Page 1

IOP PUBLISHING

NANOTECHNOLOGY

Nanotechnology 19 (2008) 025701 (12pp)

Electromechanical interactions in a carbon

nanotube based thin film field emitting

diode

doi:10.1088/0957-4484/19/02/025701

N Sinha1, D Roy Mahapatra2, Y Sun1, J T W Yeow1, R V N Melnik3

and D A Jaffray4

1Department of Systems Design Engineering, University of Waterloo, Waterloo, ON,

N2L3G1, Canada

2Department of Aerospace Engineering, Indian Institute of Science, Bangalore 560012, India

3M2NeT Laboratory, Wilfrid Laurier University, Waterloo, ON, N2L3C5, Canada

4Department of Radiation Physics, Princess Margaret Hospital, Toronto, ON, M5G2M9,

Canada

E-mail: jyeow@engmail.uwaterloo.ca

Received 13 September 2007

Published 6 December 2007

Online at stacks.iop.org/Nano/19/025701

Abstract

Carbon nanotubes (CNTs) have emerged as promising candidates for biomedical x-ray devices

and other applications of field emission. CNTs grown/deposited in a thin film are used as

cathodes for field emission. In spite of the good performance of such cathodes, the procedure to

estimate the device current is not straightforward and the required insight towards design

optimization is not well developed. In this paper, we report an analysis aided by a

computational model and experiments by which the process of evolution and self-assembly

(reorientation) of CNTs is characterized and the device current is estimated. The modeling

approach involves two steps: (i) a phenomenological description of the degradation and

fragmentation of CNTs and (ii) a mechanics based modeling of electromechanical interaction

among CNTs during field emission. A computational scheme is developed by which the states

of CNTs are updated in a time incremental manner. Finally, the device current is obtained by

using the Fowler–Nordheim equation for field emission and by integrating the current density

over computational cells. A detailed analysis of the results reveals the deflected shapes of the

CNTs in an ensemble and the extent to which the initial state of geometry and orientation angles

affect the device current. Experimental results confirm these effects.

(Some figures in this article are in colour only in the electronic version)

1. Introduction

Field emission from CNTs was first reported in 1995 by

three research groups [1–3]. With significant improvement in

processing techniques, applications of CNTs in field emission

devices (e.g., field emission displays, x-ray tube sources,

electron microscopes, cathode-ray lamps, nanolithography

systems etc) have been successfully demonstrated [4–6]. The

cathodes in these devices are better electron field emitters than

Spindt-type emitters and diamond structures [7]. The field

emission performance of a single isolated CNT is found to be

remarkable, and this is due to structural integrity, high thermal

conductivity, chemical stability and geometry of the CNTs.

However, the situation becomes highly complex for cathodes

comprising an ensemble of CNTs, where the individual CNTs

are not always aligned normalto thesubstrate surface. Figure 1

shows an SEM image in which the CNT tips are oriented

in a random manner.This is the most common situation,

which can evolve from an initially ordered state of uniformly

distributed and vertically oriented CNTs. Such an evolution

process must be analyzed accurately from the viewpoint of

long-term performance of the device. The authors’ interests

towards such an analysis and design studies stem from the

problem of precision biomedical x-ray generation.

In this paper, we focus on a diode configuration, where

the cathode contains a CNT thin film grown/deposited on a

0957-4484/08/025701+12$30.00

© 2008 IOP Publishing LtdPrinted in the UK

1

Page 2

Nanotechnology 19 (2008) 025701 N Sinha et al

Figure 1. SEM image showing randomly oriented tips of CNTs in a

thin film.

metallic substrate and the anode is a magnesium–aluminum

alloy plate acting as emission current collector. Here, the most

important requirement is to have a stable field emission current

without compromising on the lifetime of the device. With

the behavior of the CNT films remaining less predictable, the

mechanisms responsible for CNT failures are also not clearly

understood so far. Failure could be due to electromechanical

deformation and stress, residual gases, ballistic transport

induced impulse and partial discharge (PD) between cathode

and anode.Several studies have reported experimental

observations confirming the degradation and failure of CNT

cathodes [8–11]. The degradation of a CNT thin film emitter is

always gradual [11] but with several small-scale fluctuations

and spikes.Such an overall gradual degradation occurs

either during initial current–voltage measurement or during

measurements under constant applied voltage over a long time

duration.

To date, from mathematical modeling and design points

of view, the detailed models and methods of system

characterization are available only for vertically aligned CNTs

grown on a patterned surface [12, 13]. In a CNT thin film, the

array of CNTs may ideally be aligned vertically. However, in

such a film, it is desired that the individual CNTs be evenly

separated, so that their spacing is greater than their height

to minimize the screening effect.

screening effect is minimized in this manner, the emission

properties as well as the lifetimes of the cathodes are adversely

affected due to a significant reduction in the density of CNTs.

Onthe otherhand, for a cathode withrandomlyorientedCNTs,

the field emission current is produced by two types of source:

(i) CNTs that point towards the current collector (anode)

and (ii) oriented CNTs subjected to electromechanical forces

causing reorientation. Therefore, a possible advantage of a

cathode with randomly oriented CNTs is that a large number

of CNTs always take part in the field emission over a longer

time duration [7]. For this reason, a thin film with randomly

oriented multi-walled carbon nanotubes (MWNTs) has been

Unfortunately, when the

considered while formulating the mathematical model. From

the modeling aspect, this becomes a general case, but is much

more challenging compared to the case of a thin film with

uniformly aligned CNTs. Although some preliminary works

have been reported (see e.g., [14–16]), a much more detailed

characterization needs to be done, especially in situations

where the array of CNTs may undergo complicated dynamics

during the process of charge transport.

accountfortherandomlyorientedCNTsandinteractionamong

them, it is necessary to consider the space charge and the

electromechanical forces responsible for their realignment.

Some of the related factors are also important from device

design considerations.

In this paper, the process of evolution, reorientation (self-

assembly of CNTs) and electrodynamic interaction among

the CNTs in the film, and the influence of these processes

on the device output current are analyzed, based on a

multiphysicsmodeldeveloped recently bythe authors(see [16]

for details). Here, first a homogeneous nucleation rate model

is employed by considering a representative volume element

(cell) to idealize the degradation and fragmentation of CNTs,

phenomenologically. The film volume is discretized by several

such cells across the planer dimensions of the film.

cell contains a specified number of CNTs with prescribed

parametric shape distribution and a certain amount of carbon

clusters as the initial description of the state. At a given time,

the evolved concentration of carbon clusters due to the process

of degradation and CNT fragmentation is obtained from the

homogeneous nucleation rate model. This information is then

used in a time-incremental manner to describe the evolved

state of the CNTs in the cells.

induced current density at the anode (cell-wise) is calculated

by using the CNT tip orientation angles and the effective

electric field in the Fowler–Nordheim equation. The diode

current is then computed by integrating the cell-wise current

density over the anode area. In this paper, while carrying out

numerical simulationsbased on the above scheme, we focus on

visualizing various details, e.g., the reorientation of the CNTs

in the ensemble. From a system perspective, such a detailed

study proves to be helpful in understanding the reason behind

the experimentally observed fluctuation in the device current.

To this end, we show experimental results and a quantitative

comparison with simulations, confirming the reorientation and

degradation of the CNTs.

Also, in order to

A

Finally, the field emission

2. Model formulation

2.1. Background

The physics of field emission from metallic surfaces is fairly

well understood. The current density (J) due to field emission

from a metallic surface is usually obtained by using the

Fowler–Nordheim equation [17], which can be expressed as

?

where E is the electric field, ? is the work function for the

cathode material, and B and C are constants. In the CNT

J =BE2

?

exp

−C?3/2

E

?

,

(1)

2

Page 3

Nanotechnology 19 (2008) 025701N Sinha et al

thin film problem, under the influence of sufficiently high

voltage at ultra-high vacuum, the electrons emitted from the

CNTs (mainly from the CNT tip region and emitted parallel

to the axis of the tubes) reach the anode. Unlike metallic

emitters, here, the surface of the cathode is not smooth. The

cathode consists of CNTs (often in curved shapes) with certain

spacings. In addition, a certain amount of impurities and

carbon clusters may be present within the otherwise empty

spaces in the film. Moreover, the CNTs undergo reorientation

due to electromechanical interactions with the neighboring

CNTs during field emission. This makes the determination of

current density more difficult.

2.2. Mathematical model

We employ a homogeneous nucleation model to include the

effect of degradation and CNT fragmentation, phenomenolog-

ically. This nucleation model is then coupled with a model of

electromechanical interaction within the cell-wise ensemble of

CNTs via the carbon cluster concentration. These two steps of

our modeling approach are discussed in the following subsec-

tions.

2.2.1. Nucleation coupled model for degradation of CNTs.

Let NTbe the total number of carbon atoms (in the CNTs and

in the cluster form) in a representative volume element (Vcell)

ofthe thinfilm, N be thenumber ofCNTsinthe cell, and NCNT

be the total number of carbon atoms present in a CNT. Hence,

NT= NNCNT+ Ncluster,

(2)

where Nclusteris the total number of carbon atoms in the cluster

form in a cell at time t, and it is given by

Ncluster= Vcell

?t

0

dn1(t),

(3)

where n1is the concentration of carbon atoms in the cluster

form in the cell.Therefore, by combining equations (2)

and (3), one can obtain the number of CNTs in the cell as

?

In order to determine n1(t), we introduce a homogeneous

nucleation model [18, 19], which is to describe the evolution

in n1(t). Here, we modify the original model (it was proposed

in the context of an aerosol and a chemical growth process)

by assuming the degradation as a reverse process of growth

and model the phenomena of CNT degradation (the nucleation

theory has been used for the growth of CNTs [20] and other

nanoparticles [21, 22] also). Based on this modified model, the

evolution equations are expressed as

N =

1

NCNT

NT− Vcell

?t

0

dn1(t)

?

.

(4)

dNkin

dt

= Jkin,

(5)

dS

dt

dM1

dt

= −JkinSg∗

n1

− (S − 1)B1An

2v1

,

(6)

= Jkind∗

p+ (S − 1)B1Nkin,

(7)

dAn

dt

=JkinSg∗2/3s1

n1

+2πB1S(S − 1)M1

n1

,

(8)

where Nkin is the kinetic normalization constant, Jkin is the

kinetic nucleation rate, S is the cluster saturation ratio, g∗

is the critical cluster size, Anis the total surface area of the

cluster, v1is the monomer volume, M1is the moment of the

cluster size distribution, d∗

the critical cluster, and s1is the surface area of a monomer.

In equations (5)–(8), the various other quantities involved are

given by

pis the diameter of the particle of

S =n1

ns,

M1=

?dmax

p

d∗

p

(n(dp,t)dp)d(dp),

(9)

Nkin=n1

Sexp(?),

B1= 2nsv1

?

?

kT

2πm1,

(10)

Jkin=βijn2

1

12S

?2

?

?

2πexp

?3

? −

4?3

27(ln S)2

?

,

(11)

g∗=

3

?

ln S

,

d∗

p=

4σv1

kT ln S,

(12)

where nsis the equilibrium saturation monomer concentration,

dpis the cluster diameter, dmax

p

is the maximum diameter of the

clusters, n(dp,t) is the particle size distribution function, ? is

the dimensionless surface tension, k is Boltzmann’s constant,

T is the temperature, m1is the mass of monomer, βij is the

collision frequency function for collisions between i-mers and

j-mers, and σ is the surface tension. The collision frequency

function (βij) is given by

?1/6?

βij=

?3v1

4π

6kT

ρp

?1

i+1

j

?

(i1/3+ j1/3)2.

(13)

The dimensionless surface tension (?) is expressed as [18]

? =σs1

kT,

(14)

where ρpis the particle mass density. In this paper, we have

considered i = 1 and j = 1 for numerical simulations; that is,

onlymonomertypeclustersare considered. Our mainintention

in this context is to find n1 using equation (5).

during the growth of clusters, the change in the particle

size (distribution) with time is governed by the saturation

ratio (ratio of atomic concentration to the concentration at

saturation). Therefore, equation (6) is required. In growth

related kinetics, the growth is measured by the moment of

cluster size distribution [23]. M1in equations (7) and (8) tells

us about the distribution of the cluster size over time. That is,

at different locations in the cell, different sizes of cluster can

form, and that also affects how the equilibrium occurs. During

growth of the clusters, a critical size nucleus is a cluster of

size such that its rate of growth is equal to its rate of decay.

The quantity Anindicates the area of the stable carbon cluster.

To know the area of the cluster, An is required. So, based

on the established kinetic theory, we need to include all the

four equations. For instance, in equation (8) for dAn/dt, the

However,

3

Page 4

Nanotechnology 19 (2008) 025701N Sinha et al

first term on the right-hand side of the equation describes an

increase/decrease in the surface area due to a newly formed

stable cluster and the second term means an increase in area

of the existing stable cluster. Equations (5)–(8) form a set

of four nonlinear coupled ordinary differential equations in

n1(t), S(t), M1(t) and An(t). This system of equations is

solved with the help of a finite difference scheme. Finally, the

average numberofCNTsinacell, andhence theaverage height

distributionwithaknownnumberofCNTsintheensemble, are

obtained with the help of equation (4).

2.2.2. EffectsofCNT geometryand orientation.

in section 1, the geometry and the orientation of CNTs play

important roles in the field emission performance of the film

and hence must be considered in the model. In section 2.2.1,

a model of degradation and fragmentation of CNTs has been

formulated. Following this model, if ?h is the decrease in the

length of the CNT over a time interval ?t due to degradation

and if dtis the diameter of the CNT, then the surface area of

the CNT decreases by πdt?h. By using the geometry of the

CNT, the decreased surface area can be expressed as

As discussed

πdt?h = Vcelln1(t)[s(s − a1)(s − a2)(s − a3)]1/2,

where a1, a2, a3are lattice constants, and s =1

The chiral vector for the CNT is expressed as

(15)

2(a1+a2+a3).

− →

Ch= n? a1+ m? a2,

(16)

where n and m are integers (n ? |m| ? 0) and the pair (n,m)

defines the chirality of the CNT. By using the fact that ? a1·? a1=

|a1|2, ? a2· ? a2= |a2|2, and 2? a1· ? a2= |a1|2+ |a2|2− |a3|2, the

circumference and the diameter of the CNT can be expressed

as [24], respectively,

?

dt=|− →

π

|− →

Ch| =

n2a2

1+ m2a2

2+ nm(a2

Ch|

1+ a2

2− a2

3),

.

(17)

By defining the rate of degradation of CNT as vburn

lim?t→0?h/?t, and by dividing both sides of equation (15)

by ?t, one has

=

πdtvburn= Vcelldn1(t)

dt

[s(s − a1)(s − a2)(s − a3)]1/2. (18)

By combining equations (17) and (18), vburncan be simplified

as

?

vburn= Vcelldn1(t)

dt

s(s − a1)(s − a2)(s − a3)

n2a2

1+ m2a2

2+ nm(a2

1+ a2

2− a2

3)

?1/2

(19)

.

Next, we calculate the effective electric field at a given time

step. The electric field at the deflected tip (see figure 2) is

approximated as

?

Ez? =

1 −x2+ y2

R2

(Present height)

(Present gap)

E0,

(20)

Figure 2. CNT array configuration.

where x and y are the deflections of the tip with respect to

its original location, 2R is the spacing between two adjacent

CNTs at the cathode substrate, and E0= V/d with V as the

applied DC voltage. As the CNTs degrade, their height also

decreases with time. On the other hand, the distance between

the tip and the anode increases with time due to reduction

in the height of the CNTs. With the above assumption, the

present height of the CNT can be written as h0− vburnt, and

hence the present distance between the tip and the anode can

be expressed as d − h0+ vburnt. Here, h0is the initial average

height of the CNTs and d is the distance between the cathode

substrate and the anode (see in figure 2). Equation (20) can

now be rewritten as

?

Subsequently, the effective electric field, which is required for

field emission calculation, can be expressed as

Ez? =

1 −x2+ y2

R2

(h0− vburnt)E0

(d − h0+ vburnt).

(21)

Ez= Ez? cosθ(t),

(22)

where θ(t) is the tip orientation angle that a CNT makes

with the Z-axis, as shown in figure 2. The current density

is calculated by substituting the value of Ez in the Fowler–

Nordheim equation. The above formulation takes into account

the effect of CNT tip orientations, and one can perform

a statistical analysis of the device current for randomly

distributed and randomly oriented CNTs. However, due to the

deformation of the CNTs under electromechanical forces, the

evolution process requires a much more detailed treatment. In

order to account for the changing orientations, we estimate the

effects of electromechanical forces as discussed next.

2.2.3.

orientation angle θ is dependent on the electromechanical

forces. Based on the studies reported in the literature, it is

reasonable to expect that a major contribution is by the Lorentz

force due to the flow of electron gas along the CNT and the

ponderomotive force due to electrons in an oscillatory electric

field. In addition, the electrostatic force and the van der Waals

force are also important.

The components of the Lorentz force along the Z- and X-

directions (see figure 2) are approximated as

Electromechanical forces.

For each CNT, the

flz= πdteˆ nE1z,

flx= πdteˆ nE1x≈ 0,

(23)

4

Page 5

Nanotechnology 19 (2008) 025701N Sinha et al

where e is the electronic charge (positive), ˆ n is the surface

electron density and E1z is the fluctuating electric field due

to electron flow in the CNTs. We now decompose the surface

electrondensityintoasteady(unstrained)partandafluctuating

part; that is,

ˆ n = ˆ n0+ ˆ n1,

where the steady part ˆ n0corresponds to the Fermi level energy

in an unstrained CNT, and it can be approximated as [25]

(24)

ˆ n0=

kT

πb2?,

(25)

whereb istheinteratomicdistanceand?istheoverlapintegral

(≈2 eV for carbon). In equation (24), the fluctuating part ˆ n1is

inhomogeneous along the length of the CNTs. Actually, ˆ n1

should be coupled nonlinearly with the deformation and the

electromagnetic field [26]. However, in a simplified form, ˆ n1is

primarily governed by the quantum-hydrodynamic continuity

equation:

˙ˆ n1+ ˆ n0∂ ˙ uz?

with uz? as the longitudinal displacement, and we employ this

simplified equation to compute ˆ n1 at a given time step. In

equation (23), the fluctuating electric field E1 = (E1x, E1z)

is computed using a Green’s function approach as discussed

in [27].

The ponderomotive force, which acts on free charges on

the surface of CNTs under oscillatory high field, tends to

straighten the bent CNTs in the Z-direction. Furthermore,

the ponderomotive forces induced by the applied electric field

stretch every CNT [28]. In order to estimate the components

of the ponderomotive force ( fpx, fpz) acting on the tip region

(see [29]), the following approximations are used:

∂z?= 0,

(26)

fpz≈

q2

2meω2E1z∂E1z

∂z

,

fpx≈ 0,

(27)

where q = (πdteˆ n)ds is the total charge on an elemental

segmentds of a CNT,meisthemass of an electron, ω = 2π/τ,

τ istherelaxationfrequency, and fpzisthe Z-componentofthe

ponderomotive force. The X-component of the ponderomotive

force fpxis assumed to be negligible.

In order to calculate the electrostatic force, interactions

among the neighboring CNTs are considered. Let us assume

two small segments of two neighboring CNTs of lengths ds1

and ds2, respectively. The charges on these two segments are

given by, respectively,

q1= eˆ n(πd(1)

where d(1)

t

(1) and (2), respectively. The electrostatic force on the segment

ds1by the segment ds2is given by

t )ds1,

q2= eˆ n(πd(2)

t )ds2,

(28)

and d(2)

t

are the diameters of two neighboring CNTs

1

4π??0

q1q2

r2

12

,

where ??0denotes the effective permittivity of the aggregate

of CNTs and carbon clusters and r12is the effective distance

between the centroids of the segments ds1and ds2. The force

on the segment ds1of one CNT due to the entire segment (s2)

of the neighboring CNT can be expressed as

?s2

Therefore, the force per unit length on s1due to s2is

1

4π??0

0

1

r2

12

(eˆ nπd(1)

t

ds1eˆ nπd(2)

t )ds2.

fc=

1

4π??0

?s2

0

(πeˆ n)2d(1)

t d(2)

t

r2

12

ds2.

(29)

The differential of the force d fc acts along the line joining

the centroids of the segments ds1 and ds2. Therefore, the

components of the force fc in the X-direction and the Z-

direction are, respectively,

?

1

4π??0

j=1

?

1

4π??0

j=1

Here, φ is the angle that the force vector d?fcmakes with the X-

axis, j is the node number, and ?s2is the length of discretized

segments(assumeduniform inthepresentstudy). Theeffective

distance (r12) between the centroids of the segments ds1and

ds2is obtained as

fcx=

d fccosφ =

1

4π??0

(πeˆ n)2d(1)

?s2

t d(2)

0

(πeˆ n)2d(1)

t d(2)

t

r2

12

cosφ ds2

≡

h0/?s2

?

t

r2

12

cosφ?s2,

(30)

fcz=

d fcsinφ =

1

4π??0

(πeˆ n)2d(1)

?s2

t d(2)

0

(πeˆ n)2d(1)

t d(2)

t

r2

12

sinφ ds2

≡

h0/?s2

?

t

r2

12

sinφ?s2.

(31)

r12= [(d1−lx2+lx1)2+ (lz1−lz2)2]1/2,

where d1is the spacing between the CNTs while in contact

with the surface of the cathode substrate, and lx1and lx2are

the deflections of the segments of two neighboring CNTs

(relative deflection considering the two end nodes of each of

the segments), respectively, which are parallel to the X-axis.

Similarly, lz1and lz2are the deflections of the two segments

which are parallel to the Z-axis.

Next, we consider the van der Waals effect. The van der

Waals force plays important roles, not only in the interaction of

the CNTswiththe substrate, but alsoin the interaction between

the walls of MWNTs and CNT bundles.

type deformation, the cylindrical symmetry of the CNTs is

no longer preserved, leading to axial–radial coupling [30].

For simplicity, let us restrict ourselves to a two-dimensional

(2D) situation described with respect to the (X, Z) coordinate

system shown in figure 2, and let us assume that the lateral

and longitudinal displacements of a CNT are ux? and uz?,

respectively. Due to their large aspect ratio, it is reasonable

to idealize the CNTs as one-dimensional elastic members (as

in an Euler–Bernoulli beam [31]). Therefore, the kinematics

can be expressed as

(32)

Under bending

u(m)

z?

= u(m)

z?0− r(m)∂u(m)

x?

∂z?,

(33)

5

Page 6

Nanotechnology 19 (2008) 025701N Sinha et al

where the superscript (m)indicates the mth wall of the MWNT

with r(m)as its radius and uz?0 as the displacement of the

center of the cylindrical cross-section. Under tension, bending

moment and shear force, the longitudinal displacement of

the cross-section of one wall relative to the cross-section

of its neighboring wall can be expressed with the help of

equation (33) as

?(m)

z?

= u(m+1)

z?

− u(m)

z?

= r(m+1)∂u(m+1)

x?

∂z?

−r(m)∂u(m)

x?

∂z?, (34)

where we assume u(m)

displacementduetopressure inthethinfilm device. The lateral

shear stress τ(m)

vs

due to the van der Waals effect can be written

as

x?

= u(m+1)

x?

= ?x? as the lateral

τ(m)

vs

= Cvs

?(m)

z?

?x?,

(35)

where Cvs is the van der Waals coefficient. Therefore, the

effective shear force per unit length of CNT can be expressed

as

?

⇒ fvs=

m

The components of the van der Waals force parallel to the Z-

and the X-directions are, respectively,

fvs=

m

?2π

0

Cvs

?(m)

z?

?x?

?r(m+1)+r(m)

πCvs[(r(m+1))2− (r(m))2]

2

?

dψ,

?

1

?x?

∂?x?

∂z?.

(36)

fvsz= fvssinθ(t),

fvsx= fvscosθ(t).

(37)

2.2.4. Modeling the reorientation of CNTs.

components acting on the CNTs parallel to the Z- and X-

directions can now be calculated as

?

?

Next, we employ these force components in the expression of

work done on the ensemble of CNTs and formulate an energy

conservation law (see, e.g., [31]).

Under the assumption of small strain and small curvature,

the longitudinal strain εzz(including thermal strain) and stress

σzzcan be written as, respectively,

The net force

fz=

( flz+ fvsz)ds + fcz+ fpz,

(38)

fx=

( flx+ fvsx)ds + fcx+ fpx.

(39)

εzz=∂u(m)

z?0

∂z?

−r(m)∂2u(m)

x?

∂z?2

+ α?T(z?),σzz= E?εzz,

(40)

where E?is the effective modulus of elasticity of CNTs under

consideration, ?T(z?) = T(z?) − T0is the difference between

the absolute temperature (T) during field emission and a

reference temperature (T0), and α is the effective coefficient of

thermal expansion (longitudinal). The longitudinal strain and

stress from equation (40) are employed in the strain energy

expression.Subsequently, the kinetic energy is expressed

in terms of the velocities ˙ ux? and ˙ uz?.

Hamilton’s principle, we obtain the governing equations

Next, by applying

in (ux?, uz?) for each CNT, which can be expressed as,

respectively,

E?A2∂4u(m)

x?

∂z?4

+ ρA0¨ u(m)

1

x? − ρA2∂2¨ u(m)

∂z?cos?θ(z?)?− flx? − fcx? = 0,

2E?A0α∂?T(z?)

∂z?

− πCvs[(r(m+1))2− (r(m))2]

− flz? − fcz? = 0,

where A0is the effective cross-sectional area, A2is the second

moment of cross-sectional area about the Z-axis and ρ is the

mass per unit length of CNT. We assume a fixed boundary

condition (u = 0) at the substrate–CNT interface (z = 0) and

a forced boundary condition at the CNT tip (z = h(t)).

x?

∂z?2

−

?

m

πCvs[(r(m+1))2

− (r(m))2]

∂2u(m)

∂z?2

?x?

−1

∂?x?

(41)

−E?A0

z?0

+ ρA0¨ u(m)

∂?x?

∂z?sin?θ(z?)?

z?0

1

?x?

(42)

2.2.5.

governing equation in temperature T(z?), the process of

heating of CNTs during field emission is idealized as follows.

During field emission, an electron flowing from the cathode

substrate towards the anode along the CNT surface acquires

energy and escapes from the tip region towards the anode

by crossing the Fermi energy.

replacement of the emitted electron is assumed to create an

intense temperature gradient in the emitter [32]. By neglecting

the radiation of heat, the resistive heating of CNTs, together

with the usual laws of heat conduction, leads to the following

differential equation:

Effects of temperature.

In order to arrive at the

The energy lost in the

(h −z?)4∂2T

∂z?2+2(h −z?)3∂T

∂z?− ¯ α2(h −z?)4˙T =¯b˙T2, (43)

where

¯b =

−I2ρc

4π2(4.18κ)(1 − cosv0)2,

where I is the time varying current, ρc is the electrical

resistivity, κ is the thermal conductivity, v0is the interior half

angle of the CNT tip, ¯ ρ is the mass density, and C is the

specific heat (see [32] for details). We assume the temperature

boundary condition, that at z = h, if |θ| ? θcthen T(t) = Tc

else T(t) = T(ti) = T(ti−1), and˙T = 0 for all time; that

is, the temperature at the tip of a CNT has a constant value Tc

when that tip is oriented within a prescribed range [−θc,+θc].

For tip orientations |θ| > θc, the temperature is tracked over

time steps (?t = ti− ti−1) based on the solution of the initial

boundary value problem governed by equations (41)–(43).

Each CNT is discretized into a number of segments with

length ?s. As in the finite element method, each of these

segments has two end nodes where the variables ux?, uz? and T

are evaluated. At each time step, the governingequations (41)–

(43) are solvedand thecurved shape s(x?+ux?,z?+uz?)of each

of the CNTs is updated. The angle of orientation θ between the

¯ α2=

¯ ρC

κ

,

(44)

6

Page 7

Nanotechnology 19 (2008) 025701N Sinha et al

5000

4000

3000

2000

1000

0

0 4080

Time (s)

n1 (m-3)

120160

Figure 3. Variation of the carbon cluster concentration with time.

Initial condition: n1(0) = 100, S(0) = 5000, M1(0) = 2.12 × 10−16,

An(0) = 0.

nodes j +1 and j at the two ends of segment ?sjis expressed

as

?

(zj+1+ uj+1

?

z

where ? is the usual coordinate transformation matrix which

maps the displacements (ux?,uz?) defined in the local (X?, Z?)

coordinate system into the displacements (ux,uz) defined in

the cell coordinate system (X, Z). For this transformation, we

employ the angle θ(t − ?t) obtained at the previous time step

and for each node j = 1,2,3,....

θ(t) = tan−1

(xj+1+ uj+1

x

) − (xj+ uj

) − (zj+ uj

?

x)

z

z)

?

?

,

(45)

uj

uj

x

?

= [?(θ(t − ?t)j)]

uj

uj

x?

z?

,

(46)

3. Results and discussions

The computational scheme involves three parts: (i) discretiza-

tion of the nucleation coupled model for degradation of CNTs,

(ii) incremental update of the CNT geometry using the elec-

tromechanical forces, and (iii) calculation of the field emission

current. At t = 0, the CNTs are assumed to be oriented ran-

domly. This random distribution is parameterized in terms of

the upper bound of the CNT tip deflection which is given by

?xmax = h/m?; m?? 1 is an integer. The angles are used

to calculate the component of electromechanical forces. In the

successive time step t ← t + ?t, the deformed state of the

CNTs is computed and the tip angle of each CNT is updated.

The CNT film considered for a representative numerical

simulation here consists of randomly oriented MWNTs. The

film has a surface area (projected on the anode) of 49.93 mm2

and average thickness of 10–14 μm. Actual experiments were

performed on various films of both types, random as well as

aligned CNTs. The films were placed in a pressure-control

chamber fitted with a micrometer to control the electrode gap,

and by connecting the diode to a resistive–capacitive circuit.

Current histories were measured under a constant DC voltage

for a controlled gap. In the simulation and analysis reported

next, we have chosen the range of DC voltage and parameters

Figure 4. CNT tip distribution showing the area coverage under the

anode. The area shown is approximately 5.08 μm × 5.08 μm.

Figure 5. Dark spots on the anode after field emission showing the

traces of clusters ejected from the CNT tips. Largest spot diameter

≈0.7 mm.

related to the CNT distribution in the film as in the actual

experimental set-up.

3.1. Degradation of CNT thin films

We have assumed that at t

minimal amount of carbon clusters.

over time (due to both fragmentation and self-assembly) and

the carbon cluster concentration in each cell also changes

(mostly increases) accordingly. Based on this assumption, the

initial conditions were set as n1(0) = 100, S(0) = 5000,

M1(0) = 2.12 × 10−16, An(0) = 0, and T

303 K. Figure 3 shows the n1(t) history over a small time

interval of 160 s. Such evolution indicates that the carbon

cluster concentration increases exponentially at the beginning

and tends to a steady state afterwards at ≈93 s, indicating

a saturation. Therefore, figure 3 indicates the time that the

degradation process takes to reach saturation (S(0) = 5000)

in the present analysis. Assuming that saturation has taken

place within the computational cell, and assuming that this

is the case for all the cells encompassing the entire volume

between the cathode substrate and the anode, one can estimate

Ncluster= 5000×(0.04993)2×14×10−6. Here, the film area

= 0, the diode contains a

The CNTs degrade

= T0 =

7

Page 8

Nanotechnology 19 (2008) 025701N Sinha et al

(a)

(b)

Figure 6. EDS results showing the exposed substrate surface made

of Cr and Cu after the field emission. The energy spectrum shown in

(b) corresponds to the region marked in (a).

is 49.93 mm2(as stated earlier) and the cathode to anode gap

is assumed to be 14 μm (a less conservative estimate since the

free space above the CNT film is excluded). From figure 4,

we obtain the effective area projected on the anode for the

degradation process to be approximately 10 × 10−4times the

totalfilm area. Multiplyingthisfactor to the number of clusters

above, we estimate Ncluster ≈ 3.495 × 10−10. The clusters

assumed to be formed in this way are partly deposited on the

cathode and partly pulled up ballistically and deposited on the

anode. Results are discussed next.

Our experiments have revealed that degradation of CNTs

into clusters is indeed one of the important processes that takes

place during field emission. Interestingly, dark spots on the

originally polished surface of the anode were found after 10 h

of field emission from a vertically aligned CNTs. The spots

were visible by the naked eye. Figure 5 shows an optical

image of these spots. The spot diameters are found to be in

the range 0.4–0.7 mm. The thin film substrate is made of

CuCr alloy. In order to find out if there is any degradation of

CNTs with any trace of exposed metal substrate and any trace

of carbon clusters on the anode, we perform energy-dispersive

x-ray spectroscopy (EDS) on the film and on the anode surface.

The EDS results in figure 6 shown for one of the sample

substrate after 10 h of field emission indicates that the carbon

clusters and the CNTs at various places were pulled up and the

substrate was exposed. Along with Ni impurities, which are

present due to catalytic processing of the CNTs, the spectral

peaks of Cr can be clearly seen in figure 6. On the other

hand, tracesofcarbonclusterareobservedontheanodesurface

(a)

(b)

Figure 7. EDS results showing the deposited carbon clusters one the

anode after field emission. The energy spectrum shown in (b)

corresponds to the region marked in (a). Optical images of the spots

are shown in figure 5.

Table 1. Trace of particles and their concentrations obtained from

EDS measurements.

Emitter regionElement Wt%At.%

Cathode C K

O K

CrK

NiK

CuK

C K

O K

MgK

AlK

ClK

87.20

0.84

11.96

0.00

0.00

32.47

29.11

1.11

31.14

6.18

96.25

0.70

3.05

0.00

0.00

45.84

30.86

0.77

19.57

2.96

Anode

(as shown in figure 5).

surface contains carbon, as indicated by the spectral peaks

in figure 7. We quantify the possible nature of degradation

process as follows. With the area of the anode exposed to

x-rays corresponding to an ≈100 μm effective diameter spot

(assumed to be smaller than the actual diameter observed,

which is diffused) containing 45.04 at.% amorphous carbon

(see table 1), we get Ncluster≈ 3.353 × 10−9. Comparing this

estimate with the previously simulated value of 3.495 × 10−10

from the degradation model, it is reasonable to assume that the

reorientation of the CNT tips actually activates or deactivates

the degradation process differently in different cells and at

different times. As a result, one may observe a much slower

EDS results reveal that the anode

8

Page 9

Nanotechnology 19 (2008) 025701N Sinha et al

Figure 8. Simulated current histories for various uniform heights,

uniform radius and uniform spacing of CNTs under a DC voltage of

600 V. The current I is in ampere units.

Figure 9. Simulated current histories for various nonuniform

heights, uniform radius and uniform spacing of CNTs under a DC

voltage of 600 V. The current I is in ampere units.

degradation in the ensemble. However, at present, it is not

possible to observe the occurrence of the spots on the anode

in a time-resolved manner.

The results analyzed above establish the fact that the

degradation of CNTs is an important process. The influence of

such degradation, along with the electromechanical forces, is

expected to play a significant role on the device current. Next,

we report the experimental results of field emission current

fluctuation and analyze such behavior based on the proposed

model and by comparing the simulationswith the experimental

data.

3.2. Device current

Here, we analyze the field emission current histories for

V = 600 V. The constants B and C in equation (1) were

taken as B = (1.4 × 10−6) × exp((9.8929) × ?−1/2) and

C = 6.5 × 107, respectively [33]. It has been reported in

the literature (see, e.g., [33]) that the work function ? for

CNTs is smaller than the work functions for metal, silicon

and graphite. However, depending on the types of CNTs

(i.e., SWNT/MWNT), there are significant variations in the

experimentally estimated values of ?, geometric parameters

and also the type of substrate materials having significant

influence on the electronic band-edge potential.

simulations have been carried out based on ? = 2.2 eV.

The following two cases are considered for numerical

simulations: (i) average height between neighboring CNTs in

the film sample h0 = 12 μm, uniform radius r = 2.75 nm

and uniform spacing d1 = 2 μm, and (ii) a random initial

height distribution: h = (h0± 2 μm) ∓ 2 μm × rand(1),

uniform radius and uniform spacing.

rand denotes random number generator. The electrode gap

(d) is maintained at 34.7 μm. Several computational runs

are performed and the output data are averaged out at each

sampling time step. The current histories for cases (i) and (ii)

are shown in figures 8, and 9, respectively. The four curves

in figures 8 and 9 correspond to four different initial states

with increasing values: ?xmax= h0/100, h0/50, h0/25, and

h0/20. Interestingly, both these figures tell us the following:

(1) as the initial state of deflections/angle increases (from

h0/100 to h/50 and higher), the average current reduces, until

h0/m?becomes large enough that during field emission, the

electrodynamic interaction among CNTs produces a sudden

Here,

Here the function

(a)(b)

Figure 10. (a) Array of 100 CNTs (nonuniform height distribution with maximum initial tip deflections of h0/20); (b) enlarged view of the

region near the pulled up CNT (marked as ‘A’ and deflected horizontally ≈2 μm). The initial shape is shown by the dashed line and the

deflected shapes at t = 100 s are shown by solid lines.

9

Page 10

Nanotechnology 19 (2008) 025701N Sinha et al

Figure 11. Vertically aligned CNTs before field emission.

Figure 12. Small reorientations and electrodynamic force induced

stretching of CNTs before long duration exposure.

pull in the deflected tips towards the anode (see the CNT tip

marked as ‘A’ in figure 10 and deflected horizontally ≈2 μm)

resultingin transients (sometimescurrent >mA, as can be seen

from the h0/20 curve); (2) the trends in the current fluctuation

for higher h0/m?in figure 8 compared to figure 9 prove that

the nonuniform distributionof height(h) gives better stabilized

current for large random orientations. Also, in this case of

nonuniform height distribution in figure 9, as one goes from

h0/25 to h0/20 curves, the trend in the average current level is

more uniform than in the uniform height case (see the h0/20

curve in figure 8, which rises >100 mA at the very beginning,

which is undesired in x-ray device applications).

other hand, the above results also support the earlier reported

experimental observation that a thin film with randomly

oriented CNTs produces larger current density compared to

a thin film with uniformly aligned CNTs. The above results

alsojustifyseveralexperimentalobservations,whichhavebeen

reported in the literature from a comparative perspective of

using a single CNT and then different types of CNT in thin

films and the effect of their alignments [4, 8, 34–36]. For

example, Collins and Zettl [34] attribute the change in currents

to interactions between neighboring CNTs.

Inthefieldemissionexperiments, wehaveexposedseveral

experimental samples having randomly oriented as well as

vertically aligned CNTs. Figure 11 shows the SEM image of

one such sample having vertically aligned CNTs. In figure 12,

On the

Figure 13. Reoriented CNTs after 10 h of field emission.

Figure 14. Fluctuation of field emission current from a baked sample

having vertically aligned CNTs.

the SEM image of the sample after a short period of exposure

is shown. Figure 13 shows the SEM of the same sample as in

figure 11 but after 10 h of field emission at 600 V. As seen

in the simulated responses in figures 8 and 9, the spikes in

the current are significant. This experimental results confirm

a strong electromechanical interaction among the CNTs along

with certain amount of degradations. The results shown in

figure 14 are for an electrode gap of 150 μm. Occasional

shortings of the electrical circuit were observed for smaller

gaps between the electrodes. Such a short circuit condition is

the result of pull up of one or more CNTs under the increased

electrodynamic force.

Most interesting is figure 13 compared to figures 12

and 11.While figure 12 depicts small reorientations and

electromechanical force induced stretching of CNTs for a

short-term exposure, figure 13 clearly shows a combined effect

of reorientations and degradation of a few CNTs (pulling

up and exposing the substrate, as discussed earlier) after

10 h of field emission. With these experimental results,

the proposed modeling framework and the simulated results

regarding CNT tip deflections in figure 10 proves to be an

important insight towards understanding the nature of the

10

Page 11

Nanotechnology 19 (2008) 025701N Sinha et al

field emission response. The results in figure 10 are also in

agreement with a previously reported study [37]. Based on the

simulations performed, the stress on the CNTs under a voltage

of 700 V was estimated using equation (40). The maximum

stress was found to be less than 20 GPa, which is generally

much below the stress at fracture (65–93 GPa) [38].

4. Concluding remarks

In this paper, we have developed a modeling approach to

analyze the evolution of CNT thin films and the effect of

such evolution on the device current. A computational scheme

has been implemented, which combines a homogeneous

nucleation model of degradation and CNT fragmentation with

successive electromechanical forcing. Numerical simulations

based on this model and the computational scheme are able

to capture the transients in the field emission current as

observed in the actual experiments. Our analysis shows how

the electromechanical interaction among the CNTs influences

the orientation of individual CNTs in an ensemble, and

consequently, how the distribution of CNTs and geometrical

factors affect thedevice current. Withthisnewfinding reported

in this paper, there are two important aspects which require

further attention: one is the physics of the CNT fragmentation

process under high field to explain the long time response of

the device, and the other is the process of electron gas flow and

ballistic transport under realistic geometric conditions.

References

[1] Rinzler A G, Hafner J H, Nikolaev P, Lou L, Kim S G,

Tomanek D, Colbert D and Smalley R E 1995 Unraveling

nanotubes: field emission from an atomic wire Science

269 1550–3

[2] de Heer W A, Chatelain A and Ugrate D 1995 A carbon

nanotube field-emission electron source Science

270 1179–80

[3] Chernozatonskii L A, Gulyaev Y V, Kosakovskaya Z Y,

Sinitsyn N I, Torgashov G V, Zakharchenko Y F,

Fedorov E A and Valchuk V P 1995 Electron field emission

from nanofilament carbon films Chem. Phys. Lett. 233 63–8

[4] Bonard J M, Salvetat J P, Stockli T, Forro L and

Chatelain A 1999 Field emission from carbon nanotubes:

perspectives for applications and clues to the emission

mechanism Appl. Phys. A 69 245–54

[5] Saito Y and Uemura S 2000 Field emission from carbon

nanotubes and its application to electron sources Carbon

38 169–82

[6] Sugie H, Tanemure M, Filip V, Iwata K, Takahashi K and

Okuyama F 2001 Carbon nanotubes as electron source in an

x-ray tube Appl. Phys. Lett. 78 2578–80

[7] Cheng Y and Zhou O 2003 Electron field emission from carbon

nanotubes C. R. Physique 4 1021–33

[8] Bonard J M, Maier F, Stockli T, Chatelain A, de Heer W A,

Salvetat J P and Forro L 1998 Field emission properties of

multiwalled carbon nanotubes Ultramicroscopy 73 7–15

[9] Avouris P, Martel R, Ikeda H, Hersam M, Shea H R and

Rochefort A 2000 Electrical properties of carbon nanotubes:

spectroscopy, localization and electrical breakdown Funda.

Mat. Res. Series: Sci. Appl. Nanotubes ed D Tomanek and

R J Enbody (New York: Kluwer Academic/Plenum

Publishers) pp 223–37

[10] Nilsson L, Groening O, Groening P and Schlapbach L 2001

Collective emission degradation behavior of carbon

nanotube thin-film electron emitters Appl. Phys. Lett.

79 1036–8

[11] Bonard J M, Klinke C, Dean K A and Coll B F 2003

Degradation and failure of carbon nanotube field emitters

Phys. Rev. B 67 115406

[12] Nicolaescu D, Filip L D, Kanemaru S and Itoh J 2004

Modeling of optimized field emission nanotriodes with

aligned carbon nanotubes of variable heights Japan. J. Appl.

Phys. 43 485–91

[13] Nicolaescu D, Filip V, Kanemaru S and Itoh J 2003 Modeling

of field emission nanotriodes with carbon nanotube emitters

J. Vac. Sci. Technol. 21 366–74

[14] Sinha N, Roy Mahapatra D, Yeow J T W, Melnik R V N and

Jaffray D A 2006 Characterization of self-assembly and

evolution in carbon nanotube thin film field emitter Proc. 6th

IEEE Conf. Nanotech. (Cincinnati, July 2006)

[15] Sinha N, Roy Mahapatra D, Yeow J T W, Melnik R V N and

Jaffray D A 2006 A coupled model of field emission from

carbon nanotube thin films Proc. 7th World Cong. Comp.

Mech. (Los Angeles, July 2006)

[16] Sinha N, Roy Mahapatra D, Yeow J T W, Melnik R V N and

Jaffray D A 2007 Carbon nanotube thin film field emitting

diode: understanding the system response based on

multiphysics modelling J. Comput. Theor. Nanosci. 4

535–49

[17] Fowler R H and Nordheim L 1928 Electron emission in intense

electric field Proc. R. Soc. Lond. A 119 173–81

[18] Friedlander S K 1983 Dynamics of aerosol formation by

chemical reactions Ann. New York Acad. Sci. 404 354–64

[19] Grishick S L, Chiu C P and McMurry P H 1990

Time-dependent aerosol models and homogeneous

nucleation rates Aerosol Sci. Technol. 13 465–77

[20] Watanabe T, Notoya T, Ishigaki T, Kuwano H, Tanaka H and

Moriyoshi Y 2006 Growth mechanism for carbon nanotubes

in a plasma evaporation process Thin Solid Films

506/507 263–7

[21] Watanabe T and Fujiwara K 2004 Nucleation and growth of

oxide nanoparticles prepared by induction thermal plasmas

Chem. Eng. Comm. 191 1343–61

[22] Watanabe T and Okumiya H 2004 Formation mechanism of

silicide nanoparticles by induction thermal plasmas Sci.

Technol. Adv. Mater. 5 639–46

[23] Farrell J E and Valls O T 1991 Growth kinetics and domain

morphology after off-critical quenches in a two-dimensional

fluid model Phys. Rev. B 43 630–40

[24] Jiang H, Zhang P, Liu B, Huang Y, Geubelle P H, Gao H and

Hwang K C 2003 The effect of nanotube radius on the

constitutive model for carbon nanotubes Comput. Mater. Sci.

28 429–42

[25] Slepyan G Y, Maksimenko S A, Lakhtakia A,

Yevtushenko O and Gusakov A V 1999 Electrodynamics of

carbon nanotubes: dynamic conductivity, impedance

boundary conditions, and surface wave propagation Phys.

Rev. B 60 17136–49

[26] Wei L and Wang Y-N 2004 Electromagnetic wave propagation

in single-wall carbon nanotubes Phys. Lett. A 333 303–9

[27] Svizhenko A, Anantram M P and Govindan T R 2005 Ballistic

transport and electrostatics in metallic carbon nanotubes

IEEE Trans. Nanotechnol. 4 557–62

[28] Glukhova O E, Zhbanov A I, Torgashov I G, Sinistyn N I and

Torgashov G V 2003 Ponderomotive forces effect on the

field emission of carbon nanotube films Appl. Surf. Sci.

215 149–59

[29] Musatov A L, Kiselev N A, Zakharov D N, Kukovitskii E F,

Zhbanov A I, Izrael’yants K R and Chirkova E G 2001 Field

electron emission from nanotube carbon layers grown by

CVD process Appl. Surf. Sci. 183 111–9

[30] Ruoff R S, Tersoff J, Lorents D C, Subramoney S and

Chan B 1993 Radial deformation of carbon nanotubes by

van der Waals forces Nature 364 514–6

11

Page 12

Nanotechnology 19 (2008) 025701 N Sinha et al

[31] Reddy J N 2002 Energy Principles and Variational Methods in

Applied Mechanics (New Jersey: Wiley)

[32] Dolan W W, Dyke W P and Trolan J K 1953 The field emission

initiated vacuum arc II. the resistively heated emitter Phys.

Rev. 91 1054–7

[33] Huang Z P, Tu Y, Carnahan D L and Ren Z F 2004 Field

emission of carbon nanotubes Encyl. Nanosci. Nanotechnol.

vol 4, ed H S Nalwa (Stevenson Ranch: American Scientific

Publishers) pp 401–16

[34] Collins P G and Zettl A 1997 Unique characteristics of cold

cathode carbon nanotube matrix field emitters Phys. Rev. B

55 9391–9

[35] Chen Y, Shaw D T and Guo L 2000 Field emission of different

oriented carbon nanotubes Appl. Phys. Lett. 76 2469–71

[36] Kuttel O M, Groening O, Emmeneger C and

Schlapbach L 1998 Electron field emission from phase pure

nanotube films grown in a methane/hydrogen plasma Appl.

Phys. Lett. 73 2113–5

[37] Zhbanov A I and Glukhova O E 2001 Lengthening of

nanotubes in strong electrical fields Proc. 14th Int.

Microelect. Conf. pp 63–4

[38] Belytschko T, Xiao S P, Schatz G C and Ruoff R S 2002

Atomistic simulations of nanotube fracture Phys. Rev. B

65 235430

12