Carbon nanotube micro-electrodes for neuronal interfacing

HIGHLIGHT www.rsc.org/materials | Journal of Materials Chemistry

Carbon nanotube micro-electrodes for neuronalinterfacingE. Ben-Jacoba and Y. Hanein*b

DOI: 10.1039/b805878b

This article highlights our recent progress in developing carbon nanotube based electrodes forneurochip applications. By integrating carbon nanotube growth with standard micro-fabrication techniques we have realized novel carbon nanotube based micro-electrode arraysfor neuronal interfacing and network engineering. The novel electrodes possess a unique setof properties that make them a promising platform for future neuronal interfacingapplications. In particular, carbon nanotube electrodes may potentially be used for neuro-prosthetic devices or as novel biosensors.

Introduction

The design and development, in recent

decades, of electronic chips capable of

interfacing with neuronal systems have

stirred immense interest.1–5 Development

of such neuronal devices is geared

towards the development of two related

device categories: one of in-vivo applica-

tions such as retinal and brain implants,6

the other of in-vitro devices for basic

investigation of neuronal systems. The

latter category of devices, often termed

‘‘brain on a chip’’ (or neurochip), may be

used to better understand biological

mechanisms or even for the realization of

Prof: Eshel Ben-Jacob

Eshel Ben J

PhD in phys

Aviv Univers

worked in the

ing to Tel A

1986, where

Maguy-Glass

Physics of

Professor Be

interests inc

tion and p

in nonlinea

DNA nano-el

computing.

aSchool of Physics and Astronomy, Tel-AvivUniversity, Tel-Aviv, 69978, Israel. E-mail:[email protected] of Electrical Engineering, Departmentof Physical Electronics, The Iby and AladarFleischman Faculty of Engineering, Tel-AvivUniversity, Tel-Aviv, 69978, Israel. E-mail:[email protected]

This journal is ª The Royal Society of Chemistry

organic computers.7 Clearly, the feasi-

bility of creating devices suited for any of

these possible applications rests in the

ability to produce high quality chips with

efficient interfacing between the chip and

the biological system.

For both application categories, the

integration between biological cells and

electronic chips is laden with challenges:

biocompatibility, chip durability in a wet

environment, packaging issues, electronic

circuitry to accommodate data collection,

and software to accommodate data anal-

ysis and compression, to name a few.

Owing to intensive research in the last

several decades, remarkable progress has

been achieved in all of these issues.6,8–10

Of the various challenges, achieving the

right properties of the actual interface

between the chip and the cells are of

utmost importance. This interface is crit-

ical in ensuring both the viability of the

cells and the effectiveness of the electrical

interface.

acob finished his

ics (1982) at Tel

ity, Israel. He then

US before return-

viv University in

he is currently

Professor in

Complex Systems.

n Jacob’s research

lude self-organiza-

attern formation

r open systems,

ectronics and DNADr Yael Hanein

2008 J

The desired properties of the interface

between the chip and neurons depend on

the transduction method used to convert

the electro-physiological activity of the

cells (i.e. ionic currents) to electrical

signals (i.e. electronic voltage or current).

The most common transduction scheme

in use today to interface with neuronal

cells is based on planar metallic

electrochemical electrode arrays (multi-

electrodes arrays, MEAs). When in

contact with ionic solution (i.e. biological

medium), these electrodes can be

regarded as electrochemical capacitors

and the coupling between the electrodes

and the cells is capacitive in nature.1,2

Consequently, a fundamental require-

ment that these interfaces have to accom-

modate is large electrode capacitance.

This parameter defines the impedance of

the electrode and determines its noise

level, how well it can record electrical

activity and, most importantly for

implantable devices, its effectiveness in

Yael Hanein received her PhD in

physics from the Weizmann

Institute of Science, Israel. After

a research associate position at

the University of Washington,

Seattle, US, she joined the

School of Electrical Engineering

of Tel-Aviv University as

a senior lecturer specialising in

bio-nanosystems. Her current

interests include BioMEMS,

micro-self-assembly, and carbon

nanotube electronic devices.

. Mater. Chem., 2008, 18, 5181–5186 | 5181

stimulating cell activity. On the other

hand, to allow high specificity, the elec-

trodes must be as small as possible. Thus,

to optimize these two contradicting

factors (i.e., small electrode projection and

large effective area), electrodes with high

specific areas are most desired. Addition-

ally, since neuronal cells are far more

particular than most cell types, special

surface treatment prior to culturing is

required to facilitate cell adhesion, proli-

feration, growth and proper activity.

Commonly, these two requirements

(proper electrical and biological coupling)

are met separately, using two independent

approaches. To achieve electrodes with

large specific area various roughening

procedures are used, and various chem-

ical modification schemes are employed

to mediate cell adhesion. Among the

roughening procedures, electroplating is

the most commonly used (in particular,

platinum black electroplating11,12), but

recently the use of highly porous sput-

tered TiN has been demonstrated as

a very effective scheme.13 Capacitance

values are in the several mF cm�2 range.14

Surface chemical modification schemes to

ensure cell adhesion and viability include

various adhesion proteins and growth

factors.15 The deposition of the needed

chemistry is typically straightforward, but

its addition onto the electrode may affect

the electrical performances of the elec-

trode as well as its stability. Restricting

the adhesion of the cells to specific

locations in order to control the geometry

of in-vitro neuronal networks is highly

desired and is often achieved by

patterning the chemical adhesion layer in

alignment with the recording sites.16–18

Even though surface chemical modifi-

cation is the most commonly used scheme

to achieve cell adhesion in neurochip

technology,19 a second alternative exists.

A great deal of research has revealed the

strong sensitivity of cells to the roughness

of the surface.20,21 In fact, it has been

shown that by increasing the roughness of

the surface to the nano-scale it is possible

to achieve cell adhesive surfaces. With the

use of lithographic methods to create

heterogeneous surfaces containing flat

and rough regions, it was possible to show

that cells would preferentially adhere to

the rough surface even if the chemistry

of the rough and flat surfaces was

identical.22,23 It should be noted that,

despite the compelling evidence linking

5182 | J. Mater. Chem., 2008, 18, 5181–5186

surface topography and surface rough-

ness to cell adhesion, the origin of this link

is not understood. We shall return to

discuss this issue in the Conclusions and

outlook section of this article.

The fact that rough surfaces can

provide a good substrate for cell adhesion

implies that a direct electrical interface

between recording electrodes and cells

can be obtained provided the surface of

the electrode is both conducting and

rough enough to allow direct cell attach-

ment to the surface. The rough surface

must also be very durable to avoid

breakage. An ideal material for these

requirements is carbon nanotube (CNT)

paper. CNT paper is an entangled,

spaghetti-like, three-dimensional matrix

of CNTs. Being electrically conductive,

these materials are excellent electro-

chemical capacitors and have been

investigated for super-capacitance appli-

cations with and without additional

chemical treatments.

CNT paper is also a promising material

in the realm of biomaterials,24 and in

particular for neurological applica-

tions,14,25–31 since CNTs possess a unique

combination of properties: they are

biologically inert, flexible yet strong, and

electrically conductive. Their electrical

conductivity renders them particularly

interesting for neuronal applications.

Extensive investigations in the last several

years have validated their biocompati-

bility and suitability to interface with

neurons. Even though various surface

modification schemes have been explored,

pristine CNTs make an excellent surface

for cell culturing.27

Our research in the last several years

was concerned with the development of

a novel class of micro-electrodes coated

with islands made of CNTs that were

specifically designed as an interface for

neuronal systems. We review below the

main progress in this endeavor.

CNT coated electrodes

To develop a superior electro-chemical

interface for neuronal recordings, we

developed a unique micro-fabrication

process that allows integration of very

high density CNT paper directly onto

micro-electrodes. The electrodes were

fabricated by synthesizing (by a thermal

chemical vapor deposition process) high

density carbon nanotube islands on

This journ

lithographically defined, passivated

titanium nitride (TiN) conductors on

a silicon dioxide substrate. The addition

of CNTs provides an elegant and efficient

solution to the two main challenges in

building efficient micro-electrodes for

neuronal applications. Foremost, CNT

coated surfaces have a very large surface

area, thus can achieve the very high

specific capacitance necessary for low

noise electrodes. Second, the roughness of

the CNT surface is suitable for anchoring

the cells to the electrode site. The inter-

action between the cells and the surface is

so strong that it dramatically affects

the overall arrangement of the cells on

the surface, thus enabling network

engineering.

The use of CNT paper as a coating

material for neuronal electrodes thus

facilitates two fundamental capabilities.

First, the electrodes act as anchoring

points onto which cells can adhere and

grow. Second, the electrodes can be used

for high fidelity electrical recording and

stimulation.

As expected, one of the most conspic-

uous effects observed when culturing

neurons on CNT coated islands is the

strong affinity of cells to these surfaces.

This affinity is best illustrated when

isolated islands of CNTs are realized on

a flat SiO2 background (Fig. 1), in which

case cells migrate from the flat surface and

adhere exclusively onto the CNT coated

areas.

However, this is not the only noticeable

effect associated with the interaction of

neurons with the CNT. In addition to the

apparent cell affinity to the CNT surfaces,

cells cultured on CNT electrodes also

self-form ordered networks (Fig. 2).

Connections between cells attached to

different islands are clearly apparent, and

interconnected networks are formed

following the exact pattern of the CNT

templates. This process typically occurs

by bridging gaps between neighboring

islands. The bridging consists of either an

axon or bundles of axons and dendrites.

In some cases, the bridge is covered with

clusters of cells.

The manner by which the cells move

towards the CNT-coated surfaces and

rearrange into a network is not obvious.32

To investigate the origin of this mobility,

we examined the dynamics of neurons by

placing cultures in an environmental

chamber under a microscope and taking

al is ª The Royal Society of Chemistry 2008

Fig. 1 Neuronal cells adhere preferentially to isolated islands of pristine CNTs patterned on SiO2

surface. The adhesion of the neuronal cells is typified by extensive process growth and branching.

Glia cells also adhere selectively to the CNT surfaces. (a) High resolution scanning electron

microscope (HRSEM) image. (b) Confocal fluorescence image of neurons (red) and glia cells (green)

on a large CNT Island. In both pictures the bar is 10 mm.

Fig. 2 Neuronal network arranges on isolated CNT crosses. Cell bodies are anchored to the CNT

coated regions and are interconnected by thick bundles of axons and dendrites. Dark crosses are

regions coated with high density CNTs.

time lapse recordings at four-minute

intervals. The most conspicuous process

revealed by investigating the temporal

development of these cultures in vitro

takes place in the first two days after cells

are plated on the surface of the substrate.

Interestingly, the network connectivity is

not the result of the dynamics of single

cells. Rather, the cells first aggregate into

clusters at random locations on the

substrate, after which they separate into

individual clusters that migrate and cling

to the CNT sites (Fig. 3). As these clusters

migrate, an isolated process bundle

remains suspended between the islands

and stretches with the increasing distance

between the cell clusters. This process is

This journal is ª The Royal Society of Chemistry

depicted in Fig. 3, where a 10 to 15 hour

time lapse shows a clear separation

between two clusters and the formation

and extension of a single connecting

bundle between them. It is important to

note that several additional processes are

taking place which contribute to the final

shape of the network, including process

navigation on the glass surface and

process stretching and fasciculation.

To utilize the CNT coating as electrical

electrodes, a special micro-fabrication

procedure was developed (Fig. 4). This

method has to accommodate both the

high-temperature chemical vapor deposi-

tion (CVD) CNT growth and the harsh

ionic environment used to culture the

2008 J

cells. The electrodes were realized either

on quartz or oxidized silicon substrates.

Titanium nitride (TiN) was found to be

the ultimate choice for realizing the

metallic tracks due to its conductivity,

durability in the high temperature CVD

process and its compatibility with the

CNT growth. TiN is an ideal barrier layer

onto which a thin nickel layer is depos-

ited. During the CVD process, the Ni

layer on the TiN separates into small

droplets acting as catalyst sites for the

subsequent CNT growth. After their

deposition, the TiN tracks are coated with

sputtered Si3N4, which passivates and

protects the TiN from the ionic solution

of the neuron culturing medium. The

Si3N4 layer, like the TiN layer, has to be

durable in the high temperature process of

the CVD growth, thus its stress level and

content must be tuned to avoid cracking.

Using proper deposition conditions of the

various layers, it is possible to build stable

chips with durability of several weeks in

the biological medium.

The three dimensional nature of the

CNT electrodes contributes to a very

large surface area, and consequently to

large electrode specific capacitance

and low frequency dependence of the

electrode impedance.33 These properties

were validated using DC and AC elec-

trochemical characterization. The values

obtained with our micro-fabricated CNT

electrodes are consistent with previously

reported values of macroscopic CNT

electrodes, validating the success of our

scheme in producing high quality,

microscopic CNT electrodes.14

The superb DC and AC properties of

the CNT MEA are compounded by their

interesting interaction with biological

cells described above. By culturing

neurons on CNT electrodes in properly

calibrated cell density it is possible to

form highly organized neuronal systems

in excellent fidelity to the position of the

electrodes (Fig. 5a). The electrodes func-

tion as focal points that attract the cells

and facilitate the organization of the

network, thus making it possible to

engineer the network geometry. For

comparison, if the entire surface of the

chip is coated with an adhesive protein

layer, cells may adhere to the rough as

well as the smooth surface of the chip and

the effectiveness of the electrodes in

positioning the cells and shaping the

network is markedly reduced (Fig. 5b).

. Mater. Chem., 2008, 18, 5181–5186 | 5183

Fig. 3 The dynamics of a network self-organization process on 100 mm patterned CNT islands, with

150 mm separation between the islands. Arrows indicate the process bundle interconnecting the two

separating clusters. (A) 30 h after plating; (B) 40 h after plating; (C) 55 h after plating; (D) 69 h after

plating. The final compact bundle (of axons) connecting the two isolated islands is depicted in

(E) (adapted from ref. 32).

Fig. 4 The micro-fabrication process scheme of the CNT-based electrodes. (a) A side view illus-

trating the patterned TiN tracks and the Si3N4 passivation layer. In the next step, holes are etched in

the Si3N4 layer and a subsequent Ni layer is deposited inside the holes directly onto the TiN layer.

Consequent thermal CVD growth procedure deposits CNTs using the Ni layer as a catalyst (b). (c)

Top view of CNT electrodes. (d) Optical microscope image of the multi- electrode array device (only

a part of the device with several of the electrodes is shown).

After several days in culture, the

networks generate spontaneous activity

marked by firing of neuronal action

potentials (spikes). The activity can be

captured by recording the extra-cellular

signals generated by cell assemblies

formed on the CNT electrodes (Fig. 6).

The recorded signals demonstrate

a typical extra-cellular recording shape.

Note that the electrodes detect the first

derivative of the actual neuronal action

potential since the interface is achieved via

capacitive coupling. Typical amplitudes

of the recorded signals are in the hundred

5184 | J. Mater. Chem., 2008, 18, 5181–5186

mV range. Stimulated electrical activity

is also achieved by the very same

electrodes. In contrast to the activity of

uniform networks, which is very

synchronous,34 in clustered networks

the overall level of synchronization at

the network level depends on the strength

of connectivity between the neuron

clusters on the different electrodes. For

loosely connected clusters, the emerged

synchronization appears to be localized to

specific clusters while global synchroni-

zation between clusters is suppressed

(Fig. 6c).

This journ

Conclusions and outlook

As we described, the most important

advantage of the CNT electrodes is their

unique ability to interface electrically

and mechanically with neurons. These

properties facilitate two-way, unmediated

interfacing between electrodes and neural

cells.

As highlighted above, multi-electrode

array devices are developed with the

foresight to make effective in-vitro and

in-vivo devices. We review below two

specific examples of the possible advan-

tages of CNT electrodes in the realm of

neurological applications.

One of the most profound open

challenges in modern science has to do

with the emergence of a functioning brain

from a collection of individual neurons.

This question translates to how an

assemblage of initially simple elements,

namely neurons, self-organizes to form

a new, extremely complex and highly

coordinated functioning system: the

neural network. Even at the level of

a simple network, unraveling the under-

lying mechanisms involved in the complex

activity of neural circuits has been, and

still remains, a fundamental challenge.35

A common and well-accepted

approach to studying these issues is

through the use of cultured neuronal

networks on MEA systems. Cultured

networks offer the proper model system

for investigating small-scale networks

under controlled conditions for a long

time (weeks) and at very high resolution.17

Extra-cellular physiological measure-

ments combined with an expanded

arsenal of staining techniques allow

effective characterization and identifica-

tion of the various biological elements.

To better define the cultured systems,

forming ordered networks of neurons

has been the focus of intense research

over many years.11 This goal has been

achieved to an extent by many groups,

using a wide range of techniques,

primarily surface modification and soft

lithography. Despite the considerable

success of some of these approaches in

forming ordered networks, none of them

can facilitate high-fidelity electrical

investigation of the individual cells in

these networks. Typically, cells are only

loosely coupled to the recording elec-

trodes and also tend to be very mobile.

Hence, these systems offer only a very

al is ª The Royal Society of Chemistry 2008

Fig. 5 Optical microscope images of neurons on 80 mm CNT islands: (a) patterned interconnected

neuronal systems formed with pristine CNT islands. Closer inspections (by electron microscope

observations or staining) of the network in this case reveal that all the cells are attached to the CNT

islands. Note that cells attached to the electrodes are not visible in the optical image since the bases of

the electrodes are not transparent. For comparison, we show in (b) a random network formed on

a CNT island array coated with an adhesive protein layer. We can clearly see here cells that are

spread between the electrodes.

Fig. 6 Spontaneous electrical activity of

neuronal clusters on CNT electrodes. (a, b) A

voltage trace of spontaneous electrical activity

recorded from a CNT electrode. (c) A raster

plot of the spiking activity in several CNT

electrodes. Activity patterns are characterized

by synchronized bursting events (SBEs)—short

time windows (several hundreds of millisec-

onds) of rapid collective neuronal firing

followed by long intervals (tens of seconds) of

sporadic firing. The events of mutual synchro-

nization between the SBEs of different

electrodes are relatively rare in this example.

limited scope to investigate properties

related to coordinated activity of

neuronal circuits. For example, contem-

porary methods do not allow contin-

uous, multi-site, high resolution

investigation of signal propagation at the

single neuron level.

This journal is ª The Royal Society of Chemistry

The success of the CNT electrodes in

patterning and interfacing with neurons

suggests the unique possibility to carry

out investigations targeting basic

questions related to the activity of small

neuronal circuits. CNT electrodes facili-

tate the construction of the circuit as

well as provide the means to perform the

electrical recordings. Moreover, prelimi-

nary results demonstrate that the

CNT MEAs are also suitable for

interfacing with single cells at a very

high level of fidelity. By using cultures of

locust cells (neurons can be as large as

50 mm) we are now able to pattern

neuronal circuits consisting of ordered,

isolated single cells.

Moreover, since only very few neurons

occupy each electrode site, it is possible to

simultaneously monitor the activity of

these very small cell populations at

well known variable conditions such as

different degrees of connectivity to other

clusters, number of cells, number of glia

cells, etc. Such measurements could help

elucidate important aspects related to

cell–cell signaling, and in particular the

important influence of glia cell signaling

on neuronal activity and synchronization.

The high viability of the neuronal cells

on CNT surfaces, as is manifested in their

rapid development, also suggests their

potential in neuronal implant applica-

tions. In these devices, one of the major

challenges is the reactivity of the tissue

with the electrode surface. In retinal

implant devices, for example, one of the

major problems is the weak coupling

between the tissue and the electrodes,

which results in very poor stimulation

2008 J

specificity. The cells are simply too far

away from the electrode to allow stimu-

lation at high efficacy, in terms of both

charge injection and site specificity.

Preliminary tests we conducted have

validated the potential of this approach.

Mouse retinal whole mounts were placed

onto CNT MEAs, with ganglion cells (the

output cells of the retina) in direct contact

with the electrodes. Spontaneous bursts

of neural impulses as well as slow oscil-

latory activity were routinely recorded,

indicating the viability of these electrodes

for recording from intact neural tissue.

Successful stimulation tests were also

performed. We anticipate that the cell-

adhesive nature of our CNT electrodes, as

compared to standard microelectrodes

or even to vertically aligned, multi-walled

carbon nanotube microelectrodes, may

offer the important advantage of

promoting neuronal proliferation and

improving tissue–electrode interaction

without the need for additional adhesion

promoting coating.31

Even though the CNT electrodes

appear to be very effective, we believe that

their properties can be further enhanced

through surface modification. Special

chemical treatments can be applied to

improve the specificity of the electrodes

and to facilitate novel interfacing

schemes. It is also important to note

that the fabrication procedure currently

used to produce the CNT MEA employs a

relatively high temperature CVD proce-

dure, which restricts the range of

substrates used while also increasing the

fabrication cost relative to conventional

MEA. This drawback may be overcome

by developing alternative low-cost,

low-temperature procedures.

Our final comment is concerned with

the origin of the cell–CNT interaction.

Despite the great excitement in this field,

the underlying mechanism that governs

the interaction between cells and rough

surfaces in general, and with carbon

nanotubes in particular, is still elusive.

Our recent investigations into this issue

suggest a strong cellular sensitivity to

the surface roughness, which results

in process twining that may lead to cell

entanglement in the three dimensional

CNT matrix. Clearly, effective exploita-

tion of these surfaces in future bio-

medical devices strongly depends on

a better understanding of the biological

mechanisms involved.

. Mater. Chem., 2008, 18, 5181–5186 | 5185

Acknowledgements

This work was supported in part by an

ISF grant and by the Tauber Fund. The

authors thank Moshe David-Pur, Tamir

Gabay, Raya Sorkin, Alon Greenbaum,

Mark Shein, and Inna Brains who

contributed to and assisted in performing

the research described herein. YH thanks

Amir Ayali, Shlomo Yitzchaik, Evelyne

Sernagor, Chris Adams, and Danny

Baranes for very stimulating discussions.

References

1 J. Ji and K. D. Wise, An implantableCMOS circuit interface for multiplexedmicroelectrode recording arrays, IEEE J.Solid State Circuits, 1992, 27, 433–443.

2 Y. Jimbo and A. Kawana, Electricalstimulation and recording from culturedneurons using a planar electrode array,Bioelectrochem. Bioenerg., 1992, 29, 193.

3 P. Fromherz, C. O. Muller and R. Weis,Neuron transistor: electrical transferfunction measured by the patch-clamptechnique, Phys. Rev. Lett., 1993, 71,4079–4082.

4 M. P. Maher, J. Pine, J. Wright andY.-C. Tai, The neurochip: A newmultielectrode device for stimulatingand recording from cultured neurons,J. Neurosci. Methods, 1999, 87, 45–56.

5 S. Tatic-Lucic, Y.-C. Tai, J. A. Wright andJ. Pine, Silicon Cultured-Neuron ProstheticDevices for In Vivo and In Vitro Studies,Sens. Actuators, B: Chem., 1997, 43,105–109.

6 J. D. Weiland, W. Liu and M. S. Humayun,Retinal prosthesis, Annu. Rev. Biomed.Eng., 2005, 7, 361–401.

7 I. Baruchi and E. Ben-Jacob, TowardsNeuro-Memory-Chip: Imprinting multiplememories in cultured neural networks,Phys. Rev. E, 2007, 75, 050901.

8 E. Hulata, R. Segev, Y. Shapira,M. Benveniste and E. Ben-Jacob,Detection and sorting of neural spikesusing wavelet packets, Phys. Rev. Lett.,2000, 85, 4637–4640.

9 J. J. Pancrazio, S. A. Gray, Y. S. Subin,N. Kulagina, D. S. Cuttino,K. M. Shaffer, K. Eisemann, A. Curran,B. Zim, G. W. Gross andT. J. O’Shaughnessy, A portablemicroelectrode array recording systemincorporating cultured neuronal networksfor neurotoxin detection, Biosens.Bioelectron., 2003, 18(11), 1339–1347.

10 Y. Hanein, C. G. J. Schabmueller,G. Holman, P. Lucke, D. D. Denton andK. F. Bohringer, High-aspect ratio

5186 | J. Mater. Chem., 2008, 18, 5181–5186

submicrometer needles for intracellularapplications, J. Micromech. Microeng.,2003, 13, S91–S95.

11 C. D. James, A. J. H. Spence,N. M. Dowell-Mesfin, R. J. Hussain,K. L. Smith and H. G. Craighead,Extracellular recordings from patternedneuronal networks using planarmicroelectrode arrays, IEEE Trans.Biomed. Eng., 2004, 51, 1640–1645.

12 http://www.med64.com/products.html.13 M. Janders, U. Egert, M. Stelzle and

W. Nisch, Novel thin film titanium nitridemicro-electrodes with excellent chargetransfer capability for cell stimulation andsensing applications, Nano Lett., 1996, 1,245–247.

14 T. Gabay, M. Ben-David, I. Kalifa,R. Sorkin, Z. R. Abrams, E. Ben-Jacoband Y. Hanein, Electro-chemical andbiological properties of carbon nanotubebased multi-electrode arrays,Nanotechnology, 2007, 18, 035201.

15 M. Matsuzawa, S. Tokumitsu, W. Knolland H. Sasabe, Effects of organosilanemonolayer films on the geometricalguidance of CNS neurons, Langmuir,1998, 14, 5133–5138.

16 J. C. Chang, G. J. Brewer andB. C. Wheeler, Modulation of neuralnetwork activity by patterning, Biosens.Bioelectron., 2001, 16, 527.

17 R. Segev, M. Benveniste, E. Hulata,N. Cohen, A. Palevski, E. Kapon,Y. Shapira and E. Ben-Jacob, Long termbehavior of lithographically prepared invitro neuronal networks, Phys. Rev. Lett.,2002, 88, 118102.

18 A. Offenhausser, S. Bocker-Meffert,T. Decker, R. Helpenstein, P. Gasteier,J. Groll, M. Moller, A. Reska, S. Schafer,P. Schulte and A. Vogt-Eisele,Microcontact printing of proteins forneuronal cell guidance, Soft Matter, 2007,3, 290–298.

19 A. Folch and M. Toner, Microengineeringof cellular interactions, Annu. Rev.Biomed. Eng., 2000, 2, 227.

20 N. M. Dowell-Mesfin, M. A. Abdul-Karim,M. Turner, S. Schanz, H. G. Craighead,B. Roysam, J. N. Turner and W. Shain,Topographically modified surfaces affectorientation and growth of hippocampalneurons, J. Neural Eng., 2004, 1, 78–90.

21 S. K. Seidlits, J. Y. Lee and C. E. Schmidt,Nanostructured scaffolds for neuralapplications, Nanomedicine, 2008, 3,183–199.

22 Y. W. Fan and et al., Culture of neural cellson silicon wafers with nano-scale surfacetopography, J. Neurosci. Methods, 2002,120, 17–23.

23 C. C. Berry, G. Campbell, A. Spadiccino,M. Robertson and A. S. G. Curtis, Theinfluence of microscale topography on

This journ

fibroblast attachment and motility,Biomaterials, 2004, 25, 5781–5788.

24 E. Bekyarova, Y. Ni, E. B. Malarkey,V. Montana, J. L. McWilliams,R. C. Haddon and V. Parpura,Applications of Carbon Nanotubesin Biotechnology and Biomedicine,J. Biomed. Nanotechnol., 2005, 1, 3–17.

25 M. P. Mattson, R. C. Haddon andA. M. Rao, Molecular functionalization ofcarbon nanotubes and use as substratesfor neuronal growth, J. Mol. Neurosci.,2000, 14, 175–182.

26 H. Hu, Y. Ni, V. Montana, R. C. Haddonand V. Parpura, ChemicallyFunctionalized Carbon Nanotubes asSubstrates for Neuronal Growth, NanoLett., 2004, 4, 507–511.

27 T. Gabay, E. Jakobs, E. Ben-Jacob andY. Hanein, Carbon nanotube basedneuro-chip for engineering, recording andstimulation of cultured networks,Proceedings of the 13th InternationalConference on Solid-State Sensors,Actuators and Microsystems, IEEE,Piscataway, NJ, 2005, 2, 1226–1229.

28 T. Gabay, E. Jakobs, E. Ben-Jacob andY. Hanein, Engineered self-organizationof neural networks using carbon nanotubeclusters, Physica A, 2005, 350, 611–21.

29 V. Lovat, D. Pantarotto, L. Lagostena,B. Cacciari, M. Grandolfo, M. Righi,G. Spalluto, M. Prato and L. Ballerini,Carbon Nanotube Substrates BoostNeuronal Electrical Signaling, Nano Lett.,2005, 5–6, 1107–1110.

30 X. Zhang, S. Prasad, S. Niyogi, A. Morgan,M. Ozkan and C. S. Ozkan, Guided neuritegrowth on patterned carbon nanotubes,Sens. Actuators, B: Chem., 2005, 106, 843–850.

31 K. Wang, H. A. Fishman, H. Dai andJ. S. Harris, Neural Stimulation withCarbon Nanotube Microelectrode Array,Nano Lett., 2006, 6, 2043–2048.

32 R. Sorkin, T. Gabay, P. Blinder,D. Baranes, E. Ben-Jacob and Y. Hanein,compact self-wiring in cultured neuralnetworks, J. Neural Eng., 2006, 3, 95–101.

33 J. Li, A. Cassell, L. Delzeit, J. Han andM. Meyyappan, Novel three-dimensionalelectrodes: Electrochemical properties ofcarbon nanotube ensembles, J. Phys.Chem. B, 2002, 106, 9299–9305.

34 R. Segev, Y. Shapira, M. Benveniste andE. Ben-Jacob, Observation and modelingof synchronized bursting in two-dimensional neuronal networks, Phys.Rev. E, 2001, 64, 011920.

35 J. J. Hopfield, Neural Networks andPhysical Systems with Emergent CollectiveComputational Abilities, Proc. Natl. Acad.Sci. U. S. A., 1982, 79, 2554–2558.

al is ª The Royal Society of Chemistry 2008