION WAFER
d) side view

ANISOTROPIC ETCHING (110) ORIENTATION WAFER
e) top view

f) side view

Figure 3.2. Isotropic etching of silicon using an etchant such as HF-Nitric-Acetic

(a) top-view (b) side view. Anisotropic etching on (100) surfaces using EDP, (c) top
view (d) side view. Anisotropic etching on (110) surface using KOH∕H2O (e) top view
(f) side view.

etching (100) orientation silicon that has a S1O2 mask. EDP etches down until

a (111) plane is reached which serves as an etch stop. Even when the opening
in the oxide is circular the resulting etching pattern is an inverted pyramid with
the (111) planes making an angle of approximately 54.7° with the (100) surface
which it intersects in perpendicular (110) planes.

There are also orientation-dependent etches that are used on (110) silicon

wafers as shown in Figures 3.2(e) and (f). KOH∕H2O [8] is the primary etchant
used in such applications. It etches silicon in the 100:110:111 directions at rates

of 60:400:1 μm∕hour [9]. Such etches attack thermal SiO2 masks at 0.2 μm∕hour,

therefore Si3N4 which is attacked extremely slowly, is used as a mask. As is seen

in Figure 3.2’using a circle for a mask will result in a structure as shown, with

(111) planes serving as etch stops. However, if a properly oriented mask is used
it is possible to etch vertical grooves completely through the wafer, bounded by

(111) planes.

While the mechanisms responsible for orientation-dependent etching have

not been studied extensively they can be qualitatively understood. The etching
of silicon by any etch proceeds via essentially the same course: (1) Si is oxidized
to Si+; (2) OH- attaches to the positively charged group; (3) hydrated silicon

reacts with complexing agents; and (4) reaction products dissolve into solution.

There are two principle ways in which differences in crystal planes can affect these
reaction rates. First, different crystal planes have different atomic densities and

water effectively screens reactive species much better for high atomic densities.
Since the atomic packing densities of the different crystal planes are (111) >

(100) > (HO), one would expect (111) to etch more slowly. Second, the energy to

move an atom is dependent upon orientation and the number of dangling bonds.

On this basis the (111) orientation, with only one dangling bond, is expected

to etch much more slowly than either the (100) or the (110) orientations, which
both have two dangling bonds.

The dopant dependence of etching is also poorly understood. When silicon

is doped to boron concentrations in excess of 5 × 1019 cm-3 the etch rate is
reduced in EDP by a factor of 50 [7], and similarly reduced in KOH∕H2O. But

for“Dash” etch [10], a mixture of HF∕HNO3∕CH3COOH the etch rate is actually
increased by a factor of 20 for such heavily doped regions. There are two factors
that could affect the etch rate: first, such heavily doped regions are in strong

tension with the smaller boron atom replacing the larger silicon atom, increasing
the binding energy, and increasing the etch rate as is seen in the case of EDP and
KOH/H2O; and second, the availability of holes would tend to make the reaction

proceed more quickly as in the case of “Dash” etch. Qualitatively such different
etch-rate effects can be explained by different steps being the rate-limiting steps
for different etchants.

Figure 3.3. Etching of (100) silicon with heavily boron-doped regions.

In the fabrication of the diving-board electrode the anisotropic etch of choice

was EDP used on (100) orientation wafers. The fabrication required an etch

31
that would attack insulators such as thermal S1O2 and SiOrNy very slowly;

would not etch heavily boron-doped regions; and had a well defined undercut
rate. Figure 3.3 demonstrates how the etch resistance of boron diffusion can be

exploited. A heavily boron-doped epitaxial region followed by a layer with lower
doping levels serves as' an etch-stop layer and surface doping defines regions that
aren’t etched by EDP.

Figures 3.4(a),(b) show how it is possible with a system such as EDP on

(100) orientation wafers to define a free standing cantilever structure. The con

vex comer of the diving board is readily attacked by EDP; only (111) planes
that intersect in concave corners are protected from EDP. However, atoms at

convex corners are exposed and readily attacked by EDP. Figure 3.4(c) shows

the undercut proceeding along the (331) planes, with
. (rate of undercut)/(etch rate in the (100) direction) ≈ 0.6 [10].

Wet chemical etching techniques with great selectivity, are extremely easy to
use. While useful anisotropic etching methods have been obtained using such

techniques as reactive-ion etching, such methods tend to have much slower etch

rates and their use is generally limited to smaller geometries.

3.3 Insulators for Saline Environments
A necessary fabrication requirement is a means of insulating the electrode

lead that is both biocompatible and capable of standing up to saline for the

duration of the experiment. In addition it may be necessary to pattem the

passivation layer. Such long-term passivation of sensors remains a difficult prob

lem [11] without a standard solution. Choosing an insulation is a compromise

between equipment availability, cost, fabrication time, and film properties.
The available insulating layers can be grouped into polymers and dielectrics

such as oxides and nitrides. The success of polymer encapsulants requires that

∖',

κv z
Jf
----- ∖

⅝,

Figure 3.4. Etching of a cantilever in (100) silicon using EDP. (a) Etch not complete
showing partial undercut of the cantilever (b) Etch complete and cantilever is free
standing over a pyramidal pit bounded by (111) planes, (c) Undercut of a corner using
EDP and (100) silicon.

33
the insulating layer adhere well to the substrate [12]. Parylene C [13] is a va

por deposited polymer with low water vapor permeability [14]. Various devices
insulated with Parylene C have survived for a matter of months. However, the

adhesion to gold is poor; Parylene coatings are susceptible to mechanical stress;
a dedicated deposition system is required; and the films are difficult to pattern.

Silicone rubber is a polymer that has been used successfully for passivating lead
wires in saline for several years [15]. Polyurethane, is stiffer, with a higher tear
strength than silicon rubber [16], but which is susceptible to stress cracking.

Polysiloxane resin Dow Coming DC648 has been used for up to 2 months [16][19]. Patterns were defined by laser deinsulation of the electrode sites, a tech
nique that is inferior to photolithography.

There are a large number of polymers with desirable properties that are

easily patterned, without requiring expensive deposition systems. Layers can be
spun on, with layer thickness determined by viscosity and spin speed. Positive
photoresists such as Shipley 1350J have been used successfully for days [20],

but scratch resistance is poor, and after several days adhesion is a problem and
insulation is compromised. Negative photoresists such as KTFR and KTI732
(KTI Chemicals, Sunnyvale) have been used to insulate electrode leads for up
to 3 months [21,22]. Dupont pyralin 2555 is a polyimide that has been used

for 25 days in saline using adhesion promoter VM-651 [23]. We have found ad

hesion to be better with an aluminum oxide adhesion layer. A thin aluminum

oxide layer is applied to the wafer either by using Hitachi PIQ-Coupler-3 or
by evaporating 30Â of aluminum. The thin layer of aluminum oxidizes at the

temperatures necessary to cure the polyimide, becoming transparent and non-

conductive. Pyralin 2555 is not photosensitive, and must be patterned with
a photoresist such as 1350J. Fabrication time is reduced by using a negative

photosensitive polyimide, MRK Selectilux HTR 3-50. Using a 30Â thick layer
of aluminum oxide both Selectilux and pyralin 2555 insulated dishes have been
used for up to 3 months without deterioration of the insulation. None of these

insulating layers are very scratch resistant, and their adhesion is poor in etches
such as EDP.
Other dielectrics such as silicon dioxide and silicon nitride can be used for
insulations. These films can be deposited via several different techniques, with

film quality being very process dependent. Many dielectrics can be deposited
by evaporation to form thin films [28]-[29]. A source is heated with either an
electron beam or a resistor, it vaporizes, and then transverses the space between

the source and the substrate to condense and form a thin film on the substrate.
Since evaporations are typically conducted at low pressures (about 1 × 10"6 torr)

from essentially point sources, film deposition is very directional, resulting in

nonconformal films with poor step coverage.
At high temperatures in either steam or O2 thermal SiO2 can be grown on

silicon [14]-[15]. This is a very high quality oxide, being attacked very slowly
by etches such as EDP, and having been used as an insulting layer in saline

environments [11]. However, the usefulness of such oxides is limited, since they

must be grown at very high temperatures on bare silicon surfaces.

Insulating oxide films can be electrolytically deposited [11] ,[24]-[25]. Tanta
lum pentoxide is the most promising of these, having been shown to be stable in
vivo and insoluble in saline. However, tantalum has disadvantages: it cannot be

thermally evaporated due to the very high temperatures required (> 3000oC),
so that e-beam evaporation is necessary; and it is not possible to directly wire

bond to tantalum.
Sputtering can also be used to deposit dielectrics [30]-[34]. An RF power

35
source is used to accelerate ions, usually Ar+, towards a target. These ions strike

the target and dislodge primarily neutral target atoms which condense to form

thin films. Sputtering has many advantages over other deposition techniques:

a wide variety of materials can be deposited including insulators, alloys, and
refractory materials; adhesion is generally good; and substrate deposition tem
peratures are low. When using an inert carrier gas such as argon the material
to be deposited must be available in sheet form. It is also possible to deposit a

compound made up of the target material reacted with the sputtering gas. For

example oxides such as S1O2 are formed by sputtering with a silicon target with
a sputtering gas that is a mixture of oxygen and argon, and nitrides such as
Si3N4 are formed using either NH3 or N2 as the active gases. While film quality

and step coverage is typically better than that of evaporated films, dielectrics

deposited by chemical vapor deposition (CVD) are generally superior.
CVD coatings are in general of high quality, and etch resistant, with good
step coverage [29], [35]-[36]. In this process gases at a certain temperature react

to form a solid. For example at 400° C
3SiH4(<,) + 4O2(3) → SiO2(5) + 12H2(3)∙

While CVD S1O2 can be deposited at temperatures low enough to be compatible
with metals such as gold, such layers tend to have poor step coverage and pinhole

defects are a problem due to particles loosely adhering to the reactor walls [29].
Such CVD S1O2 layers have been used to insulate electrodes for several weeks [37].
At 700oC-1150oC and atmospheric pressure ammonia and monosilane react to

form silicon nitride
3SiH4(3) + 4NH3(3) → Si3N4(3) + 12H2(<,).
Silicon nitride layers produced in this manner are excellent diffusion barriers to

water and sodium, and have been used to encapsulate an MOS capacitor im

mersed in saline for one year with no change in device characteristics [11]. Film
uniformity and step coverage due to gas streaming in CVD films produced at

atmospheric pressure can be improved by depositing at low pressures (LPCVD).

However, the high temperatures required for film deposition are undesirable:
differences in thermal expansion coefficients result in large stresses, impurities
diffuse, materials can form undesirable alloys, and metals such as gold and alu

minum cannot be used. Such fabrication difficulties can be overcome using metals
such as tantalum, and very large stresses due to thermal expansion mismatch
by depositing alternating layers with thermal expansion coefficients chosen to

give the desired stress [38].

However, this complicates fabrication consider

ably, and it is difficult to reliably have two very large stresses (greater than
1 × 101° dynes/cm2) cancel each other.

Using a plasma, rather than thermal energy, to provide the activation en

ergy for reactions facilitates deposition of high quality films at low tempera

tures [36],[39]-[41]. An RF discharge accelerates electrons which excite ions and
neutral particles through collisions which react to form a solid on the substrate

surface. The electrons and ions are not in thermal equilibrium, with the elec
trons being at much higher temperatures than ions, neutral particles, and the

substrate. The reaction is very complicated, and the resulting films are amor
phous and contain large amounts of hydrogen. Consequently PECVD silicon
nitride films are less resistant to etches than CVD nitride produced at 900°C.

None the less, PECVD Si3N4 layers provides excellent scratch resistance, good
step coverage, and are an effective barrier to water and sodium diffusion. In ad
dition sihcon oxinitride can be deposited by reacting silane with a combination

of ammonia and nitrous oxide. This allows excellent control over the stress in the
deposited layer. Since PECVD films can be deposited at temperatures down to

200° C dielectrics deposited in this manner are compatible with any metallization
layer.
PECVD silicon oxinitride was chosen as the insulating layer for the diving-

board electrode based upon: machine availability, slow etch rates in EDP, low

deposition temperatures, superior adhesion and insulting properties, and abil
ity to control the stress. PECVD. of thin films is a complicated process with
many variables: temperature, relative flow rates of the different gases, power

density and frequency, and pressure. Film properties such as stress, growth rate,
refractive index, uniformity, and adhesion are machine dependent.
PECVD silicon-oxynitride films are deposited in a Pacific Western Systems
450 Vertical Parallel Plate Plasma Reactor as shown in Figure 3.5 [42]. Sub

strates are held vertically in a graphite boat to reduce contamination from par
ticles falling from the reactor walls. Graphite is used as the electrically con

ductive boat due to its low thermal expansion coefficient and low reactivity. A
custom graphite boat can hold 4" wafers, 3" wafers, 2" wafers, and substrates as

small as 1" × 1". The reactor is a vacuum sealed quartz tube contained within a
three-zone furnace capable of attaining temperatures up to 500°C. RF power to

the boat consists of pulses of 440 KHz with a pulse separation of 8 msec and a
variable duration depending on the desired average power level.
Depositions are typically performed at 1-2 Torr with total gas flow rates in

the range 1000-2000 seem. A roots blower pump [43] backed by a vane pump is
used to move such large volumes at low pressures. High flow capacity particulate
filters and dilution of process gas with nitrogen protect the pumps from damage.
Pressure is maintained using a Vacuum General Controller that compares the

desired pressure to the actual pressure and adjusts a throttle valve between
the reaction chamber and the vacuum system accordingly. Pressure is detected

PRESSURE SENSOR

VIEW
GLASS

Figure 3.5. Schematic diagram of PWS 450 used in deposition of PECVD siliconoxynitride layers.

by a capacitance manometer (MKS Baratron) that gives an absolute pressure
insensitive to the chemical composition of the gas.

The flow rates of eight gases are controlled by automatic flow controllers.
Typically SiH4, N¾, and N2O, are reacted to form SiNrOjz. Nitrogen is used

to backfill the system, and CF4 and O2 are used to clean the system.

3.4 Electrode Fabrication
The diving-board electrode is made by a five-mask process as shown in

Figure 3.6 (see Appendix for detailed procedure). The first mask defines the

39
cup structure. The outer rim of this cup forms the seal with the cell, and the
cup must be deep enough to keep the metal surface away from the surface of

the cell. The silicon is etched to a depth of 3 μm using an isotropic silicon etch:

100 HNO3 /100 CH3COOH /15 HF [5]. Once this cup structure is defined it is

necessary to use a thick photoresist with good step coverage; a 3 μτn thick layer
of Shipley 1400-37 photoresist is compatible with 3 μm cup structures and 2 μm
line widths.
The second mask defines a pedestal to which the wire bond is attached, and
which is glued to the bottom of the culture dish. While the pedestal should be

mechanically strong, it cannot be thicker than a cell, 10 μm-50 μm. The height
is determined by a boron etch-stop process [44]—[46]; therefore working with very

thin substrates is unnecessary. A 1 μm thick thermal oxide layer is grown and
then patterned with buffered HF to define the intended pedestal area. This

is followed by boron diffusion for 10 hours at 1170oC in a 5% O2 ∕ 95% N2

atmosphere that heavily dopes the substrate in what is to be the pedestal region
to a concentration exceeding 5 × 1019cm-3 to a depth of 10 μm. The boron-glass
layer on the surface of the wafer is removed by oxidizing for one hour at 1100oC

in dry O2 and etching with buffered HF.
A 1000Â thermal oxide is grown at 1000oC in steam for 10 minutes, and

the third mask defines the contact hole in this layer. In addition to being impor

tant for its insulating properties, this thin oxide is important because it touches
the cell. This surface is a high quality glass, which seems important for high
resistance seals comparable to those of patch pipettes.

Next 100 Â of chrome, S00 Â of gold, 100 Â of chrome are successively evap
orated on the wafer, and mask four defines the metal pattern. The top layer

of chrome is etched with Transene Chromium Mask Etchant for 15 seconds; the

1. Mask 1
Define Cup

2. Mask 2
Boron Diffusion

3. Mask 3
Silicon Dioxide
Insulation

4. Mask 4
Metalization

5. Mask 5
Silicon oxinitridβ
Insulation

6. Wire Bond and Final
Etch
gold wire

7. Insulate Wire bond
and Platinize

polyimide tube

negative photoresist

platinum black

Figure 3∙β. Fabrication procedure for the diving-board electrode.

Figure 3.7. Diving board electrode after removal from EDP. There would norm
be a 25 μτn diameter wire bonded to the gold pad at the back of the pedestal.

gold is removed in 30 seconds in Transene TFA Gold Etchant that has been
diluted 1:4 with water [47]; followed by another 15 seconds in the chrome etch.

Then 1 μm of silicon oxinitride is deposited by PECVD to form the top in

sulating layer [48], [49]. This layer determines the mechanical properties of the
diving board, as well as defining tabs that later hold the electrodes in place, to
protect them from damage. The final mask patterns this top insulating layer,
which is etched in a 85% CF4∕i5% O2 plasma [50]. A 7μm thick layer of Ship-

ley AZ4770 photoresist is used for this step to insure protection of the cap region
during the plasma etch.
The wafer is scribed and broken into 4 mm × 20 mm pieces, each of which is

glued to a piece of glass. Then a wire bond is made from a large gold bonding pad
on the piece of glass to each of about 40 devices. The bonds are made of gold so

that they will survive the next etching step. Now the wafer is ready to be etched

in an ethylenediamine, pyrocatechol, and water (EDP) solution [6], [7]. This

etch attacks different silicon planes at different rates, but etches (111) planes
very slowly. Furthermore it does not attack regions that are doped with boron

to levels in excess of 5 × 1019cm-3. After three hours in EDP at 100° C the
undoped silicon is completely etched from beneath each electrode. Figure 3.7

show the device after the EDP etch. Each electrode is temporarily held to the

substrate by a silicon-oxinitride grill.
Next each device is separated from the wafer and the wire bond is insulated.
Each lead wire is manually loaded into a polyimide tube (Polymicro Technologies

i.d. 120 μm, o.d. 150 μm) and the wire bond is insulated by painting with negative

photoresist. This poly imide tube can be held by tweezers and manipulated into
place without compromising the insulation. The final step is to electroplate the
electrode at the bottom of the diving board with platinum black, to produce a

Figure 3.8. A finished diving-board electrode, (a) top view of the cup structure that
is to fit above the neuron, (b) bottom view of the cup structure, (c) overview of the
electrode.

44
low impedance contact to the electrolyte [51]. By electroplating with a current

of 2 nA (about 10mA∕cm2) for 15 seconds the electrode impedance at 1 kHz
drops from 25 MΩ to below 1 MΩ, while the platinum remains thin enough that
it does not interfere with the seal.

Figure 3.8 shows scanning electron micrographs of the cup structure, and
an overview of a finished electrode. It takes an average of about 45 minutes to

fabricate and test each device. Defective electrodes are eliminated on the basis

of optical observations and impedance measurements made before platinization.
Satisfactory electrodes can be reused several times.

3.5 Mechanical Properties of Diving-Board Electrode

In addition to considering the electrical properties of the diving-board elec
trode, it is necessary to take into account its mechanical properties. A relatively

flat diving board is necessary to make it possible to simultaneously contact the
pedestal with the bottom of the culture dish and the electrode tip with the
neuron, and to make the force exerted by the electrode on the cell downward

without a lateral force that would tend to force the cell out from beneath the
diving board.

Figure 3.9. Bending due to stress in a two-layer cantilever.

However, there are stresses in each layer which may bend the diving board.

This is illustrated in Figure 9, which shows a cantilever of length I that is com
posed of two layers of thickness d3 and d2. The Young’s modulus of the thick
layer is given by K1, and the Poisson’s ratio is ι∕1. A stress σ in layer 2 bends the

beam an amount δ. For small deflections and d1

d2 [52], we can write

3Z2(1 — ι∕ι)σd2

s = —μ---

In our electrodes, the thick layer is silicon oxinitride, 200 μm long and 1 μm thick.
Young’s modulus, Fi, 1 × 109dynes∕cm2 [53] and the Poisson ratio ui is 0.2. If
d2 = 0.2 μm a stress typical of insulating dielectrics of σ = 5 × 1012 dynes∕cm2,

would cause an unacceptably large beam deflection of 100 μm. If the stress is
compressive it will cause the cantilever to bend up as in the diagram, while if

the stress is tensile the cantilever will bend down.
The present fabrication technique minimizes the deflection of the cantilever
in two ways. First, the silicon-oxinitride layer is made very rigid compared to

the other two layers by making it much thicker than the gold and the oxide layer,

and by making the surface areas of the gold and the silicon dioxide layers small
compared to the surface area of the oxinitride layer. Second, the stress of the

silicon oxinitride layer is empirically adjusted to minimize the beam deflection.
Consider the three layer structure of the diving board as represented in

Figure 3.10. Layer 1 is silicon oxinitride, layer 2 is gold, and layer 3 is silicon
dioxide. The stresses of the films on the silicon substrate are σ1, σ2, and σ3.

When the cantilever has been freed from the substrate, layer 1, which is much

thicker than the other two layers, relaxes to a stress-free state, σ* = 0. The
stresses in the other two layers become σ⅛ = cr2 — (Y2∕^1)σ1 and σ⅛ = σ3 —

(Y3∕Y1)σ1 for dι 3> d2 and di

d3, and ιq ≈ v2 ≈ z√3. The surface areas of

a. SIDE UIEID

b. TOP UIEID

Figure 3.10.

(a) Side view of three layer cantilèver. (b) Top view of three layer

cantilever.

layer 2 and layer 3 relative to layer 1 are given by ∕2 = w2∕w1 and ∕3 = w⅛∣w-i.
The beam deflection is given by:

δ=

3Z2(l-ι∕1) (σ

2d2∕2 + σ3d3 /3^ — δ2 -⅛- δ3

(2)

47
So that if it is possible to vary the stress in any one of the layers sufficiently to
have δ2 = — δ3 then the cantilever can be made flat.

It is difficult to change the stress in the gold and silicon-dioxide films in our

electrodes, but we can vary the stress in the silicon-oxinitride films by adjusting
the ratio of oxygen to nitrogen in the film. This is done by varying the flow
rates of N20 and NH3 during the PECVD process (Figure 3.11). By depositing

the silicon oxinitride layer at 350oC at a pressure of 1.5 torr, with flow rates of
NH3∕N2O∕SiH4 being 750/1150/250 [N2O]∕[N2O+NH3]=O.67), the deposited
layer had a tensile stress of 1.4 × 109 dynes∕cm2 and the resulting diving-board

electrodes are flat to a few microns.

STRESS VS GAS FLOW RATE

Figure 3.11. Controlling the stress in oxinitride films by varying the flow rates of
N2O and NH3 [54]. The films are deposited at 350oC on silicon. The x-axis is the ratio
of the N2O flow rate to the total combined flow rate of N2O and NH3. When the ratio
is zero the film is silicon nitride, and when the ratio is one the film is silicon dioxide.

49
References

[1] K. E. Peterson, “Silicon as a mechanical material”, Proc. IEEE, vol. 70, no. 5,
pp. 420-457, 1982.
[2] J. B. Angell, S. C. Terry, and P. W. Barth, “Silicon micromechanical devices,”
Scientific American, pp. 44-55, April 1983.
{3] K. D. Wise, “Silicon micromachining and its application to high performance
integrated sensors,” Micromachining and Micropackaging of Transducers, eds.
C. D. Fung, P. W. Chueng, W. E. Ko, and D. G. Fleming, Elsevier Science
Publishing Inc., New York, 1985.
[4] C. D. Fung, P. W. Chueng, W. E. Ko, and D. G. Fleming, eds., Microma
chining and Micropackaging of Transducers, Elsevier Science Publishing Inc.,
New York, 1985.
[5] B. Schwartz and H. Robbins, “Chemical etching of silicon, IV. The system
HF, HNO3, HC2H3O2, J∙ Electrochem. Soc., vol. 106, p. 505, 1959.
[6] R. M. Finne, and D. L. Klein J.Electrochem. Soc., “A water-amine-complexing
agent system for etching silicon,” vol. 114, pp.965-970, Sept. 1967.
[7] A. Reisman, M. Berkenblit, S. A. Chan, F. B. Kaufman, and D. C. Green,
“The controlled etching of silicon in catalyzed ethylenediamine-pyrocatecholwater solutions,” J. Electrochem. Soc., vol. 126, no. 8, pp.1406-1415, 1979.
[8] D. L. Kendell, and G. R. de Gruel, “Orientations of the third kind: the coming
of age of (110) silicon,” Micromachining and Micropackaging of Transducers,
eds. C. D. Fung, P. W. Chueng, W. E. Ko, and D. G. Fleming, Elsevier Science
Publishing Inc., New York, 1985.
[9] E. Bassous, “Fabrication of novel three-dimensional microstructures by the
anisotropic etching of (100) and (110) silicon,” IEEE Trans. Electron. Devices,
vol. ed-25, no. 10, pp. 1178-1184, Oct. 1978.
[10] K. E. Bean, “Anisotropic etching of silicon,” IEEE Trans. Electron. Devices,
vol. ed-25, no. 10, pp. 1185-1193, Oct. 1978.
[11] L. Bowman, and J. D. Meindl, “The packaging of implantable integrated
sensors,” in IEEE Trans. Biomed. Eng., vol. BME-33, no. 2, Feb. 1986, pp. 248255.
[12] P. E: K. Donaldson, “Experimental visual prosthesis,” IEEE Proc., vol. 120,
pp. 281-298, Feb. 1973.
[13] Union Carbide Corp., Bound Brook, NJ.
[14] G. E. Loeb, M. J. Bak, M. Salcman, and E. M. Schmidt, “Parylene as a chron
ically stable, reproducible microelectrode insulator,” IEEE Trans. Biomed.
Eng., vol. BME-24, pp. 121-128, Mar. 1977.
[15] G. E. Loeb, C. L. Byers, S. J. Rebscher, D. E. Casey, M. M. Fong, R. A.
Schindler, R. F. Gray, and M. M. Merzenich, “Design and fabrication of an
experimental cochlear prosthesis,"Med. Biol. Eng. Comput.,vol.21, pp.241254, May 1983.
[16] G. W. Gross, “Simultaneous single unit recording in vitro with a photoetched
laser deinsulated gold multimicroelectrode surface,” IEEE Trans. Biomed.

Eng., vol. 26, pp. 273-279, 1979.
[17] G. W. Gross, A. N. Williams, and J. H. Lucas, “Recording of spontaneous
activity with photoetched microelectrode surfaces from spinal cord neurons
in culture,” J. Neurose. Methods, vol. 5, pp. 13-22, 1982.
[18] G. W. Gross, W. Wen, and J. Lin, “Transparent indium-tin oxide electrode
patterns for extracellular, multisite recording in neuronal cultures,” J. Neurosci. Methods vol. 15, pp. 243-252, 1985.
[19] M. H. Droge, G. W. Gross, Μ. H. Hightower, and L. E. Czisny, “Multielec
trode analysis of coordinated, multisite, rhythmic bursting in cultured CNS
monolayer networks,” J. Neurosci., vol. 6, no. 6, June 1986.
[20] C. A. Thomas, Jr., P. A. Springer, G. E. Loeb, Y. Berwald-Netter, and
L. M. Okun, “A miniature microelectrode array to monitor the bioelectric
activity of cultured cells,” Exp. Cell Res., vol. 74, pp. 61—66, 1972.
[21] M. Kupperstein, and D. A. Whittington, “A practical 24 channel microelec
trode for neural recording in vivo, IEEE Trans. Biomed. Eng., vol. BME-28,
no 3, March 1981.
[22] D. A. Israel, W. H. Barry, D. J. Edell, and R. G. Mark, “An array of
microelectrodes to stimulate and record from cardiac cells in culture,” Am. J.
Physio., vol. 247, pp. H669-H674, 1984.
[23] J. L. Novak, B. C. Wheeler, “Recording from the Aplyδiα abdominal ganglion
with a planar microelectrode array,” IEEE Trans. Biomed. Eng., vol. BME-33,
no. 2, Feb. 1986.
[24] F. A. Lowenheim, “Deposition of inorganic thin films from solution,” in Thin
Film Processes, ed. J. L. Vossen, and W. Kem, Academic Press, New York,
1978, pp. 209-255.
[25] P. E. K. Donaldson, “The stability of tantalum pentoxide films in vivo,”Med.
Biol. Eng., vol. 12, pp. 131-135, Jan. 1974.
[26] L. E. Katz,“Oxidation,” in VLSI Technology, ed. S. M. Sze, p. 194, 1983.
[27] A. S. Grove, Physics and Technology of Semiconductor Devices, Wiley, New
York, 1967.
[28] R. Glang, “Vacuum evaporation,” in Handbook of Thin Film Technology,
ed. L. I. Maissel and R. Glang, McGraw-Hill Book Company, New York, pp. 11-1-300, 1970.
[29] A. C. Adams,“Dielectric and polysilicon film deposition,” in VLSI Technol
ogy, ed. S. M. Sze, pp. 93-129, 1983.
[30] L. Maissel, “Application of sputtering to the deposition of films,” in Hand
book of Thin Film Technology, ed. L. I. Maissel and R. Glang, McGraw-Hill
Book Company, New York, pp. 4-1-4-44, 1970.
[31] J. L. Vossen, J. J. Cuomo, “Glow Discharge Sputter Deposition,” in Thin
Film Processes, ed. J. L. Vossen, and W. Kem, Academic Press, New York,
1978, pp. 12-62.
[32] J. A. Thornton, and A. S. Penfield, “Cylindrical Magnetron Sputtering,” in
Thin Film Processes, ed. J. L. Vossen, and W. Kern, Academic Press, New
York, 1978, pp. 76-110.

51
[33] D. B. Fraser, “ The sputter and S-Gun magnetrons,”in Thin Film Processes,
ed. J. L. Vossen, and W. Kern, Academic Press, New York, 1978, pp. 115-128.
[34] R. K. Waits, “Planar Magnetron Sputtering,” in Thin Film Processes, ed.
J. L. Vossen, and W. Kem, Academic Press, New York, 1978, pp. 131-170.
[35] W. Kern and V. S. Ban, “Chemical vapour deposition of inorganic thin
films,” in Thin Film Processes, ed. J. L. Vossen, and W. Kern, Academic
Press, New York, 1978, pp. 258-320.
[36] T. Sugano, Applications of plasma processes to VLSI technology, John Wiley
& Sons Inc., New York, 1985.
[37] J. Pine, “Recording action potentials from cultured neurons with extracel
lular microcircuit electrodes,” J. Neurose. Methods, vol. 2, pp. 19-31, 1980.
[38] K. Najafi, K. D. Wise, and T. Mochizuki, “A high-yield IC-compatible mul
tichannel recording array,” IEEE Trans. Electron. Devices, vol. ed-32, no. 7,
pp. 1206-1211, July 1985.
[39] J. R. Hollahan, and R. S. Rosier, “Plasma deposition of inorganic thin films,”
in Thin Film Processes, ed. J. L. Vossen, and W. Kern, Academic Press, New
York, 1978, pp. 335-358.
[40] H. Yasuda, “Glow discharge polymerization,” in Thin Film Processes, ed.
J. L. Vossen, and W. Kem, Academic Press, New York, 1978, pp. 361-396.
[41] A. R. Reinberg, “Plasma deposition of inorganic thin films,” Ann. Rev.
Mater. Sei., vol. 9, pp. 341-72, 1979.
[42] W. R. Snow, “PWS 450 COYOTE: A vertical parallel plate plasma reac
tor for silicon nitride deposition,” Pacific Western Systems Application Note,
March 1981.
[43] R. Glang, “High-vacuum technology,” in Handbook of Thin Film Technology,
ed. L. I. Maissel and R. Glang, McGraw-Hill Book Company, New York, pp. 27-2-8, 1970.
[44] Sohio Carborundum PDS Planar Diffusion Sources: Technical Information,
1986.
[45] D. Rupprecht, and J. Stach, “Oxidized boron nitride wafers as an in-situ
boron dopant for silicon diffusions,” J. Electrochem. Soc., vol. 120, no. 9,
pp. 1266-1271, 1973.
[46] K. Najafi, K. D. Wise, and T. Mochizuki, “A high-yield IC-compatible mul
tichannel recording array,''IEEE Trans. Electron. Devices, vol. ed-32, no. 7,
July 1985.
[47] R. Glang and L. V. Gregor, “Generation of patterns in thin films,” in Hand
book of Thin Film Technology, ed. L. I. Maissel and R. Glang, McGraw-Hill
Book Company, New York, p. 7-37, 1970.
[48] C. M. Melliar-Smith and C. J. Mogab, “Plasma assisted etching techniques
for pattern deliniation” in Thin Film Processes, ed. J. L. Vossen and W. Kem,
Academic Press, New York, pp. 497-552, 1978.
[49] R. C. Gesteland, Bτ Howland, J. Y. Lettvin, and W. M. Pitts, “Comments
on microelectrodes,” Proc. IRE, vol. 47, pp. 1856-1862, 1959.

[50] D. S. Campbell, “Mechanical properties of thin films,” in Handbook of Thin
Film Technology, ed. L. I. Maissel and R. Glang, McGraw-Hill Book Company,
New York, p. 12-29, 1970.
[51] K. E. Peterson and C. R. Guarnieri, “Young’s modulus measurements of thin
films using micromechanics,” J. Appl. Phys., vol. 50, pp. 6761-6766, 1979.
[52] Figure provided courtesy of W. R. Snow and V. Dunton of Pacific Western
Systems.

53
Chapter 4

Stimulating and Recording with Current-Clamped

Loose-Patch Electrodes

4.1 Equivalent Circuit

The operation of the diving-board electrode is very much like that of a
loose-patch electrode [6]. Figure 4.1(a) shows a schematic of an electrode in

contact with a cell and an electrical equivalent circuit. The shunt resistance
Rsfl is greater than 1 G Ω and shunt capacitance Cah is approximately 10 pF, so
that for a low seal-resistance the shunt impedance can be ignored. The electrode
impedance Ze depends upon the type of electrode used. In the case of a glass

loose-patch electrode it is essentially resistive, and varies between 100KΩ and
1MΩ. For the metal diving-board electrode, Ze the impedance varies as the

area of the exposed metal and is primarily capacitive. It is desirable to have

a low electrode-impedance for three reasons. First, it facilitates measuring the
seal resistance. Second, for recording, a low impedance electrode reduces the
Johnson noise and increases the signal-to-noise ratio. Third, when stimulating,

a low impedance electrode helps avoid the problem of gas evolution.

The patch region beneath the electrode in Figure 4.1 is made up of a layer of
membrane with a capacitance estimated by assuming a membrane capacitance

of approximately lμF∕cm2∙, therefore Cmι is lpF for a 100 μm2 patch. Spanning
this membrane are n different channel types with electrochemical driving forces
½...Vn. The resistances Rχ...Rn corresponding to the each particular channel
type are voltage and time dependent. The net resistance depends upon the
channel densities, which is in general nonuniform over the cell and unknown

in the patch region. The whole-cell impedance Zm2 is also voltage and time

Figure 4.1. (a) Schematic diagram of an electrode in contact with a cell and an
approximate equivalent circuit for (b) recording and (c) stimulating.

dependent. At the resting potential of the cell Zτ∏2 can be approximated by

a capacitor and a resistor in parallel. For Helisoma neurons B19 and B5 the

resistance is typically about 50 MΩ and the capacitance 500 pF.

The seal resistance Rseaι should be as large as possible. It is determined by
the conductance of the thin film of tissue culture medium between the rim of the

electrode and the cell. For Helisoma neurons seals of 1-5 MΩ were obtained.
4.2 Recording

Figure 4.1 (b) is an equivalent circuit of a diving-board electrode being used

to record from a neuron, fι...fn are the currents passing through each particular

type of channel, and ic is the current through the membrane capacitance. The
recorded voltage Vout is given by:

½,ui = (ÿc ~t~

ichan^ Rseal

(1∙)

chan=l

rr
fr, dVceu
Vout = Cml ——— +

d,t

v~λ

∕->
cħan≈l

(KeZ∕ - Kftαn)∖ p

—-------- ) Λseαi
Λc∕ιan

Vout ~ Vcap + Vionic

(n∖

(2j

(^)

Vionic = VjVa+ ÷ Vκ+

+ ⅛ ÷ Σ 1'chan

(4)

chan=i

Figure 4.2 shows qualitatively the relative contribution of different channels
and the capacitive current to the signal recorded by the loose-patch electrode.

Vcap is the differentiated intracellular voltage seen through the capacitance of the

membrane under the electrode. For this signal to be large the action potential
must be fast. For the rising phase of the action potential the value of dVcell∕dt
is approximately 100mV∕msec for Helisoma B19 neurons, and 20mV∕msec for
Heli30ma B5 neurons. Assuming a membrane area of 100 μm2, and a seal resis

tance of 2MΩ for Helisoma B19 neurons Vcap ≈ 200μV, and for Helisoma B5
neurons Vcap ≈ 40μV.

VNa+ is the voltage drop across the seal resistance due to current flowing

through sodium channels in the patch. The sodium conductance

is voltage

and time dependent, and the shape of this curve is given in Figure 4.2 (c). The

shape of the contribution of current flowing through the sodium channels is

given by, <7τvα+(KeZZ — ½vo+)∙ The contribution per ion channel can be given by

Figure 4.2. (a) intracellular potential (b) capacitive signal through cell membrane

(c) signal through sodium channel (d) signal through potassium channel (e) signal
through non-ion-selective leakage channels.

assuming a conductance of 10 pS per sodium channel and a maximum driving
force of -50 mV, so the maximum voltage contribution per open sodium channel

is about -lμV for βseα(=2MΩ.
Vjtf+ is the voltage drop across the seal resistance due to current flowing
through potassium channels in the patch. The sodium conductance gχ+ is volt

age and time dependant, and the shape of this curve is given in Figure 4.2(d).
The shape of the contribution of current flowing through the potassium channels
is given by, <7κ+(KeZZ — Vκ+')∙ The contribution per ion channel can be given by

assuming a conductance of 10 pS per potassium channel and a maximum driv
ing force of +50 mV, so the maximum voltage contribution per open potassium

channel is about +1 μV for 72sea∕=2MΩ.
Vι is the voltage drop across the seal resistance due to current flow through
. For a leakage resistance Rι = 1 GΩ ½ = 100 μV. This signal is the intracellular

signal attenuated by an amount R3∕(Rs + β∕). By applying large current pulse

to the patch electrode it is possible to temporarily make this signal very large
by putting holes in the patch beneath the membrane and greatly decreasing the

leakage resistance. Ideally one would like this signal to be as large as possible, so
that patch recording would make it possible to record post synaptic potentials.

This must be balanced with the fact that for seal resistances less than a gigaohm
Ri can not be made too small or the cell will be severely damaged. For any given

cell type an acceptable maximum leakage resistance Rι + Rs that will not harm
the cell can be empirically determined.
While the contributions of sodium channels, potassium channels, and nonselective holes has been shown explicitly, there is also a contribution from the

other channels contained in the membrane, such as calcium channels and chlo

ride channels. In general the capacitive contribution has been found to dominate,

except when the leakage resistance has been reduced by a high voltage pulse.
Figure 4.3 compares the intracellular potential, the derivative of the intra

cellular potential, and the recording from a glass loose-patch electrode. For the
Helisoma B19 neuron of Figure 4.3(a) the signal is essentially a derivative, but

the magnitude of the signal is larger by a factor of 10 than one would predict by
using the measured derivative of the action potential, the measured seal resis
tance, and the predicted area of membrane beneath the patch electrode. This is

a consequence of these neurons having a convoluted rather than a smooth mem
brane, resulting in a very much larger capacitance in the membrane beneath the

patch, which in turn results in very much larger extracellular signals. In addition

to the large capacitive signals smaller ionic signals were sometimes seen, as in
Figure 4.3 (b). The small negative signal superimposed on the capacitive signal

during the rising phase of the action potential are presumably due to sodium
channels contained in the patch beneath the membrane.

The signal-to-noise ratio can be determined by considering the noise contri
butions. The primary source of noise is Johnson noise

Kιotae(rm∙s ) — y∕4(Raeaι + Re)kTB.

(6)

Where the Rseaι is the seal resistance, Re is the real part of the electrode
impedance, k is Boltzmann’s constant, T is the temperature, and B is the band
width in Hz. For a 1 kHz bandwidth, a seal resistance of 2MΩ, and an electrode

with a real impedance Re = 500kΩ, the corresponding rms Johnson noise is 5μV.
Therefore, for Rs6ai = 2Λ∕Ω, Vcap∕Vnoi3e(rms') ≈ 50 for Helisoma B19 action
potentials, and Vcap∕Vnoi3e(rms') ≈ 10 for Helisoma B5 action potentials.

Recording subthreshold signals will be difficult since the signals are typ
ically small and slow. Furthermore signals are difficult to interpret since the

59
(a) Helisoma B19

LΩ

5 ms

(b) Helisoma B5

cxj 20 ms

Figure 4.3. Neurons stimulated with a current pulse passed through an intracellu

lar electrode (40-50 MΩ), and the response of the cell recorded with the intracellular
electrode, the derivative of the intracellular signal, and the loose-patch electrode (di
ameter ≈ 10 μm). (a) Helisoma B19 neuron (Raeaι = 3.0 MΩ, bandwidth 10 Hz-1 kHz).
(b)A Helisoma B5 neuron (Raeaι — 2.0 MΩ, bandwidth 10 Hz-1 kHz).

intracellular signal is distorted by passing through the patch membrane. In or
der to record subthreshold signals very good seals are required. If a gigaohm

seal could be obtained then it would be practical to electrically break down the

cell membrane beneath the cup and have essentially a chronic whole-cell patch
recording. Diving-board electrodes have been used in this manner, but so far

the seal resistances have not been large enough to prevent a deterioration of the
resting potential of the cell.
4.3 Stimulation

The equivalent circuit for stimulation is shown in Figure 4.1(c). It is gen
erally true that a membrane depolarization of approximately 15 mV will fire a
neuron. As a current pulse is applied through the electrode most of the current

passes out beneath the cup, with a small fraction passing into the cell through the

patch and then out through the rest of the cell. For a stimulus current pulse Istim

the voltage applied to the patch is Vpatch = Λpositive current pulse thus hyperpolarizes the patch membrane, and depolarizes

the rest of the cell membrane. For stimulus pulses short compared to the time

constant of the cell (which is typically ≥ 5msec.) the internal voltage change is
approximated by

ΔKc(( =

_ ∆f(√τnι d∙Vpatcfι

-∙" - ~c∑,

(6)

~c^

∆i

(*c 4^
' ^c⅛c
^m2 v
chan=l

ΔVccι↑ ~

Λif

∆t(z)

∆f

∖~,

dΓ + ‰ IA
cΛan=l

(^zceZi

(?)

Fpαicfc

⅛~
an

IzcΛaπ.)'∖

zoλ

(8)

61
Since channel type and density beneath the patch is not known it is difficult to
predict the required stimulus current.

There are two constraints limiting the amount of current that can be used
for stimulating. First, if a voltage of greater than about 1V is applied to a metal
electrode, gas is evolved that kills the cells. An electrode can be approximated
by a capacitor with a capacitance Ce. A current pulse with amplitude Istim and

duration ∆f results in an electrode potential ΔVe = Istim × Δt∕Ce, which limits

the total charge ∕∆f useful for stimulating. Second, Iatim must be kept small

enough that Vpαtch remains less than about 300 mV, or the membrane in the
patch may be electrically broken down. This would damage the cell unless Rseaι
was much larger than the few megohms thus far achieved.

62
References

[1] A. Marty, and E. Neher, “Tight-Seal Whole-Cell Recording”, in Single Chan
nel Recording, ed. B. Sakmann, and E. Neher, Plenum Press, New York,
pp. 107-121, 1983.
[2] C. A. Thomas, Jr., P. A. Springer, G. E. Loeb, Y. Berwald-Netter, and
L. M. Okun, “A miniature microelectrode array to monitor the bioelectric
activity of cultured cells,” Exp. Cell Res., vol. 74, pp. 61-66, 1972.
[3] G. W. Gross, A. N. Williams, and J. H. Lucas, “Recording of spontaneous
activity with photoetched microelectrode surfaces from spinal cord neurons
in culture,” J. Neurose. Methods, vol. 5, pp. 13-22, 1982.
[4] J. Pine, “Recording action potentials from cultured neurons with extracellular
microcircuit electrodes,” J. Neurose. Methods, vol. 2, pp. 19-31, 1980.
[5] D. W.· Tank, C. S. Cohan, and S. B. Kater, “Cell body capping of array elec
trodes improves measurements of extracellular voltages in micro-cultures of
invertebrate neurons,” IEEE Conference on Synthetic Microstructures, Airlie,
Virginia, 1986.
[6] D. A. Israel, W. H. Barry, D. J. Edell, and R. G. Mark, “An array of microelec
trodes to stimulate and record from cardiac cells in culture,” Am. J. Physio.,
vol. 247, pp. H669-H674, 1984.
[7] W. Stuhmer, W. M. Roberts, and W. Aimers, “The loose patch clamp”, in
Single Channel Recording, ed. B. Sakmann, and E. Neher, Plenum Press, New

York, pp. 123-132, 1983.

63
Chapter 5
Diving-Board Electrode Tests

5.1 Establishing the Chronic Connection

An underwater-setting biocompatible adhesive was required for gluing the

diving-board to the bottom of the tissue culture dish. Various glues were screened
for their ability to set and adhere under water. Of these many were not suitable

due to their viscosity and the lack of a thinner that would allow a very small drop
of glue to be applied. While several epoxies set and adhere under water, limited
working times make them difficult to work with. Electro-Lite Corporation 4481

is an adhesive that sets when exposed to ultraviolet light and adheres well under

water to all plastics tested, but not to glass, or substrates coated with polylysine
or collagen. The adhesive can be thinned with isopropyl alcohol facilitating
formation of small drops.

Once a glue with the desired curing and adhesion characteristics was found,
it was necessary to test the biocompatibility of the gluing procedure as shown

in Figure 5.1. Twenty-four hours after plating a glue drop was placed by one

of two Helisomα B19 neurons contained in the same dish. The adhesive was
set using a 1 minute exposure to a mercury arc lamp through a 20X objective,
using an iris to restrict the illumination to the glue drop. The cell continued to

grow in the same manner as the control. The growth of neurons (n=10) tested
in this manner was found to be no different than that of the controls. However,

processes illuminated with levels of u.v. light necessary to set the glue quit
growing within five minutes of illumination. Similar tests were performed with
two-week old SCG neurons, which continued to look healthy for many days after

the gluing process.

T=0 hrs

'-- ri------ !-- T⅝l lr⅜" »— »;■ ∙'--ι

fe>⅛
R⅛.≥⅛ff∙
‘ i A '.

√fe√"
'⅛O⅛-∖

T=O hrs

V∖⅛rΛ∙-*'*
⅛⅛V∙'

^,jιl... >..⅛ι..l⅜⅛ J.⅛.⅛.⅛,

-.......... J

T=5 hrs

T=5 hrs
ffκ..........

tsj⅛

W √

-~'r C '

■f$.

fM^^gM^M^M⅝¾llg⅜⅝M^⅜⅛⅝

T=1O hrs

T≈1O hrs

100um

Figure 5.1. Helisoma B19 neurons. At t = 0 hours a drop of adhesive was placed

near one neuron. Pictures taken at t = 0 hours, t = 5 hours, and t = 10 hours.

65
Intracellular penetrations were performed to check the electrophysiology of

neurons with nearby drops of glue compared to controls. In no case did the

gluing process have any effect on either Heli30ma or SCG neurons.
The diving-board electrodes are manipulated into place while being viewed
through an inverted microscope equipped with phase and epifluorescence optics.

The polyimide handle of an electrode is held by fine forceps mounted in a holder
that provides pressure to hold the electrode firmly until it has been glued to the

bottom of the culture dishes. Then the pressure is reduced and the electrode
is gently released, without damaging the glue joint. The holder is mounted in

a special manipulator that is in turn mounted in a Leitz micromanipulator. In
addition to the standard x, y, z degrees of freedom provided by the micromanip

ulator, this special manipulator provides θ, and φ control needed to make the
diving board flat relative to the surface of the tissue culture dish.

Once the electrode has been made flat relative to the culture dish a metal
needle is used to scratch the plastic in a region where the pedestal is to be glued.

Then a 50 μm diameter drop of glue is applied to the bottom of the tissue culture
dish through a 10 μm diameter glass electrode using pressure to control the drop

size. Then the electrode is placed with the pedestal in contact with the glue and
the diving-board tip in contact with the cell. When a seal resistance of several
Megohms has been obtained the drop of glue is exposed to ultraviolet light from
the mercury arc lamp of the epifluorescence illuminator.

Electrical contact to the electrode is maintained during manipulation and
gluing so that the seal resistance can be monitored. The culture dish has wires
extending to the interior through holes in the side. At the end of each wire is

a flexible 25 μm gold wire-bond wire. When the electrode has been positioned

near the cell a conductive graphite paint, TV Tubecoat, is used to connect the

Figure 5.2. Diving board electrode in chronic electrical contact with a Helisoma B19

neuron. The pedestal is glued to the bottom of the culture dish and the electrode tip
is in contact with the cell body.

wire from the diving board electrode to one of the wires connected to the side
of the dish, which is in turn connected to external electronics. After the device
has been glued it is possible to position another diving-board electrode. Once

the electrodes are mounted, the lid is put back on the tissue culture dish and the
diving-board electrodes are ready for long-term experiments. Figure 5.2 shows
a diving-board electrode in contact with a cultured Helisoma neuron.

To understand device operation two-electrode experiments were performed.
In preliminary experiments a 12 μm glass pipette was substituted for the divingboard electrode. An intracellular electrode and a patch pipette recording from
the same cell. In this way it was possible to compare the intracellular response
with the response of the patch pipette. Once stimulation and recording were

accomplished, and understood with the patch pipette, a diving-board electrode
was used. Other than the inherent electrical differences between Equid filled and

metal electrodes, diving-board electrodes behave like glass patch-electrodes of

the same tip diameter for both stimulation and recording.
5.2 Recording from Identified Helisoma Neurons
With the seals obtained it is possible to record action potentials with good
signal-to-noise ratios, but in order to record subthreshold signals the seal re

sistance needs to be greatly improved. In the experiment of Figure 5.3 with
diving board electrode mounted on top of a neuron an intracellular electrode

was used to stimulate using a current pulse, and the diving-board electrode was

used to record the resulting action potential. Signal-to-noise ratios for action
potentials recorded from Helisoma B19 neurons were typically 20-100:1, and for

Helisoma B5 neurons 4-10:1.
Diving-board electrodes have been used to record spontaneous activity from

Helisoma B19 neurons for up to four days. Figure 5.4(a) shows a diving-board
electrode monitoring the spontaneous activity of a Helisoma B19 neuron for 2

days. Successive pictures at t = 0 hours, t = 4 hours, and t = 24 hours show
normal cell growth. Figure 5.4 (b) is an example of the spontaneous activity
recorded from this cell.

∏u>02

Figure 5.3. A Helisoma B19 neuron stimulated with a current pulse passed through
an intracellular electrode, and the response of the cell recorded with the intracellular
electrode, and the diving-board electrode (R3eaι = 4.0 MΩ, bandwidth 10Hz-lkHz).

SPONTANEOUS ACTIVITY

10s

Figure 5.4.(a) Successive pictures of a diving-board being used to record from a Helisoma B19 neuron at t = 0 hours, t = 4 hours, and t = 24 hours, (b) An example of
spontaneous activity recorded at t = 2 hours (Rseaι = 3MΩ).

70
5.3 Stimulating Identified Helisoma Neurons

Diving-board electrodes have been used to stimulate Helisoma neurons. In
the experiment of Figure 5.5 a current pulse passed through the diving-board
electrode was used to stimulate the neurons while an intracellular electrode was

used to monitor the potential of the cell.

20 ml)

jθ 0

csc

Stimulation of a Helisoma B19 neuron using a current pulse passed
through a diving-board electrode to stimulate and recording using an intracellular elec
trode. {Isti-m, — 90 nA, ∆t = 5 msec, Vpatc∣l — 200 mV)
Figure 5.5.

Helisoma cells have been stimulated intermittently for up to 4 days. Fig

ure 5.6 shows a diving-board electrode that was placed on a cell at time t =
0 hours. At 24 hours the cell was penetrated with an intracellular electrode to
monitor the activity of the cell. Then the current pulses were passed through the

diving board and the cell was stimulated at 1 Hz and 10 Hz. The intracellular

electrode was then removed and the cell was stimulated at 1 Hz for 24 hours.
At that time the cell was penetrated once again to verify that the cell was still
.being stimulated and that the threshold had not changed.

' 24h

0.5 sec

48h

Figure 5.6. A diving-board electrode in electrical contact with a Helisoma B19 neuron.

Calibration bar is 50μm. (I3tim — 140 nA, ∆t — 0.5 msec, Vpatcfl = 170 mV) Twentyfour hours after the device was placed on the neuron an intracellular electrode was used
to monitor the activity of the cell and current pulses passed through the electrode were
used to stimulate the cell at 1 Hz and at 10 Hz. The stimulus current passed through the
diving-board electrode is shown below the intracellular voltage record. The intracellular
electrode was then removed while the diving board electrode continued to stimulate
the cell at 1 Hz. At time t = 48 hours an intracellular electrode was used to verify that
the stimulus threshold had not changed.

72
It is necessary to measure the evoked action potential to verify that the cell

is being stimulated. While simultaneous intracellular recording has been used,
a less invasive method is preferred. It is desirable to be able to measure the
evoked action potential with the same diving-board electrode used to stimulate

the cell. The primary difficulty is the large stimulus artifact, which is on the

order of 200 mV, much larger than the recorded action potential, which is on the
order of 100 μV. There are two ways of solving this problem. The first is to raise
the stimulus pulse very slowly; at threshold a late action potential occurs many

milliseconds after the stimulus pulse can be recorded. Once threshold has been

determined the stimulus pulse can be increased by 10% to obtain reliable stimu
lation. In Figure 5.7 Heli30ma B19 is stimulated by a diving-board electrode and

the response of the late action potential is recorded by the same electrode. The
second solution, which has not been done yet, is to remove the stimulus artifact

by computer-processing the signal.
5.4 Preliminary Studies on Vertebrate Neurons

Preliminary studies have also been performed with dissociated rat superior
cervical ganglion neurons (SCG’s). Several factors make working with SCG’s

more difficult than with Heli30ma neurons: they are smaller (a diameter of
about 25 μm for SCG’s compared to about 50 μm for Heli30ma neurons); the

extracellular solution resistivity is lower by about a factor of four resulting in

lower seal resistances; and SCG’s are in general more delicate than Helisoma
neurons. Despite these difficulties it has been possible to stimulate SCG’s, and
to record action potentials with a signal-to-noise ratio of 10:1 (Rseal = 0.4MΩ).

Long-term stimulation and recording studies will be performed in the near fu

ture. The success of the diving-board electrodes on SCG’s demonstrates their
versatility, and indicates that they can be used with neurons 20 μm in diameter

Figure 5.7. A Hélisoma neuron B19 stimulated by a diving-board electrode. (Iatim =
100 nA, Δ∕ = 0.2 msec, Vpatc∣l = 210 mV) Since the stimulus is near threshold the
resulting action potential is many milliseconds after the stimulus pulse and it can be
recorded on the diving-board electrode. (Raeaι ≈ 1.5 MΩ, bandwidth 100Hz-lkHz).

and larger. The devices can be scaled down for use with smaller neurons.

Figure 5.8. Diving-board electrode on a rat SCG neuron

75
Chapter 6

Electrode Capping By
Cultured Invertebrate Neurons

6.1 Introduction
Multielectrode dishes have been used to record from a variety of in vitro

preparations [l]-[5]. In general it is possible to record small extracellular signals
resulting from current flow associated with an action potential. However, it may

be difficult to interpret these signals, since the signal-to-noise ratio is often poor,
and it is not always possible to get a one-to-one electrode-neuron correspondence.

It is also possible to stimulate neurons using these electrodes [2]. Such stimula

tion, in which a current pulse is passed through an electrode, relies on creating

a voltage drop in the media sufficient to fire nearby axons and cell bodies. Such
stimulation techniques suffer from the disadvantage that it is difficult to verify

the stimulus using the extracellular array, therefore an independent means of
stimulus verification such as voltage-sensitive dyes [6],[7] or intracellular record

ing is required. These problems are a result of the extracellular media (typically
50-400 Ω-cm) shunting the stimulus current and extracellular signals produced
by the cell. For cultured vertebrate neurons there is no easy way to reliably

overcome this problem.
However, it is possible to seal a dish electrode to a cultured invertebrate

neuron in much the same manner as a loose-patch electrode. Figure 6.1 (a) is an
example of a Heli30ma neuron sealing over several electrodes. This results in a
one-to-one electrode-neuron correspondence, facilitating recording action poten
tials from each neuron in a simple network, stimulating neurons, and recording

action potential with the stimulating dish electrode.

76
(a)

(b)

Polyimide

Figure 6.1. (a) Helisoma B19 neuron growing on a multielectrode dish. Loose-patch
type recordings can be obtained from the electrodes completely covered by the neuron,
(b) Schematic of a neuron sealing over a dish electrode.

6.2 Dish Fabrication

The multielectrode arrays are fabricated using conventional integrated cir

cuit technology. Figure 6.2 (a) is a picture of a completed array. The electrode

leads are indium-tin oxide (ITO) [3], [8]; the insulation is photosensitive poly

imide; and the electrode tip is electroplated platinum black. The electrode pat

tern consists of a hexagonal array of 61 electrodes, 12 μm in diameter, separated
by 70 μm.

(a)

(b)

Figure 6.2. Masks used in fabrication of microelectrode array, (a) Lead Mask (b) In
sulation Mask.

Fabrication begins with a glass substrate (0.016" thick), coated with a layer

of ITO (lOOOÂthick, and with a sheet resistance of 100Ω∕square) obtained from
Donnelly Corporation (Midland, MI). After cleaning, the leads are patterned
using Shipley 1350J photoresist, with a 125mJ∕cm2 exposure for the central
pattem and a 800 mJ∕cm2 exposure for the peripheral part of the wafer to remove

the thick edge bead that forms at the wafer edge when photoresist is spun on,

that would result in shorted leads. The ITO is etched for 4 minutes in a freshly
prepared solution of 50 H2O∕5O HC1∕1 HNO3 at 40°C. The photoresist is removed
and a 30Â thick layer of aluminum is evaporated on to the wafer, resulting in

a thin layer of aluminum oxide which is an effective adhesion promoter. Next

photosensitive polyimide ( MRK Selectilux HTR 3-50) is spun on at 5000 rpm
for 30 seconds, soft baked for 5 minutes at 85°C, patterned with the electrode

mask, developed, and cured for 12 hours at 200° C. To reduce the electrode
impedances to less than 500 KΩ the electrodes are platinized in a solution of

1% chloroplatinic acid in 0.0025% HC1, plus 0.01% lead acetate, using a current

density of 20mA∕cm2 for 10 seconds. Then the bottom of a small tissue culture
dish with a 5 mm diameter hole is glued to the top of the substrate. Figure 6.3 (a)
shows a completed ITO dish. It takes an average of about 1 hour to fabricate

each dish, and they have been routinely reused several times, for a total time in
saline of 60 days.

6.3 Dish Electronics
Figure 6.3 (b) shows a dish mounted in mounted in the interfacing printed-

circuit board, which in turn mounts to the microscope stage. Sixty-three leads
go to the inputs of eight eight-to-one multiplexers, with outputs designated as

channels A-H. Each channel is amplified with a gain of 11, and is differential

with respect to two large dish ground electrodes. A JFET switch connects each
channel to an externally generated stimulus. Two 34-wire flat .cables connect the

printed circuit board to interface electronics, which generates five control signals
for each channel. Figure 6.4 (a) shows one of eight identical channels, channel A.

AEN enables multiplexer A; AO, Al, A2 selects the input to multiplexer A, and
ASTIMON switches the STIMAIN signal on. The eight stimuli are turned on
and off using two quad-JFET switches (Figure 6.4 (b)). Figure 6.4 (c) shows the

reference amphfier.
Electrodes are addressed as in Figure 6.4(c). By sending the correct control

voltages any one of electrodes A-H can be accessed, making it possible to record
and stimulate simultaneously using eight electrodes. A control box is used having

eight switches that set the binary code to address the multiplexers A-H, and eight

switches for stimulus enable A-H (Figure 6.4(e)).
6.4 Experimental Procedure

Identified neurons were plated on polylysine coated multielectrode dishes.

Figure β.3. (a) Completed multielectrode culture dish, (b) Dish mounted on a circuit
board.

(a)

(b)

REFOUT

(θ)

(d)

Alco Switch
0Dβ

βGβ

„A7

„B7

„07

.07

βES

„G1

,E7

„F7

„C6

•B6

0E3

.E3

„F3

„A2

„C1

„FS

„AS

„B5

„G4

„E4

,F4

„C4

„H5

„E5

»CS

„A3

„H4

„F2

„B1

,Hβ

.C3

„E2

,A1

.B3

„02

„C2

„El

„GS

eD3

»H3

„B2

„FS

„ES

„DS

βC3

„F1

.01

„H7

βAS

„F7

»M7

„C7

„AS

„BS

.D4

„B4

„A4

.GS

„D5

+5V

ASTIMON

Figure 6.4. (a) Channel A (b) Stimulus Switches (c) Reference Electrodes (d) Dish
addresses (e) Control box circuitry

81
Cell bodies moved an average of about 100 μm after plating, making neuron

placement on a specific electrodes extremely difficult. However, by using an
array with electrodes spaced 70 μm apart it was possible to have virtually every

neuron sealed to an electrode. Cell growth dictated the seal resistance. When
cells grew as in Figure 6.1, with a very large lamellipodia seals were obtained on

several electrodes.

6.5 Recording
Dish electrodes have been used to record action potentials from Heli30ma

B19 neurons with signal-to-noise ratios of 100:1, and from B5 neurons with
signal-to-noise ratios of 20:1. In Figure 6.5 is an example of spontaneous activity

recorded using a multielectrode dish.
Using multielectrode dishes it is possible to record spontaneous activity

from small networks of invertebrate neurons with good signal-to noise ratios and
one-to-one electrode-neuron correspondence. In Figure 6.5 A is a Helisoma B5

neuron; B, and C are Heli30ma B19 neurons. The action potentials recorded
from Helisoma B5 neurons are in general much smaller than those recorded
from Heli30ma B19 neurons which have a much faster action potential. Dish

recordings are primarily derivatives of the action potentials as seen through the
capacitance of the cell membrane, with small ionic contributions. Multielectrode

dishes have been used to monitor spontaneous activity for up to 13 days from
Helisoma neurons.
By stimulating with an intracellular electrode and simultaneously recording
with an intracellular electrode and a dish electrode the signals can be compared,
and the frequency dependance can be determined as in Figure 6.6. Recording

from a dish electrode under the soma of a Helisoma B19 neuron fired at 0.1 Hz

the signal recorded by the dish electrode was essentially a derivative of the action

1r,∙sV

Figure 6.5. Spontaneous activity recorded from Helisoma neurons using a multielec

trode array (bandwidth 10Hz-lkHz). The electrodes are 70μm apart.

potential. As the frequency of stimulation increased to 10 Hz the signal recorded

on the dish electrode changed in two ways: first the amplitude decreased due

to action potential broadening; second, the signal had an ionic component in
addition to the capacitive component. It would be possible to determine the

channel type responsible for this ionic component using a patch pipette and
different channel blockers, but these experiments have not been performed.

Experiments with multielectrode dishes also suggest ways in which record

ings might be enhanced. If a cell is used to seal over processes or the entire cell
body then extracellular signals similar to those obtained in vivo are possible.
Figure 6.7 (a) shows two cells: cell A is a B19 neuron, and cell B is a B5 neuron.

Figure 6.6. Simultaneous recording from dish electrode and intracellular electrode,

with the neuron stimulated at 0.1 Hz, 1 Hz, 2 Hz, and 10 Hz (bandwidth 10Hz-l kHz).

Figure 6.7 (b) shows spontaneous activity recorded from dish electrodes 1 and 2.

By stimulating cell A and cell B we see that the spontaneous activity previously
recorded is not from cell B which is located directly over the electrode, but rather

is from cell A. An axon stump from neuron A is located under neuron B. As a
result the current flow associated with the action potential results in a much

larger voltage drop. Figure 6.7 (b) shows signals recorded from a dish electrode
that is not sealed and signals from an electrode with a neuron sealed over the

axon stump, demonstrating the improvement in signal-to-noise ratios facilitated
by obtaining a seal, over that obtained with an extracellular electrode.

These results suggest a technique that might be implemented to obtain im-

Cell B

10 ms

y I
∖_______

Cell A

Figure 6.7. (a) Picture of cells growing on multielectrode array, (b) Spontaneous
activity recorded by the multielectrode array, (c) Intracellular penetration of cell A.
(d) Intracellular penetration of cell B (bandwidth 10Hz-lkHz).

proved recording. Once cells have grown out and formed connections a confluent
layer of cells such as glia could be grown over the cells, sealing the electrodes,

preventing bath electrolyte from shunting the signal, and allowing recordings

processes as well as cell bodies. This method could be applied to smaller verte

brate cells, since the “sealing” need not be done by the neurons, but would be
done by other cell types.

6.6 Stimulation

Neurons can be stimulated by passing current pulses through dish electrodes.

The fact that the neuron is sealed to the electrode makes it possible to use the
multielectrode array to stimulate a neuron, and to record the resulting action
potential. In Figure 6.8(a) a dish electrode is used to stimulate a neuron and

to record the resulting action potential. In Figure 6.8 (b) a current pulse passed
through a dish electrode is used to stimulate a neuron and another dish electrode

is used to record the resulting activity. A large stimulus current is required to
fire the cell because the stimulating electrode is not sealed beneath the cell.
However, it would not be possible to record the resulting action potential were

the neuron not sealed to the dish electrode.
Dish electrodes are promising for chronic stimulation experiments in which

a means of noninvasively verifying stimulation is required. However it is impor
tant to monitor the seal resistance and adjust the stimulus current so that the
breakdown voltage of the cell membrane is not exceeded. The procedure for long
term stimulation experiments is: (1) measure the seal resistance, (2) determine

the firing threshold of the cell, (3) monitor the seal resistance and the resulting
action potential throughout the experiment, (4) adjust the stimulus current to

take into account changes in the seal resistance.

Figure 6.8. (a) A current pulse of duration 0.5 msec passed through a dish electrode
stimulates a cell. The same dish electrode is used to record the resulting action po
tential, and an intracellular electrode is used to verify the response, (b) A current
pulse passed through one dish electrode stimulates the cell, and another dish electrode
is used to record the resulting action potential. An intracellular electrode is used to
verify the response.

References

[1] C. A. Thomas, Jr., P. A. Springer, G. E. Loeb, Y. Berwald-Netter, and
L. M. Okun, “A miniature microelectrode array to monitor the bioelectric
activity of cultured cells,” Exp. Cell Res., vol. 74, pp. 61-66, 1972.
[2] J. Pine, “Recording action potentials from cultured neurons with extracellular
microcircuit electrodes,” J. Neurose. Methods, vol. 2, pp. 19-31, 1980.
[3] G. W. Gross, A. N. Williams, and J. H. Lucas, “Recording of spontaneous
activity with photoetched microelectrode surfaces from spinal cord neurons
in culture,” J. Neurose. Methods, vol. 5, pp. 13-22, 1982.
[4] D. W. Tank, C. S. Cohan, and S. B. Kater, “Cell body capping of array elec
trodes improves measurements of extracellular voltages in micro-cultures of
invertebrate neurons,” IEEE Conference on Synthetic Microstructures, Airlie,
Virginia, 1986.
[5] D. A. Israel, W. H. Barry, D. J. Edell, and R. G. Mark, “An array of microelectrodes to stimulate and record from cardiac cells in culture,” Amencan
Physiological Society, pp. H669-H674, 1984.
[6] Rayburn,H., J. Gilbert, C-B .Chien, and J. Pine, Noninvasive techniques for
long term monitoring of synaptic connectivity in cultures of superior ganglion
cells, Soc. of Neurosci. 14th Annual Meeting, Abstract 171.1, 1984.
[7] Chien, C-B., W. D. Crank, J. Pine. Noninvasive techniques for measurement
and long-term monitoring of synaptic connectivity in microcultures of sym
pathetic neurons, Soc. of Neurosci. 17th Annual Meeting, Abstract 393.14,
1987.

Chapter 7
Future Work
Using both the diving-board electrode and multielectrode dishes it is pos

sible to establish a chronic two-way electrical connection to cultured neurons.

There are many experiments that could be conducted with the devices in their
present form, and by improving device performance the range of experiments
would be increased significantly. Such integrated-circuit techniques will allow

many experiments to be conducted that are difficult or impossible using what is
now conventional technology.

7.1 Future Experiments with Diving-Board Electrodes
There are several obvious ways in which diving-board electrodes might be

improved. By increasing the seal resistance, whole-cell intracellular connections

could be obtained. Preliminary experiments should be performed using glass

patch electrodes of the appropriate diameter, since diving-board electrodes are
valuable and difficult to work with. The first step is to insure that the device
and the cell surface are extremely clean. The next step is to treat the surface to

make it “stickier”. Once large seal resistances are obtained using glass electrodes
without suction, diving-board electrodes can be used. Also, the manipulation
and gluing procedure could be improved: the manipulator could be made smaller,

with a better release mechanism; the contact between the electrode wire and the
wire mounted to the dish could be made more reliably. Such improvements

to the diving-board electrode should make it an attractive tool for researchers
interested in long-term communication with cultured neurons.

Even without the ability to obtain intracellular connections, there are many

experiments that can be done with diving-board electrodes that involve moni

89
toring spontaneous activity and long-term stimulation. Dishes are easier to use,

with performance comparable to diving-board electrodes, for large invertebrate

neurons. This is not the case with vertebrate neurons: diving-board electrodes

offer improved signal-to-noise, as well as the ability to record the evoked action
potential with the stimulating electrode. Long-term electrical connections to cul

tured SCG neurons should be demonstrated, leading to long-term stimulation
experiments on small systems of neurons.
7.2 Future Dish Experiments
Dish fabrication and dish electronics have progressed to the point where

many experiments can be performed using cultured invertebrate neurons. Using

multielectrode dishes facilitates stimulating either the processes or the cell body

through a “sealed” electrode, and long-term recording of action potentials.
An example of experiments employing dishes is to study the collective behav
ior of small networks composed of identified Aplysia neurons. First, determine

if stereotypical patterns of connections develop, and if so what the attributes
of these connections are. Second, observe the effect of imposed activity on the

system. Once the capability of long-term stimulation and recording has been
established, connection strengths between neurons comprising a simple network
can be mapped. Each cell in the culture will be stimulated by dish electrodes
to fire an action potential, and the response of other neurons recorded with the

array. Dish electrodes can also be used to monitor spontaneous activity of the
nervous network. If dish electrodes fail to provide sufficient signal-to-noise ra
tios to resolve subthreshold signals voltage-sensitive dyes can be used to measure

the resulting monosynaptic potentials from all the other cells. Using such dyes
action potentials can be recorded without signal averaging with signal-to-noise
ratios of 40:1 [2], indicating that subthreshold activity can be easily resolved.

Aplysia neurons are particularly suited to this study for four reasons. First,

extensive work has been done using these neurons to study cellular mechanisms

of learning [1]. Associative learning, activity-dependent facilitation, Hebbsian
learning, sensitization, and habituation have all been studied in this preparation.

Second, the neurons are large and technically easier to work with than vertebrate
neurons. Third, preliminary results indicate that voltage sensitive dyes can be
used to map the connection strengths of a system of such neurons, and extra

cellular electrodes can be used for long-term stimulation. Fourth, spontaneous

activity can be continuously monitored, and the activity correlated with the
connection strengths between the neurons.
Experiments similar to those proposed here are currently being performed

using SCG’s [3],[4]. These results indicate that mapping of small systems of
neurons using voltage-sensitive dyes is an effective technique. Using cultured

Aplysia neurons will provide several advantages over this vertebrate preparation
including: larger signal-to-noise ratios using voltage sensitive dyes, the ability to
easily record spontaneous activity using dish electrodes, the ability to stimulate

the cell body and avoid complications associated with stimulating the processes,
and more robust cells that are more convenient for conventional physiology.

7.3 Future In Vitro Applications of Microdevices
In addition to the diving-board electrode and multielectrode arrays, many

types of devices could be fabricated and used for in vitro physiology using inte

grated circuit technology.

The dish structure could be modified to enhance electrode performance.
One way is to confine neurons to wells as shown in Figure 7.1 (b). Neurons are

placed in wells when they are young, growing to fill the well and seal to the
electrode, with the forces exerted by the processes unable to pull the neuron

91
SIDE VIEW

TOP VIEW
(a)

Insulation

∠2
«Ä........

■I·
'∙⅛÷⅜s⅛<∙*∙y∙w∙'∙'∙v ..
·:·:·:·:·:·:·:;:;: : Λ: :>: : :·:·:·:·:·
⅞OB⅛β⅛¾
÷÷ι÷⅛÷⅛∙⅛⅛∙ >⅜÷⅛⅛
⅛⅛8⅛s —.....

— Lead

. ⅛%⅛⅛½⅞¾
<%¾¾¾⅛⅛γ
..⅛∙⅞⅞∙⅛
⅛∙⅞⅛⅛∙⅛

.Electrode

Seal

Overhanging
Bars

(b)

Insulation
Electrode

Lead
Electrode Tip
Insulation

i⅛∙÷⅛÷!⅛x÷ι÷ιTi⅛∙⅛∙⅛÷÷x⅛⅛∙÷i÷!
w ⅞⅞⅝¾⅞⅞^

Lead

(d)

Figure 7.1. (a) Standard dish electrode (b) Cells in wells (c) Dish electrode capable
of providing suction (d) spike electrodes to record from a slice.

away from intimate electrode contact [5],[6]. Techniques developed to improve
seals to diving-board electrodes can be applied to well electrodes.

One serious limitation to both diving-board electrodes and dish-electrodes is

the inability to provide suction to obtain a gigaohm seal. This problem could be

overcome by building a special dish with hollow channels leading to the electrode
sites as shown in Figure 7.1 (c). Suction could be applied through this tube, fa

cilitating chronic whole-cell recording from many cells simultaneously. There are

problems that must be overcome. First there is a conflicting surface requirement:
while the surface to which the seal must be made should ideally be very clean,
it is usually necessary to coat the dish with substances such as polylysine and

collagen in order to get good outgrowth and healthy cells. Second, dializing the

interior of the cell limits the duration of whole-cell recording. While appropriate
choice of solution in the patch pipette helps to alleviate this problem, it doesn’t
solve the problem entirely. Once these problems are overcome such a device has

great potential for providing more sensitive connections than can be obtained
with conventional dishes.

Another application of the arrays is to stimulate and record from brain slices.
Slices are typically prepared by dissecting out a brain, and cutting it into 500 μτn
thick sections, which are placed in tissue culture dishes. Such preparations offer

the advantage of maintaining architecture found in the brain, while allowing
manipulation of the biochemical environment and increased flexibility. While

the prospect of using dishes to record from and to stimulate slices is attractive
it is extremely difficult: cells within 50 μm of the surface are typically dead in

slices, and anything of interest is located well away from the surface of the dish.
What is. needed is an array of electrodes sticking up into the slice, as is shown in
Figure 7.1 (d) that could be manufactured using integrated circuit technology.

93
7.4 Impact of Integrated-Circuit Based Microdevices on Neurobiology

What would it take for these microdevices to have an impact on neurobi
ology, and to become a standard tool? First, their usefulness must be demon

strated. Second, the devices must be made available to the neurobiologists who
want to use them.
Once fabrication and gluing procedures have been improved and further

studies have been made demonstrating the general usefulness of the diving-board
electrode, making this device widely available will be a problem, since the fab

rication procedure requires equipment and expertise well beyond the limits of
most neurobiology labs.

Somehow the laboratory developing the technology

must make the devices available. Since it takes about 80 hours to fabricate 100

devices it is unreasonable to expect one lab to supply devices to more than sev
eral other labs. One solution is to supply partially fabricated devices, and have
fabrication completed in the labs where they are to be used. By supplying wafers

after they have been wire bonded but before they have been etched, no special
facilities would be required other than a fume hood for the EDP. In this way a
fabrication facility could easily supply enough bonded wafers to meet the needs

of many labs.

While multielectrode dishes have been used since 1972, and their usefulness
has been demonstrated, technical barriers prevent widespread use: the devices

are not generally available to researchers; and the electronics for recording from

many electrodes simultaneously, while not particularly sophisticated, must be
custom made. Only a few very “high tech” groups are prepared to invest the
time, money, and resources to develop such a system.
Two ways to make this technology generally available are: to have such
groups supply the dishes, with those wanting the dishes paying for them; or

else once a market is established to have private enterprise take over device and
electronics production.

In the near future devices made utilizing integrated circuit technology will

make considerable contributions to neurobiology.

References

[1] Abrams,T.W., “Cellular studies of an associative mechanism for classical con
ditioning in Aplysia,” in Model Neural Networks and Behavior, A. S. Selverston ed., 1985.
[2] B. M. Salzberg, Private Communication, 1987.
[3] Rayburn,H., J. Gilbert, C-B .Chien, and J. Pine, “Noninvasive techniques for
long term monitoring of synaptic connectivity in cultures of superior ganglion
cells,” Soc. of Neurosci. 14th Annual Meeting, Abstract 171.1, 1984.
[4] Chien, C-B., W. D. Crank, J. Pine. “Noninvasive techniques for measurement
and long-term monitoring of synaptic connectivity in microcultures of sym
pathetic neurons,” Soc. of Neurosci. 17th Annual Meeting, Abstract 393.14,
1987.

96
Appendix

Detailed Diving-Board Electrode Fabrication Procedure

Diving board electrodes are fabricated using integrated circuit technology,
and micromachining. Figure A.l shows the five masks used in the fabrication.

Clean wafers.

Use < 100 > orientation silicon wafers with any

doping but p++.

a. 5 minute ultrasonic clean in trichloroethylene (TCE).
b. 5 minute ultrasonic clean in acetone.

c. 5 minute ultrasonic clean in ethanol.
d. 5 minutes in running water.

e. 5 minute clean in 3 parts sulfuric acid 1 part hydrogen peroxide (Pi
ranha etch).

f. 5 minutes in running water.

g. 5 minutes in buffered HF.
h. Rinse in water and blow dry.

Wafers are cleaned in TCE, acetone, and in ethanol to remove organic con
taminants. Then the surface of the wafer is oxidized in the sulfuric acid, hydrogen

peroxide mixture. This surface oxide layer is removed in buffered HF, exposing
fresh silicon. After briefly rinsing in water the wafer should be hydrophobic,

indicating a very clean surface.
2. Cup
a. Grow 0.2μm thick oxide: 15 minutes in steam 1100oC.
b. Take directly from furnace, allow to cool and spin on 1400-37 photore
sist at 4000 rpm.

c. Pattern cup mask 85 mJ∣eτsλ2.

d. Develop in Shipley 455 developer for 1 minute.
e. Hard bake for 30 minutes at 125°C.

f. 2 minute clean in O2 plasma.

! Mask 1

Cup

! Mask 2

Metallization

8B
3I
! Mask 3 Pedestal

« Mask 3

Mask 5 Upper Insulating Layer

Lower Insulating Layer

100 um

Figure A.l. The five masks used in the fabrication of diving board electrodes.

g. Etch in buffered HF for 3 minutes.
h. Remove photoresist in acetone, ethanol, water.

i. Etch for in 3 minutes in 10Ö parts HNO3, 100 parts CH3COOH, and
15 parts HF [1].
j. remove masking oxide: etch in buffered HF for 3 minutes.

k. As in Figure A.2 measure the diameter of the overhanging layer of oxide

<¾ and the diameter of the silicon cup d1 to determine the depth of the
cup, d(,ttp — (d3

d2)∣2.

Figure A.2. Schematic view of the cup.

The cup is etched in silicon to a depth of 3 μm. An isotropic etch is used to
make the resulting cup as circular as possible, with no sharp corners that would
tend to reduce the seal resistance. It was necessary to use thermal S1O2 [2,3] to

mask the etch since no commonly available photoresist could stand up to any
isotropic silicon etch well enough to etch a 3 μm deep cup. The oxide mask also

made it easy to accurately determine the depth of the cup. By measuring the
diameter of the masking oxide d3 and the diameter of the top of the cup ⅛ as in

Figure A.2, the depth of the cup can be easily determined, dcup = (d1 — ⅛)∕2.

3. Pedestal.

a. Grow 1 μτn thick oxide: 3 hours in steam 1100oC.
b. Spin on 1400-37 photoresist for 30 seconds at 4000 rpm.

c. Soft bake for 30 minutes at 85° C.
d. Pattem pedestal: 85mJ∕cm2. Use the alignment marks to correctly

orient the mask.
e. Develop in Shipley 455 developer for 1 minute.
f. Hard bake 30 minutes at 125°C.

g. Buffered HF etch for 12 min, rinse, and dry.
h. Dry in 130° C oven for 5 minutes.

i. Boron diffusion 16 hours 1180oC, O2 100 seem N2 2000 seem.
j. Boron Nitride Preparation: Clean and dry wafers as per data sheet

[4], oxidize for 30 min at 1100°C. Set to desired temperature and gas
flow and allow to stabilize for 12 hours before beginning diffusion. Take
wafers out and set to 1100oC with oxygen at 200 seem allow to stabilize

and oxidize the surface for 1 hour.

k. Remove boron glass: 20 minutes buffered HF etch.
The pedestal is defined by a boron diffusion, since silicon doped to levels of

greater than 5 × 1019 cm-3 is etched very slowly by EDP. The pedestal should
be at least 10 μm thick, therefore a very deep boron diffusion is necessary. The

diffusion is carried out at the highest temperature that is practically possible with
a quartz furnace (1180oC) so that the diffusion coefficient of boron in silicon

is as large as possible. Oxidized boron nitride wafers are used as the boron
dopant [5]. Since boron diffuses much faster in heavily boron doped silicon [6]
than in intrinsic silicon, a pedestal 12 μm thick requires a diffusion time of only

10 hours. By carrying out the diffusion at such high temperatures for such long

periods of time the solid solubility of boron in silicon is exceeded; resulting in a
distinctive surface. The depth of the diffusion can be estimated by measuring

the distance to the faint line seen in the surface of the silicon, after the diffusion.

When doing a boron diffusion a layer of boron glass forms on the surface

of the silicon. To remove this layer it must first be oxidized so that it can then

removed in buffered HF along with the masking oxide.
4. Define lower insulating layer

100

a. Grow 1000 Â thick S1O2 layer: 10 minutes 1000° C in steam.
b. Spin on 1400-37 photoresist for 30 seconds at 4000 rpm.

c. Soft bake for 30 minutes at 85°C.
d. Pattern pedestal: 85 mJ∕cm2.

e. Develop in Shipley 455 developer for 1 minute.
f. Etch for 2 minutes in buffered HF.

g. Remove photoresist: 5 minutes acetone, 5 minutes ethanol, 5 minutes
running water, blow dry, and 2 minutes oxygen plasma.

The lower SiO2 insulating layer is a high quality insulator that is etched
at less than lOOÂ/hour in EDP and provides a pinhole free, high quality glass
surface with which to contact the cell. It is desirable to have as much of the

top of the cup etched as possible to reduce the electrode impedance. This is
simplified since the resist is thinner on the top of the cup so that only a very
small exposure is required to completely etch the top of the cup. There is a limit
on the thickness of this layer. As a layer of oxide is grown boron at the surface

of the wafer tends to segregate into the oxide layer, reducing the concentration
of boron in the surface layer of silicon.

For long oxide growths the surface

concentration of boron is low enough that the silicon is no longer protected from
the EDP etch, and the pedestal is attacked.

5. Define conducting layer.
a. Evaporate 100 Â Cr, 800 Â Au, 100 Â Cr.
b. Pattern contact lead with 1400-37 photoresist.

c. Clean in O2 plasma for 2 minutes.
d. Etch: Transene chrome etch 15sec, Transene gold etch:water(4:l) 30 sec,
Transene chrome etch 15sec.

e. Remove photoresist and clean.
The conducting layer is defined with an etch since adhesion of the conducting
layers was consistently better than with a liftoff. Since gold does not stick well
to oxides chrome was used as an adhesion layer. The chrome oxidizes and sticks

well to the silicon dioxide. If the gold is evaporated on to chrome that has not
been oxidized it alloys and sticks. However, chrome oxidizes very easily so that

the evaporation must be carried out at a low pressures (less than 2 × 10-6iorr).

101

Evaporation conditions are kept as similar as possible to keep the film stress and

etching properties reproducible.
The chrome layers are etched in 9 parts cerric sulfate, 1 part HNO3 [7], and

the gold layer is etched in a solution of 400 g KI, 100 g I2, and 400 ml of water
diluted with water 1:4 [8].
6. Define top insulating layer.

a. Deposit 1.5 μm oxinitride: NH3∕N2O∕SiH4 (flow rate 750/1150/250
seem) are reacted at 300° C at a pressure of 2.0 torr and a power density
of 11.7mW∕cm-2.

b. Spin on AZ 4770 photoresist for 30 seconds at 4000 rpm.

c. Soft bake for 100 minutes at 85° C.
d. Pattern pedestal: 400mJ/cm2.

e. Develop in AZ developer for 4 minutes.

c. Plasma etch: O2∕65∏1torr, CF4∕34θmtorr, net power 80 watts, for 70
minutes.
d. Remove photoresist and clean wafer.
The oxinitride layer is deposited using plasma enhanced chemical vapour

deposition [9], [10]. This allows the deposition of a film with a well defined,

selectable stress, at temperatures low enough to prevent undesirable alloying of

the chrome with the gold. The silicon oxinitride layer is etched in the O2, CF4
plasma [11]. Etch completion can be determined by the rough surface of the
etched silicon and the removal of the upper chrome layer on the bonding pad as

seen in Figure A.3.
7. Make wire connections to each device.

a. Scribe the wafer and glue to a microscope slide, along with a gold pad
for the other end of the wire bond. Use RTV and allow to cure for at

least 24 hours before etching.
b. Wire bond with 25 μm gold wire: use Westbond ultrasonic wirebonder

with power and time set to maximum. Only wire bond those devices
that pass optical inspection.

102

Figure A.3. Wafer after the oxinitride layer has been plasma etched.

Wire bonding is done with gold wire that will stand up to the EDP. The
wire sticks well to clean gold, but sticks poorly or not at all to other surfaces.

8. Free the devices.

a. Dip in buffered HF diluted 3:1, and rinse in water.
b. EDP 100°C for 3 hours.

c. Rinse successively in clean beakers of water.
d. Dry in 85° C oven.
e. Inspect under optical microscope and discard any defective devices.

The wire-bonded wafers are cleaned of their native oxide and then the silicon
is etched out from beneath each device [12,13]. At this point the devices are

held to the wafer only by temporary grill work. Great care must be taken not
to damage the devices. They must be cleaned in successive water baths.

9. Insulate the wire and wire bond.
a. prepare polyimide tube

b. using number 3 tweezers load each device into a tube, putting the wire
through first, and then pulling the wire through from the other end.

c. bend the pedestal to make a 60 degree angle with the polyimide tube.
d. use a drop of negative photoresist to insulate the wire bond.

103
e. bake for at least 4 hours at 150oC

Special holders facilitate quick and easy loading of the polyimide tube and
insulation of the wire bond. The polyimide tubes are temporarily held in place
by photoresist.

10. Platinize and test.

a. Etch: Transene chrome etch 1 minute.

b. Measure electrode impedance.

c. Platinized in a solution of 1% chloroplatinic acid in 0.0025% HC1, plus

0.01% lead acetate, using a current density of 20mA∕cm2 for 10 sec
onds.

d. Measure electrode impedance.

For an exposed area of 20μm2 the impedance in saline at 1000 kHz is
25 megohms [14]. This value can be reduced to about 500 kilohms, by elec
troplating with platinum.

104
References

[1] B. Schwartz and H. Robbins, “Chemical etching of silicon, IV. The system

HF, HNO3, HC2H3O2, J. Electrochem. Soc., vol. 106, p. 505, 1959.
[2] L. E. Katz,“Oxidation,” in VLSI Technology, ed. S. M. Sze, p. 194, 1983.
[3] A. S. Grove, Physics and Technology of Semiconductor Devices, Wiley, New

York, 1967.
[4] Sohio Carborundum PDS Planar Diffusion Sources: Technical Information,
1986.

[5] D. Rupprecht. and J. Stach, “Oxidized boron nitride wafers as an in situ
boron dopant for silicon diffusions,” J. Electrochem. Soc., vol. 120, no. 9,
pp. 1266-1271, 1973.

[6] J. Stack, and A. Turley, “Anomalous boron diffusion in silicon from planar
boron nitride sources,” J. Electrochem. Soc., vol. 121, no. 5, pp. 722-724,

1974.
[7] R. Glang and L. V. Gregor, “Generation of patterns in thin films,” in Hand
book of Thin Film Technology, ed. L. I. Maissel and R. Glang, McGraw-Hill

Book Company, New York, p. 7-36, 1970.
[8] R. Glang and L. V. Gregor, “Generation of patterns in thin films,” in Hand
book of Thin Film Technology, ed. L. I. Maissel and R. Glang, McGraw-Hill

Book Company, New York, p. 7-37, 1970.
[9] W. R. Snow, “PWS 450 COYOTE: A vertical parallel plate plasma reactor for
silicon nitride deposition,” Pacific Western Systems Application Note, March

1981.
[10] T. Sugano, Applications of plasma processes to VLSI technology, John Wiley

& Sons Inc., New York, 1985.
[11] C. M. Melliar-Smith and C. J. Mogab, “Plasma assisted etching techniques

for pattern deliniation” in Thin Film Processes, ed. J. L. Vossen and W. Kem,

Academic Press, New York, pp. 497-552, 1978.
[12] R. M. Finne, and D. L. Klein J.Electrochem. Soc., “A water amine complex

ing agent system for etching silicon,” vol. 114, pp.965-970, Sept. 1967.
[13] A. Reisman, M. Berkenblit, S. A. Chan, F. B. Kaufman, and D. C. Green,

“The controlled etching of silicon in catalyzed ethylenediamine-pyrocatechol-

water solutions,” J. Electrochem. Soc., vol. 126, no. 8, pp.1406-1415, 1979.

105

[14] R. C. Gesteland, B. Howland, J. Y. Lettvin, and W. M. Pitts, “Comments
on microelectrodes,” Proc. IRE, vol. 47, pp. 1856-1862, 1959.