Carbon Nanotube

C A R B O N 9 1 ( 2 0 1 5 ) 3 0 4 – 3 1 0

.sc iencedi rect .com

ScienceDirect

journal homepage: www.elsev ier .com/ locate /carbon

A 46-inch diagonal carbon nanotube field emissionbacklight for liquid crystal display

http://dx.doi.org/10.1016/j.carbon.2015.04.0930008-6223/� 2015 Elsevier Ltd. All rights reserved.

* Corresponding authors.E-mail addresses: [email protected] (Y.C. Kim), [email protected] (I.T. Han).

Yong Churl Kim *, Shang Hyeon Park, Chang Soo Lee, Tae Won Chung, Eunseog Cho,Deuk Seok Chung, In Taek Han *

Samsung Advanced Institute of Technology, Samsung Materials Research Complex, Youngtong, Suwon, Kyunggi 443-803, Republic of Korea

A R T I C L E I N F O

Article history:

Received 18 February 2015

Accepted 28 April 2015

Available online 8 May 2015

A B S T R A C T

A large area carbon nanotube field emission backlight that is built with a new cathode

structure is reported. The approach involves unique gate insulator formation by glass etch-

ing, highly populated multi wall nanotube tips, and gate electrode assembly by anodic

bonding. Impressive lamping performances are noted through dynamic control of the built

in fine 4320 local dimming blocks as small as 1 cm2. The liquid crystal display lit by the

backlight demonstrates superior luminance uniformity 97%, and a native contrast ratio

400,000:1, whilst showing less than 10% decay in emission current of the sealed panel dur-

ing continuous operation over 8000 h.

� 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Exploiting carbon nanotube (CNT) field emitters in cathodolu-

minescent devices has widely been demonstrated [1–7], such

as field emission display, backlight unit (BLU) for liquid crystal

display (LCD), and household lamp. Among the numerous

applications of CNTs, these luminescent devices can be the

most promising because the commercial market is huge,

and the impact is supposed to be very direct to the consumers

to experience nanotechnology. CNTs provide the most ideal

geometry for field emission, known as great aspect ratio over

1000. In addition, they possess superior chemical stability

thanks to strong covalent bonds, leading to more robust and

resistant to electromigration under strong electric field than

traditional metallic structures. Though CNT emitters have

shown their promising properties including extraction of high

currents (a few mA per emitter) under such an ideal condition

as ultra high vacuum (UHV, <10�8 Torr), the details of time-

dependent emission behaviors in the sealed large area

devices are strongly affected by emitter distribution and

evolution of the internal vacuum condition. Under an

extreme condition (10�10 Torr), CNTs can emit electrons sta-

bly at even high temperature heated by Joule heating, up to

1600 K. This high temperature stability is sustainable by

self-surface cleaning through desorption process [8].

However, such a condition, meaning monolayer formation

time �104 s at 10�10 Torr, is hardly attainable in the normal

consumer electronic devices, where getter materials are used

for in-situ pumping. Practically, low 10�5–10�6 Torr of vacuum

is readily available in the getter installed devices. Moreover,

sustaining vacuum condition becomes harder in large-area

flat panels due to its limited conductance. In an ensemble

of CNT arrays that appears in usual cathode structures, the

field enhancement factor (b) of the emitters depends on the

spacing between neighboring emitters (d) due to the field

screening effect. Nilsson et al. [9] showed that the overall b

can be maximized at d P 2l, where l is emitter length.

Assuming all of the prepared emitters having equal height

l = 1 lm, the attainable emitter density with minimum field

screening is estimated less than 2.5 · 107 cm�2. However, in

C A R B O N 9 1 ( 2 0 1 5 ) 3 0 4 – 3 1 0 305

the sealed devices under moderate vacuum condition with a

range of tip height distribution, preparation of the highly pop-

ulated emitter tips (P109 cm�2) at a little sacrifice of b is desir-

able to ensure the lifetime stability of the emitters via

minimizing average current load (average Joule heating) on

the field emitting CNTs.

Meanwhile, a ‘‘local dimming’’ technology is implemented

in the current LCDs lit by an array of light emitting diode (LED)

arranged at the edge of the frame or at the direct backplane of

the panel. The local dimming can dim the desired area of the

screen, while keeping the bright parts of the screen bright to

improve contrast ratio, in which however, only a broad areal

local dimming is available. The edge-lit LEDs supply dimming

units as large as a 100 cm2 through horizontal or vertical

direction of the panel, while even in the direct back-lit LEDs,

individual control of the equipped LEDs are not technically

probable, but they do zone dimming in an area as large as a

few tens of cm2. Therefore, if a zone is lit during the rest

zones are not lit, resulting in a bloom (halo) as that part of

the image becomes unnaturally brighter than its neighboring

zones (known as blooming artefact). By contrast, LCD lit by

CNTBLU suggested a possible fine local dimming block as

small as 1 cm2 by cost-effect design with a simple matrix

operation, providing high contrast and fast response time

[10].

What is missing till now in the CNTBLU is a cost-effective

architecture in the cathode expendable to large substrates

with good reliability in the emission current and brightness

uniformity. Here, we detail the fabrication and optimization

of our 46 inch diagonal CNTBLU based on a new cost-

effective cathode frame and its reliability for over 8000 h.

2. Experimental

The fabrication flow of our cathode structure and unique fea-

tures of the CNTBLU are illustrated in Fig. 1, starting with

forming groove lines (60 lm in depth and 765 lm in pitch)

Fig. 1 – Process flow schematic of a CNT cathode (left), and an il

can be viewed online.)

through spray etching on a soda-lime glass (1.8 mm t, and

1088 · 653 mm2 in dimension) patterned with thick printable

photo-resist. In the groove lines, a series of screen printing

and firing was followed successively to build cathode lines

by stacking triple layers of barrier (glass paste, Bi2O3–B2O3–

SiO2)/cathode electrode (Ag paste)/emitter (CNT paste).

Corresponding firing temperatures in air ambient were

550 �C, 500 �C, and 450 �C, respectively.

A scanning electron microscope (SEM) image of the pre-

pared cathode line is shown in Fig. 2a. To populate CNT emit-

ters, we activate the printed surface via applying an adhesive

tape with an optimized glue thickness of 60 lm. The activated

CNT emitters showed very uniform, highly dense

(�5 · 109 cm�2), and vertically aligned morphology as shown

in Fig. 2b. The preparation detail of our optimized CNT paste

is presented in the Supporting Information. Such formed

cathode lines were set to have maximum �25 lm in height

including CNT emitters. The barrier layer was introduced to

protect emitters from a damage during the following anodic

bonding [11] of a mesh grid (aluminum, 100 lm t) on the sub-

strate. In the course of bonding process, we found that the

CNT tips were seriously damaged when formed on the origi-

nal double-layered Ag/CNT structure, which was supposed

to originate from the injection of energetic O�2 into the CNT

layer. Briefly, soda-lime glasses contain mobile ions such as

Na+ and O�2 that drift under an appropriate condition (strong

electric field and elevated temperature) by hoping conduction

[12]. We heated the glass substrate up to 250 �C so as to

shorten process time. While the bottom of substrate was cov-

ered with a replaceable aluminum film to form an electrode

biased to the ground, the mesh grid was connected to a power

source (1 kV/3 A). At the initial bonding stage running for a

few seconds, constant ionic currents (3 A) flowed through

the system with a built-in potential �500 V, which was obvi-

ously developed in the glass itself. After elapse of this short

period, the potential increased up to over 950 V in about

100 s with decrease in the ionic currents below 1 A. The extra

lustration of a CNTBLU (right). (A color version of this figure

Fig. 2 – SEM images showing a cathode line in the groove (a), a cross section of the cathode with highly populated CNTs (b),

and a mesh grid mounted on the substrate (c).

306 C A R B O N 9 1 ( 2 0 1 5 ) 3 0 4 – 3 1 0

450 V was developed as aluminum oxide formed in the inter-

face of the glass and grid. A typical SEM image of the grid

installed on a cathode is presented in Fig. 2c.

The mesh grid consisted eighty strips for the gate elec-

trodes with dimensions of 573 mm in length and 12.7 mm in

pitch. Simultaneously, fifty-four addressing electrodes were

prepared through horizontal direction of the panel containing

thirteen cathode (Ag) lines for each, such that our CNTBLU

built 4320 (80 · 54) local dimming blocks (�1 cm2 each) to

operate impulsively with frequency 120 Hz. The brightness

of each dimming block could independently represent 256

grayscales (8 bits) by pulse-width modulation, synchronized

with the image operation in the LCD. In the preparation of

anode plate, P22 phosphors; red (Y2O3:Eu), green (ZnS:Cu,

Al), and blue (ZnS:Cu, Al) were paste-mixed with an appropri-

ate proportion to produce white color (color temperature

40,000 K) matching with the current color standard of the

LCD-TV. After passing through the color filters in the current

LCDs, it was shifted to �10,000 K. After printing and firing

phosphors on ITO coated soda-lime glasses, thin aluminum

films (t = 150 nm) were laminated to improve luminance by

light reflection. The cathode and anode plates were then her-

metically sealed with posting spacers (0.8 cm in height and

2 cm in spacing). Finally the inside of the panel was pumped

down to 1 · 10�7 Torr through an exhausting tube connected

to a turbo-molecular pump system. Baking (350 �C for an

hour), and successive aging were carried to eliminate

loosely-bound gaseous species on the internal components

such as emitters, grid, and phosphors. The aging process

was conducted under the conditions of anode voltage (Va)

10 kV and anode current (Ia) 25 mA for two hours. As soon

as the aging started, the chamber pressure increased rapidly

to two orders of higher, then slowly dropped and stabilized

to the base pressure. Sealed panels after tipping-off and getter

flushing showed initial vacuum level �1 · 10�6 Torr, and sta-

bly sustained during operation, measured using an SRG-3

spinning rotor gauge (MKS instrument, USA). To investigate

the surface morphologies and microstructures of the materi-

als, a transmission electron microscope (TEM, H-9000NA,

Hitachi) with an acceleration voltage of 300 kV was used. A

micro-Raman spectrometer (633 nm Ar + laser, Renishaw)

was chosen to examine CNT raw materials.

3. Results and discussion

Highly crystallized multi-wall CNT (MWCNT) produced by arc

discharge process (JFE Engineering Corp., Japan) was selected

for the formulation of emitter paste (see Supporting

Information), having ten to twelve graphitic walls (Fig. 3a).

A very low integrated D (disordered, 1351 cm�1) to G (graphi-

tic, 1588 cm�1) intensity ratio (ID/IG) �0.33 in the Raman scat-

tering data (Fig. 3b) was observed from the raw material

named MW3, which is unusually lower value than the other

comparative MWCNTs (MW1 and MW2) available in the mar-

ket (ID/IG = 0.7–0.9, spectra not shown here). Also much supe-

rior field emission characteristics (I�V, Fig. 3c) were observed

in the highly crystalline MWCNTs. When compared with

available other nanotubes including single, double, and few

(3–5) wall tubes, the chosen MWCNTs provided not only better

field emission performances with low enough turn-on voltage

with no extra purification processes, but also even better tol-

erance against possible damages by tape-activation, mesh-

bonding and high temperature (up to 480 �C) manufacturing

steps, leading to promising reliability in the sealed panels.

Here, I�V measurement for the screening of raw materials

was conducted in a vacuum chamber (1 · 10�6 Torr) under a

diode configuration with a gap 500 lm between the anode

and cathode.

As shortly introduced in the previous section, the anodic

bonding process caused serious damage on the prepared

CNT emitters. Fig. 4a compares Ia�Vg behaviors between

two cathode groups with and without barrier layer, respec-

tively. Also corresponding SEM images as presented in

Fig. 4b clearly show a distribution of CNT emitters mostly

shortened in height during the anodic bonding in the cathode

without barrier layer. Another modification should be noted

in the cathode was based on two unexpected emission behav-

iors. One was temperature dependent Ia�Vg behavior of the

Fig. 3 – (a) TEM images of the MWCNTused in this study. (b) A Raman spectrum showing high degree of crystallization of the

MWCNT (MW3). (c) Comparison of field emission behaviors under diode mode among three different MWCNT pastes. Here,

I�V measurements for each CNT are doubly checked. Inset: a Fowler–Nordheim plot for the I�V of MW3. (A color version of

this figure can be viewed online.)

Fig. 4 – (a) Comparison of the field emission behaviors

between two cathode groups processed with barrier or not.

(b) SEM images showing degradation of the emitters in the

anodic bonding process. (A color version of this figure can be

viewed online.)

Fig. 5 – (a) Strong Ia�Vg dependency on ambient

temperatures of a CNTBLU for which the BLU panel was

operated in a thermostat container. (b) Lamping behaviors

in terms of elapsed time after switching-on in a CNTBLU. (A

color version of this figure can be viewed online.)

C A R B O N 9 1 ( 2 0 1 5 ) 3 0 4 – 3 1 0 307

BLU panel (Fig. 5a). The Ia was increased monotonically from

23.5 mA up to 36.2 mA at the same applied Vg = 135 V with

increase of the panel temperature from 20 �C up to 60 �C, vice

versa. Here, the test was performed in a thermoset container.

The other curious phenomenon was a large delay in the

lamping (Fig. 5b). Delay of tens to hundreds of seconds in

turning-on and -off the backlight was observed irrespective

308 C A R B O N 9 1 ( 2 0 1 5 ) 3 0 4 – 3 1 0

of the tested ambient temperature in the range of 20–60 �C. To

the best knowledge of the authors, neither of these two phe-

nomena have been reported nor expected in the field emis-

sion materials. In principle, field emission is immune to

ambient fluctuations in the wide range of temperature

(�269–450 �C) as originally proposed by Ken Shoulders [13],

and cathodoluminescent phosphors possess quite fast decay

behavior less than a few milliseconds [14]. Among controver-

sies, we speculated that these two eccentric symptoms were

strongly coupled with each other, and probably caused by

charge accumulation on the insulating surfaces exposed near

emitter patterns. In case of surface charging on the insulator

(barrier and/or glass in this case), it can distort normal field

emission behaviors, and even worse induce a delay in switch-

ing induced by slow charge accumulation on and flushing

from insulator surfaces.

To support our conjecture, electric field distributions in the

original and modified cathode structures were simulated

Fig. 6 – (a) Strong electric field developed on the edges of the

old cathode that is fully covered with CNT layer (upper left),

and its magnified emission patterns on the anode (upper

right). Electric field lines developed in the modified cathode

with expanded silver electrode where strong fields are

concentrated on the edges of the silver pattern (lower left),

and its emission patterns on the anode (lower right). Here,

the representation of the vertical direction in the modeling

is somewhat exaggerated on purpose for the clear visibility.

Both of the luminescent images inserted in (a) were taken at

Va = 8 kV and Ia = 25 mA. (b) Invariant Ia�Vg behaviors

against ambient temperatures from the modified cathode

with extended silver electrodes. (A color version of this

figure can be viewed online.)

using a commercial software (OPERA 3D) as shown in

Fig. 6a. A strong electric field was developed on the edges of

the CNT film in the original structure, leading to a possible

charge accumulation mainly on the barrier surface by the

emitted electrons from those CNTs populated even at the pat-

tern edges (see Fig. 2b). Meanwhile, the electric field was

rather concentrated on the edges of the extended cathode

electrode (Ag) in the modified structure, where electron emis-

sion was not expected. Based on this model, we simply

extended the width of the silver layer so as not to be fully cov-

ered by CNT paste (see Fig. S4). Using such a modified cathode

structure, we could stabilize the field emission behaviors

immune to ambient temperature in the dedicated range

(Fig. 6b), and instant response to on–off switching signals.

Though about 20 V of the applied Vg was increased by this

modification (compare I�V plots in Figs. 5a and 6b) owing

to the elimination of edge-assisted emissions, luminance pat-

terns on the anode plate were fairly identical at the same

anode conditions (Va = 8 kV and Ia = 25 mA) as shown parallel

in Fig. 6a.

Fig. 7a demonstrates a uniform luminous image of the

fully sealed 46 inch diagonal CNTBLU with a luminance

6000 cd m�2, for which Vg = 153 V with a duty ratio 1/120

(0.83%) was applied under the conditions of Va = 15 kV and

Ia = 25 mA. Since the other parts of the LCD panel consume

95% of the light output from the backlight, this full-white

luminance 6000 cd m�2 is represented as 300 cd m�2 when

lit LCD-TV. To evaluate the brightness uniformity (BU), nine

groups on the panel were chosen to measure. Each group con-

sisted nine adjacent dimming blocks (3 · 3), and thus the

totally 81 blocks were used for the evaluation using the defini-

tion of BU (%) = (1 � r/m) · 100, where r and m are standard

deviation and mean brightness, respectively [15]. The mea-

sured BU of the CNTBLU shown in Fig. 7a was as high as

93.2% without an additional light diffusion film (97% with a

diffusion film). The reliability of our panel measured for

8500 h is shown in Fig. 7b. The starting Ia (�26 mA) was stably

maintained for over 6000 h, and for the following 2500 h only

2 mA (�7.7% of the initial value) was reduced. Such unusual

emission stability was ascribed mainly to a couple of note-

worthy aspects in our technology. First, highly crystallized

MWCNT was used. Second, totally 25 mA of Ia required for

the full-white mode from the 4320 dimming blocks means

that average current 5.7 lA (peak current 695 lA) is consumed

per dimming block, which yields average 23 lA (peak 2.76 mA)

per cm2 of the printed CNT area (printing coverage � 0.

25 cm2/block). This value is so fairly low emission current

density that can be stably extracted from highly dense CNT

arrays as in our emitters. For the last to add, an optimized

aging process before the panel isolation from the vacuum sys-

tem quite contributed to the stable lifetime. While aging, the

CNTBLU was operated at a constant current mode fixed at

Ia = 25 mA. Within 4 h, an increase of DVg � 25 volts (23 V of

increase in initial 2 h) was observed (inset of Fig. 7b). Thus

our aging process could probably eliminate enough volatile

gaseous species adsorbed on the CNT emitters, grid mesh,

and phosphor particles. Fig. 7c demonstrates a CNTBLU

under local dimming operation and corresponding LCD image

lit by the CNTBLU. About 35% of the area was locally operated

here for the image representation while the other part of the

Fig. 7 – (a) A photograph of a CNTBLU under full-white mode with a luminance 6000 cd m�2. (b) emission current (Ia) behavior

of a sealed 46 inch CNTBLU showing only 7.7% decay after continuous operation of 8500 h. Inset shows Vg stabilization under

constant current operation during aging process. (c) A captured image of a CNTBLU under local dimming operation (left). Note

the grayscale representation of luminance in the backlight image. Corresponding LCD-TV image lit by the CNTBLU (right). (A

color version of this figure can be viewed online.)

C A R B O N 9 1 ( 2 0 1 5 ) 3 0 4 – 3 1 0 309

panel was in off-state. The brightest blocks (an area in the

woolen muffler in the figure) showed luminance 16000 cd m�2

(800 cd m�2 on that of LCD, Lmax). The off-mode brightness

should be virtually zero because the underneath CNTBLU

did not illuminate at all, but the luminance meter (LX-100,

Konica Minolta) indicated its lower limit 0.002 cd m�2 (Lmin).

Thus, the native contrast ratio (Lmax/Lmin) represented by

our CNTBLU was as high as 400,000:1.

4. Conclusions

A 46-inch diagonal CNT field emission BLU that is built with a

new cathode structure is reported. The approach involves

unique gate insulator formation by glass etching, highly pop-

ulated MWCNT tips, and gate electrode assembly by anodic

bonding. Impressive lamping performances are noted

through dynamic control of the built in fine 4320 local dim-

ming blocks as small as 1 cm2. The LCD-TV lit by the backlight

demonstrates superior luminance uniformity 97%, and a

native contrast ratio 400,000:1, whilst showing less than 10%

decay in emission current of the sealed panel during contin-

uous operation over 8000 h. Scaling up of the proposed device

inclusive of cathode and anode are entirely compatible with

the existing processes for large area plasma panel. Though

to commercialize the CNT BLU for TV in mass as major back-

lights, further reducing in manufacturing time and cost

mainly limited by the initial exhausting and ageing steps

are still challenging, we expect that the CNTBLU is a promis-

ing technology in the high performance TVs and monitors

with motion-blur-free (short response time � 5 ms) and high

contrast images.

Appendix A. Supplementary data

Supplementary data associated with this article can be found,

in the online version, at http://dx.doi.org/10.1016/j.carbon.

2015.04.093.

R E F E R E N C E S

[1] Saito Y, editor. Carbon nanotube and related field emitters:fundamentals and applications. Weinheim,Germany: Wiley-VCH; 2010.

[2] Dijon J, Fournier A, Levis M, Meyer R, Bridoux C, MontmayeulB, et al. 37.5L: Late-News Paper: 600 Colour CNT FEDDemonstrator with High Peak Brightness. SID SymposiumDigest of Technical Papers 2007;38:1313.

[3] Liu P, Wei Y, Liu K, Liu L, Jiang K, Fan S. New-type planar fieldemission display with super aligned carbon nanotube yarnemitter. Nano Lett 2012;12:2391.

[4] Kim YC, Yoo EH. Printed Carbon Nanotube Field Emitters forBacklight Applications. Japan J Appl Phys 2005;44:L454.

[5] Wu HC, Youh MJ, Lin WH, Tseng CL, Juan YM, Chuang MH,et al. Fabrication of double-sided field-emission light sourceusing a mixture of carbon nanotubes and phosphorsandwiched between two electrode layers. Carbon2012;50:4781.

[6] Hao H, Liu P, Tang J, Cai Q, Fan S. Secondary electron emissionin a triode carbon nanotube field emission display and itsinfluence on the image quality. Carbon 2012;50:4203.

[7] Chung KJ, Pu NW, Youh MJ, Liu YM, Ger MD, Cheng K, et al.Improvement of field-emission-lamp characteristics usingnitrogen-doped carbon nanocoils. Diamond Rel Mater2015;53:1.

310 C A R B O N 9 1 ( 2 0 1 5 ) 3 0 4 – 3 1 0

[8] Purcell ST, Vincent P, Journet C. Measuring the physicalproperties of nanostructures and nanowires by fieldemission. Europhys News 2006;37:26.

[9] Nilsson L, Groening O, Emmenegger C, Kuettel O, Schaller E,Schlapbach L, et al. Scanning field emission from patternedcarbon nanotube films. Appl Phys Lett 2000;76:2071.

[10] Choi YC, Lee JW, Lee SK, Kang MS, Lee CS, Jung KW, et al. Thehigh contrast ratio and fast response time of a liquid crystaldisplay lit by a carbon nanotube field emission backlight unit.Nanotechnology 2008;19:235306.

[11] Schjølberg-Henriksen K, Poppe E, Moe S, Storas P, Taklo MMV,Wang DT, et al. Anodic bonding of glass to aluminium.Microsyst Technol 2006;12:441.

[12] Kahnt H. Ionic transport in glasses. J Non-Cryst Solids1996;203:225.

[13] Shoulders KR. Microelectronics using electron beamactivated machining techniques. Adv Comput 1961;2:135.

[14] Ozawa L. Cathodoluminescence and photoluminescence:theories and practical applications. FL, USA: CRC Press,Taylor & Francis Group; 2007.

[15] Kim YC, Nam JW, Hwang MI, Kim IH, Lee CS, Choi YC, et al.Uniform and stable field emission from printed carbonnanotubes through oxygen trimming. Appl Phys Lett2008;92:263112.