JIST - Society for Imaging Science and Technology

JIST 

Vol. 51, No. 5 

September/October 

2007 

Journal of 

Imaging Science 

and Technology 

imaging.org 

Society for Imaging Science and Technology

Editorial Staff 

Melville Sahyun, editor 

sahyun@infionline.net 

Donna Smith, production manager 

dsmith@imaging.org 

Editorial Board 

Philip Laplante, associate editor 

Michael Lee, associate editor 

Nathan Moroney, associate editor 

Mitchell Rosen, color science editor 

David S. Weiss, associate editor 

David R. Whitcomb, associate editor 

JIST papers are available for purchase 

at www.imaging.org and through 

ProQuest. They are indexed in 

INSPEC, Chemical Abstracts, Imaging 

Abstracts, COMPENDEX, and ISI: 

Science Citation Index. 

Orders for subscriptions or single 

copies, claims for missing numbers, 

and notices of change of address 

should be sent to IS&T via one of the 

means listed below. 

IS&T is not responsible for the accuracy 

of statements made by authors and 

does not necessarily subscribe to their 

views. 

Copyright ©2007, Society for Imaging 

Science and Technology. Copying 

of materials in this journal for internal 

or personal use, or the internal or personal 

use of specific clients, beyond 

the fair use provisions granted by the 

US Copyright Law is authorized by 

IS&T subject to payment of copying 

fees. The Transactional Reporting Service 

base fee for this journal should be 

paid directly to the Copyright Clearance 

Center (CCC), Customer Service, 

508/750-8400, 222 Rosewood Dr., 

Danvers, MA 01923 or online at 

www.copyright.com. Other copying 

for republication, resale, advertising or 

promotion, or any form of systematic 

or multiple reproduction of any material 

in this journal is prohibited except with 

permission of the publisher. 

Library of Congress Catalog Card 

No. 59-52172 

Printed in the USA. 

Society for Imaging Science and 

Technology 

7003 Kilworth Lane 

Springfield, VA 22151 

www.imaging.org 

info@imaging.org 

703/642-9090 

703/642-9094 fax 

Manuscripts should be sent to the 

postal address above as describe at 

right. E-mail PDF and other files as requested 

to dsmith@imaging.org. 

Guide for Authors 

Scope: The Journal of Imaging Science and Technology (JIST) is dedicated to the advancement of imaging science knowledge, the 

practical applications of such knowledge, and how imaging science relates to other fields of study. The pages of this journal are 

open to reports of new theoretical or experimental results, and to comprehensive reviews. Only original manuscripts that have not 

been previously published nor currently submitted for publication elsewhere should be submitted. Prior publication does not refer 

to conference abstracts, paper summaries, or non-reviewed proceedings, but it is expected that Journal articles will expand in scope 

the presentation of such preliminary communication. Please include keywords on your title and abstract page. 

Editorial Process/Submission of Papers for Review: All submitted manuscripts are subject to peer review. (If a manuscript appears better 

suited to publication in the Journal of Electronic Imaging, published jointly by IS&T and SPIE, the editor will make this recommendation.) 

To expedite the peer review process, please recommend two or three competent, independent reviewers. The editorial staff, will 

take these under consideration, but is not obligated to use them. 

Manuscript Guidelines: Please follow these guidelines when preparing accepted manuscripts for submission. 

• Manuscripts should be double-spaced, single-column, and numbered. It is the responsibility of the author to prepare a succinct, 

well-written, paper composed in proper English. JIST generally follows the guidelines found in the AIP Style Manual, available 

from the American Institute of Physics. 

• Documents may be created in Microsoft Word, WordPerfect, or LaTeX/REVTeX. 

• Manuscripts must contain a title page that lists the paper title, full name(s) of the author(s), and complete affiliation/address for 

each author. Include an abstract that summarizes objectives, methodology, results, and their significance; 150 words maximum. 

Provide at least four key words. 

• Figures should conform to the standards set forth at www.aip.org/epub/submitgraph.html. 

• Equations should be numbered sequentially with Arabic numerals in parentheses at the right margin. Be sure to define symbols 

that might be confused (such as ell/one, nu/vee, omega/w). 

• For symbols, units, and abbreviations, use SI units (and their standard abbreviations) and metric numbers. Symbols, acronyms, 

etc., should be defined on their first occurrence. 

• Illustrations: Number all figures, graphs, etc. consecutively and provide captions. Figures should be created in such a way that 

they remain legible when reduced, usually to single column width (3.3 inches/8.4 cm); see also 

www.aip.org/epub/submitgraph.html for guidance. Illustrations must be submitted as .tif or .eps files at full size and 600 dpi; 

grayscale and color images should be at 300 dpi. JIST does not accept .gif or .jpeg files. Original hardcopy graphics may be sent 

for processing by AIP, the production house for JIST. (See note below on color and supplemental illustrations.) 

• References should be numbered sequentially as citations appear in the text, format as superscripts, and list at the end of the document 

using the following formats: 

• Journal articles: Author(s) [first/middle name/initial(s), last name], “title of article (optional),” journal name (in italics), ISSN 

number (e.g. for JIST citation, ISSN: 1062-3701), volume (bold): first page number, year (in parentheses). 

• Books: Author(s) [first/ middle name/initial(s), last name], title (in italics), (publisher, city, and year in parentheses) page reference. 

Conference proceedings are normally cited in the Book format, including publisher and city of publication (Springfield, VA, for all 

IS&T conferences), which is often different from the conference venue. 

• Examples 

1. H. P. Le, Progress and trends in ink-jet printing technology, J. Imaging Sci. Technol. 42, 46 (1998). 

2. E. M. Williams, The Physics and Technology of Xerographic Processes (John Wiley and Sons, New York, 1984) p. 30. 

3. Gary K. Starkweather, “Printing technologies for images, gray scale, and color,” Proc. SPIE 1458: 120 (1991). 

4. Linda T. Creagh, “Applications in commercial printing for hot melt ink-jets,” Proc. IS&T’s 10th Int’l. Congress on Adv. In 

Non-Impact Printing Technologies (IS&T, Springfield, VA 1994) pp. 446-448. 

5. ISO 13655-1996 Graphic technology: Spectral measurement and colorimetric computation for graphic arts images (ISO, 

Geneva), www.iso.org. 

6. Society for Imaging Science and Technology website, www.imaging.org, accessed October 2003. 

Reproduction of Color: Authors who wish to have color figures published in the print journal will incur color printing charges. 

The cost for reproducing color illustrations is $490 per page; color is not available to those given page waivers, nor can color page 

charges be negotiated or waived. Authors may also choose to have their figures appear in color online and in grayscale in the printed 

journal. There is no additional charge for this, however those who choose this option are responsible for ensuring that the captions 

and descriptions in the text are readable in both color and black-and-white as the same file will be used in the online and 

print versions of the journal. Only figures saved as TIFF/TIF or EPS files will be accepted for posting. Color illustrations may be 

also submitted as supplemental material for posting on the IS&T website for a flat fee of $100 for up to five files. 

Website Posting of Supplemental Materials: Authors may also submit additional (supplemental) materials related to their articles 

for posting on the IS&T Website. Examples of such materials are charts, graphs, illustrations, or movies that further explain the 

science or technology discussed in the paper. Supplemental materials will be posted for a flat fee of $100 for up to five files. For 

each additional file, a $25 fee will be charged. Fees must be received before supplemental materials will be posted. As a matter of 

editorial policy, appendices are normally treated as supplemental material. 

Submission of Accepted Manuscripts: Author(s) will receive notification of acceptance (or rejection) and reviewers’ 

reports. Those whose manuscripts have been accepted for publication will receive correspondence informing them of the issue for 

which the paper is tentatively scheduled, links to copyright and page charge forms, and detailed instructions for submitting accepted 

manuscripts. A duly signed transfer of copyright agreement form is required for 

publication in this journal. No claim is made to original US Government works. 

Page charges: Page charges for the Journal is $80/printed page. Such payment is 

not a condition for publication, and in some circumstances page charges are 

waived. Requests for waivers must be made in writing to the managing editor prior 

to acceptance of the paper and at the time of submission. 

Manuscripts submissions: Manuscripts should be submitted both electronically 

and as hardcopy. To submit electronically, send a single PDF file attached to an e- 

mail message/cover letter to jist@imaging.org. To submit hardcopy, mail 2 singlespaced, 

single-sided copies of the manuscript to: IS&T. With both types of submission, 

include a cover letter that states the paper title; lists all authors, with complete 

contact information for each (affiliation, full address, phone, fax, and e-mail); identifies 

the corresponding author; and notes any special requests. Unless otherwise 

stated, submission of a manuscript will be understood to mean that the paper has 

been neither copyrighted, classified, or published, nor is being considered for 

publication elsewhere. Authors of papers published in the Journal of Imaging 

Science and Technology are jointly responsible for their content. Credit for the 

content and responsibility for errors or fraud are borne equally by all authors. 

JOURNAL OF IMAGING SCIENCE AND TECH- 

NOLOGY ( ISSN:1062-3701) is published bimonthly 

by The Society for Imaging Science and Technology, 

7003 Kilworth Lane, Springfield, VA 22151. Periodicals 

postage paid at Springfield, VA and at 

additional mailing offices. Printed in Virginia, 

USA. 

Society members may receive this journal as part of 

their membership. Forty-five dollars ($45.00) of 

membership dues are allocated to this subscription. 

IS&T members may refuse this subscription by written 

request. Domestic institution and individual nonmember 

subscriptions are $195/year or $50/single 

copy. The foreign subscription rate is $205/year. 

For online version information, contact IS&T. 

POSTMASTER: Send address changes to JOURNAL 

OF IMAGING SCIENCE AND TECHNOLOGY, 

7003 Kilworth Lane, Springfield, VA 22151.

JIST 

Vol. 51, No. 5 

September/October 

2007 

Journal of 

Imaging Science 

and Technology ® 

Feature Article 

391 Digital Color Image Halftone: Hybrid Error Diffusion Using the Mask 

Perturbation and Quality Verification 

JunHak Lee, Takahiko Horiuchi, Ryoichi Saito, and Hiroaki Kotera 

Special Section: Digital Printing Technologies, including papers presented at 

IS&T NIP 

402 Human Perception of Contour in Halftoned Density Step Image 

Phichit Kajondecha, Hongmei Cheng, and Yasushi Hoshino 

407 Advanced Color Toner for Fine Image Quality 

Akihiro Eida, Shinichiro Omatsu, and Jun Shimizu 

413 Preparation of Microemulsion Based Disperse Dye Inks for Thermal 

Bubble Ink Jet Printing 

S. Y. Peggy Chang and Y. C. Chao 

419 Effects of Molecular Substituents of Copper Phthalocyanine Dyes on 

Ozone Fading 

Fariza B. Hasan, Michael P. Filosa, and Zbigniew J. Hinz 

424 Relationship between Paper Properties and Fuser Oil Uptake in a 

High-speed Digital Xerographic Printing Fuser 

Patricia Lai, Ning Yan, Gordon Sisler, and Jay Song 

431 Simulation of Traveling Wave Toner Transport Considering Air Drag 

Masataka Maeda, Katsuhiro Maekawa, and Manabu Takeuchi 

438 High Speed Imaging and Analysis of Jet and Drop Formation 

Ian M. Hutchings, Graham D. Martin, and Stephen D. Hoath 

445 Effects of Thin Film Layers on Actuating Performance of Microheaters 

Min Soo Kim, Bang Weon Lee, Yong Soo Lee, Dong Sik Shim, and Keon Kuk 

continued on next page 

imaging.org 

Society for Imaging Science and Technology

IS&T BOARD OF DIRECTORS 

President 

Eric G. Hanson 

Department Manager 

Hewlett Packard Company 

Immediate Past President 

James R. Milch Jim 

Director Research & Innovation Labs. 

Carestream Health, Inc. 

Executive Vice President 

Rita Hofmann 

Chemist, R&D Manager 

Ilford Imaging Switzerland GmbH 

Conference Vice President 

Robert R. Buckley Rob 

Research Fellow 

Xerox Corporation 

Publication Vice President 

Franziska Frey 

Assist. Prof., School of Print Media 

Rochester Institute of Technology 

Secretary 

Ramon Borrell 

R & D Director 

Xaar Plc 

Treasurer 

Peter D. Burns 

Principal Scientist 

Carestream Health, Inc. 

Vice Presidents 

Choon-Woo Kim 

Inha University 

Laura Kitzmann 

Marketing Dev. & Comm. Manager 

Sensient Imaging Technologies, Inc. 

Michael A. Kriss 

MAK Consultants 

continued from previous page 

Special Section: Digital Fabrication, including papers presented at IS&T DigiFab2006 

452 Hybrid Stacked RFID Antenna Coil Fabricated by Ink Jet Printing of Catalyst with 

Self-assembled Polyelectrolytes and Electroless Plating 

Chung-Wei Wang, Ming-Huan Yang, Yuh-Zheng Lee, and Kevin Cheng 

456 Top Contact Organic Thin Film Transistors with Ink Jet Printed Metal Electrodes 

Kuo-Tong Lin, Chia-Hsun Chen, Ming-Huan Yang, Yuh-Zheng Lee, and Kevin Cheng 

461 Ring Edge in Film Morphology: Benefit or Obstacle for Ink Jet Fabrication of 

Organic Thin Film Transistors 

Jhih-Ping Lu, Ying-pin Chen, Yuh-Zheng Lee, Kevin Cheng, and Fang-Chung Chen 

465 Digital Fabrication Using High-resolution Liquid Toner Electrophotography 

Atsuko Iida, Koichi Ishii, Yasushi Shinjo, Hitoshi Yagi, and Masahiro Hosoya 

IS&T Conference Calendar 

For details and a complete listing of conferences, visit www.imaging.org 

IS&T/SID’s Fifteenth Color Imaging 

Conference cosponsored by SID 

November 5–November 9, 2007 

Albuquerque, New Mexico 

General chairs: Jan Morovic 

and Charles Poynton 

Electronic Imaging 

IS&T/SPIE 20th Annual Symposium 

January 27–January 31, 2008 

San Jose, California 

General chair: Nitin Sampat 

CGIV 2008: The Fourth European Conference 

on Color in Graphics, Image and Vision 

June 9–June 13, 2008 

Terrassa, Spain 

General chair: Jaume Pujol 

Archiving 2008 

June 24–June 27, 2008 

Bern, Switzerland 

General chair: Rudolf Gschwind 

Ross N. Mills 

CTO & Chairman 

Imaging Technology International 

Jin Mizuguchi 

Professor, Yokohama National Univ. 

David Weiss 

Scientist Fellow, 

Eastman Kodak Company 

Chapter Director 

Franziska Frey – Rochester 

Patrick Herzog – Europe 

Takashi Kitamura – Japan 

Executive Director 

Suzanne E. Grinnan 

IS&T Executive Director 

ii J. Imaging Sci. Technol. 515/Sep.-Oct. 2007

Journal of Imaging Science and Technology® 51(5): 391–401, 2007. 

© Society for Imaging Science and Technology 2007 

Digital Color Image Halftone: Hybrid Error Diffusion 

Using the Mask Perturbation and Quality Verification 

JunHak Lee, Takahiko Horiuchi, Ryoichi Saito and Hiroaki Kotera 

Graduate School of Science and Technology, Chiba University, 1-33, Yayoi-cho, Inage-ku, Chiba-Shi, Chiba, 

263-8522, Japan 

E-mail: leejunhak@graduate.chiba-u.jp 

Abstract. Error diffusion is widely used in digital image halftones. 

The algorithm is very simple to implement and very fast to calculate. 

However, it is known that standard error diffusion algorithms, such 

as the Floyd Steinberg error diffusion, produce undesirable artifacts 

in the form of structure artifacts, such as worms, checkerboard patterns, 

diagonal stripes, and other repetitive structures. The boundaries 

between structural artifacts break the visual continuity in regions 

of low intensity gradients and therefore may be responsible for 

false contours. In this paper, we propose a new halftone method to 

reduce the structural artifacts and to improve the gray expression, 

called hybrid error diffusion, by using the concept of “error diffusion 

by perturbing the error coefficient with a mask.” The proposed algorithm 

consists of two steps in each pixel position. In the first step, a 

perturbation is calculated using the internal pseudorandom number 

and a selected 44 mask, similar to a dither mask. In the second 

step, error diffusion weights are calculated with the criterion for each 

pixel value. The proposed hybrid method can reduce the structural 

artifacts while keeping the advantage of the error diffusion. This 

paper discusses the performance of the proposed algorithm with 

experimental results for natural test images. Then, objective assessment 

results are given using statistical tools and the structural similarity 

measure for color images. © 2007 Society for Imaging Science 

and Technology. 

DOI: 10.2352/J.ImagingSci.Technol.200751:5391 

INTRODUCTION 

Halftoning is a method of producing the pseudocontinuous 

tone images using only a finite number of gray levels. Because 

of the inherent characteristic of the human visual system 

with regard to observing average gray level over an area, 

the human observer perceives intermediate tones. Generally, 

halftoning is considered as a simple “on” and “off” modulation 

technique, where the sensation of intermediate tones 

is created by the presence and absence of a pixel. The digital 

halftone technology plays an important role in transforming 

a continuous tone (gray or color) image to an image with a 

reduced number of gray levels for display devices. 

For halftone technologies, each pixel value is determined 

to be white or black when compared to the threshold 

value, and the quantization error is then fed back and added 

to adjacent pixels. 1,2 The conventional error diffusion algorithm 

has the advantages of simple implementation and fast 

calculation speed. It uses the concept of overflow and diffusion 

of the quantized error, and then resets the diffusion. 

Received Oct. 28, 2006; accepted for publication Jun. 5, 2007. 

1062-3701/2007/515/391/11/$20.00. 

However, the conventional error diffusion algorithm introduces 

distortion, reducing the visibility, worms, and false 

textures or additive noise. In order to solve these problems, 

many digital halftone algorithms have been proposed. Examples 

include using variable thresholds, 3 and variable filter 

weights 4–6 with input data. There were also approaches that 

considered the color channel correlation 6–8 for improving 

the color halftone visibility in color images. However, these 

methods require a complex process and long calculation 

time or many lookup tables. 

In this paper, we propose a well-organized halftone algorithm, 

hybrid error diffusion (HED) to improve the conventional 

halftone artifacts and enhance the visibility of 

color. The proposed algorithm is very simple, easy to implement, 

and can reduce the structural artifacts, keeping with 

the advantage of the error diffusion algorithms. We use the 

concept of a perturbing error filter weight by using the mask 

value, which is perturbed with a pseudorandom number. 

The proposed algorithm is basically the same as using the 

four-tap style error filter similar to that of the Floyd– 

Steinberg error diffusion. The basic procedure of the proposed 

algorithm is as follows: 

1. Determine the mask value for each color plane and 

added to pseudorandom number. 

2. Calculate the error filter weights for each color 

plane. 

The mask that is used is similar to the ordered dither mask. 

The mask value is selected by pixel, line, and each color 

plane. We also use the different error filter weights for each 

color plane to enhance the color visibility. The error filter 

weights were calculated by using a different mask for each 

color plane. The results of the proposed algorithm show 

good performance for reducing artifacts, worms, and false 

textures. 

The paper is organized as follows. In the next section, 

we review the conventional error diffusion algorithm and 

investigate its general problem. In the Hybrid Error Diffusion 

Algorithm section, we explain the proposed HED algorithm 

in detail. In Experimental Results, we introduce the 

conventional halftone evaluation tools, pair correlation, and 

radially averaged power spectrum density. As another 

method of halftone evaluation, a structural similarity mea- 

391

Lee et al.: Digital color image halftone: Hybrid error diffusion using the mask perturbation and quality verification 

Figure 1. General block diagram of conventional error diffusion. 

sure between color images is also derived. Then, the simulation 

results of the algorithms are given with natural images. 

We also compare our algorithm with conventional methods 

using the objective assessment, halftone statistical analysis, 

and color image structural similarity measure. Finally, there 

is the Conclusion. 

CONVENTIONAL ERROR DIFFUSION ALGORITHM 

There are many error diffusion algorithms for improving the 

halftone quality. 1–12 Almost of the conventional algorithms 

are designed based on the Floyd–Steinberg error diffusion 

algorithm. In this section, we investigate the Floyd–Steinberg 

algorithm 1 and the Jarvice et al. 2 error diffusion algorithm, 

as a representative conventional error diffusion algorithm, to 

simulation results. In Figure 1, each signal can be defined as 

follows: 

=1, if jx,y T, 

bx,y 1 

0, otherwise, 

jx,y = ix,y − a jk ex − j,y − k, 

ex,y = bx,y − jx,y, 

2 

3 

Figure 2. Simulation results of conventional error diffusion algorithm. 

Figure 3. Block diagram of hybrid error diffusion. 

where ix,y is the input image and the bx,y is the output 

image of the halftone process. The signal jx,y represents 

the modified input, a jk are the error filter weights, a jk =1, 

and T is the threshold value. The signal ex,y is the accumulated 

error value that will be diffused to adjacent pixels. 

The conventional error diffusion algorithm has the advantage 

of simple implementation and fast calculation. However, 

the conventional error diffusion introduces the distortion 

that reduces image quality and produces worms, false 

textures, and additive noise. The simulation results of the 

Floyd–Steinberg error diffusion algorithm and the Jarvice et 

al. 2 error diffusion algorithm are given in Figure 2. The input 

image is the “gradation ramp” of increasing gray levels 

from 0 to 128 gray levels with the slope of 4 pixels per gray 

level. From the simulation results, we can see the prominent 

discontinuity and a white dot in the middle of the gradation. 

In the low gray area, we can also see diagonal stripes. The 

worm patterns are also in the middle of the gradation ramp. 

The reduction of these kinds of conventional artifacts is the 

objective of the proposed algorithm. At the end of this paper, 

we compare the results of conventional error diffusion 

to the results of the Floyd–Steinberg error diffusion, vector 

color error diffusion, 7 and Shiau–Fan error diffusion 10 with 

natural images. 

HYBRID ERROR DIFFUSION ALGORITHM 

We use the concept of perturbing error filter weight by using 

the mask value, which is perturbed with a pseudorandom 

number. The concept of the proposed hybrid error diffusion 

is shown in Figure 3, in which each signal can be defined as 

follows: 

=1, if jx,y T 

bx,y 4 

0, otherwise, 

jx,y = ix,y − ha jk ex − j,y − k, 

ex,y = bx,y − jx,y, 

Ma jk = MaskValue for each color/line/pixel , 

ha jk = RndNum + Ma jk , 

5 

6 

7 

8 

392 J. Imaging Sci. Technol. 515/Sep.-Oct. 2007


Figure 5. Determination of the error weight ha jk . 

Figure 4. Results of gray level 28: a R, G, B same mask and b R, G, 

B different mask. 

ha jk =1, ha jk 0, 9 

where ix,y is the input image and bx,y is the output 

image of the halftone process. jx,y represents the modified 

input, and ex,y is the accumulated error value that will be 

diffused to adjacent pixels. Ma jk is the selected mask value, 

which is dependent on the pixel position and color plane. 

ha jk are the error filter weights, and T is the threshold 

value. The four-tap style error filter weight of the proposed 

algorithm is similar to that of the Floyd–Steinberg error diffusion 

algorithm. However, the internal pseudorandom 

number generator and mask selector is newly added to that. 

The error filter weights ha jk are varied with the internally 

calculated pseudorandom number and the mask value. The 

mask that is used is similar to the ordered dither mask. This 

mask value is selected by pixel, line, and color plane. As a 

result, the error filter weight varies with the pixel position, 

line, and color plane. Finally, the error filter weight ha jk 

values are determined pixel by pixel by the criterion of 

Eq. (9). 

The procedure of calculating the error carry is as follows. 

First, the mask value is determined based on the color 

plane, the pixel position, and the line position. For example, 

if the color plane is red R, the (pixel, line)(3, 3), the 

mask value will be 9 in the R of Fig. 3. Next, a pseudorandom 

number is added to the predetermined mask value. 

Finally, the error filter weights are determined by a normalizing 

process for each pixel and color plane. Then, the diffusion 

process is carried out, which is similar to conventional 

error diffusion. In view of hardware implementation, 

for example, the mask values and pseudorandom number 

Figure 6. System block diagram for CISM. 

seeds can already be stored in the internal RAM (random 

access memory) area. The generation and reading process 

can be carried out in pixel calculation duration. 

We use a different mask for each color plane to improve 

the color visibility. The results are compared in Figure 4. The 

result of a comes from using the same R, G, and B masks, 

while the result of b is from using different R, G, and B 

masks as shown. We can see that the white dots in the result 

of a are replaced by the mixing of R, G, and B dots in the 

case of b. We can confirm the increasing effect of halftone 

carry density in these results. The green (G) and blue (B) 

masks are generated from the red mask. The red mask is 

generated with the relation of check board weight. That is, 

the green mask is generated by moving 1 pixel to left of the 

red mask. The blue mask is generated by moving 2 pixels to 

left of the red mask. In Figure 5, the example is given in the 

case of the RndNum=7 for each color plane. 

It is possible to control the error diffusion pattern by 

controlling the mask value. Because the error diffusion pattern 

mainly depends on the error filter shape, the scan direction, 

and error filter weights, we can control the error 

diffusion pattern by controlling the R, G, and B masks. 

EXPERIMENTAL RESULTS 

Introduction to Halftone Evaluation 

In this section, we introduce assessment tools used to evaluate 

the HED algorithm. In order to evaluate the halftone, we 

use conventional halftone statistics. However, it is difficult to 

verify the similarity of the structural pattern using conventional 

verification measures. In this paper, we use the color 

image similarity measure for evaluating the structural 

pattern. 

J. Imaging Sci. Technol. 515/Sep.-Oct. 2007 393


Figure 7. System block diagram for color HVS part in CISM. 

Figure 8. System block diagram for the structural similarity measure part in CISM. 

Halftone Statistics: Point Process 

We introduce the conventional point process statistics to 

evaluate the halftone image. The candidates are pair correlation 

and radially averaged power spectrum density 

(RAPSD). 11 Pair correlation, the first candidate, is the influence 

that the point at y has at any x in the spatial annular 

ring. The pair correlation is a strong indicator of the 

interpoint relationships for a given pattern. The pair correlation 

Rr is known as 

Rr = ER yry 

, 10 

ER y r 

where y is the point that is influenced by x. is the sample 

of point process. Spectral analysis was first applied to stochastic 

patterns by Ulichney 12 to characterize patterns created 

via error diffusion. To do so, Ulichney developed the 

radially averaged power spectra along with a measure of anisotropy. 

The radially averaged power spectrum is as follows: 

Pˆf = 1 K DFT 2D i 2 

, 11 

K i=1 N i 

where DFT 2D represents the two-dimensional, discrete 

Fourier transform of the sample , N is the total number 

of pixels in the sample , and K is the total number of 

periodic area being averaged to form to estimate. Finally, the 

RAPSD is defined as follows: 

Pf = 

1 

Pˆf, 

NRf fRf 

12 

where Rf is the series of annular rings and NRf is the 

number of frequency samples in Rf . 

Color Image Similarity Measure 

In this section, we introduce the evaluation method, color 

image similarity measure (CISM), used to evaluate the HED 

algorithm. The color image similarity measure is largely 

composed of two blocks. The first one is the color consid- 



Figure 9. Results of gray level 28 with the Floyd–Steinberg algorithm: 

Pair correlation/RAPSD. 

ering block with the human visual system (HVS) and color 

image structural similarity calculation block as shown in 

Figure 6. As introducing the color HVS model, we could 

consider the interrelation for each color channel in structural 

similarity measure. A color HVS model takes into account 

the correlation among color planes. The HVS model 

is based on a transformation to CIELab color space and 

exploits the spatial frequency sensitivity variation of the luminance 

and chrominance channels. Having a model of the 

HVS allows us to measure the distortion seen by a human 

viewer. 

Color HVS block 

The color HVS block is composed of several subblock, color 

space conversion, discrete Fourier transform, and human visual 

filters as shown in Figure 7. We carry out the color space 

conversion to use the human visual frequency response 

model. The RGB image was transformed to CIEXYZ, and 

then to CIELab color space. We denote the L, a * , b * as the 

Y y , C x , C z for the convenience of equation, respectively. 

As a luminance HVS filter, we use the model that is 

proposed by Sullivan et al. 13 and Nasanen. 14 The contrast 

sensitivity of the human viewer to spatial variations in 

chrominance falls off faster as a function of increasing spatial 

frequency than does the response to spatial variations in 

luminance. This HVS chrominance filter is based on the 

experimental results obtained by Mullen. 15 The details of the 

HVS model are in Appendix I (available as Supplemental 

Material on IS&T website, www.imaging.org). 

The flow of the color HVS block is as follows. Let 

x R,G,B m,n and y R,G,B m,n denote the continuous tone 

image and distorted image, respectively. x Yy ,C x ,C z 

m,n and 

y Yy ,C x ,C z 

m,n are obtained by transforming x R,G,B m,n 

and y R,G,B m,n to the Y y C x C z color space, 

X Yy ,C x ,C z 

k,l = DFTx Yy ,C x ,C z 

m,n, 

Y Yy ,C x ,C z 

k,l = DFTy Yy ,C x ,C z 

m,n, 

13 

14 

Figure 10. Results of gray level 28 with the hybrid error diffusion algorithm: 

Pair correlation/RAPSD for each RGB channel. 

H HVS k,l = H Yy k,l,H Cx k,l,H Cz k,l, 

P XYy 

,C x ,C z k,l = X Y y ,C x ,C z 

k,lH HVS k,l, 

P YYy 

,C x ,C z k,l = Y Y y ,C x ,C z 

k,lH HVS k,l, 

15 

16 

17 

x Yy ,C x ,C z m,n = DFT −1 P XYy k,l, 18 

,C x ,C z 

y Yy ,C x ,C z m,n = DFT −1 P YYy k,l, 19 

,C x ,C z 

where DFT is the discrete Fourier transform and DFT −1 is 

the inverse discrete Fourier transform. The HVS filters are 

applied to the luminance and chrominance components in 

the spatial frequency domain. Finally, the output of the color 

HVS part is x R,G,B 

m,n and y R,G,B 

m,n, which is transformed 

to RGB color space, taking HVS into account. This is 

the input to the structural similarity measure block. 

Structural similarity measure block 

The system block diagram for the structural similarity 

(SSIM) measure block in CISM is shown in Figure 8. The 



Figure 11. Inputs of the structural similarity measure and the color image similarity measure. 

SSIM algorithm is expanded to RGB color. The structural 

similarity method was proposed by Wang et al. 16,17 The 

SSIM compares local patterns for pixel intensities that have 

been normalized for luminance and contrast between a reference 

image and a distorted image. The MSSIM is the mean 

structural similarity measure for entire image. The details of 

the SSIM and MSSIM are in Appendix II (available as 

Supplemental Material on IS&T website, www.imaging.org). 

In this paper, we expand the concept of MSSIM to color 

images. The input of the structural similarity measure part is 

x R,G,B 

m,n and y R,G,B 

m,n, which was already processed 

with consideration of the HVS. The final output of CISM is 

the weighted sum of the MSSIM value for each channel in 

RGB color as shown in 

CISMx,y = w i MSSIMx,y, 

i 

20 

where w i is the weight for each channel in RGB color. In this 

paper, we used the value of w i =1/3. 

Figure 12. Comparison of the structural similarity measure and the color 

image similarity measure: Floyd–Steinberg error diffusion raster scan. 

Evaluation and Experimental Results 

In order to verify the proposed algorithm, we compared the 

conventional error diffusion algorithm, Floyd–Steinberg error 

diffusion algorithm, 1 vector color error diffusion, 7 and 

Shiau–Fan error diffusion 10 with the proposed algorithm. 

Results of Halftone Statistics 

The results of pair correlation and RAPSD are shown in 

Figures 9 and 10. The input image resolution is a 128 

128 pixel image with a gray level of 28 as shown in Fig. 4. 

Figure 9 is the result of the Floyd–Steinberg algorithm, and 

Fig. 10 is the result of hybrid error diffusion. For the result 

of pair correlation, Rr=0 for r3.5 is a consequence of 

the inhibition of points within a distance of 3.5 of each 

other. The more frequent occurrence of halftone result is in 

the distance of 4.5r7.5 with the condition of Rr1. 

There is no area for the case of Rr=0 in Fig. 10. This 

means that the hybrid error diffusion can offer the chance of 

occurrence in the RGB mixed model. 

For the result of RAPSD, it has a power spectrum that is 

composed entirely of high frequencies, in the case of the 

Floyd–Steinberg algorithm in Fig. 9. However, the power 

spectrum of hybrid error diffusion is extended to the lower 

Figure 13. Objective evaluation results: Color image similarity measure. 

frequency area and the high frequency components are suppressed. 

This means that the density of a dot is increased 

and spread with a good pattern profile. 

Results for the Gradation Characteristic 

First, we try to show the capability of MSSIM and CISM to 

assess the halftone image quality. We used the Floyd– 

Steinberg error diffusion (raster scan) as a test halftone 



Table I. Numerical data: Color image similarity measure CISM. 

Table I. 

Continued. 

Input 

Gray 

F/S E.D. 

Raster Scan 

Shiau–Fan 

E.D. 

Proposed 

HED 

Vector Color 

E.D. 

Input 

Gray 

F/S E.D. 

Raster Scan 

Shiau–Fan 

E.D. 

Proposed 

HED 

Vector Color 

E.D. 

0 1.0000 1.0000 1.0000 1.0000 

1 0.9556 0.9556 0.9556 0.9556 

2 0.7842 0.7842 0.7842 0.7867 

3 0.5526 0.5526 0.5526 0.5779 

4 0.3760 0.3760 0.3760 0.4222 

5 0.2615 0.2615 0.2718 0.3143 

6 0.1886 0.1886 0.2793 0.2661 

7 0.1410 0.1410 0.2448 0.2307 

8 0.1089 0.1089 0.2751 0.2290 

9 0.0863 0.0863 0.3149 0.2324 

10 0.0705 0.0701 0.3280 0.2643 

11 0.1362 0.0883 0.2964 0.2610 

12 0.2079 0.2076 0.2693 0.2654 

13 0.1956 0.1956 0.2798 0.2658 

14 0.1793 0.1793 0.3135 0.2778 

15 0.1654 0.1654 0.3217 0.2902 

16 0.1534 0.1534 0.3268 0.2956 

17 0.1429 0.1429 0.3303 0.2963 

18 0.1338 0.1338 0.3150 0.2969 

19 0.1301 0.1269 0.3305 0.3084 

20 0.1677 0.1621 0.3421 0.3242 

21 0.1908 0.1768 0.3544 0.3302 

22 0.1972 0.1969 0.3651 0.3252 

23 0.1883 0.1883 0.3619 0.3353 

24 0.1800 0.1800 0.3753 0.3509 

25 0.1724 0.1724 0.3813 0.3644 

26 0.1654 0.1654 0.3933 0.3639 

27 0.1599 0.1591 0.3886 0.3733 

28 0.1782 0.1734 0.4098 0.3836 

29 0.1947 0.1925 0.4162 0.3998 

30 0.2076 0.2063 0.4266 0.3994 

31 0.2015 0.2015 0.4403 0.4037 

32 0.1951 0.1951 0.4449 0.4155 

33 0.1890 0.1890 0.4494 0.4232 

34 0.1881 0.1849 0.4626 0.4283 

35 0.2044 0.2015 0.4685 0.4377 

36 0.2166 0.2148 0.4741 0.4529 

37 0.2206 0.2211 0.4898 0.4608 

38 0.2155 0.2155 0.4954 0.4701 

39 0.2103 0.2099 0.4990 0.4704 

40 0.2150 0.2112 0.5138 0.4835 

41 0.2269 0.2275 0.5202 0.4876 

42 0.2361 0.2371 0.5257 0.4983 

43 0.2350 0.2357 0.5348 0.5041 

44 0.2304 0.2304 0.5397 0.5160 

45 0.2348 0.2319 0.5523 0.5264 

46 0.2459 0.2455 0.5540 0.5297 

47 0.2516 0.2535 0.5645 0.5396 

48 0.2520 0.2516 0.5714 0.5501 

49 0.2527 0.2487 0.5815 0.5567 

50 0.2593 0.2594 0.5873 0.5658 

51 0.2668 0.2682 0.5938 0.5725 

52 0.2689 0.2694 0.5994 0.5774 

53 0.2700 0.2667 0.6072 0.5850 

54 0.2759 0.2765 0.6147 0.5956 

55 0.2829 0.2846 0.6214 0.5936 

56 0.2848 0.2847 0.6223 0.6044 

57 0.2869 0.2870 0.6304 0.6112 

58 0.2936 0.2952 0.6393 0.6215 

59 0.2992 0.3012 0.6459 0.6268 

60 0.3005 0.3005 0.6531 0.6405 

61 0.3062 0.3084 0.6607 0.6345 

62 0.3121 0.3155 0.6627 0.6458 

63 0.3146 0.3163 0.6686 0.6475 

64 0.3220 0.3265 0.6780 0.6646 

65 0.3231 0.3240 0.6805 0.6640 

66 0.3286 0.3274 0.6794 0.6772 

67 0.3327 0.3319 0.6911 0.6753 

68 0.3371 0.3393 0.6970 0.6796 

69 0.3423 0.3418 0.7011 0.6922 

70 0.3465 0.3468 0.7049 0.6892 

71 0.3518 0.3527 0.7124 0.6953 

72 0.3559 0.3554 0.7172 0.7046 

73 0.3613 0.3618 0.7217 0.7085 

74 0.3661 0.3663 0.7300 0.7145 

75 0.3703 0.3699 0.7317 0.7230 

76 0.3762 0.3764 0.7370 0.7297 

77 0.3800 0.3805 0.7352 0.7341 

78 0.3853 0.3866 0.7420 0.7338 

79 0.3903 0.3914 0.7495 0.7333 

80 0.3948 0.3961 0.7538 0.7430 

81 0.4001 0.4018 0.7533 0.7423 

82 0.4043 0.4074 0.7612 0.7474 

83 0.4097 0.4143 0.7646 0.7281 

84 0.4146 0.4191 0.7675 0.7570 

85 0.4199 0.4230 0.7739 0.7248 

86 0.4219 0.4213 0.7736 0.7264 

87 0.4290 0.4275 0.7817 0.7065 

88 0.4333 0.4331 0.7842 0.7808 

89 0.4390 0.4379 0.7883 0.7991 



Table I. 

Continued. 

Table I. 

Continued. 

Input 

Gray 

F/S E.D. 

Raster Scan 

Shiau–Fan 

E.D. 

Proposed 

HED 

Vector Color 

E.D. 

Input 

Gray 

F/S E.D. 

Raster Scan 

Shiau–Fan 

E.D. 

Proposed 

HED 

Vector Color 

E.D. 

90 0.4437 0.4430 0.7902 0.7993 

91 0.4490 0.4486 0.7945 0.8058 

92 0.4538 0.4531 0.7995 0.8049 

93 0.4591 0.4586 0.7993 0.8039 

94 0.4638 0.4631 0.8047 0.8022 

95 0.4694 0.4690 0.8071 0.8051 

96 0.4743 0.4734 0.8132 0.8034 

97 0.4795 0.4789 0.8179 0.8061 

98 0.4847 0.4840 0.8142 0.8030 

99 0.4898 0.4890 0.8236 0.8107 

100 0.4952 0.4948 0.8257 0.8131 

101 0.4999 0.4996 0.8257 0.8099 

102 0.5055 0.5052 0.8305 0.8227 

103 0.5106 0.5088 0.8342 0.8284 

104 0.5157 0.5152 0.8346 0.8295 

105 0.5213 0.5207 0.8379 0.8275 

106 0.5262 0.5256 0.8417 0.8309 

107 0.5315 0.5312 0.8460 0.8361 

108 0.5369 0.5363 0.8489 0.8406 

109 0.5417 0.5412 0.8477 0.8444 

110 0.5472 0.5467 0.8512 0.8397 

111 0.5527 0.5521 0.8534 0.8407 

112 0.5577 0.5569 0.8557 0.8400 

113 0.5628 0.5622 0.8601 0.8449 

114 0.5686 0.5678 0.8627 0.8441 

115 0.5734 0.5726 0.8633 0.8369 

116 0.5783 0.5778 0.8665 0.8336 

117 0.5843 0.5836 0.8701 0.8411 

118 0.5891 0.5884 0.8713 0.8350 

119 0.5939 0.5934 0.8731 0.8203 

120 0.5989 0.5982 0.8762 0.8230 

121 0.6043 0.6035 0.8790 0.8226 

122 0.6092 0.6085 0.8804 0.8277 

123 0.6143 0.6137 0.8826 0.8299 

124 0.6194 0.6187 0.8873 0.8472 

125 0.6251 0.6236 0.8869 0.8352 

126 0.6306 0.6434 0.8894 0.8440 

127 0.6368 0.6395 0.8894 0.9160 

128 0.6401 0.6357 0.8909 0.8827 

129 0.6447 0.6318 0.8917 0.8286 

130 0.6494 0.6486 0.8925 0.8077 

131 0.6548 0.6539 0.8982 0.8820 

132 0.6603 0.6590 0.8989 0.8629 

133 0.6656 0.6645 0.8986 0.8762 

134 0.6705 0.6696 0.9034 0.8641 

135 0.6759 0.6746 0.9044 0.8604 

136 0.6810 0.6795 0.9045 0.8611 

137 0.6859 0.6848 0.9066 0.8692 

138 0.6911 0.6893 0.9109 0.8713 

139 0.6959 0.6945 0.9112 0.8768 

140 0.7009 0.6996 0.9126 0.8751 

141 0.7059 0.7047 0.9131 0.8752 

142 0.7105 0.7093 0.9147 0.8802 

143 0.7151 0.7138 0.9158 0.8828 

144 0.7200 0.7186 0.9188 0.8798 

145 0.7245 0.7231 0.9209 0.8933 

146 0.7293 0.7279 0.9228 0.8984 

147 0.7338 0.7326 0.9216 0.8969 

148 0.7382 0.7370 0.9247 0.9007 

149 0.7429 0.7418 0.9246 0.9018 

150 0.7473 0.7465 0.9288 0.8984 

151 0.7521 0.7510 0.9283 0.9070 

152 0.7568 0.7571 0.9299 0.9031 

153 0.7612 0.7607 0.9307 0.9020 

154 0.7654 0.7649 0.9312 0.8913 

155 0.7700 0.7695 0.9365 0.9016 

156 0.7747 0.7739 0.9354 0.9045 

157 0.7786 0.7781 0.9349 0.9021 

158 0.7832 0.7822 0.9357 0.9049 

159 0.7873 0.7868 0.9404 0.9027 

160 0.7913 0.7911 0.9415 0.9157 

161 0.7956 0.7953 0.9391 0.9194 

162 0.8000 0.7992 0.9431 0.9201 

163 0.8038 0.8040 0.9431 0.9225 

164 0.8077 0.8077 0.9446 0.9274 

165 0.8119 0.8114 0.9447 0.9292 

166 0.8159 0.8157 0.9460 0.9268 

167 0.8200 0.8192 0.9454 0.9168 

168 0.8248 0.8242 0.9486 0.8530 

169 0.8332 0.8321 0.9489 0.8756 

170 0.8336 0.8312 0.9499 0.8611 

171 0.8360 0.8328 0.9502 0.8943 

172 0.8393 0.8348 0.9540 0.8480 

173 0.8429 0.8399 0.9526 0.8902 

174 0.8461 0.8440 0.9542 0.8993 

175 0.8498 0.8482 0.9557 0.9003 

176 0.8531 0.8514 0.9541 0.9073 

177 0.8564 0.8551 0.9568 0.9158 

178 0.8596 0.8588 0.9570 0.9254 

179 0.8622 0.8625 0.9595 0.9313 



Table I. 

Continued. 

Table I. 

Continued. 

Input 

Gray 

F/S E.D. 

Raster Scan 

Shiau–Fan 

E.D. 

Proposed 

HED 

Vector Color 

E.D. 

Input 

Gray 

F/S E.D. 

Raster Scan 

Shiau–Fan 

E.D. 

Proposed 

HED 

Vector Color 

E.D. 

180 0.8649 0.8662 0.9587 0.9290 

181 0.8675 0.8699 0.9594 0.9261 

182 0.8702 0.8731 0.9623 0.9246 

183 0.8726 0.8764 0.9609 0.9273 

184 0.8754 0.8791 0.9630 0.9370 

185 0.8782 0.8827 0.9634 0.9316 

186 0.8800 0.8860 0.9643 0.9432 

187 0.8838 0.8888 0.9655 0.9429 

188 0.8847 0.8920 0.9658 0.9475 

189 0.8865 0.8974 0.9679 0.9477 

190 0.9059 0.9015 0.9689 0.9424 

191 0.9025 0.9034 0.9709 0.9466 

192 0.9006 0.9030 0.9708 0.9201 

193 0.9035 0.9051 0.9719 0.9375 

194 0.9043 0.9075 0.9718 0.9325 

195 0.9073 0.9099 0.9737 0.9457 

196 0.9105 0.9117 0.9727 0.9461 

197 0.9145 0.9159 0.9750 0.9470 

198 0.9132 0.9168 0.9738 0.9409 

199 0.9132 0.9193 0.9760 0.9496 

200 0.9166 0.9220 0.9765 0.9498 

201 0.9182 0.9245 0.9772 0.9595 

202 0.9202 0.9275 0.9769 0.9547 

203 0.9237 0.9303 0.9788 0.9594 

204 0.9263 0.9333 0.9785 0.9589 

205 0.9279 0.9352 0.9798 0.9637 

206 0.9309 0.9381 0.9799 0.9637 

207 0.9327 0.9404 0.9799 0.9620 

208 0.9359 0.9427 0.9809 0.9630 

209 0.9386 0.9458 0.9812 0.9664 

210 0.9425 0.9483 0.9822 0.9655 

211 0.9574 0.9569 0.9817 0.9628 

212 0.9460 0.9528 0.9824 0.9644 

213 0.9478 0.9536 0.9823 0.9612 

214 0.9486 0.9556 0.9836 0.9662 

215 0.9520 0.9583 0.9829 0.9665 

216 0.9533 0.9605 0.9841 0.9633 

217 0.9547 0.9616 0.9845 0.9728 

218 0.9561 0.9618 0.9853 0.9676 

219 0.9595 0.9632 0.9861 0.9703 

220 0.9627 0.9651 0.9851 0.9718 

221 0.9641 0.9664 0.9856 0.9729 

222 0.9658 0.9674 0.9855 0.9710 

223 0.9686 0.9671 0.9860 0.9737 

224 0.9698 0.9679 0.9865 0.9743 

225 0.9785 0.9738 0.9861 0.9764 

226 0.9732 0.9700 0.9863 0.9751 

227 0.9726 0.9705 0.9861 0.9775 

228 0.9754 0.9704 0.9863 0.9754 

229 0.9737 0.9732 0.9867 0.9783 

230 0.9739 0.9744 0.9866 0.9758 

231 0.9744 0.9765 0.9860 0.9760 

232 0.9777 0.9789 0.9863 0.9744 

233 0.9746 0.9784 0.9860 0.9713 

234 0.9749 0.9775 0.9862 0.9731 

235 0.9752 0.9742 0.9859 0.9682 

236 0.9756 0.9745 0.9850 0.9669 

237 0.9757 0.9757 0.9855 0.9696 

238 0.9772 0.9748 0.9841 0.9672 

239 0.9727 0.9718 0.9855 0.9622 

240 0.9714 0.9736 0.9836 0.9487 

241 0.9720 0.9710 0.9843 0.9528 

242 0.9689 0.9692 0.9838 0.9414 

243 0.9711 0.9733 0.9838 0.9394 

244 0.9630 0.9640 0.9834 0.9379 

245 0.9596 0.9622 0.9831 0.9370 

246 0.9553 0.9599 0.9836 0.9379 

247 0.9529 0.9560 0.9825 0.9432 

248 0.9512 0.9520 0.9838 0.9505 

249 0.9508 0.9517 0.9856 0.9578 

250 0.9555 0.9561 0.9870 0.9644 

251 0.9694 0.9957 0.9896 0.9794 

252 0.9964 0.9957 0.9923 0.9813 

253 0.9964 0.9957 0.9957 0.9918 

254 0.9964 0.9957 0.9957 0.9976 

255 0.9964 0.9956 0.9956 1.0000 

algorithm. The input image (128128 pixels) is the constant 

valued continuous tone image having 0 to 255 gray 

level, and its corresponding halftone image as shown in Figure 

11. The MSSIM and CISM take the value between 0 and 

1. When there is no difference between reference image and 

halftone image, the value is 1. The MSSIM was applied to 

assess the image quality, such as a jpeg image, noise added 

image, video source etc. 16,17 From Fig. 12, we can see that the 

MSSIM must be modified for assessing the halftone visibility. 

The MSSIM value is close to 0 throughout the gray level 

besides the high and low gray area. But in the result of 

CISM, the luminance distortion is mainly shown in the low 

gray area. The contrast distortion of the halftone image is 

mainly shown in the high gray area. Although the Gaussian 

weight (1111, =1.5) is applied to the image in the spatial 

domain, the MSSIM values do not fully comprise the 



Figure 14. Results of the “elliptical ramp.” 

concept of the HVS. The color HVS filters contribute to 

detecting the disturbance of gradation and color correlation 

of the RGB channel. 

In Figure 13, we try to show the visibility characteristics 

throughout the gray level of various kinds of halftone methods: 

Floyd–Steinberg error diffusion (raster scan), Shiau–Fan 

error diffusion, vector color error diffusion, and proposed 

algorithm. In the cases of conventional error diffusion, the 

CISM values are lower than that of the proposed algorithm 

throughout the gray level of 0–255. Especially, the discontinuity 

of the gradation characteristic is shown in the case of 

the vector color error diffusion. From the results, the visibility 

characteristic of HED is outstanding compared to that of 

the conventional halftone method. The numerical data of 

CISM is given in Table I. 

Results for Natural Images 

We compare the results of Floyd–Steinberg error diffusion 

(raster scan), Shiau–Fan error diffusion, vector color error 

diffusion, and the proposed algorithm with the source of the 

“elliptical ramp” image in Figure 14 and the “closed rose” 

image in Figure 15. The size of source images is 256256 

pixels. In Figs. 14(b)–14(d), false textures are prominent in 

the middle of elliptical ramp gradation. But as shown in Fig. 

Figure 15. Results of the “closed rose.” 

Table II. Numerical data for natural images: Color image similarity measure CISM. 

F/S E.D. 

Raster Scan 

Shiau–Fan 

E.D. 

Vector Color 

E.D. 

Proposed 

HED 

Elliptical ramp 0.69 0.69 0.87 0.90 

Closed rose 0.74 0.61 0.69 0.75 

14(e), there is no structural pattern caused by the error diffusion 

in the case of the proposed algorithm. In addition, 

the proposed algorithm does not suffer from the directional 

artifacts, such as diagonal worms, which appear in the elliptical 

ramp edge and highlight area. The gradation of color 

rendition is also better for the proposed algorithm. The 

white dots in Figs. 14(b) and 14(c) are replaced by the mixture 

of red, green, and blue, which is less visible. The numerical 

data of CISM are given in Table II. 

CONCLUSION 

In this paper, we proposed a new error diffusion algorithm. 

The proposed algorithm is very simple, easy to implement, 

and can reduce the structure artifacts while keeping the advantages 

of the error diffusion. We use the concept of perturbing 

error filter weight using the mask, which is selected 



with a pseudorandom number. The results of the proposed 

method and conventional error diffusion were compared to 

the natural image. In addition, the proposed algorithm was 

evaluated with the objective assessment tools, halftone statistics, 

and CISM. We improved the gray expression using 

the mask and pseudorandom number. The color visibility 

was also improved by using the mixed error diffusion weight 

method. The proposed algorithm has good performance for 

improving the gradation characteristics and reducing the 

structural pattern induced by conventional error diffusion. 

For future work, we will try to investigate the possibility of 

adaptation to the flat panel display. 

REFERENCES 

1 R. W. Floyd and L. Steinberg, “An adaptive algorithm for spatial gray 

scale”, Proc. SID 17, 75 (1976). 

2 J. F. Jarvis, C. N. Judice, and W. H. Ninke, “A survey of techniques for 

the display of continuous tone pictures on bilevel displays”, Comput. 

Graph. Image Process. 5, 13 (1976). 

3 N. Damera-Venkata and B. L. Evans, “Adaptive threshold modulation 

for error diffusion halftoning”, IEEE Trans. Image Process. 10, 104 

(2001). 

4 P. Li and J. P. Allebach, “Tone dependent error diffusion”, IEEE Trans. 

Image Process. 13, 201 (2004). 

5 R. Eschbach, “Reduction of artifacts in error diffusion by means of input 

dependent weights”, J. Electron. Imaging 2, 352 (1993). 

6 V. Monga and B. L. Evans, “Tone dependent color error diffusion”, Proc. 

IEEE Int. Conf. on Acoustics, Speech, and Signal Proc. (IEEE, Piscataway, 

NJ, 2004) Vol. 3, p. 101. 

7 N. Damera-Venkata and B. L. Evans, “Design and analysis of vector 

color error diffusion halftoning systems”, IEEE Trans. Image Process. 10, 

1552 (2001). 

8 J. H. Lee, “Plasma display apparatus and image processing method 

thereof”, US Patent Application, 0,253,782A1 (2005). 

9 D. J. Lieberman and J. P. Allebach, “Model based direct binary search 

halftone optimization with a dual interpretation”, Proceedings of the 

IEEE Int. Conf. on Image Proc. (IEEE, Piscataway, NJ, 1998) Vol. 2, p. 44. 

10 J. N. Shiau and Z. Fan, “Set of easily implementable coefficients in error 

diffusion with reduced worm artifacts”, Proc. SPIE 2658, 222 (1996). 

11 D. L. Lau and G. R. Arce, Modern Digital Halftoning (Marcel Dekker, 

New York, 2001), pp. 27–45. 

12 R. Ulichney, Digital Halftoning (MIT Press, Cambridge, MA, 1987). 

13 J. Sullivan, L. Ray, and R. Miller, “Design of minimum visual 

modulation halftone patterns”, IEEE Trans. Syst. Man Cybern. 21, 33 

(1991). 

14 R. Nasanen, “Visibility of halftone dot textures”, IEEE Trans. Syst. Man 

Cybern. 14, 920 (1984). 

15 K. T. Mullen, “The contrast sensitivity of human color vision to 

red-green and blue-yellow chromatic gratings”, J. Physiol. (London) 359, 

381 (1985). 

16 Z. Wang, A. C. Bovik, and E. P. Simoncelli, “Structural Approaches to 

Image Quality Assessment”, in Handbook of Image and Video Processing, 

2nd ed. (Academic Press, New York, 2005), Chapter 8.3. 

17 Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, “Image quality 

assessment: From error visibility to structural similarity”, IEEE Trans. 

Image Process. 13, 600 (2004). 




Human Perception of Contour in Halftoned Density Step 

Image 

Phichit Kajondecha, Hongmei Cheng and Yasushi Hoshino 

Department of Systems Engineering, Nippon Institute of Technology, Miyashiro, Saitama, 354-8501, Japan 

E-mail: s3054602@sstu.nit.ac.jp 

Abstract. Contour is an important element of an image, usually 

expressed by the difference in density between two areas of an 

image. Objects are usually recognized from observation of their contours. 

In some cases, emergence of a false contour is a problem in 

image coding and decoding. Halftoning is an essential image process 

for digital printing of continuous tone. Contour perception is 

studied experimentally under various halftoning conditions, observation 

distances, and illumination conditions. The results show that in 

the human visual system, contour perception becomes easier with a 

decreasing stimulus of dots, which serve as the component of a 

halftone image. © 2007 Society for Imaging Science and 

Technology. 


INTRODUCTION 

In image processing, contours or boundaries in an image are 

important factors because our perception and recognition of 

an object depend on contour. In a sense, therefore, our eyes 

can be said to search for contour in an image. 1 Contour is 

the boundary between two areas of different density, color, 

or texture. If the dots are sufficiently small, the eye cannot 

detect the dot pattern 2 and we can recognize contour. In 

order to achieve high-quality image printing, the contour 

expression characteristics of an image are considered important. 

One way to understand and evaluate image quality is 

use of the human visual system (HVS). 3–5 

Digital halftoning is essential for processing in digital 

printing and can be achieved by various methods. A conventional 

halftone is a fundamental method that has a wide 

application in printing. Digital halftoning has been evaluated 

from various viewpoints, but we have not yet gained a sufficient 

understanding of the relationship between the ease of 

perceiving contour, which is one of most fundamental elements 

of an image, and half-toning methods. 

In this study, images with contours of two different density 

values are processed to halftoned images of various halftone 

cycles and halftone angles of 0° and 45°. These 

halftoned images are observed under several conditions of 

illumination and observation distance. With respect to halftone 

frequency, images of relatively large cycles are prepared 

for accurately expressing the density of the image. If the 

observation is conducted at long distances, the condition of 

view angle from the eye is the same as that of a small cycle. 

The perception of contour is examined under various conditions, 

and the dependence of perception ratio on the difference 

in density is obtained. As the stimuli of the halftone 

cycle decrease relative to the density difference between regions, 

the contour is perceived more easily. 

EXPERIMENTAL 

As shown in Figure 1, two types of images were produced, 

each of which differed with respect to the direction of the 

contour line: (a) Images with perpendicular contour lines 

and (b) images with diagonal contour lines. The images were 

printed at a size of 10 cm10 cm with a resolution of 

300 dpi. The image printing conditions are shown in Table 

I, and as can be seen in the table, in some cases the density 

level of area A was fixed at 192 while the density level of area 

B was varied between 191 and 180. In other cases, the density 

level of area A was fixed at 64 while that of area B was 

varied between 63 and 52. 

Figure 2 illustrates the halftone dot arrangement, screen 

dot shape, and screen angle. The screen dot shape of the 

halftone is circular and produced at two screen angles (0° 

and 45°). As shown in Figure 3, the cycle of halftone dots 

was prepared in four levels: 1mm, 2mm, 3mm, and 

4mm. 

The observation experiments are carried out as illustrated 

in Figure 4. The halftoned images were presented under 

three illumination conditions at four different observation 

distances. The perception ratios are obtained as the 

ratio of testees who perceived contour (responded “yes”) 

verses total testees. Testees who could not perceive contour 

responded “no.” 

RESULTS AND DISCUSSIONS 

The perception ratios obtained in the experimented conditions 

are plotted against the density difference between areas 

Received Feb. 2, 2007; accepted for publication May 7, 2007. 

1062-3701/2007/515/402/5/$20.00. 

Figure 1. Example density difference patterns: a perpendicular contour 

pattern and b diagonal contour pattern. 

402

Kajondecha, Cheng, and Hoshino: Human perception of contour in halftoned density step image 

Table I. Preparation of observation samples. 

Density 

Difference 

Density Values of Areas A and B 

A-B 

A-B 

1 64-63 192-191 

2 64-62 192-190 

3 64-61 192-189 

4 64-60 192-188 

5 64-59 192-187 

6 64-58 192-186 

7 64-57 192-185 

8 64-56 192-184 

9 64-55 192-183 

10 64-54 192-182 

11 64-53 192-181 

12 64-52 192-180 

Figure 3. Patterns of dot cycle at screen angle 45° at density level 189- 

192: a 1 mm cycle, b 2 mm cycle, c 3 mm cycle, and d 4mm 

cycle. 

Figure 2. Explanation of halftone dot arrangement: a angle of halftone 

dots and b cycle of halftone dots. 

A and B. The results show that the ratio increases with the 

density difference. The ratios are obtained while four factors 

are varied: cycle of halftone dot, screen angle, observation 

distance, and illumination. We discuss three points: effect of 

halftone dot cycle, effect of halftone screen angle, and effect 

of illumination. 

Effect of halftone Dot Cycle 

Figure 5 shows the perception ratio dependence on density 

difference of four distances and four cycle conditions when 

the screen angle is 45°. From Fig. 5, for all observation distances, 

perception ratios generally increase with a decreasing 

dot cycle. Concerning observation distance, when the dot 

cycle is 4mm, the contour is very difficult to perceive at 

0.5 m, but the contour becomes easier to perceive at 3m.At 

every observation distance, the difference in perception ratio 

among the dot cycles of 1–4 mm decreases with an increase 

in observation distance. 

Similarly, at every observation distance, the increase in 

the perception ratio is considered from the viewpoint of the 

human visual system (HVS). As shown in Figure 6, within 

the employed experimental range, the sensitivity of the HVS 

decreases as the dot cycle increases. As the stimulus of halftone 

dot decreases, the perception ratio increases. In the 

observers’ opinions, there are two cases: perception of only 

Figure 4. Schematics of experiment. 

dots in the halftone image and perception of contour. As a 

whole, when the stimulus of the dots is strong, the observer 

has difficulty in perceiving the contour. In addition, the decrease 

in the differences between the dot cycle at certain 

observation distances derives from the decrease in dot sensitivity 

in the HVS. 

The results above are consistent with our subjective estimation. 

Figure 5 shows that the perception ratio increases 

as observation distance increases. This is considered to be 

due to the decrease in the stimulus of dots as observation 

distance increases. 1 

Effect of halftone Screen Angle 

Figure 7 shows the perception ratio dependence on density 

difference at observation distances of 0.5 m and 1.0 m with 



Figure 5. Perception ratio of average patterns diagonal and perpendicular pattern vs density differences 

192-180 at four distances and a dot cycle of 1–4 mm when the halftone angle is 45°. The plates show the 

conditions where the observation distance is a 0.5 m, b 1m,c 2m,andd 3m. 

halftone frequencies of 1mm, 2mm, 3mm, and 4mm.In 

a digital system, the halftone screen angle is simulated by 

placement of the dots within the halftone cells. Figure 7(a) 

shows a comparison between a screen angle of 0° and 45° at 

an observation distance of 0.5 m relativetoaveragevalueof 

diagonal and perpendicular pattern under an illumination of 

1.010 3 Lux. The result shows that at a halftone cycle 

1mm, the halftone image with the screen angle 45° can be 

detected more easily than the halftone image with the screen 

angle 0°. 

According to the results shown in Fig. 7, the perception 

ratio of the halftone image with a screen angle of 45° is 

higher than that with a screen angle of 0° at a halftone cycle 

of 1mm. The sensitivity of our human visual system is reported 

to decrease at angles of 45° and 135°. 6 Therefore, the 

stimuli of the halftone dot when the halftone angle is 45° 

decreases, and we perceive the contour more easily when the 

angle is 0°. However,atcyclesof2mm, 3mm, and 4mm, 

the ratios do not show as clear a difference as when the dot 

cycle is 1mm, possibly because the dot stimulus is too 

strong. Therefore, in some cases, the opposite relationship 

can arise for all of the other cycles (2 mm, 3mm, and 

4mm). 

Effect of Illumination 

One factor that affects contour perception of a halftoned 

image is illumination. Figure 8 shows the perception ratio 

dependence on density difference at a halftone angle of 45°. 

The results show a comparison of when the illumination 

conditions are 1.010, 1.010 2 , and 1.010 3 Lux and observation 

distance is 1mand 3m, under halftone cycles of 

1mmand 4mm, respectively. Figure 8(c) shows that the 

perception ratio increases with illumination when the dot 

cycle is 1mmat an observation distance of 3m. Figures 

8(b) and 8(d) show that the perception ratio decreases when 

the dot cycle is 4mmat observation distances 1mand 3m 

with increasing illumination. In summary, the perception 



Figure 6. Relationship between discrete Fourier transform of halftone frequency and modulation transfer function 

of HVS: a halftone frequency 1 dot/mm and b halftone frequency 0.33 dot/mm. 

Figure 7. Perception ratio of average pattern diagonal and perpendicular dependence on density differences. 

Density values ranged between 192 and 180, the observation distance was a 0.5 m and b 1.0 m, 

and illumination was 1.010 3 Lux. 

ratio increases with decreasing illumination at a dot cycle 

4mm, whereas the perception ratio increases when illumination 

increases at a dot cycle of 1mmat an observation 

distance of 3m. 

In the perception of contour, competition arises between 

Stevens effect and the decrease in sensitivity in a highfrequency 

area under low illumination. When the dot cycle is 

1mm, the Stevens effect becomes dominant, and the perception 

ratio increases with illumination. However, when the 

dot cycle 4mm, the perception ratio increases with a decrease 

in illumination. The reason for the increasing in the 

perceptionratioatadotcycleof4mm under decreased 

illumination is considered to be that the MTF of the HVS 

shifts to the low frequency range, resulting in a decrease in 

the sensitivity of the halftone dot. 7 

CONCLUSIONS 

Contour perception was investigated under various halftone 

and observation conditions. Perception ratio was observed 

to increase with density difference. At large halftone dot 

cycles of 3mm, the ease of contour perception increases as 



Figure 8. Perception ratio dependence on density difference between 64-52 and an angle of 45°, with a 

cycle condition and observation distance of a 1mmand1m,b 4mmand1m,c 1mmand3m,and 

d 4mmand3m. 

observation distance increases. This is due to the decrease in 

dot stimulus relative to the contour signal in HVS. The perception 

ratio of halftone images with screen angles of 45° 

was found to be higher than that observed when the screen 

angle is 0° with a cycle of 1mmat observation distances of 

0.5 m and 1m. For illumination, contour perception was 

found to improve when illumination was decreased at the 

large halftone dot cycle of 3mm. 

ACKNOWLEDGMENTS 

The authors would like to thank Shigeru Kitakubo of the 

Nippon Institute of Technology, Japan, Kazuhisa Yanaka of 

Kanagawa Institute of Technology, Japan, and the members 

of the Hoshino Laboratory for their kind contributions and 

efforts, support, and valuable comments. 

REFERENCES 

1 P. Kajondecha, H. Cheng, S. Huang, and Y. Hoshino, “Contour 

perception of halftoned density step image”, Proc. IS&T’s NIP 22 (IS&T, 

Springfield, VA, 2006) pp. 351–355. 

2 H. R. Kang, Digital Color Half-toning (SPIE, Bellingham, WA, 1999), 

p. 61. 

3 J. Dusek and K. Roubik, “Testing of new models of the human visual 

system for image quality evaluation”, Proc. IEEE International 

Symposium on Signal Processing and its Applications, Vol. 2 (IEEE, Los 

Alamitos, CA, 2003) pp. 621–622. 

4 S. Kitakubo and Y. Hoshino, “Some characteristics on human visual 

sensitivity or spatial frequency of digital halftone images”, Proc. IS&T’s 

NIP 21 (IS&T, Springfield, VA, 2005) pp. 118–121. 

5 M. Ikeda, Image Processing in Human Vision (Heibonsha, Japan, 1999), 

p. 137. 

6 D. W. Heeley and B. Timmey, “Meridional anisotropies of orientation 

discrimination for sine wave gratings”, Vision Res. 28, 337–344 (1988). 

7 F. L. Van Nes and M. A. Bouman, “Spatial modulation transfer in the 

human eye”, J. Opt. Soc. Am. 57, 401–406 (1967). 




Advanced Color Toner for Fine Image Quality 1 

Akihiro Eida, Shinichiro Omatsu and Jun Shimizu 

Performance Chemicals Research Laboratories, Kao Corporation, Wakayama, 640-8580 Japan 

E-mail: eida.akihiro@kao.co.jp 

Abstract. Application of full color printers has spread rapidly in the 

office environment, and demands on the full color printer have been 

also increasing; especially, high-speed printing, oilless fusing and 

fine image quality are strongly requested. Recently, several types of 

chemical toner have been launched. Because they have small and 

narrow particle size distributions, they have an advantage in fine 

image quality. However, the resin of chemical toners are usually 

styrene-acrylic, which is not suitable for high-speed printing and 

glossy images for pictorial applications because the styrene-acrylic 

has to be high molecular weight to make toner have sufficient durability. 

On the other hand, polyester has better properties for highspeed 

printing than styrene-acrylic because of the existence of a 

polar group that enhances a compatibility with the cellulose of paper. 

Polyester has good durability even if it has a low molecular 

weight and thus can provide a glossy image. Thus polyester has 

several advantages over styrene-acrylic. In general, it is difficult to 

use polyester as the binder resin of chemical toner. On the other 

hand, it is easy to make polyester color toner by the pulverization 

method, but it is difficult to make pulverized toner having small and 

narrow particle size distribution for fine image quality, and it is also 

difficult to make pulverized toner having both oilless fusing ability 

and sufficient durability. In this article, design of oilless fusable, pulverized 

color toner with small and narrow particle size distribution 

was investigated. This advanced color toner named “MC toner” can 

provide high-speed printing, oilless fusing, and also fine image 

quality. © 2007 Society for Imaging Science and Technology. 


INTRODUCTION 

In recent years, as color printers have become widely used, 

demands on the color printer have also been increasing: for 

instance, high image quality, high speed, and low cost performances 

are required for the toner. These properties imply, 

for example, small particle size, narrow particle size distribution, 

good transfer property, low temperature fusing, etc. 

To achieve these requirements, several types of chemical 

toner were developed and launched. It is said that performance 

of chemical toner is better than pulverized toner with 

respect to various points of performance. 1,2 

Demands on color printing and requirements for color 

toner in order to achieve these demands are listed below. 

Comparison of typical chemical toner and pulverized toner 

and a comparison of styrene-acrylic and polyester resins 

with respect to these requirements are also discussed. 

Fine Image Quality 

For fine image quality, faithful dot reproduction of latent 

image, faithful transfer property, glossy image for pictorial 

image, wide color gamut, are required. 

Faithful Dot Reproduction 

Figure 1 shows a dependency of raggedness of line defined as 

TEP on toner particle size. This indicates that the raggedness 

increases in proportion to toner particle size. Thus, for faithful 

dot reproduction, toner is required to be a small size. 

From this point of view, chemical toner has an advantage 

because chemical toners are made by a buildup method. 

Figure 2 shows toner size distribution of conventional pulverized 

toner and typical chemical toner. The toner size distribution 

of the chemical toner is smaller and narrower. 

Faithful Transfer 

In general, it is said that a spherical toner shows a good 

transfer efficiency. 3,4 Indeed, Figure 3 shows transfer efficiency 

of chemical toner having a spherical shape 

circularity=0.98 is much better than irregular-shaped 

toner circularity=0.92, where, transfer efficiency is represented 

by E, which is the color difference between blank 

tape on paper and tape removing residual toner from the 

organic photoconductor (OPC). (For more detail, see Experimental 

section.) On the other hand, scatter of spherical 

toner is much worse than for irregularly shaped toner, which 

indicates that spherical toner tends to roll in the transfer 

process because of high flowability. Moreover, cleaning properties 

of spherical toner are also bad. Thus, it is not clear 

whether spherical shape is beneficial for fine image quality. 

Recently, some chemical toners have adopted moderately 

spherical shape (i.e., circularity is 0.96; for example, potato 

shape, spindle shape, or dimple shape). Transfer efficiency, 

1 Presented, in part at IS&T NIP21 Baltimore. 

Received Jan. 9, 2005; accepted for publication Jun. 10, 2007. 

1062-3701/2007/515/407/6/$20.00. 

Figure 1. Dependency of image quality TEP on toner particle size. 

407

Eida, Omatsu, and Shimizu: Advanced color toner for fine image quality 

Figure 4. Comparison of pigment dispersion of pulverized toner and 

typical chemical toner. 

Figure 2. Comparison of particle size distribution and image quality of 

conventional pulverized toner and typical chemical toner. 

Figure 5. Offset band of St/Ac toner, polyester toner, and crystalline 

polyester containing toner. 

Figure 3. Comparison of performance of spherical shape toner and irregular 

shape toner. 

scattering property, and cleaning property of toner with such 

a moderately spherical shape are well balanced. 

Glossy Image 

For pictorial image printing, a glossy image is preferable. To 

print an image with high gloss, it is necessary that a toner 

resin have a low molecular weight, because in the fusing 

process, viscosity of the resin having low molecular weight is 

low and the surface of fused image becomes smooth, and the 

image having such smooth surface specularly reflects light 

well and shows high gloss. 

From this point of view, polyester resin has an advantage 

over styrene-acrylic resin. Polyester resin has sufficient 

durability owing to hydrogen bonded polar groups, while 

styrene-acrylic resin has poor durability for the same low 

molecular weight distribution as polyester. 

Wide Color Gamut 

For wide color gamut, fine dispersion of the pigment is necessary. 

Figure 4 shows dispersion of the pigment of pulverized 

toner and chemical toner. In general, dispersion of a 

pulverized toner pigment is better than for a chemical toner, 

because the kneading process enhances dispersion of the 

pigment. In general polyester resin has better pigment 

dispersability than styrene-acrylic resin, because the polar 

groups of polyester resin enhance dispersion of the pigment. 

Therefore, for a wide color gamut, pulverized toner and 

polyester resin have advantages over chemical toner and 

styrene-acrylic resin. 5 

High-Speed Printing 

For high-speed printing, low-temperature fusing ability is 

required for color toner. As described above, polyester resin 

has sufficient durability with low molecular weight. Moreover, 

polyester has polar groups that enhance compatibility 

with cellulose of paper. Thus, polyester resin has an advantage 

for high-speed printing over styrene-acrylic resin. Recently 

a crystalline polyester resin that has ideal viscoelasticity 

for low-temperature fusing has been proposed (Figure 5) 

for higher speed printing. 6–8 

Oilless Fusability for Low Cost Machine 

For low machine cost, oilless fusing ability is required for 

color toner, so that the oil applying unit can be removed 



Resin 

Table I. Properties of the experimental polyester resin. 

Acid Value a 

mg KOH/g 

T1/2 b 

°C 

T g 

c 

°C 

Polyester 20 115 63 

a The acid value was measured according to ASTM D1980-67. 

b The softening point T1/2 was measured according to ASTM E-28-67. 

c The glass transition temperature T g was measured by a differentital scanning calorimeter “DSC 

Model 200” manufactured by Seiko Instruments Inc., at a heating rate of 10°C/min. 

Toner 

Wax 

% 

Table II. Toner samples. 

Kneading 

Machine 

Preconditioning 

Target Size 

µm 

A 7.0 Twin-screw None 8 

B 7.0 Open-roll None 8 

C 6.0 Open-roll None 5 

D 6.0 Open-roll Silica=1.0% 5 

E 6.0 Twin-screw Silica=1.0% 5 

• to contain plenty of wax with high durability, for oil-less 

fusing 

In this article, the design of an oilless fusable polyester pulverized 

color toner with small and narrow particle size distribution 

is presented. Factors that enable toner to be pulverized 

with a small and narrow particle size distribution 

were investigated in detail. 

Figure 6. Open-roll-type kneader. 

from the printer system. Generally, a resin for color toner 

has a low and narrow molecular weight distribution for 

glossy image and high transparency, but such a resin cannot 

give a wide nonoffset range in fusing because elasticity of 

such resin fusing conditions is low and offset problems tend 

to occur. Therefore, to prevent offset problems, under a conventional 

oil applying unit may be attached to the fuser unit, 

which is costly. Thus, oilless fusable toner is described for 

design of a low cost machine. To achieve oilless fusing, it is 

necessary to add wax to the toner. In the fusing process, the 

wax melts and diffuses out from inside the toner and works 

as a releasing agent instead of oil applied to the fusing roller. 

However, it is difficult to disperse sufficient wax by a twinscrew 

extruder, such as is conventionally used in the kneading 

process. Also, thereby the dispersed size of wax domains 

in the toner becomes very large, and such poor dispersion of 

wax causes poor durability of the toner. 9,10 On the other 

hand, chemical toner can include plenty of wax, which does 

not exist on the toner surface; thus, chemical toner is oilless 

fusable without a durability problem. 

As described above, both chemical toner and pulverized 

toner have good points and bad points. But polyester resin 

has a certain advantage over styrene-acrylic resin. In general, 

it is difficult to use polyester resin in chemical toner. On the 

other hand, it is easy to make polyester color toner by the 

pulverization method, but it is difficult to make pulverized 

toner with small and narrow particle size distribution for 

fine image quality. To summarize, the ideal toner is expected 

as follows: 

• to use polyester for glossy image and high-speed 

printing 

• to be small and narrow size distribution, for fine image 

quality 

EXPERIMENTAL 

Preparation of Polyester Resin 

Bisphenol-A propylene oxide adducts, ethylene oxide adducts, 

terephthalic acid, C12-succinic anhydride, and trimellitic 

anhydride were allowed to react for condensation polymerization 

at 230°C with a small amount of the catalyst in 

a glass flask, which was equipped with a thermometer, a 

stainless steel stirring rod, a reflux condenser, and nitrogen 

inlet tube. Physical properties of this polyester resin are 

listed in Table I. 

Preparation of Toner Samples: For Investigation 

of a Factor of Oilless Fusable Toner 

Toner samples A and B comprised the resin described above, 

wax, charge control agent, and colorant. The colorant was 

Quinacridone (Pigment Red 122); the melting point of the 

wax was 80°C. Content of the colorant was 6.5 wt. %; content 

of the wax was 7.0%, which is the amount that enables 

oilless fusability. The materials were premixed in a batch 

mixer, and then they were kneaded. We used two types of 

kneader, a conventional twin-screw extruder for toner A and 

an open-roll-type kneader (Figure 6) for toner B. The toners 

were pulverized and classified to 8 m. Toner samples are 

listed Table II. 

Preparation of Toner Samples: For Investigation 

of a Factor of Efficient Pulverizing 

Toner samples C, D, and E also comprised the resin described 

above, wax, charge control agent, and colorant. The 

colorant was Quinacridone (Pigment Red 122); the melting 

point of the wax was 80°C. Content of the colorant was 

6.5 wt. %, and content of the wax was 6.0%, which is the 

amount that enables oilless fusability. 

The materials were premixed in a batch mixer, and then 

they were kneaded. Toners C and D were kneaded by the 

open-roll-type kneader, and toner E was kneaded by the 

twin-screw extruder. Before pulverizing, toner chips for tonners 

D and E were mixed with 1.0 wt. % of hydrophobic 



Figure 7. Comparison of wax dispersion and durability. 

silica; we call this process “preconditioning.” Then, they 

were pulverized and classified. The target size was 5 m. 

Toner samples are listed Table II. 

MEASUREMENTS 

Durability was tested by using a toner cartridge of a color 

laser printer for which the developing system was a single 

component. Toner was put into the cartridge, and the developer 

roll was rotated at 60 rpm without developing the 

toner to the organic photoconductor. Durability was defined 

as time when filming of the toner to the doctor blade occurred 

and a streak appeared in the toner layer on the developer 

roller. 

The particle size distribution was measured by a Coulter 

Multisizer II with a 100 m size aperture. The size of dispersed 

wax domains was observed by TEM. For convenience 

of evaluation, a cross section of the toner chip before pulverizing 

was observed. 

Toner was developed in a nonmagnetic singlecomponent 

printer whose resolution was 600 dpi. A dot image 

on the paper before fusing was observed under the microscope. 

Dispersion state of silica and free silica rate were 

measured by Particle Analyzer FT-1000 (Horiba). Circularity 

was measured by FPIA 2100 (Sysmex). 

Transfer efficiency was defined as the amount of residual 

toner on the photoconductor after the transfer process. 

Areal density on the photoconductor was controlled at 

0.40–0.45 mg/cm 2 . The residual toner was removed by 

clear tape and placed on white paper. The amount of the 

residual toner is defined as the difference of hue E between 

the residual toner sample and the blank tape on the 

same paper. 

RESULTS AND DISCUSSION 

Wax Dispersion and Durability 

Figure 7 shows wax dispersion in toners A and B. The size of 

the dispersed wax domains in toner A that was kneaded by 

conventional twin-screw extruder is very large. This indicates 

the twin-screw extruder cannot disperse wax finely 

enough. On the other hand, the dispersed size of wax of 

Figure 8. Result of pulverizing of toners C, D, and E. 

toner B was very small. Toner B was kneaded by an openroll-type 

kneader, and this type of kneader can knead toner 

at a lower kneading temperature, which means toner is 

kneaded more strongly at a higher resin viscosity compared 

to the twin-screw extruder. 

Durability of toner B is over six times that of toner A, 

and is enough for practical use. This result indicates that 

large domains of wax decrease durability of toner because, it 

is thought, that such a large wax domain tends to (i) exist on 

the toner surface after pulverizing and (ii) stick to the doctor 

blade and developer roller. To avoid this problem, it is necessary 

to disperse wax finely in the toner; thus, toner 

kneaded by the open-roll-type kneader can contain plenty of 

wax with finely dispersed size and can yield both efficient 

fusing ability and durability for oilless fusing. 

Result of Pulverizing of Toners C, D, and E 

Figure 8 shows the result of pulverizing. Toner D was pulverized 

to 5 m size with narrow distribution successfully. 

Toner C, which was not preconditioned, was not pulverized 

to a small size, and it contains large-size particles. Toner E, 

which was preconditioned but kneaded by the twin-screw 

extruder, also was not pulverized to 5 m size. Dot image 

quality with toner D is as good as with typical chemical 

toner shown in Fig. 2. On the other hand, the dot images 

with toners C and E are worse because of toner scatter. 

These results indicate that small and narrow size distribution 

is necessary for fine image quality. The reason why toners C 

and E were not pulverized successfully is discussed below. 

Difference between Toners C and D 

In general, as toner particle size in the pulverizer becomes 

smaller, the adhesion force between particles increases; small 

toner particles tend to agglomerate. Such agglomerated particles 

cannot be given pulverizing energy efficiently, and it is 

difficult to pulverize toner to small size, as was observed for 

toner C. But in the case of toner D, silica is present in the 



Figure 9. Effect of the preconditioning silica. 

Figure 11. Difficulty of pulverizing in case of large domain of wax. 

Figure 10. Wax dispersion in toners D and E. 

pulverizer, and the silica decreases the cohesive force of the 

toner particles (Figure 9). Such dispersed particles can be 

pulverized efficiently, and the efficiency of classifying is 

also improved. Thus, the preconditioning method is suitable 

for pulverizing toner to a small and narrow size 

distribution. 

Difference between Toners D and E 

Figure 10 shows wax dispersion in toners D and E. Wax 

domain size in toner E is much bigger than in toner D 

because of the difference of kneader, as described above. It is 

thought that this difference of wax dispersion also affects 

pulverizing efficiency. Figure 11 illustrates the influence of 

wax domain size on pulverizing. When the wax domain is 

small, the effect of preconditioned silica is sufficient. On the 

other hand, in the case of large wax domains, such as toner 

E, the toner chip tends to be pulverized at the surface of a 

wax domain. Consequently, the large domains of wax exist at 

toner surface and cause decreasing flowability and sticking 

to the pulverizer, even if silica preconditioning is employed. 

Thus, such large domains of wax disrupt efficient pulverizing. 

These results indicate that both preconditioning and 

fine wax dispersion are necessary to pulverize toner to small 

size with narrow distribution. 

Effect of Silica Preconditioning 

Toner pulverized by the preconditioning method already has 

silica added during the pulverizing process. Thus, the preconditioning 

method is a technique of both pulverizing and 

surface treatment. The properties of silica added by the preconditioning 

method are analyzed by a particle analyzer 

Figure 12. Comparison of silica dispersion of toner surface. 

(Horiba). Figure 12 shows a comparison of the state of silica 

on the toner surface. Silica dispersion of toner added by the 

preconditioning method is better than conventional silica 

dispersion by addition in the mixer, and free silica rate is 

less. It is supposed that silica added in the pulverizing process 

is strongly bound to the particles, and extra free silica 

agglomeration is avoided or eliminated on classification. In 

general, free silica agglomeration is a cause of filming; thus, 

toner made by the preconditioning method has good 

durability. 

In summary, toner pulverized by the preconditioning 

method has a small and narrow size distribution that is suitable 

for fine image quality and also has good silica dispersion 

on the surface for uniform charging ability and good 

durability. Accordingly, we named the preconditioning 

method “MechanoChemical” technology, and we call toner 

made by mechanochemical technology, “MC toner.” 



(Figure 14). Transfer efficiency E of this potatolike shape 

toner 4.5 m was 5.2. 

CONCLUSIONS 

The present investigation leads to the following conclusions: 

Figure 13. Dependency of circularity on toner particle size. 

Figure 14. SEM photograph of small-size toner made by preconditioning 

method. 

Tendency of Toner Shape on Particle Size 

Figure 13 shows circularity of toner pulverized by the preconditioning 

method. Circularity increases as toner particles 

become small. This indicates that small particle size toner 

tends to be rounded by frequent collision in the pulverizer 

1. To achieve both oilless fusability and high durability, 

fine wax dispersion is necessary. An open-roll-type 

kneader is suitable for kneading the requisite 

amount of wax because low kneading temperature is 

possible. 

2. To pulverize toner to small size with narrow distribution, 

both preconditioning and fine wax dispersion 

are necessary. 

3. Distribution of silica added to toner surface by preconditioning 

is uniform, and the resulting toner has 

less free-silica agglomerations. 

4. Shape of toner pulverized to small size become 

rounded because of frequent collisions in the 

pulverizer. 

REFERENCES 

1 Y. Matsumura, P. Gurns, and T. Fuchiwaki, Proc. IS&T’s NIP17 (IS&T, 

Springfield, VA, 2001) p. 341. 

2 M. Yamazaki, M. Uchiyama, and K. Tanigawa, Proc. of Japan Hardcopy 

2000 Fall Meeting (Imaging Soc. Japan, Tokyo, 2000) p. 1. 

3 C. Suzuki, M. Takagi, S. Inoue, T. Ishiyama, H. Ishida, and T. Aoki, Proc. 

IS&T’s NIP19 (IS&T, Springfield, VA, 2003) p. 134. 

4 T. Aoki, Proc. IS&T’s NIP19 (IS&T, Springfield, VA, 2003) p. 2. 

5 A. Eida, S. Omatsu, and J. Shimizu, Proc. IS&T’s NIP20 (IS&T, 


6 E. Shirai, K. Aoki, and M. Maruta, Proc. IS&T’s NIP18 (IS&T, 


7 E. Shirai, K. Aoki, and M. Maruta, Proc. IS&T’s NIP19 (IS&T, 


8 T. Kubo, E. Shirai, and K. Aoki, Proc. IS&T’s NIP20 (IS&T, Springfield, 

VA, 2004) p. 73. 

9 A. Eida and J. Shimizu, Proc. IS&T’s NIP16 (IS&T, Springfield, VA, 2000) 

p. 618. 

10 J. Shimizu, S. Omatsu, and Y. Hidaka, Proc. IS&T’s NIP19 (IS&T, 





Preparation of Microemulsion Based Disperse Dye Inks 

for Thermal Bubble Ink Jet Printing 

S. Y. Peggy Chang and Y. C. Chao 

Institute of Organic and Polymeric Materials, National Taipei University of Technology, No. 1, Sec. 3, 

Chung-Hsiao E Rd., Taipei, 106, Taiwan 

E-mail: ycchao@ntut.edu.tw 

Abstract. This study describes the preparation of microemulsion 

inks based on commercially available disperse dyes for thermal 

bubble ink jet printing. The approach to make the inks is by formulating 

to form water-soluble emulsions (o/w), which were then optimized 

to reach a dynamically stable isotropic condition. The dyes in 

three primary colors—cyan, magenta, and yellow—were systematically 

investigated. Different species and amounts of dispersants, 

emulsifiers, and dyes were selected to prepare microemulsions after 

intense stirring. The system, D50/963H/ Pannox140/glycerin/water, 

performed excellently in particle size reduction and size stabilization. 

The surface tension, pH value, viscosity, and stability of the 

microemulsion inks meet the requirement of the applied printhead to 

provide both good printing quality and excellent printing consistency 

without clogging in the nozzle. The interaction between dyes and the 

microemulsion system can be comprehensively understood based 

on the results of particle size analysis, lightfastness, water fastness, 

and the characterization of printing quality. The low particle sizereducing 

efficiency can be attributed to the compatibility between 

dye and dispersant structures as well as particle aggregation. The 

lightfastness could be enhanced with smaller particle size but decreased 

with dye content. However, the water fastness of the preformed 

sample was identically high without being affected by the 

dye particle size in the range between 80 nm and 96 nm. The printing 

performance was found to be closely correlated with the dye 

structure and ink concentration. © 2007 Society for Imaging Science 



INTRODUCTION 

With superiority in cost and manufacturing technique, the 

thermal bubble ink jet technology shares a valuable ink jet 

printing market with the piezo ink jet. To formulate water 

miscible inks with water insoluble colorants, the dyes or pigments 

must be processed via microparticle reduction and 

intensive dispersing to form stable jet inks. The present article 

focuses on applying microemulsion technology to render 

the ink of disperse dye base a homogeneous and thermally 

stable system for the thermal bubble printhead. As 

previously defined, 1 the microemulsion is formed by mixing 

oil phase, surfactant, cosurfactant, and water to provide a 

large oil-and-water interfacial area and low viscosity. Different 

surfactants and composition determine the microscopic 

structure of oil-in-water droplets, bicontinuous systems, and 

 

Present address: 10F, No. 3, Linsen North Road, Taipei City, Taipei, 

Taiwan. E-mail: sychang168@gmail.com 

Received Jan. 16, 2007; accepted for publication Mar. 29, 2007. 

1062-3701/2007/515/413/6/$20.00. 

water-in-oil droplets. These small structures enable not only 

the reaction of oil-soluble phase and water-soluble phase but 

also the synthesis of ceramic pigments 2 as well as phase 

transfer catalysis. 3 Both categories of normal microemulsion 

and reverse microemulsion have been reported to have very 

low water solubility, and consequently, low solid content in 

the preparation of pigment containing inks, 4 although the 

nanoparticles were produced with narrow distribution and 

no agglomeration. That is because the microemulsion phase 

transforms to another state when the water amount exceeds 

a certain level. An effective ink, comprising high solid content 

and merits of microemulsions, should be formulated 

with a choice of a microemulsion or reverse microemulsion 

system, optimization of preparation conditions, and application 

of bicontinuous structure. However, most prior research 

work 5–10 focused only on the preparation and characteristics 

of well-dispersed nanoparticles by the reverse micoremulsion 

method, without experiments on practical application 

in the ink jet printer. Moreover, comparatively little information 

has been published on how ink jet printability of dyestuffs 

correlate with dyestuff structure and molecular weight, 

ink composition, concentration, and solvent used. Only by 

systematically exploring inherent possibilities and limitations 

of the technique can strategies be envisioned for printing 

with a large number of different compounds in future. 11 

In order to prepare an isotropic and highly disperse dye 

containing ink jet ink, this work first optimized the dye dispersion 

formulation to obtain well-dispersed and storage 

stable preink, followed by studying effects of dye structure 

on the microemulsion region in the quasi-ternary phase diagram. 

The particle site in extremely purified disperse dye 

presscake should be reduced, and the dye well suspended 

and free of impurities to be capably incorporated into a 

microemulsion. Then, conditions for the maximization of 

water-solubility were identified as the basis for the ink 

preparation. Precisely speaking, the microemulsion of disperse 

dyes in this work was prepared by forming droplets of 

water, containing dye, suspended and stabilized in an appropriate 

mixture of cosurfactant and cosolvent. Finally, the 

preferred ink physicochemical properties were determined 

according to printability tests. In addition to the effect of dye 

structure, the particle size of dyes and the concentration of 

inks influenced printing consistency and the light stability. 

413

Chang and Chao: Preparation of microemulsion based disperse dye inks for thermal bubble ink jet printing 

Table I. Preparations and particle sizes of stabilized disperse dye dispersions dye content: 20%. 

EXPERIMENTAL 

Preparation of Dispersion of Disperse Dye 

Disperse dyes in the form of presscake with high purity is a 

prerequisite in the ink formulation to ensure an accurate 

reaction system unaffected by side products of the synthesized 

dye. The dyes applicable for ink jet ink should be exhibited 

with neutral color shade and demonstrated dyeing 

fastness. In the interest of our investigation, we selected the 

dyes resembling the process colors in commercial ink jet 

printing and analyzed the effects of different dyes on the ink 

formulation. We selected C.I. Disperse Blue 60 (bright 

greenish blue, C.I. Constitution No. 61104, Widetex, Taiwan) 

for cyan by comparing to C.I. Disperse Blue 79 (dark reddish 

blue, C.I. Constitution No. 11345, Widetex, Taiwan), 

C.I. Disperse Red 60 (bright bluish red, C.I. Constitution 

No. 60756, Widetex, Taiwan) for magenta in contrast with 

C.I. Disperse Red 109 (bright yellowish red, C.I. Constitution 

No. 11192, Widetex, Taiwan), and C.I. Disperse Yellow 

77 (bright greenish yellow, C.I. Constitution No. 70150, 

Widetex, Taiwan) for yellow by analogy with C.I. Disperse 

Yellow 99 (bright greenish yellow, C.I. Constitution No. 

48420, Widetex, Taiwan). A combination of dispersants and 

wetting agents applicable for disperse dyes in ink formulation 

were divided into two systems of alkyl polynaphthalene 

formaldehyde sulfonate (D50, Pleasant and Best Chemical 

Co.) with the sodium salt of di-butyl naphthalene sulfonic 

acid (MT 830L, Pleasant and Best Chemical Co.), and maleic 

anhydride diisobutylene copolymer (PT 245L, Pleasant and 

Best Chemical Co.) with Triton X-100 (PerkinElmer). Those 

disperse dye dispersions with particles of small size, i.e., precursors 

of microemulsion, were prepared by grinding with a 

planetary micromill (Pulverisette 7, Fritsch) where zirconium 

oxide grinding bowls and 0.5 mm beadswereadopted 

to effectively degrade particle size. Each sample was ground 

at 400 rpm for different times to obtain the derived particle 

size. Then, the dispersion liquid was filtered to remove contamination, 

such as particles derived from unwanted wear of 

the grinding elements, and finally stabilized with an 

antiprecipitating agent, NL428 (Jintex Co., Taiwan). Preferred 

dye dispersing methods are listed in Table I, where 

systems have been optimized to achieve extraordinary stability 

when particle size dropped below 200 nm. Although 

such preink obtained was well dispersed and stable, its physicochemical 

properties did not meet the demands of ink jet 

printing and immediately clogged printing nozzles because 

of high dye content (20%), high surface tension 

50–60 dyne/cm, and high viscosity 7–15 cps, whereas 

the relative values for thermal bubble ink jet inks were 

3–10% of dye content, 20–40 dyne/cm surface tension, 



Table II. Composition of disperse dye microemulsion microemulsion system: Sinopol 

80:Sinopol 963H:Pannox 140:Glycerin:Deionized water=2.4%:4.8%:7.2%: 

26.4%:49.2%–54.2%. 

Ink 

Number 

Maximum 

Amount of 

Dye in 

Microemulsion 

% 

Balanced 

Amount 

of Deionized Water 

% 

Particle 

Size 

nm 

Particle Size after 

Stability test 

three cycles of 5°C for 

3 h and 70°C for 3 h 

nm 

B60-am 8.0 50.2 a 80.9 80.1 

B60-bm 5.0 53.2 a 129.7 131.4 

B79-am 6.0 52.2 a 88.6 88.1 

B79-bm 6.0 52.2 a 128.3 130.2 

R60-am 10.0 49.2 49.5 48.5 

R60-bm 8.0 51.2 89.5 91.6 

R109-am 7.0 52.2 86.2 86.8 

R109-bm 6.0 53.2 96.5 98.4 

Y77-am 5.0 54.2 97.5 99.3 

Y77-bm 10.0 49.2 99.6 98.7 

Y99-am 8.0 51.2 83.8 84.2 

Y99-bm 5.0 54.2 102.2 104.5 

a The addition of 1% diethylene glycol monoethyl ether. 

and 2–5 cps viscosity. The results of surface tension and 

viscosity were shown in Table I. 

Determination of Microemulsion System 

The microemulsions of disperse dyes were obtained by 

intensively mixing the dye dispersion with an emulsifier 

(Sinopol 963H, polyoxyethylene nonyl phenyl ether, 

HLB=12.0, Sino-Japan Chemical), a coemulsifier (Pannox 

140, nonyl phenol polyethylene glycol ether, HLB=17.7, Pan 

Asia Chemical), a cosolvent (glycerin, Ferak), sorbitan oleate 

(Sinopol 80, HLB=4.3, Sino-Japan Chemical), and deionized 

water. The mixture was homogenized (1500 rpm for 

10 min) to disperse the dye into very small emulsion droplets, 

whereby the microemulsion spontaneously formed. The 

system achieved its stability due to unique size compatibility 

and adhesion of the dye particles inside the oil phase of the 

microemulsion. Microemulsions of magenta and yellow dyes 

were thermodynamically stable for an indefinite period of 

time and an survive freezing and thawing cycles. However, 

the cyan dye samples were observed precipitating in the storage 

stability test (three cycles of 5°C for 3hand 70°C for 

3h) whereas others did not. Despite the stabilization of 

microemulsion structure by coemulsifier, hydrotropic 

amphiphiles (Diethylene glycol monoethyl ether, Riedel–de 

Haen AG Seelze–Hannover) were needed for cyan dyes to 

surround each microemulsion droplet in order to prevent 

the microscopic droplets from coalescing. In order to determine 

the maximum amount of water content for each system, 

the quasiternary phase diagram was used to establish 

the largest microemulsion regime, where the ratios of emulsifier, 

coemulsifier, and cosolvent formed the upper region of 

the boundary between the total mass of dye phase, water 

Figure 1. Good dispersion of B79-a and R109-b in comparison to 

B79-b under an electronic microscope Scope: 100.25. 

phase, and emulsifiers. The optimized compositions for each 

dye microemulsion are summarized in Table II. 

Preparation of Disperse Dye Inks 

The dye microemulsions met the viscosity requirements for 

thermal ink jet printing, i.e., 2–6 cps., measured by Brookfield, 

LVDV-III. However, other additives, e.g., fungicide and 

buffer solution (ammonium solution), employed in ink to 

optimize ink performance, decreased the viscosity without 

affecting other properties of the microemulsion. In general, 

the lower the viscosity the greater the velocity and amount of 

fluid propelled forward, which usually leads to the formation 

of long tails behind the head of the drop. Glycerin was therefore 

added additionally about 0.5–2 wt. % to restore the 

viscosity and optimize printing consistency. The characteristics 

of the inks were determined using a Horbita pH meter 

for pH measurement, Face CBVP-A3 for surface tension 

measurement, and ZetaPlus particle analyzer for particle 

size evaluation. Phase separation was determined by visual 

observation. The printing quality of line sharpness, edge 

acuity, successiveness, and nozzle clogging was evaluated by 

printing on a Lexmark Z43 printer on paper. The lightfastness 

of inks printed on fabric (100% polyester) was evaluated 

according to AATCC16-1998A. 


Effects of Disperse Dye on Dispersion Preparation 

The effect of different disperse dyes on particle size were 

displayed in Table I. The dispersion formulations with D50/ 

MT830L have revealed smaller particle size with shorter 

grinding time except for Y77-a. The samples thereof, B79-a 

and R109-a, achieved average particle size of 105 nm, and 

moreover, R60-a achieved the nanoparticle size of 53.2 nm, 

which might be due to D50, polynaphthalene formaldehyde 

sulfonate, having better wetting and deagglomeration than 

PT245L, maleic anhydride copolymer, under the mechanical 

force of micromilling. B79-b presented the largest particle 

size with agglomerates, indicating an unsuitable combination 

with PT245L/Triton X-100 (Figure 1), and then further 

failed to form microemulsion. By comparing B79-b to 

R109-b, more alkyl groups in the diazo structure decreased 

the efficiency of particle size reduction and consequently 

required more time or more active surfactants to accomplish 

the desired result. Among the systems of PT245L/Triton 

X-100, only the structure of Y77-b had more satisfactory 

compatibility with the copolymer allowing reduction in particle 

size to 106.5 nm in 120 h. The slight difference in 



Table III. The physicochemical properties of microemulsion based disperse dye inks. 

Ink- 

Dye Content 

% 

Surface 

Tension 

dyne/cm 

Viscosity 

cps 

pH 

Value 

Particle 

size 

nm 

Lightfastness 

AATCC16- 

1998A 

Printabiliy 

h 

B60-am-4 34.9 4.28 7.7 82.6 4 8 

B60-am-8 33.7 4.58 7.6 80.8 5 8 

R109-am-3.5 36.0 3.51 7.7 90.2 4 8 

R109-am-7 41.2 3.57 7.6 84.4 5 6 

Y77-bm-5 41.4 5.80 7.3 90.7 5 8 

Y77-bm-10 41.1 3.96 7.3 95.5 4 6 

Y99-am-5 41.5 4.50 7.5 82.9 4 8 

Y99-am-8 40.6 4.43 7.6 83.2 4 8 

Figure 2. Quasi-ternary phase diagrams of the microemulsion systems, 

where E+C+S represents the total mass of emulsifier, coemulsifier and 

cosolvent, O is the mass of oil phase Sinopol 80, andWisthemassof 

water. Lines 1–5 represent the ratio of emulsifier to coemulsifier and 

cosolvent in mass 2:3:11, 2:1:11, 3:2:11, 2:3:7, and 2:3:13. I is 

the system of Sinopol 963H HLB=12.0/Pannox 140 

HLB=17.7/Glycerin. II is the system of Sinopol 960H 

HLB=10.0/Pannox 119 HLB=15.8/Glycerin. III is the system of 

Sinopol 966H HLB=14.1/Pannox 150 HLB=18.0/Glycerin. 

particle-reducing results between the two dispersion systems 

at both 120 h and 168 h occurred in samples of C.I. Disperse 

Blue 60 and C.I. Disperse Red 109 probably due to 

lower molecular weight and reduced steric hindrance of the 

dye structures. The antiprecipitating additive, NL428 was 

found to successfully prevent reaggregation in both dispersing 

systems. 

Determination of Disperse Dye Microemulsion System 

The microemulsion system of Sorbitan trioleate/ 

polyoxyethylene nonyl phenyl ether/nonylphenol polyethylene 

glycol ether/glycerin/water had been chosen by comparing 

to other compositions in the quasi-ternary phase 

diagrams at 25°C, shown in Figure 2. The experimental results 

shown in Table II indicated that the system had good 

compatibility with most dispersing systems and could be 

optimized to prepare the disperse dye inks of different colors. 

Figure 2 demonstrated that the microemulsion could be 

formed if the HLB values of emulsifiers lay between 10.0 and 

18.0. The microemulsion area, formed in the upper region 

of the boundary line, depends on the ratios of emulsifier to 

coemulsifier and cosolvent. A maximum ratio was found as 

2:3:11 and decreased in the following sequences: 2:1:11, 

3:2:11, 2:3:7, and 2:3:13. When the cosolvent was the ratio of 

11, additional emulsifier (emulsifier:coemulsifier2:3) enlarged 

the microemulsion area more than those 

(emulsifier:coemulsifier2:1) with the three different HLB 

value combination systems. However, more emulsifier of 

HLB value 14.1 than coemulsifier of HLB value 18.0 was 

required in system (III) to form the largest microemulsion 

area. It could be inferred that the emulsifiers of HLB values 

of 10.0 and 12.0 had more affinity to the oil phase and 

emulsified effectively at the lower ratio of 2 than the emulsifiers 

of HLB value 14.1, but required more coemulsifier of 

HLB value 15.8 to form the largest microemulsion area, i.e., 

maximum water dissolving capacity. When the ratio of 

emulsifier to coemulsifier was accordingly set at 2:3, the 

cosolvent at ratio of 13 increased water solubility more than 

at the ratio of 7, especially in the system (I), where the HLB 

value was 12.0 for emulsifier and 17.7 for coemulsifier. Nevertheless, 

the cosolvent at the ratio of 11 was most preferred 

to form the maximum water solubility in three systems. 

Relatively, the ratio of 2:3:11 (emulsifier:coemulsifier: 

cosolvent) in system (I) presented the largest microemulsion 

area among other two systems. The Sinopol 963H/Pannox 

140/Sinopol 80/Glycerin microemulsion system had a maximum 

microemulsion area, implying maximum solubility. 

This ratio dependence indicated that the interfacial area generated 

by the emulsion drops can be expanded with the increased 

amount of coemulsifier and cosolvent. The maximum 

dye amount of each dispersion formulation dissolved 

in the given microemulsion system was determined by particle 

size variation 2.5% after the storage stability test, as 

shown in Table II. The resulting microemulsions were referred 

to as disperse dye inks, listed in Table III. In short, the 

composition corresponding to the maximum solubility was 

Dye:Sinopol 80:Sinopol 963H:Pannox 140:Glycerin:Deionized 

water5%:2.4%:4.8%:7.2%:26.4%:54.2%. 

Effects of Disperse Dye on Microemulsion System 

In Table II, the maximum amount of disperse dyes in a given 

microemulsion system was determined as the amount at 

which 2.5% or larger variation of the particle size was detected 

after the storage stability test. The samples of B60-am, 

R60-am, R60-bm, Y77-bm, and Y99-am could be incorporated 

more effectively inside the microemulsion droplets to 

allow the highest dye content, up to 8–10%. Other samples 

imply the maintenance of at least 5% of the dye, and all the 

required ink properties can also be met. Notably, to avoid 

the dye agglomeration mentioned above, 1% amount of 

amphiphile (diethylene glycol monoethyl ether) was added 

into samples of B60-a, B60-b, B79-a, and B79-b. The particle 



Table IV. The water fastness of microemulsion based disperse dye inks. 

Water-fastness Test AATCC-61-2003-2A 

Figure 3. The printing quality of inks, B60-am-8%, R109-am-7%, and 

Y77-bm-5%, which were ink jet printed with Lexmark Z43 printer. 

size of dispersed dye was further reduced in the microemulsion 

after ball milling. In the storage stability test, the particle 

size of B79-am and R60-am decreased further while the 

R109-am, B60-am, Y77-bm, and Y99-am did not change 

apparently (nearly 0%). It could be concluded from the 

above analyses that when the ratio of Sinopol 963H/Pannox 

140/Sinopol 80/Glycerin was 2:3:1:11, the system of disperse 

dye/D50/MT830L/water maintained good stability when the 

temperature changed from 5°C to 70°C. 

Effects of Ink Concentration on Physicochemical 

Properties 

With the superiority in particle stability, the samples of 

R109-am, B60-am, Y77-bm, and Y99-am were selected for 

further study on the effect of ink concentration. Since the 

process color of light cyan and light magenta are usually 

formulated in half concentration of cyan and magenta, the 

concentration effect on lightfastness, particle size, and printability 

therefore becomes important. In this work, sample 

B60-am-8% was investigated with the concentration of 4%, 

while the sample R109-am-7% is investigated with the concentration 

of 3.5%. Y77-bm-10% and Y99-am-8% are performed 

with the lowest dye content of 5% in microemulsion 

according to the above analyses. To be applicable as inks for 

the thermal bubble ink jet printing, each dye microemulsion 

was formulated with the added fungicide, the buffer solution 

to stabilize pH value of the system, and a little glycerin (the 

cosolvent applied in the above microemulsion system) to 

optimize viscosity. The above agents could all be incorporated 

into formulations without disrupting microemulsion 

stability. The inks for thermal ink jet printing were produced 

with acceptable physicochemical properties, summarized in 

Table III. The surface tension and viscosity of each sample 

met the demands of ink drop performance, to enable efflux 

through the capillary tube of the printer nozzle on to be able 

to print continuously and present good printing quality, 

sharp line, and accurate edge, either on paper or polyester 

fabric, without nozzle clogging and edge feathering (Figure 

3). However, the inks of R109-am-7% and Y77-bm-10% had 

printing continuity of 8 h, but it can be extended much 

longer when the dye content was reduced, e.g., R109-am- 

3.5% and Y77-bm-5%. By comparing the particle size to 

other samples, the structure or formulation of R109-am-7% 

would not tolerate the temperature or vibration variation 

inside the thermal bubble printhead. The failure of Y77-bm- 

10%, however, could be attributed to the largest particle size 

causing inferior printability. The samples of B60-am-4% and 

R109-am-3.5% exhibited an increase in particle size and a 

slight decrease in lightfastness when the concentration was 

Ink- 

Dye Content 

% 

Color 

Change 

Staining on 

Cotton Fiber 

Staining 

on Acrylic 

Fiber 

B60-am-4 5 5 5 

B60-am-8 5 5 5 

R109-am-3.5 5 5 5 

R109-am-7 5 5 5 

Y77-bm-5 5 5 5 

Y77-bm-10 5 5 5 

Y99-am-5 5 5 5 

Y99-am-8 5 5 5 

diluted in half. This implies that excessive amount of surfactant 

and solvent breaks up the microemulsion system and 

induces the primary particles to generate larger particles, i.e., 

agglomerates. Moreover, the greater amount of small dye 

particles on the medium surface would reflect more light 

and protect prevent the printed fabric from the effects of 

light exposure. Such particle effects are also seen in the yellow 

samples, although the differences are insignificant. The 

anthraquinone dyes performed better in lightfastness than 

the diazo ones, and their lightfastness grade can reach 5 by 

decreasing the particle size. No formulation in Table III performs 

less than grade 4, which demonstrates that all the 

formulated inks meet the requirement of light stability for 

thermal bubble ink jet printing. 

The water fastness of each sample in Table IV was identically 

at grade 5, according to AATCC-61-2003-2A. There 

was no color change in shade and no staining on cotton and 

acrylic fiber. The cotton and acrylic fiber are usually easily 

stained by all kinds of dyes in the water-fastness test. It could 

be inferred that the microemulsion ink drops containing 

dyes (from 3.5% to 10%) were jetted on demand by the 

thermal bubble ink jet nozzle and fixed well on the polyester 

filament. Excessive dye that could be carried away by water 

did not occur on the fiber surface. Moreover, those dye particle 

sizes in the range between 80 nm and 96 nm did not 

affect the fastness to the fiber in water but slightly in the 

light. The high water fastness is thus another advantage of 

the studied microemulsion ink for the application of textile 

ink jet printing. 

In the future, more systematic study of ink rheology in 

the dyestuff ink jet printability will be carried out to investigate 

the relationship between dye structure, ink composition 

and fluid viscoelasticity during ink jet printing. 

CONCLUSIONS 

The microemulsion system of Sinopol 80/Sinopol 963H/ 

Pannox 140/Glycerin/Deionized water was determined to be 

compatible with the best dispersing composition of disperse 

dyes and exhibited excellent behavior in the quasi-ternary 

phase diagrams. It can be concluded that only C.I. Disperse 



Yellow 77 had better compatibility with other ingredients to 

achieve printability of more than 8h and lightfastness of 

grade 5. Moreover, the samples of B60-am, R109-am, Y77- 

bm, and Y99-am in the given microemulsion presented best 

stability without particle size variation in the storage stability 

test. Nevertheless, the samples of B79-am and R60-am 

demonstrated slightly decreased particle size in the 

microemulsion. 

After the microemulsions were formulated into inks, the 

samples of R109-am-7% and Y77-bm-10% could not be 

printed with thermal bubble ink jet for more than 8h, 

which is probably due to poor heat resistance of diazo structure 

or larger particle clogging in the nozzle after the ink has 

been frequently heated in the thermal bubble print head. 

When the ink concentration was diluted in half, B60-am-4% 

and R109-am-3.5% exhibited a particle size increase and 

slight decrease in lightfastness, which is due to the generation 

of agglomerates in the unbalanced system with higher 

amounts of surfactants and solvent. Notably, it was found 

that anthraquinone dyes yield higher lightfastness than diazo 

dyes to attain the lightfastness grade of 5. 

The microemulsion inks of this study had many advantages 

over other inks containing just the dye dispersions. 

One advantage was that microemulsion ink was a 

microheterogeneous system that provided a large interfacial 

area to accommodate dye content up to 8–10% and low 

viscosity to effectively meet the demand of ink drop performance 

without adding other viscosity modifiers. Another 

advantage was that microemulsion inks produced printed 

dot sizes that were nearly independent of the surface properties 

of the medium; they could present good printing quality 

either on paper or 100% polyester fabric without nozzle 

clogging and edge feathering. The most important advantage 

was that dyes in the given microemulsion exhibited smaller 

particle size than in dispersion formulations and, thereby, 

good storage stability with very little particle size variation. 


The authors thank S. P. Rwei in our department for helpful 

discussions. The authors also thank Widetex, Pleasant and 

Best Chemical, Jintex Co., Pan Asia Chemical, and Sino- 

Japan Chemical for offering materials. Nevertheless, we 

gratefully acknowledge Taiwan Textile Research Institute for 

support. 

REFERENCES 

1 K. J. Lissant, Emulsions and Emulsion Tech., Part III (Marcel Dekker, 

New York, 1984) pp. 140–162. 

2 A. García, M. Llusar, S. Sorli, J. Calbo, M. A. Monros, and G. Page, J. 

Eur. Ceram. Soc. 23, 1829–1838 (2003). 

3 M. Häger et al., Colloids Surf., A 183-185, 247–257 (2001). 

4 R. Guo, H. Qi, Y. Chen, and Z. Yang, Mater. Res. Bull. 38, 1501–1507 

(2003). 

5 C. Mo, M. Zhong, and Q. Zhong, J. Electroanal. Chem. 493, 100–107 

(2000). 

6 K. Aikawa, K. Kaneko, T. Tamura, M. Fujitsu, and K. Ohbu, Colloids 

Surf., A 150, 95–104 (1999). 

7 R. Guo, H. Qi, Y. Chen, and Z. Yang, Int. J. Inorg. Mater. 18, 645–652 

(2003). 

8 C. A. Miller, R. N. Huan, W. J. Benton, and T. Fort, Jr., J. Colloid 

Interface Sci. 61, 554–568 (1977). 

9 V. K. Bansal, D. O. Shah, and J. P. O’Connell, J. Colloid Interface Sci. 75, 

462–475 (1980). 

10 R. Guo, H. Qi, Z. Yang, and Y. Chen, Ceram. Int. 30, 2259–2267 (2004). 

11 B. D. Gans, P. C. Duineveld, and U. S. Schubert, Adv. Mater. (Weinheim, 

Ger.) 16, 203–213 (2004). 




Effects of Molecular Substituents of Copper 

Phthalocyanine Dyes on Ozone Fading 

Fariza B. Hasan, Michael P. Filosa and Zbigniew J. Hinz 

ZINK Imaging Inc., Waltham, Massachusetts 02451 

E-mail: fariza.hasan@zink.com 

Abstract. Copper phthalocyanine dyes, widely used in various imaging 

systems, are susceptible to fading under ambient conditions. 

One of the main factors responsible for such fading is the presence 

of ozone. Addition of suitable antiozonants has been shown to be 

effective in improving ozone stability. Although the exact mechanism 

of such stabilization is not fully understood, the electronic structures 

of the additives have shown to have significant impact on their effectiveness. 

In order to increase the ozone stability of copper phthalocyanine 

dyes without any additives, several copper phthalocyanine 

dyes containing substituents of varying electronic structures 

were synthesized and tested for ozone stability. The structure of a 

copper phthalocyanine dye with significantly improved stability to 

ozone is described in this paper. Possible mechanisms leading to 

such stability are also discussed. © 2007 Society for Imaging Science 



INTRODUCTION 

Copper phthalocyanine CuPC dyes are extensively used in 

various imaging systems because of several advantages, including 

high spectral absorptivity, solubility in various common 

solvents, and relatively greater lightfastness compared 

to several other classes of dyes. However, CuPC dyes are also 

susceptible to ozone fading under various ambient conditions, 

which can cause deterioration of images. A large number 

of papers related to the effects of ozone exposure on 

image stability have been published. Some of the recent publications 

are referred to in this article. 1–11 

Several types of materials commonly used as antiozonants 

in other industries, such as p-phenylenediamines for 

reducing degradation of rubber due to exposure to ozone, 

are not suitable for imaging systems because of the highly 

colored products formed due to their reactions with ozone. 

These reactions have been shown to proceed through electron 

transfer mechanisms. 12 Other methods used for minimizing 

degradation of rubber due to exposure to ozone, 

such as coating with impermeable materials, are not easily 

applicable to imaging products, except in a limited number 

of imaging systems where polymeric materials can be transferred 

over the printed images. However, in such systems the 

efficiency of these materials depends on the continuity of the 

transferred polymeric layer, and any defect present in this 

Received Jan. 11, 2007; accepted for publication Mar. 1, 2007. 

1062-3701/2007/515/419/5/$20.00. 

layer would allow ozone to diffuse through and cause degradation 

of the dyes in images. 

Several aminoanthraquinone dyes having similar functional 

groups as p-phenylenediamines are efficient antiozonants, 

but they do not form highly colored reaction products, 

due to the presence of electron withdrawing carbonyl 

groups in the fused ring systems. 13,14 Aminoanthraquinone 

dyes, as well as p-phenylenediamines, are essentially ozone 

scavengers and thus reduce the availability of ozone to react 

with CuPC dyes. It is expected that the presence of molecular 

substituents, which can act as ozone scavengers would 

also render the CuPC dye molecules more resistant to fading 

and discoloration due to exposure to ozone. In order to test 

this hypothesis, and since the reaction of olefins with ozone 

is well known, a CuPC dye with allyl substituents was synthesized. 

To verify the mechanism, two other CuPC dyes 

with different substituents were also synthesized and tested. 

EXPERIMENTAL PROCEDURES 

Substituted Copper Phthalocyanine Dyes 

The structures of the synthesized CuPC dyes with varying 

substituents, labeled as CuPC-B, CuPC-C, and CuPC-D, are 

shown in Figure 1. A commercially available CuPC dye, Solvent 

Blue 70 (CAS No. 12237-24-0) was used as the control 

for the tests. Although the exact structure of this dye is not 

known, it is possibly a mixture of several isomers containing 

various ring substituents, and can be represented as 

CuPC-A. 

Synthesis of Copper Phthalocyanine Dyes 

The synthetic method used for the substituted copper phthalocyanine 

dyes is described in Scheme 1. 15 Copper 

phthalocyaninetetrasulfonic acid tetrasodium salt (Aldrich), 

containing four −SO 3 Na groups as ring substituents was 

suspended in a 4:1 mixture of sulfolane and 

dimethylacetamide. A 32-fold excess of phosphorus oxychloride 

was added, and the reaction mixture was allowed to 

reflux for 4h, after which the reaction mixture was cooled to 

room temperature and poured into rapidly stirring ice water. 

The blue precipitated solid was collected, washed with water, 

and dried in a vacuum desiccator at room temperature for 

24 h. The crude wet cake, with some water present, was 

dissolved/suspended in methylene chloride and cooled to 

4°C. A large excess of diallyl amine (for CuPC-B), 

dibutylamine (for CuPC-C) or diethoxyethylamine (for 

419

Hasan, Filosa, and Hinz: Effects of molecular substituents of copper phthalocyanine dyes on ozone fading 

without any polymeric binder from a donor sheet to a porous 

receiver. 16 The material used as donor sheets for these 

experiments was 4.5 m thick PET with backcoat containing 

lubricant suitable for thermal printing. The receiver consisted 

of a coating of fumed silica with a hydrophilic polymeric 

binder system on a polyester base. The coverage of the 

dyes coated on donor sheets was maintained constant 

285 mg/m 2 . No protective coating was transferred over 

the images. Standard-type thermal printheads for monochrome 

prints were used for transferring the images, at 

maximum energy of 2.5 J/cm 2 . 

Figure 1. Structures of copper phthalocyanine dyes tested. 

Scheme 1. 

CuPC-D) was slowly added to the cooled solution and then 

allowed to warm to room temperature and stirred for 16 h. 

The crude reaction mixture was placed onto a plug of dry of 

silica gel and eluted with methylene chloride. A greenish 

colored material came off and was discarded. The silica with 

the product adhering to it was sucked down until nearly dry. 

The elution was continued with a mixture of ethyl acetate/ 

hexanes, 1:1 by volume. This process removed a dark brown 

impurity from the product. The eluent mixture was changed 

to one containing 4:1 ethyl acetate/hexanes. This fraction 

contained the desired copper phthalocyanine dye. 

Imaging Method 

Printed images of the dyes were generated by thermally 

transferring the coated dyes and suitable thermal solvents, 

Spectra of Copper Phthalocyanine Dyes 

Absorption spectra of the dyes in solution were obtained, 

after dissolving the dyes in methylene chloride, using a 

HP8452A diode array spectrophotometer. The reflection 

spectra of the dyes were measured at the D max regions of 

printed images of the dyes, using a Gretag SPM500 densitometer. 

The conditions for the measurements were: 

illuminationD50, observer angle2°, density 

standardANSI A, reflection standardwhite base, and no 

filter. 

Ozone Stability Test Method 

To evaluate the ozone stability of the synthesized CuPC dyes 

containing various substituents, as well as the commercially 

available CuPC dye, the D max regions of the transferred images 

were exposed in an ozone chamber constructed from a 

Pyrex jar 1.2 ft 3 and a mercury-argon lamp. 17 Ozone was 

produced in situ by the direct photolysis of oxygen in the 

ambient air within the chamber. A fan within the chamber 

ensured that all samples were uniformly exposed. The temperature 

in the ozone chamber was between 21°C and 23°C 

and relative humidity was 47–50%. The images were exposed 

to ozone for definite periods of time. For each set of 

experiments, a “control” image, which was an image of the 

control commercially available dye, CuPC-A with cyan reflection 

density of very close to unity, was exposed for 1h 

with the tests samples. A comparison of the extent of change 

of each of these “control” images was used as a method to 

calibrate the effective ozone concentration in the chamber 

for each experiment. However, during the short period of 

time in which these experiments were conducted, the 

changes of the control images were practically identical, 

which indicated that the concentration of ozone in the 

chamber during these experiments remained unchanged. 

Since the aim of these experiments was to compare the 

ozone resistance of the CuPC dyes under identical conditions, 

it was not necessary to measure the exact concentration 

of ozone in the chamber. The ozone stability of each of 

these dyes was compared by overall spectral changes and 

quantified by the changes of the reflection densities (cyan, 

magenta, and yellow) and colorimetric parameters (L * , a * , 

and b * ) of the images after ozone exposure, using a Gretag 

SPM500 densitometer. The conditions for the measurements 

are as described in the previous paragraph. 



Figure 3. Comparison of spectra of thermally transferred images of 

CuPC dyes. 

Figure 4. Ozone induced spectral changes on thermally transferred images 

of CuPC-A and CuPC-B. 

Figure 2. Comparison of spectra of CuPC dyes in methylene chloride 

solution. 


Figure 2 shows the absorption spectra of the commercially 

available and the three synthesized copper phthalocyanine 

dyes in methylene chloride. Although the solution spectra of 

these dyes are not identical, they are similar to each other. 

These results indicate that the electronic structures of the 

substituents have minor or insignificant effects on the chromophore. 

However, a comparison of the reflection spectra of 

the transferred images of the three dyes, shown in Figure 3, 

indicates significant differences between the spectra of the 

dyes. Although all three dyes show blueshifts of the transferred 

images compared to their spectra in solution, which is 

an indication of H-aggregation, CuPC-B shows greater shift 

than the other two synthesized dyes. It is likely that the 

planar allyl groups facilitate the stacking of the dye molecules 

more than the nonplanar butyl and diethoxyethyl 

groups, causing such aggregation. 

Figure 4 shows the reflection spectra of images of 

CuPC-A and CuPC-B, before and after exposure to ozone 

for 1h. The control CuPC-A image showed large changes, 

including a significant decrease of cyan density and increase 

of yellow density, which is due to the chemical degradation 

of the chromophore by ozone. The CuPC-B image after exposure 

to ozone under identical conditions remained almost 

unchanged. In order to determine if the observed ozone stability 

of CuPC-B is due to the steric hindrance caused by the 

allyl groups on the reaction of ozone with the chromophore, 

two other dyes, CuPC-C and CuPC-D, in which the allyl 

groups were replaced by butyl and diethoxyethyl groups, respectively, 

were also tested. Figures 5 and 6 show larger spectral 

changes for both CuPC-C and CuPC-D, compared to 

the changes for CuPC-B. These results indicate that the 

higher stability of CuPC-B to ozone is unlikely to be due to 

simple steric hindrance. 

The changes in cyan (C), magenta (M), and yellow (Y) 

densities, and L * , a * , and b * values of images from each of 

the four dyes after exposure to ozone are listed in Tables I 

and II, respectively. 



Table I. Effects of ozone exposure on C, M, Y densities of printed images of CuPC dyes 

Dyes Density DC DM DY 

CuPC-A Initial 1.22 0.23 0.10 

After exposure 0.46 0.15 0.26 

Ratio, final/initial % 38 65 260 

CuPC-B Initial 1.20 0.31 0.25 



CuPC-C Initial 1.20 0.30 0.20 


Ratio, final/initia% 68 80 140 

Figure 5. Ozone induced spectral changes in thermally transferred images 

of CuPC-A and CuPC-C. 

CuPC-D Initial 1.15 0.30 0.20 



Table II. Effects of ozone exposure on L * , a * , and b * values of printed images of CuPC 

dyes. 

Dyes L * a * b * 

CuPC-A Initial 68.82 −45.86 −38.88 

Final 81.12 −27.49 −1.29 

Change 12.30 18.37 37.59 

CuPC-B Initial 62.78 −46.42 −30.19 

Final 62.75 −46.43 −29.81 

Change −0.03 −0.01 0.38 

Figure 6. Ozone induced spectral changes in thermally transferred images 

of CuPC-A and CuPC-D. 

CuPC-C Initial 63.73 −45.35 −36.36 

Final 70.45 −41.70 −16.59 

Change 6.72 3.65 19.77 

CuPC-D Initial 64.56 −44.71 −34.32 

Final 70.88 −40.31 −17.96 

Change 6.32 4.40 16.36 

A comparison of gradual cyan density loss with time of 

printed images of CuPC dyes due to exposure to ozone over 

an extended period of time is shown in Figure 7. The 

CuPC-A image shows 15% retention after 8.5 h of exposure, 

whereas the most stable CuPC-B image retains 73% of the 

initial cyan density under identical conditions. The other 

two dyes, CuPC-C and CuPC-D show much less density 

retention than CuPC-B, but are slightly more stable than 

CuPC-A. These results show the same trend in ozone stability 

of the CuPC dyes as observed after 1h of ozone 

exposure. 

The data indicate that the ozone stability of CuPC-B, 

containing allyl substituents, are considerably greater than 

that of the butyl or diethoxyethyl substituted CuPC-C and 

CuPC-D dyes. As discussed before, a comparison of the 

spectra of the four dyes shows that greater extent of 

H-aggregation as indicated by the blue shift in the spectrum 

of CuPC-B, may be associated with the observed ozone stability. 

The spectra of the control CuPC-A dye and the synthesized 

dyes, dissolved in methylene chloride, do not show 

any such difference. It is possible that the planar allyl groups 

facilitate the required rearrangement of the molecules to 

cause such aggregation more than the other substituents and 

result in a more stable form of dye in solid state, as in a 

transferred image. In addition, the allyl groups may also 

react with ozone and act as ozone scavengers, thus protecting 

the chromophore. 

CONCLUSIONS 

Difference in molecular substituents of copper phthalocyanine 

dyes produces large effects on ozone fading of the 

printed images. Images formed with the allyl substituted dye 

show greater ozone stability than those with butyl or 

diethoxyethyl substituted dyes. The allyl substituted dye also 



Figure 7. Rates of cyan density loss due to ozone exposure of thermally 

transferred images of CuPC dyes with varying substituents. 

exhibits a larger extent of H-aggregation and blueshift in 

transferred images than the dyes containing other molecular 

substituents, when compared to their absorption spectra in 

solution. An increase in stability to exposure to ozone is 

likely to be due to the reaction of the allyl substituents with 

ozone, which accordingly act as ozone scavengers. Another 

possible factor may be the ease of formation of H-aggregates 

that are more stable than the monomeric dyes. 

REFERENCES 

1 M. Berger and H. Wilhelm, Proc. IS&T’s NIP19 (IS&T, Springfield, VA, 

2003) pp. 438–443. 

2 M. Thornberry and S. Looman, Proc. IS&T’s NIP19 (IS&T, Springfield, 

VA 2003) pp. 426–430. 

3 K. Kitamura, Y. Oki, H. Kanada, and H. Hayashi, Proc. IS&T’s NIP19 

(IS&T, Springfield, VA, 2003) pp. 415–419. 

4 D. Bugner, R. Van Hanehem, P. Artz, and D. Zaccour, Proc. IS&T’s 

NIP19 (IS&T, Springfield, VA, 2003) pp. 397–401. 

5 J. Geisenberger, K. Saitmacher, H. Macholdt, and H. Menzel, Proc. 

IS&T’s NIP19 (IS&T, Springfield, VA, 2003) pp. 394–395. 

6 S. Wakabayashi, T. Tsutsumi, M. Sakakibara, Y. Nakano, and Y. Hidaka, 

Proc. IS&T’s NIP19 (IS&T, Springfield, VA, 2003) pp. 203–206. 

7 R. A. Barcock and A. J. Lavery, J. Imaging Sci. Technol. 48, 153–159 

(2004). 

8 D. Brignone, D. W. Fontani, and A. Sismondi, Proc. IS&T 13th 

International Symposium on Photofinishing Technology (IS&T, 

Springfield, VA, 2004) pp. 47–50. 

9 C. Halik and S. Biry, Asia Pac. Coatings J. 16, 20–21 (2003). 

10 S. Kiatkamjornwong, K. Rattanakasamsuk, and H. Noguchi, J. Imaging 

Sci. Technol. 47, 149–154 (2003). 

11 A. Kase, H. Temmei, T. Noshita, M. Slagt, and Y. Toda, Proc. IS&T’s 


12 X. Zhang and Q. Zhu, J. Org. Chem. 62, 5934–5938 (1997), and 

references cited therein. 

13 F. B. Hasan and S. E. Rodman, Proc. IS&T’s NIP20 (IS&T, Springfield, 

VA, 2004) pp. 729–733. 

14 F. B. Hasan, J. Imaging Sci. Technol. 49, 667–671 (2005). 

15 US Patent, pending. 

16 US Patent 6,537,410. 

17 Designed by M. A. Young, ZINK Imaging, LLC. 




Relationship between Paper Properties and Fuser Oil 

Uptake in a High-speed Digital Xerographic Printing 

Fuser 

Patricia Lai and Ning Yan 

University of Toronto, 33 Willcocks Street, Toronto, Ontario M5S 3B3, Canada 

E-mail: ning.yan@utoronto.ca 

Gordon Sisler 

Xerox Research Center of Canada, 2660 Speakman Drive, Mississauga, Ontario L5K 2L1, Canada 

Jay Song 

Cincinnati Technology Center, International Paper Company, 6283 Tri-Ridge Blvd, 

Loveland, Ohio 45140-8318 

Abstract. The fuser roll in high-speed xerographic printers is typically 

coated with a silicon-based oil to facilitate clean splitting of the 

fused image and paper as it exits the fusing nip. If oil uptake by 

paper is excessive, the fuser roll will become insufficiently coated 

with oil. Consequently, both image quality and fuser roll life will be 

negatively impacted. A variety of commercial papers were characterized 

and evaluated for their oil uptake performance. Using partial 

least-squares regression, key paper properties correlated with oil 

uptake were identified. Surface energy of paper was found to have 

the strongest correlation with oil uptake. Furthermore, based on 

contact mechanic theories, a contact area index (CI) was used to 

describe the degree of contact between paper and the fuser roll at 

the nip. A positive correlation between oil uptake and CI suggests 

that roughness and stiffness also have significant influences on oil 

uptake. © 2007 Society for Imaging Science and Technology. 


INTRODUCTION 

High-speed digital xerographic printing presses are versatile 

with the capability to produce high resolution prints on a 

variety of media at rapid speeds. The focus of this study is 

on the fusing section, where the image is fixed onto paper. 

Hot roll fusing is the most common fixing method for these 

printers, where toner is fused onto the substrate by heat and 

pressure within a nip formed by two rotating rolls, which are 

typically a fuser roll and a pressure roll 1 (Figure 1). 

Clean splitting between the fuser roll and paper with a 

toner image must be achieved to produce a sharp image 

without offset at high reproduction speeds. To facilitate this, 

the fuser roll is typically lubricated with a release agent, in 

most cases a silicon-based oil, that is applied continuously 

using a donor roll. The printing media takes away oil at the 

fusing nip and over time, with high volume production, excessive 

oil uptake may occur, resulting in oil depletion within 

the system. Consequently, the fuser roll will become insufficiently 

coated, resulting in toner offset, poor print quality, 

and a reduction in fuser roll life. 

Over the years, extensive research in the area of fusing in 

xerographic printing has focused on the interaction between 

toner and paper in terms of the effect of paper properties, 

toner properties, and operational conditions on fusing fix. 2–4 

However, an important aspect in the fusing process that has 

not been well studied is the effect of fuser roll lubrication on 

print quality. It has been observed that the oil depletion rate 

varies with different types of papers in high-speed xerographic 

printing, but thus far the specific paper properties 

that control this oil depletion rate are not clear. 

This study aims to bridge this gap in the literature by 

identifying the key paper properties that control fuser oil 

uptake and by obtaining a mechanistic understanding of the 

interactions between oil, paper, and the fuser roll material. 

An understanding of the interactions between fuser oil and 

paper is of fundamental importance in improving fuser roll 

life and print quality. Meanwhile, understanding these interactions 

will help papermakers optimize paper properties to 

minimize fuser oil consumption and eliminate downstream 

problems, such as oil streaks on the printed copy. 

 

IS&T Member. 

Received Jan. 22, 2007; accepted for publication May 10, 2007. 

1062-3701/2007/515/424/7/$20.00. 

Figure 1. Schematic of the fusing section in a high-speed xerographic 

printing press. 

424

Lai et al.: Relationship between paper properties and fuser oil uptake in a high-speed digital xerographic printing fuser 

Table I. Properties measured for all paper samples. 

Property Range Test Method 

Basis weight 76–269 g / m 2 TAPPI T410 

Caliper 0.10–0.24 mm TAPPI T411 

Porosity 12–3500 Gurley sec TAPPI T460 

Bending stiffness MD 0.84–32.6 mN TAPPI T543 

rms roughness 1.20–5.83 µm WYKO NT-2000 

Surface free energy 35.1–51.2 mJ/ m 2 Wu’s geometric mean 

method 

Contact angle of oil 

on paper 

57.9°–67.2° See Materials and Method 

Section 

MATERIALS AND METHODS 

Characterizing the Test Paper Set 

The test paper set consisted of 16 commercial papers 

(samples 1–16), five of which were uncoated papers (samples 

1, 4, 5, 7, 8), and the remaining were coated papers with 

varying coating compositions. A typical commercial siliconbased 

fuser oil with a viscosity of 5.75 cm 2 /s 575 cSt at 

25°C and surface tension of 20.6 mJ/m 2 at 25°C was used in 

this investigation. 

All samples were characterized for their physical, topographical 

and surface chemistry properties. Physical properties 

(basis weight, porosity, caliper, bending stiffness) were 

determined for each sample according to the standard test 

methods listed in Table I. 

Surface topography of each sample was evaluated in 

terms of the root-mean-square (rms) roughness. Measurements 

were obtained using WYKO NT-2000, an optical 

noncontact surface profiler system. Surface profile measurement 

data were collected for each sample. The data were 

median smoothened, and the tilt in the data was removed to 

ensure that a completely horizontal surface was analyzed. 

The surface chemistry of paper was characterized in 

terms of contact angle measurements of fuser oil on paper, 

surface free energy of paper, and work of adhesion between 

paper and oil. Static contact angles of fuser oil on all the 

papers were measured using the Fibro DAT contact angle 

machine. All samples were conditioned for at least 24 h in a 

controlled room at 23°C and 50% relative humidity (T402 

TAPPI Standard). The contact angle measurements were also 

carried out in the conditioned room. For each drop of oil on 

the paper substrate, the static contact angle was measured as 

a function of time due to spreading on the substrate. For 

each paper sample, 16 contact angle-time profiles were obtained 

and, from these profiles, the contact angles at 1 sec 

were averaged. This resulted in an average contact angle of 

oil on paper at 1 sec for each sample. 

The surface free energy of paper and the work of adhesion 

between paper and oil were also determined for all 

samples. Wu’s geometric mean method, with diiodomethane 

and water as test liquids, was used to determine the surface 

free energy of the papers. 5 Literature values for the surface 

tension, dispersive and polar components for diiodomethane 

( D =50.8 mJ/m 2 ; D D =49.5 mJ/m 2 ; D P =1.3 mJ/m 2 ) 

and water ( W =72.2 mJ/m 2 ; W D =22 mJ/m 2 ; 

W P =50.2 mJ/m 2 ) were used. 5 The work of adhesion between 

paper and oil at ambient conditions W PO,AMB was 

calculated using the Young–Dupré equation. 6 A W PO,AMB 

value was calculated for each paper using ambient contact 

angle measurements of fuser oil on paper at 1 sec. The surface 

tension of fuser oil at ambient temperature, OIL,AMB 

=20.6 mJ/m 2 , was used. 7 

Evaluating the Oil Uptake Performance of the Test Paper 

Set 

All paper samples were evaluated for their oil uptake performance 

by quantifying the amount of oil that each paper 

sorbed in the fusing nip. The amount of oil uptake for all 

blank paper samples was obtained using a prototype highspeed 

xerographic fusing system and inductively coupled 

plasma mass spectroscopy (ICP-MS). The detailed experimental 

procedure is described in a previous paper. 8 

Partial Least-Squares Regression Modeling 

After characterizing the papers and obtaining oil uptake 

amounts for each sample, these two data sets were analyzed 

using partial least-squares (PLS) regression to determine the 

correlation structure between paper properties (input or x 

variables) and oil uptake (output or y variable). Partial leastsquares 

(PLS) regression is an established multivariate statistical 

method that can extract the correlation structure between 

input and output variables from experimental data. 9 

The dimensions of large data sets are reduced by projecting 

the relevant information onto low-dimension subspaces defined 

by principal components. 10 Using graphical methods, 

the data can then be analyzed easily to understand processes 

and to identify key factors that predict response variables. 

In this study, partial least-squares (PLS) regression is 

used to model the relationship between paper properties x 

and oil uptake y. Traditionally, multiple linear regression 

(MLR) is used in relating a response variable y toasetof 

x variables, but PLS is more robust because unlike MLR, PLS 

can handle large numbers of x variables, highly dependant 

variables, and noisy data. 10 Simca-P software by Umetrics 

was used to analyze the data. 

The PLS model performance was assessed using the 

loading plot and the parameters, R 2 Y and Q 2 . R 2 Y is a multiple 

correlation coefficient between the measured response 

y and the predicted response ŷ PRED values. The Q 2 parameter 

denotes the model’s ability to predict the response. 

An ideal model 10 has R 2 Y and Q 2 values of 0.5 while minimizing 

, the difference between R 2 Y and Q 2 . Further details 

of the PLS regression method used in this study are 

given in a previous paper. 8 


Test Paper Set 

All papers were characterized and the samples cover a broad 

range of the assessed properties as summarized in Table I. 

Paper samples range from text weight to cover stock for both 

coated and uncoated papers. 



Figure 2. Oil uptake amounts for all paper samples. Sheet size is 8.5 

11 in. 

Oil Uptake Data 

Oil uptake data for all paper samples are shown in Figure 2. 

It is important to note that the experimental technique used 

to quantify oil uptake results in higher oil uptake values than 

observed when printing on a commercial press. Therefore, 

the results should be interpreted in terms of relative oil uptake 

amounts among the paper samples. 

In general, coated papers appear to have higher oil uptake 

amounts than uncoated papers. However, some coated 

papers, such as sample 6, have small oil uptake values that 

are similar to that of uncoated papers. To gain insight into 

the relationships between paper properties and oil uptake, 

the oil uptake data are correlated with the properties of all 

the samples using PLS regression. 

Figure 3. Loading plot for the PLS model. Key paper properties x correlated 

with oil uptake y are identified. R 2 Y=0.72 and Q 2 =0.61 with 

=R 2 Y−Q 2 =0.11. 

Figure 4. Observed oil uptake data y measured vs predicted oil uptake 

ŷ PLS predicted values from the PLS model. R 2 Y=0.72 and Q 2 =0.61. Each 

data point represents a different paper sample. 

Key Paper Properties Identified by the PLS Regression 

Model 

PLS regression was performed on the oil uptake y and 

paper properties x data sets. All the paper properties listed 

inTableIwereusedasx-variable inputs into the initial PLS 

regression model. The data were mean centered and scaled 

to unit variance. One principal component, calculated 

through cross-validation, was found to sufficiently explain 

the variance in the data, resulting in R 2 Y and Q 2 values of 

0.72 and 0.61 respectively with =R 2 Y−Q 2 =0.11. 

The correlation structure between paper properties and 

oil uptake are identified in the loading plot (Figure 3). Variables 

with a PLS weighting factor, w * c, of the same magnitude 

and sign as that of oil uptake have strong positive correlations, 

while variables with negative w * c values have 

negative correlations with oil uptake. rms roughness and 

contact angle measurements, for example, are furthest from 

zero in the negative direction w * c−0.5, and thus have a 

strong negative correlation with oil uptake. Surface free energy 

has the highest PLS weight w * c0.5, and therefore 

has the strongest positive correlation with oil uptake. Basis 

weight, bending stiffness, and caliper have moderate positive 

correlations with oil uptake, with PLS weight values between 

0.2 and 0.4. With the lowest PLS weight w * c0.2, porosity 

based on Gurley seconds has the weakest positive correlation 

with oil uptake. Therefore, the loading plot reveals that papers 

with high roughness and high surface energy correspond 

with large oil uptake amounts. 

For this PLS regression model, Figure 4 compares the 

observed y measured and the predicted values ŷ PLS predicted of 

oil uptake. The model’s ability to predict oil uptake well is 

shown in Fig. 4, where a good scatter of the data points 

around the 45° line is present. 

The equation for the final PLS regression model is 

ŷ PLS predicted = 0.22x SFE + 0.16x BWT + 0.13x STF + 0.11x CAL 

+ 0.08x POR − 0.22x RMS − 0.20x CA + 6.4, 1 

where SFEsurface free energy of paper, BWTbasis 

weight, STFbending stiffness, CALcaliper, 

PORporosity in Gurley seconds, RMSrms roughness, 

and CAambient contact angle measurements of oil on paper. 

The coefficients are based on mean centered and unit 

variance scaled data. To accurately predict absolute values 

for oil uptake, this model must be validated with an external 

data set, consisting of a large number of new and well characterized 

paper samples. 



Table II. Ambient work of adhesion between fuser oil and paper for all paper samples. 

Paper Sample No. 

W PO,AMB = OIL 1 + cos 

mJ/m 2 

Figure 5. Oil uptake vs surface free energy of paper, R 2 =0.65. 

Mechanistic Understanding of Fuser Oil-Paper 

Interactions in the Fusing Nip 

Although PLS regression reveals positive and negative correlations 

between x and y variables, PLS regression does not 

ascertain cause and effect relationships between the two data 

sets. Therefore, it is necessary to gain a mechanistic understanding 

of how these interacting paper properties affect the 

oil uptake process. The key properties identified by PLS suggest 

that oil uptake is highly related to two effects: the wetting 

behavior of oil on paper and the nip contact area between 

paper and the fuser roll. These two mechanisms are 

further examined to explain how oil is physically transferred 

to paper within the fusing nip. 

Wetting Behavior of Fuser Oil on Paper 

From the PLS regression model, a negative correlation between 

oil uptake and ambient contact angle measurements 

of oil on paper was observed. As the contact angle of oil on 

paper decreases, wetting increases and the result is more oil 

uptake. Therefore, contact angle measurements of oil on paper 

at ambient conditions can be used as a tool to predict 

the wetting behavior of oil on paper during uptake in a 

fusing nip. 

The wetting behavior of oil on paper was further investigated 

by looking at the effect of surface free energy (SFE) 

of paper on oil uptake. From the PLS model, a positive 

correlation between oil uptake and SFE is present. The surface 

tension of the fuser oil at room temperature is low 

20.6 mJ/m 2 , and the oil will preferentially wet papers that 

exhibit higher surface energies. As the surface energy of paper 

increases, further wetting of oil occurs and the result is 

an increase in oil uptake. This result follows the same approach 

by Sanders et al., 11 where it was found that higher 

energy surfaces promoted wetting and spreading of molten 

toner on paper. The positive correlation between oil uptake 

and SFE can also be seen in the independent linear regression 

plot of these two variables (Figure 5), where an increase 

in SFE of paper results in an increase in oil uptake, as the 

surface tension of oil remains constant. With a good positive 

correlation R 2 =0.65 between these two variables, SFE of 

paper is shown to be an important paper property in explaining 

the effect of wetting on oil uptake; therefore, it is a 

Uncoated papers 1 28.80 

4 29.58 

5 28.70 

7 29.02 

Coated papers 2 31.45 

3 31.40 

6 31.51 

9 30.80 

10 31.15 

11 31.49 

12 31.20 

13 31.56 

14 31.51 

15 31.14 

16 30.01 

mechanism by which the oil uptake trend can be partially 

explained. It is important to note that while the correlation 

between oil uptake and SFE of paper is determined at ambient 

conditions, it is expected that the surface energy of the 

paper will not change significantly with fusing temperature 

(185°C) due to a short residence time of the paper within 

the fusing nip 30 ms. 

The surface chemistry effect of paper on oil uptake is 

also investigated by looking at the ambient work of adhesion 

between paper and oil W PO,AMB . The calculated values for 

all paper samples are listed in Table II. In general, the work 

of adhesion between oil and uncoated papers is slightly 

lower than the work of adhesion between oil and coated 

papers. Figure 6 shows a positive correlation between oil 

uptake and W PO,AMB . This indicates that to minimize oil 

uptake the work necessary to separate oil from paper must 

be minimized, which is achieved by lowering the surface free 

energy of the paper surface. 

Figure 6. Oil uptake vs ambient work of adhesion between paper and 

oil, W PO,AMB . 



Relative Contact Area Between Paper and the Fuser Roll 

in the Nip 

Oil uptake is also examined with respect to the contact area 

between paper and oil within the fusing nip. This contact 

area is examined on the microscopic scale, where surfaces 

exhibit asperity peaks and valleys. As a result, when two 

surfaces touch, the real contact area is less than the nominal 

contact area. The relative contact area between paper and the 

fuser roll at the nip is modeled using theory from contact 

mechanics and in relation to the identified key nonsurface 

chemistry related properties (basis weight, caliper, stiffness, 

and rms roughness). The transfer of oil from the fuser roll to 

paper is then examined as a function of this relative contact 

area at the nip. 

Greenwood and Williamson developed a theory for the 

elastic contact of rough surfaces that takes into account surface 

topography and elastic deformation based on Hertz 

theory. 12 The model assumed that contact occurs at the asperity 

peaks of the surface, and that the total contact area is 

the sum of the contacting areas of the peaks. According to 

Greenwood and Williamson, 12 for a negative exponentially 

distributed rough surface, the relative contact area parameter, 

A re , is defined as 

A re =k E, 

P 2 

where is the root-mean-square (rms) roughness value, E 

is a complex modulus defined in Eq. (3), and k is the rms 

value of the peak curvature and is defined in Eq. (4), 12 

1 1− 

E = paper 

+ 

E paper 

2 

1− 2 roll 

. 3 

E roll 

A re is a dimensionless parameter describing the relative 

contact area between two rough surfaces or the ratio of real 

contact area to nominal contact area. This parameter describes 

the contact area between surfaces as a function of 

surface texture and compressibility of the substrate. The surface 

texture parameter was also defined as a dimensionless 

parameter, R f =k, by Yan and Aspler 13 ; and k are obtained 

from surface profile measurement data using Eq. (4), 

where z i−1 , z i , and z i+1 are three consecutive height measurements; 

h is the horizontal step interval between the height 

measurements; and N A is the total number of peaks measured. 

The rms roughness and k values for the test paper set 

are shown in Table III, and these values are in agreement 

4 

Table III. rms roughness and curvature parameter k values for all paper samples. 

Paper 

Sample 

No. 

rms 

Roughness 

µm 

Standard 

Deviation 

for rms Roughness 

with those published by Yan and Aspler. 13 As expected, rms 

roughness values for uncoated papers are greater than those 

of coated papers. Curvature parameters appear to be greater 

for uncoated papers than for coated papers. 

Based on the work by Greenwood and Williamson 12 as 

well as Yan and Aspler, 13 a contact area index (CI) is developed 

to describe the degree of contact between the paper 

and the fuser roll at the nip [Eq. (5)]. This index represents 

the relative contact area as a function of rms roughness , 

asperity peak curvature k, and complex modulus E. E 

combines the modulus of paper and that of the fuser roll. In 

this study, the fuser roll was silicon rubber and for Eq. (3), 

E roll =2 MPa, and Poisson’s ratio, roll =0.5, at 175°C were 

used 7 as estimated values for the fuser roll at fusing temperature. 

These values remained constant throughout the study. 

E paper was calculated using the bending stiffness of paper S 

and Eq. (6). 14 Changes in Poisson’s ratio among the paper 

samples were assumed to be negligible, and a constant Poisson’s 

ratio of paper =0.4 was used. 15 Pressure P within the 

fusing nip was estimated as a function of caliper (C), where 

P=C and is a constant, which depends on the spring 

constant of the nip, 

CI = 

A re 

= 1 C 

k E , 

E paper = 12S 

C 3 . 

Average k 

1/µm 

Standard 

Deviation 

for k 

Uncoated papers 1 5.83 0.38 1.35 0.05 

4 3.24 0.11 0.74 0.04 

5 4.84 0.17 1.21 0.03 

7 4.10 0.38 1.02 0.14 

8 4.73 0.05 1.16 0.12 

Coated papers 2 1.25 0.01 0.17 0.02 

3 1.20 0.11 0.20 0.02 

6 1.40 0.04 0.20 0.01 

9 1.61 0.11 0.30 0.04 

10 1.44 0.03 0.24 0.01 

11 1.61 0.06 0.25 0.01 

12 1.81 0.13 0.29 0.02 

13 1.80 0.16 0.26 0.01 

14 1.58 0.09 0.24 0.01 

15 1.50 0.02 0.28 0.004 

16 1.26 0.05 0.17 0.01 

Although CI is not a measurement for the actual contact 

area between paper and the fuser roll at the nip, it is an 

5 

6 



Table IV. Contact area index, CI, values for all paper samples. 

Paper Sample No. CI 10 −7 StDev 10 −7 

Uncoated papers 1 0.21 0.007 

4 0.36 0.12 

5 0.19 0.08 

7 0.35 0.02 

8 0.17 0.002 

Coated papers 2 2.23 0.03 

3 0.96 0.08 

6 1.99 0.01 

9 0.63 0.06 

10 0.86 0.02 

11 0.79 0.03 

12 0.70 0.04 

13 0.73 0.03 

14 0.82 0.04 

15 0.78 0.002 

16 0.90 0.05 

Figure 7. Positive correlation between oil uptake and contact area index 

CI is illustrated. 

indication of the degree of contact that would result when 

paper contacts the fuser roll. A paper sample with a lower CI 

value indicates a lower relative contact area with the roll than 

that of a paper with a higher CI value. A CI value was 

calculated for each sample, which allowed for a relative comparison 

among papers for their degree of contact with the 

fuser roll. The results are summarized in Table IV. 

When two surfaces touch, the actual contact is assumed 

to occur at the tips of the paper surface asperity peaks. 12 In 

the case of coated or smooth papers, the asperity peaks are 

broader than the asperity peaks on uncoated or rougher papers. 

As a result, the peak surface area in contact with the 

fuser roll is larger for coated papers, and CI will be greater 

for coated papers. The trend in Table IV, that coated papers 

have greater CI values than uncoated papers, is in good 

agreement with this explanation. Furthermore, as peak contact 

area increases with the fuser roll or as CI increases, 

further oil uptake by paper is expected. Figure 7 shows a 

positive linear trend between oil uptake and CI. While R 2 is 

not too high, the positive trend indicates that surface texture 

(roughness) and stiffness of paper (which encompasses effects 

from basis weight and caliper) have an affect on oil 

uptake. 

The approach in applying contact mechanics to examine 

oil uptake by paper has been simplified, where nip residence 

time and oil thickness were considered constant due 

to steady-state conditions in the fusing apparatus. A good 

basis for understanding oil uptake by paper during fusing 

has been established, and further study is warranted to investigate 

the affect of oil rheology on the oil uptake process. 

Simplified Model for Oil Uptake 

Overall, the surface chemistry of paper and the degree of 

contact that occurs between paper and the fuser roll at the 

Figure 8. Oil uptake measured y measured vs oil uptake predicted by the 

MLR model ŷ MLR predicted , R 2 =0.8. 

nip are shown to be significant parameters in predicting and 

understanding oil uptake. These parameters characterize two 

separate mechanisms by which paper takes up oil from the 

fuser roll. Surface free energy characterizes the dominant 

surface chemistry effect of paper within the fusing nip, 

where wetting of oil on paper is promoted as paper surface 

energy increases. CI, on the other hand, accounts for the 

effects of physical paper properties on oil uptake, where an 

increase in nip contact area between paper asperity peaks 

and the fuser roll results in an increase in oil uptake. Therefore, 

with CI and SFE as dominant factors in describing oil 

uptake by paper, a multiple linear regression (MLR) model is 

developed using CI and SFE as independent parameters. After 

scaling CI and SFE to unit variance, the MLR model is 

ŷ MLR predicted = 2.25x SFE + 1.27x CI − 1.65, 

with R 2 =0.8. Figure 8 compares y measured with ŷ MLR predicted , 

which shows that SFE and CI describe oil uptake well. 

Figure 9 is a three-dimensional (3D) contour plot summarizing 

the relationship between oil uptake, surface free 

energy (SFE) of paper, and contact area index (CI). The 

linear relationship between oil uptake and CI, and between 

oil uptake and SFE are evident from this plot. 

Of the paper properties included in the PLS model, further 

study is required to ascertain the relationship between 

7 



Figure 9. The 3D contour plot describes the relationship between oil 

uptake, surface free energy of paper, and contact area index CI. 

porosity and oil uptake. It is expected that porosity may play 

a minor role in the initial transfer of oil from the fuser roll to 

paper. The effects of pore size, volume, and distribution on 

oil uptake all require thorough investigation. In addition, 

work of adhesion between oil and paper at fusing conditions 

has a good potential for describing the oil uptake process 

within the fusing nip. At ambient conditions, W PO,AMB was 

found to have a positive correlation with oil uptake; however, 

with work of adhesion values calculated at conditions 

similar to fusing conditions W PO,fusing , the correlation between 

W PO,fusing and oil uptake may improve. Therefore, 

W PO,fusing may also help to account for the unexplained variance 

in the oil uptake data. Moreover, surface tension of oil 

and accurate contact angle measurements of oil on paper at 

fusing conditions require further investigation in order to 

accurately calculate W PO,fusing . 

CONCLUSIONS 

A systematic study was carried out to investigate the relationship 

between fuser oil uptake and paper properties in 

high-speed xerographic printing. A partial least-squares 

(PLS) regression model was developed to identify the key 

paper properties that are significantly correlated with the 

uptake of a typical silicon-based oil; rms roughness and ambient 

contact angle measurements of the oil on paper were 

found to have strong negative correlations with oil uptake. 

Caliper, basis weight, bending stiffness, and surface free energy 

had positive correlations with oil uptake. 

Contact angle measurements and SFE were both included 

in the PLS model. Although both parameters describe 

the surface chemistry of paper, the contact angle measurements 

were more convenient to obtain as it involved 

only one fluid, fuser oil. The PLS results showed that both 

parameters indicate the same trend with oil uptake, providing 

the experimenter with confidence to use the more convenient 

method to determine the correlation between oil 

uptake and surface chemistry of paper. Furthermore, the 

predictive ability of this PLS model will improve with validation 

using a new paper set. 

Surface free energy (SFE) of paper was found to have 

the strongest positive correlation with oil uptake in the PLS 

regression model, and this positive correlation was also confirmed 

in a linear regression of the two parameters with an 

R 2 =0.65. Therefore, SFE was determined to be a dominant 

property that describes the wetting behavior of oil on paper. 

A contact area index (CI) was used to compare the degree 

of contact between paper and the fuser roll at the nip. 

CI was developed as a function of surface texture (rms 

roughness and surface curvature), nip pressure as estimated 

as a function of caliper, and elastic moduli of paper and the 

fuser roll. CI and oil uptake were found to have a positive 

correlation, suggesting that surface texture and stiffness of 

the paper have significant influences on oil uptake. 

A multiple linear regression model was developed by 

regressing oil uptake against CI and SFE. Both CI and SFE 

were treated as independent factors affecting oil uptake. 

These two factors were found to be dominant parameters in 

predicting oil uptake, as the MLR model accounted for 80% 

of the variance in the oil uptake data. Therefore, oil uptake 

can be explained mechanistically in terms of the wetting 

behavior of oil on paper and by a contact area index to 

describe the degree of contact between paper and the fuser 

roll at the nip. Other factors that potentially may have significant 

affects on oil uptake are surface porosity of paper 

and the work of adhesion between oil and paper at fusing 

conditions. Both parameters require further investigation. 


Financial support from NSERC Canada and the member 

companies of the Surface Science Consortium at the University 

of Toronto Pulp and Paper Centre are gratefully acknowledged. 

As well, the support from Xerox and International 

Paper in carrying out experimental work is gratefully 

appreciated. 

REFERENCES 

1 L. Leroy, V. Morin, and A. Gandini, “Electrophotography: Effects of 

printer parameters on fusing quality”, Proc. 11th International Printing 

and Graphic Arts Conference (TAPPI, Bordeaux, France, 2002). 

2 M. Alava and K. Niskanen, “The physics of paper”, Institute of Physics 

Publishing: Rep. Prog. Phys. 69, 669 (2006). 

3 J. S. Aspler, “Interactions of ink and water with the paper surface in 

printing”, Nord. Pulp Pap. Res. J. 8, 68 (1993). 

4 M. B. Lyne and J. S. Aspler, “Ink-Paper Interactions in Printing: A 

Review”, in Colloids and Surfaces in Reprographic Technology, edited by 

M. Hair and M. D. Croucher (ACS, Washington DC, 1982), p. 385. 

5 S. Wu, Polymer Interface and Adhesion (Marcel Dekker, New York, 1982). 

6 A. V. Pocius, Adhesion and Adhesives Technology (Hanser/Gardner 

Publications, New York, 1997). 

7 Xerox Research Centre of Canada, private communication. 

8 P. Lai and N. Yan, “The relationship between paper properties and fuser 

oil uptake in high-speed xerographic printing”, Proc. 92nd Paptac 

Annual Meeting (PAPTAC, Montreal, Quebec, 2006). 

9 R. Shi and J. F. Macgregor, “Modeling of dynamic systems using latent 

variable and subspace methods”, J. Chemom. 14, 423 (2000). 

10 L. Eriksson, E. Johnansson, N. Kettaneh-Wold, and S. Wold, Multi- and 

Megavariate Data Analysis: Principles and Applications (Umetrics 

Academy, Sweden, 2001). 

11 D. J. Sanders, D. F. Rutland, and W. K. Istone, “Effect of paper properties 

on fusing fix”, J. Imaging Sci. Technol. 40, 175 (1996). 

12 J. A. Greeenwood and J. B. P. Williamson, “Contact of nominally flat 

surfaces”, Proc. R. Soc. London, Ser. A, 295, 300 (1966). 

13 N. Yan and J. Aspler, “Surface texture controlling speckle-type print 

defects in a hard printing nip”, J. Pulp Pap. Sci. 29, 357 (2003). 

14 J. Levlin and L. Söderbjelm, Pulp and Paper Testing: Papermaking Science 

and Technology, Book 17 (Fapet Oy, Finland, 1999). 

15 N. Stenberg and C. Fellers, “Out-of-plane Poisson’s ratios of paper and 

paperboard”, Nord. Pulp Pap. Res. J. 17, 4 (2002). 




Simulation of Traveling Wave Toner Transport 

Considering Air Drag 

Masataka Maeda 

Graduate School of Science and Engineering, Ibaraki University, 4-12-1 Nakanarusawa-cho, Hitachi, Ibaraki 

316-8511, Japan 

Katsuhiro Maekawa 

The Research Center for Superplasticity, Ibaraki University, 4-12-1 Nakanarusawa-cho, Hitachi, Ibaraki 

316-8511, Japan 

Manabu Takeuchi 

Department of Electrical and Electronic Engineering, Ibaraki University, 4-12-1 Nakanarusawa-cho, Hitachi, 

Ibaraki 316-8511, Japan 

Abstract. A simulation method to analyze the behavior of a toner 

cloud driven by a traveling electrostatic wave with air drag is proposed. 

The two-fluid flow model, which regards air and toner cloud 

as incompressible and viscous fluids, is employed. A method of the 

calculation of the electric potential using the moving particle semiimplicit 

(MPS) method, which is often used in the study of fluid dynamics, 

is developed in this article. In the present method, all of the 

motion of the toner, the airflow, and the electric potential are calculated 

by using the moving particle semi-implicit method. Common 

calculation points are used in the electrostatic calculation, the toner 

motion calculation, and the airflow calculation in order to avoid the 

complexity of data exchange among those calculations. The validity 

of the calculation of the electric potential using the MPS method is 

confirmed by comparing the results to those of a model that has a 

known strict solution. Modeling of the behavior of toner cloud in 

toner transport with a traveling electrostatic wave is performed. An 

increase in the maximum synchronous frequency due to air drag is 

demonstrated. © 2007 Society for Imaging Science and 

Technology. 


INTRODUCTION 

Since Masuda et al. 1 demonstrated that powders could be 

transported by means of a traveling wave of an electric field, 

the method has been studied in various applications. In electrophotography, 

theoretical and experimental studies of 

toner transport have been made using this method. Melcher 

et al. 2–4 and Schmidlin 5,6 provided much insight into the 

understanding of the different modes of toner transport with 

a traveling electrostatic wave. Numerical analysis of toner 

transport has also been improved. In early studies, the numerical 

analyses focused on the motion of a single particle 

to avoid the complexity of the interaction between particles. 

In recent years, the motion of many particles has been analyzed. 

By using the particle-in-cell method for many particle 

systems, Gartstein and Shaw 7 showed that the velocity of 

Received Nov. 8, 2006; accepted for publication May 1, 2007. 

1062-3701/2007/515/431/7/$20.00. 

particle flow could vary all the way from zero to the phase 

velocity of the traveling wave. Kober 8 numerically solved the 

equations of toner motion by superimposing the electric 

field due to toner charge and image charge on the electric 

field applied by the electrodes. 

These analyses provided insight into the behavior of 

large ensembles of toner particles. However, the calculation 

of the motion of the toner and the electric field was complicated 

by the dual approach of using both mesh and particle. 

The motion of the toner is calculated using particles, 

but the electric field is calculated using a mesh. The variables 

are calculated independently and must be interpolated between 

the mesh and the particles at each time step. Moreover, 

in a many-particle system, the motion of air dragged by 

the toner particles is not negligible. In many previous numerical 

analyses of the traveling wave toner transport, the air 

was supposed to be quiescent and the motion of the air 

caused by particle motion was not taken into account. A 

calculation method that not only simplifies the calculation of 

electric field but also includes the air effect is required. 

In order to avoid the complexity of the calculation, we 

use calculation points that move with the toner or air rather 

than fixed meshes or grids. This technique makes the calculation 

of electric field associated with the moving charge and 

moving air more manageable. The smoothed particle hydrodynamics 

(SPH) method and the moving particle semiimplicit 

(MPS) method, which are often used in the study of 

fluid dynamics, are well-known modeling methods that use 

moving particles. 

In this paper, we propose a simulation method that can 

solve the problem of toner and air motions and electric field, 

simultaneously, using a particle method. Using the present 

method, we calculate the toner motion in traveling wave 

toner transport and show an overview of the behavior of 

toner cloud that is affected by airflow as the speed of traveling 

wave increases. 

431

Maeda, Maekawa, and Takeuchi: Simulation of traveling wave toner transport considering air drag 

MODEL 

Physical Model 

Let us consider the motion of charged toner particles driven 

by a traveling electrostatic wave taking into account the effect 

of air motion. We examine the effect of electrostatic 

force and airflow as a major factor of basic driving mechanism 

of toner transport with traveling wave. 

The electric potential formed by toner charge is given 

· = Q, 

where is the permittivity of media and Q is the space 

charge density due to toner charge. The electric field can be 

expressed as 

E =−. 

In the analysis of a solid-gas flow, such as motion of 

toner particles in air, the flow structure is different according 

to whether the particulate phase is dispersed in a fluid phase 

or separated from a fluid phase. It was reported that toner 

particles are transported in bands that are spaced by the 

wavelength of traveling wave. 4 Therefore, the objective of our 

method to be developed is to obtain an overview of the 

behavior of a cluster of toner particles rather than of a discrete 

toner particle. We regard a cluster of toner particles in 

a band as a separated phase and as an incompressible fluid. 

In other words, we employ the two-fluid flow model of air 

and toner cloud in this study. 

From mass and momentum conservation and incompressibility, 

the equations for air are 

D a 

Dt =0, 

1 

2 

3 

Du a 

=− 1 P a + a 2 u a + g − 1 F at , 4 

Dt a a 

air by pressure and viscosity of toner cloud. F ta is the interaction 

force that is exerted on toner cloud by pressure and 

viscosity of air. D/Dt is denotes the Lagrangian time derivative. 

Since we discretize these equations using a calculation 

point, as discussed later, these are described as calculation 

points for air and toner, not for a volume. Thus, the concentration 

of toner cloud is not expressed in the equations. 

Numerical Model 

In order to avoid the complicated procedure of interpolating 

between grid and particle, only calculation points are employed 

in the calculation of the electric field, motion of 

toner, and airflow. Mesh and grid are not employed in this 

study. The calculation points are placed evenly in an analysis 

area. A toner particle corresponds to one calculation point. 

The calculation points assigned for toner particles move 

with the motion of the toner cloud. In order to calculate the 

electric field, the airflow, and the flow of toner cloud, we 

examined the particle methods, such as the SPH method 

and the MPS method, which are usually used for fluid analysis. 

The SPH method was used to calculate electric potential 

in our previous study, 9 but this method requires special 

schemes to treat incompressible fluid. 10 In the present study, 

the MPS method, 11 which is often used in studies of incompressible 

flow, is employed the calculation of the electric 

field. The calculation points assigned for air also move with 

the airflow calculated using the MPS method. 

In the MPS method, calculation points (also called particles 

in the MPS method) with the physical quantity f are 

placed evenly in an analysis area. The gradient of the physical 

quantity f is modeled as the average slope weighted with 

a weighting function between the object calculation point i 

and its neighboring calculation points j, and is expressed at a 

position of calculation point i as 

f i = d 

n ji f j − f i r ij 

wr 0 ij , 

r ij 

r ij 

F at =−− 1 a 

P a + a 2 u at. 5 

r ij = r j − r i , r ij = r ij , 9 

The equations for toner cloud are 

D t 

Dt =0, 

Du t 

Dt =− 1 P t + t 2 u t + 1 QE + g + 1 F ta , 

t t t 

6 

7 

where r i is the position vector of calculation point i, and d is 

the spatial dimension. The weight function w uses the following 

function: 

e 

wr =r r −1 0 r r e 

, 10 

0 r e r 

F ta =− 1 t 

P t + t 2 u ta, 8 

where u is velocity, is density and is viscosity of air and 

toner cloud. Subscript a, t denote air and toner cloud, respectively. 

Q is volume charge density of toner cloud, and E 

is electric field. F at is the interaction force that is exerted on 

where r e is the size of weight function that defines interacting 

calculation points and n 0 is the number density of calculation 

points in the initial configuration, and is defined by 

n 0 = wr ij . 

ji 

The Laplacian is modeled as 

11 



where 

2 f i = 2d 

f 

n 0 j − f i wr ij , 

ji 

12 

wrr 2 d 

=V 

. 13 

V 

wrd 

In order to discretize Eq. (1) for the electric potential 

using the MPS Laplacian model, Eq. (12), we use the identity 

· = 1 2 2 − 2 + 2 , 

and we can obtain the discretization of Eq. (1) as 

· i = 

d 

 

n 0 j + i j − i wr ij =−Q i . 

ji 

14 

15 

Solving Eq. (15) simultaneously for unknown with 

the charge density and the boundary potential we obtain the 

potential at each calculation point. Here, the charge densities 

Q i are given as a property of the calculation points for toner, 

and the boundary potentials are given as a property of the 

calculation points at the boundary. From Eq. (9), the electric 

field is obtained by 

E i =− i =− d 

n ji j − i r ij 

0 wr ij . 16 

r ij 

Two-fluid flow of toner cloud flow and airflow is 

discretized by calculation points with toner mass and calculation 

points with air mass, respectively, in MPS. The continuity 

equations [Eqs. (3) and (6)] are satisfied by keeping 

the number of calculation points constant during the motion 

in the analysis area. Keeping the number density of the 

calculation points constant, n 0 satisfies the incompressibility. 

The Navier–Stokes equations [Eqs. (4) and (7)] are calculated 

as follows. 12 

In a time step k, the terms with the exception of the 

pressure term of Eqs. (4) and (7) are explicitly calculated, 

and temporal velocities u * are obtained as 

u * a = u k a + g + t, 17 

a 

a 

2 u aa 

r ij 

+ a 

a 

2 u at 

u * t = u k t + t 

2 u 

t 

tt 

+ t 

2 u 

t 

ta 

+ g + QE 

 

t. t 

18 

In Eqs. (17) and (18), a and t denote interactions from 

air and toner, respectively. Temporal positions r * are obtained 

as 

r a * = r a k + u a * t 

r t * = r t k + u t * t 

where t is time step. The viscous term is calculated as 

2 u i = 2d 

 

n 0 u j − u i wr ij . 

ji 

19 

20 

21 

Since temporal velocities u * do not take the pressure 

term into account, the velocities of the next step k+1, u k+1 

should be corrected as follows: 

u a k+1 = u a * + u a , 

u t k+1 = u t * + u t , 

22 

23 

where u are the correction velocities. The correction velocities 

u due to the pressure term can be written as 

1 

P a 

a 

k+1tt, 24 

u a =− 1 a 

P a 

k+1a 

+ 

u t =− 1 P t 

t 

+ 1 P t 

 

k+1t t 

k+1at. 25 

The first and second terms on the right-hand side of Eq. 

(24) denote pressure gradients due to interactions neighboring 

calculation points for air and toner, respectively. Adding 

both sides of Eqs. (24) and (25) gives 

u a 

1−c 

t + c t u t 

a t =− 1 P k+1 , 

a 

P k+1 = 1−cP a a + P a t + cP t t + P t a , 

26 

where c is concentration of toner. 

Since the temporal position r * does not satisfy the incompressibility 

constraint, that is to say, the temporal number 

density of calculation point at the temporal position n * is 

not n 0 . The temporal number density of calculation point n * 

should be corrected to n 0 as 

n 0 = n * + n, 

27 

where n is the corrected value of the calculation point number 

density. The correction value of calculation point number 

density n is related to the velocity correction u through 

the mass conservation equation. 

n 

n 0 t + ·1−cu a + c t 

u t 

a 

=0. 28 

With Eqs. (26)–(28), Poisson equations of pressure are 

obtained as 



From Eq. (12), we obtain 

2 P a k+1 = a 

t 2 n * − n 0 

n 0 , 29 

2 P t k+1 = t 

t 2 n * − n 0 

n 0 . 30 

2d 

n 0 ji 

P j − P i wr ij =− m n * i − n 0 

t 2 n 0 

m = a,t. 

31 

Solving the simultaneous equations [Eq. (31)] for unknowns 

P gives the pressure at each calculation point. With 

Eq. (9), the gradient of the pressure is obtained as 

P i =− d 

n ji P j − P i r ij 

0 wr ij . 32 

r ij 

Finally, the velocities of air and toner at next step k1 

are obtained using Eqs. (22)–(25) and (32). Here, the equations 

for air and toner cloud are described separately in order 

to point out the interaction between toner cloud and air 

explicitly. However, the program code can be written in the 

same way with one flow model using two kinds of densities 

and two kinds of coefficients of viscosity. As described 

above, we can obtain the motion of the toner, airflow, and 

the electric field using only calculation points. 

TEST CALCULATIONS 

Electric Potential 

In order to confirm the validity of the application of the 

Laplacian model in the MPS method for the calculation of 

electric potential, we performed a calculation of two cases of 

potential distribution with and without charge. The first case 

is an analysis of the gap between parallel plates with sinusoidal 

potential and ground potential as shown in Figure 1. 

The analysis area is of width L=1 mm, and height 

h=0.2 mm. The potentials applied to the upper and lower 

plate are 

r ij 

upper x =0, 

lower x = V 0 sin 2 x . 

33 

34 

The peak voltage is V 0 =200 V and the wavelength is 

=L. The calculation points are placed in a 10020 matrix. 

The top and bottom rows of the calculation points are set to 

Figure 1. Calculation geometry of parallel plates. 

Figure 2. Comparison between numerical solutions by the MPS Laplacian 

model and analytical solution of the electric potential formed between 

the parallel plates with applied sinusoidal voltage. 

be the potentials upper and lower , respectively, as the 

Dirichlet boundary condition. The periodic boundary condition 

is applied in the x direction. The size of the weight 

function is r e =2.1l 0 ,wherel 0 is the distance between neighboring 

calculation points. Two rows of dummy calculation 

points are added at the outside of the plates to avoid the 

error of number density of calculation points near the 

boundary. 

Figure 2 shows a comparison between the numerical 

results by the present method and the analytical solution. As 

a whole, the calculated potential is in good agreement with 

the analytical solution. The maximum error, which is generated 

near the lower plate, is 2.5%. The error is caused by 

the fact that in order to simplify the handling of the boundary 

condition, the boundary condition is located not between 

calculation points but at the calculation points themselves. 

The second case is an electric potential analysis that 

involves toner charge. A toner layer, with depth of 

0.035 mm, dielectric permittivity of 2.0, and a charge density 

of 1.5 C/m 3 , is placed on the lower plate in the geometry 

as shown in Fig. 1. The potential distribution formed 

between the parallel plates with potentials of −200 V and 

0V was calculated using the Laplacian model in MPS as 

represented in Eq. (15). Figure 3 shows a comparison between 

the numerical results by the present method and the 

analytical solution. The left vertical dotted line denotes the 

surface of the toner layer. These results show that the Laplacian 

model in MPS correctly calculates the electric potential 

with or without space charge. 

Airflow 

A test calculation of the airflow caused by the motion of a 

toner cloud and the force exerted on the surface of toner 

cloud by air in motion were performed using the present 

method. To verify the interaction between the airflow and 

the surface of the toner cloud, insofar as the motion of the 

toner generates airflow, we used a toner cloud moving with a 

uniform velocity and a uniform shape. The calculation 

points representing the toner cloud are placed in the shape 

of a trapezoid and move at a constant velocity of 0.1 m/s in 



Figure 3. Comparison between numerical solutions by the MPS 

Laplacian model and the analytical solution of electric potential with toner 

charge. 

Figure 5. Comparison of results calculated between by the present 

method and by the FEM method, a x-component of viscous force, b 

y-component of viscous force, and c pressure that exerted on the surface 

of a toner cloud in the shape of trapezoid by airflow. 

Figure 4. Distribution of velocity of air calculated a by the present 

method and b by the FEM method. 

the x direction. The analysis area is of width L=1 mm, and 

height h=0.2 mm. The periodic boundary condition, 

whereby the calculation points leaving from one boundary 

are returned through the other, is applied in the x direction. 

The air density and viscosity are =1.205 kg/m 3 and 

=1.8210 −5 Pa s, respectively. The system is initially at 

rest. The time step used is t=110 −6 s. The steady state 

of the same problem is calculated using a commercial finite 

element method (FEM) code for comparison. In this FEM 

calculation, the toner cloud is modeled by a boundary of a 

trapezoidal shape. The boundary condition is the inflow/ 

outflow condition with the velocity of the toner cloud instead 

of moving the trapezoidal geometry to simulate the 

motion of toner cloud. The no-slip condition is applied at 

the upper and lower plates. 

Figure 4(a) shows the distribution of the velocity of air 

at t=50 s calculated by the present method. Figure 4(b) 

shows the result of the steady-state condition calculated by 

the FEM code. The viscous force and pressure acting on the 

surface of toner cloud by air in motion was calculated using 

the present method and are shown in Figure 5. The results 

calculated by FEM are also plotted for comparison. 

The velocity field calculated using the present method 

including the vortex over the trapezoidal shape is in good 

agreement with the result using the FEM code for the most 

part. There is a small discrepancy between the present 

Figure 6. Initial configuration of calculation points for analysis of traveling 

wave toner transport. 

method and the FEM code in calculating the viscous force 

and the pressure. However, the results are adequate for our 

qualitative estimate. 

TRAVELING WAVE TONER TRANSPORT 

The results of the verification tests described above show 

that the present method is adequate for the calculation of 

the electric field, the airflow caused by the motion of toner 

and the viscous force, and the pressure acting on the surface 

of toner cloud by air in motion. We employed the present 

method for the analysis of the behavior of a toner cloud 

transporting with a traveling electrostatic wave. 

The initial configuration of the calculation points is 

shown in Figure 6, and the parameters of the calculation are 

listed in Table I. The periodic boundary condition is applied 

in the x direction. The traveling potential applied to the 

lower plate is as follows: 

0 x,t = V 0 sin 2 x − vt , v = f, 35 

where v is the phase velocity. 

In our first simulation by the present method, we performed 

calculations to obtain the frequency characteristics 

of the transport velocity of the toner. The frequency of the 

traveling wave is swept from 0Hzto 3.0 kHz in the period 



Table I. Parameters of toner transport model. 

Analysis Area 

Width L m 

Height h m 

Pitch of calculation-point l 0 m 

1.0 10 −3 

0.2 10 −3 

0.01 10 −3 

Traveling Electrostatic Wave 

Wavelength m 

1.0 10 −3 

Pressure 4.0l 0 

Applied Voltage V 0 V 200 

Figure 7. Distribution of toner particle velocity vs phase velocity of traveling 

wave. 

Frequency kHz 0–3.0 

Toner 

Charge Density Q C/m 3 0.3 

Relative Permittivity 1.2 

Bulk Density t kg/ m 3 300 

Viscosity µ t Pa s 0 

Air 

Relative Permittivity 1.0 

Density a kg/ m 3 1.2 

Viscosity µ a Pa s 

1.82 10 −5 

Time Step 

Step s 

1.0 10 −6 

Span s 

75 10 −3 

Size of Weight Function 

Electric Potential 2.1l 0 

Number Density 2.1l 0 

from t=0 ms to 75 ms in the calculation. In other words, 

the phase velocity is varied from 0msto 3.0 ms at the wavelength 

of traveling wave =1 mm. We focus on the effect of 

electrostatic force and air drag. The viscosity of toner cloud 

flow is taken to be zero to exclude the effect of friction 

between toner and the plate, to which the traveling wave 

potential is applied, and the effect of friction among toner 

particles. 

Figure 7 shows the velocity distribution of the toner 

particles with airflow caused by the motion of toner as a 

function of the phase velocity of traveling wave. The solid 

line in Fig. 7 denotes the phase velocity of the traveling wave, 

and the dotted lines denote the critical velocity at which the 

Stokes’ force is equal to the electrostatic force of one particle 

in quiescent air. In Fig. 7, as the wave phase velocity of the 

traveling wave increases from zero, the toner particles move 

synchronously with the wave phase velocity until a certain 

wave phase velocity. Beyond this phase velocity, the number 

of toner particles that move synchronously with the wave 

phase velocity decreases gradually with an increase in phase 

velocity. Finally, the velocities of all toner particles do not 

synchronize with the wave phase velocity. 

Figure 8. A snapshot of position of toner particles and velocity field of 

airflow: a Initial starting step, b the step when the toner particles 

rearrange in response to the electric field, c the step when the toner 

cloud deforms due to air force, d the step when part of toner particles 

separate from the toner cloud due to air force, and e the step when all 

the toner particles are out of phase with the traveling wave and the initial 

shape of toner cloud disappears. 

The motions of toner particles and air were observed as 

an animation created using the numerical results. Figure 8 

shows snapshots of the toner particle position and airflow 

velocity from the viewpoint of a coordinate moving with the 

traveling wave. The contour line of the electric potential, 

which is formed by the traveling wave potential applied on 

the plate and toner charge, is superimposed. Figure 8(a) 

shows the positioning of calculation points for toner, which 



Figure 9. a Maximum synchronous frequency as a function of number 

of toner particles to be transport and b average velocity of airflow as a 

function of number of toner particles to be transported. 

are placed in the shape of a chord, geometrically, at the 

initial step. Figure 8(b) shows the snapshot of the step when 

the toner particles rearrange in response to the electric field. 

Figure 8(c) shows the snapshot of the step when the toner 

cloud deforms due to air force. In Fig. 8(c), the cluster of 

toner, which is around the trough of potential wave at low 

velocity in Fig. 8(b), moves to negative direction of x axis 

due to air resistance. As shown in Fig. 8(d), at higher phase 

velocity, the airflow around the surface of the cluster of toner 

causes a separation of the toner particles at the surface of the 

cluster of toner and the separated toner particles do not 

move in synchronization with the phase velocity of the traveling 

wave any longer. Finally, all the toner particles are out 

of phase of the traveling wave and the initial shape of cluster 

of toner disappears, as shown in Fig. 8(e). 

In our second simulation, we performed calculations of 

various numbers of toner particles and considered the influence 

of air drag. Figure 9(a) shows the relationship between 

the maximum frequency at which the particles move synchronously 

with the wave phase velocity and the number of 

toner particles to be transported. Figure 9(b) shows the average 

airflow velocity in x direction as a function of the 

number of toner particles. As the number of toner particles 

increases, the velocity of airflow increases and the maximum 

synchronous frequency increases. 

As mentioned above, we confirmed that air drag caused 

by toner motion increased the maximum synchronous frequency, 

at which the toner particles moved synchronously 

with a phase velocity of a traveling wave. 

CONCLUSION 

A simulation method to analyze the behavior of a toner 

cloud driven by a traveling electrostatic wave, taking into 

account the effect of air motion, is proposed. Several calculations 

of traveling wave toner transport using this method 

were performed. The results demonstrate that airflow 

dragged by toner particles plays an important role in the 

behavior of toner particles transporting with a traveling electrostatic 

wave. 

In the traveling wave toner transport, the friction between 

toner and plate, to which traveling potential is applied, 

or the friction among toner particles in addition to 

electrostatic force and air resistance are important. These 

frictions can be simulated using the viscosity of toner in our 

two-flow model. Actually, adhesion of charged toner to electrode 

and cohesion between toner particles occur. We will 

approach these subjects in the future. 

However, the present method calculates the electric potential 

and the motion of the toner with common calculation 

points to avoid the complexity of data exchange between 

the electric potential analysis using meshes or grids 

and toner motion analysis using particles. This present 

method allows a direct calculation of the electric potential 

distribution that varies with the motion of charged toner 

particle. Moreover, using the same calculation points, airflow 

can also be calculated. This makes it possible to calculate the 

behavior of the charged particles in a fluid. The present 

method can be used to simulate not only toner transport but 

also in eletrophotographic liquid development in which the 

charged toner moves in a carrier fluid or an electrophoretic 

paperlike display. 

ACKNOWLEDGMENT 

The authors are grateful to Chin-Che Tin of Auburn 

University for his critical reading, kind suggestion, and improvement 

of the English of this article. 

REFERENCES 

1 S. Masuda, K. Fujibayashi, K. Ishida, and H. Inaba, “Confinement and 

transport of charged aerosol clouds by means of electric curtains”, Trans. 

Inst. Electr. Eng. Jpn., Part B 92-B, 9–18 (1972) [in Japanese]. 

2 J. R. Melcher, E. P. Warren, and R. H. Kotwal, “Theory for puretraveling-wave 

boundary-guided transport of tribo-electrified particles”, 

Part. Sci. Technol. 7, 1–21 (1989). 

3 J. R. Melcher, E. P. Warren, and R. H. Kotwal, “Theory for finite-phase 

traveling-wave boundary-guided transport of triboelectrified particles”, 

IEEE Trans. Ind. Appl. 25, 949–955 (1989). 

4 J. R. Melcher, E. P. Warren, and R. H. Kotwal, “Traveling-wave delivery 

of single-component developer”, IEEE Trans. Ind. Appl. 25, 956–961 

(1989). 

5 F. W. Schmidlin, “A new nonlevitated mode of traveling wave toner 

transport”, IEEE Trans. Ind. Appl. 27, 480–487 (1991). 

6 F. W. Schmidlin, “Modes of traveling wave particle transport and their 

applications”, J. Electrost. 34, 225–244 (1995). 

7 Y. N. Gartstein and J. G. Shaw, “Many-particle effects in traveling 

electrostatic wave transport”, J. Phys. D 32, 2176–2180 (1999). 

8 R. Kober, “Simulation of Traveling Wave Toner Transport”, Proc. IS&T’s 


9 M. Maeda, S. Ozawa, and M. Takeuchi, “Calculation of potential 

distribution in the vicinity of photoconductive drum and rotating high 

resistive roller with particle method”, J. Imag. Soc. Jpn. 45, 90–96 

(2006). 

10 J. P. Morris, P. J. Fox, and Y. Zhu, “Modeling low Reynolds number 

incompressible flows using SPH”, J. Comput. Phys. 136, 214–226 (1997). 

11 S. Koshizuka and Y. Oka, “Moving-particle semi-implicit method for 

fragmentation of incompressible fluid”, Nucl. Sci. Eng. 123, 421–434 

(1996). 

12 H. Gotoh, Computational Mechanics of Sediment Transport (Morikita 

Publishing, Tokyo, 2004), p. 110 [in Japanese]. 




High Speed Imaging and Analysis of Jet 

and Drop Formation 

Ian M. Hutchings, Graham D. Martin and Stephen D. Hoath 

Inkjet Research Centre, Institute for Manufacturing, Department of Engineering, 

University of Cambridge, Mill Lane, Cambridge CB2 1RX, United Kingdom 

E-mail: gdm31@cam.ac.uk 

Abstract. New techniques have been developed for analyzing, in 

detail, the shape and development of ink jets and drops. By using 

flash illumination of very short duration (ca. 20 ns), high quality, 

single-event digital images of jets and drops can be captured. A 

computer program, PEJET, has been written to automate the processing 

of such images and to generate quantitative data about the 

whole ink stream. From this data, it is then possible to compute the 

variation in fluid volume, volume flow, and velocity as a function of 

both position and time. The method has been shown to have high 

accuracy. The results can be used to study the influences of nozzle 

design, drive waveform, and fluid properties on jet and drop formation, 

as well as to provide accurate data for comparison to the results 

of computational modeling. Examples of results from a dropon-demand 

system are presented that illustrate the potential of the 

method to compare quantitatively the performance of print systems 

and inks. © 2007 Society for Imaging Science and Technology. 


INTRODUCTION 

As the range of applications of ink jet systems expands, so 

the need grows to understand their performance with an 

ever-widening range of fluids, which are sometimes difficult 

to print. These may be fluids with complex rheologies or 

fluids containing high loadings of solid particles. 

The visualization of jet and drop formation close to the 

print head can lead to insights into how the system is 

behaving. 1–3 From detailed measurement of the shape and 

size of jets as they develop over time, the volumes and velocities 

of the jets and the drops which form from them can 

be computed. These results can then be used to compare the 

performance of print head designs and printing fluids, and 

check the validity of numerical models. 

This paper describes techniques that have been developed 

to make such quantitative observations and gives examples 

of different ways in which this information can be 

used to explore the performance of inks and print heads. 

We have examined model inks jetted from a Xaar dropon-demand 

print head as part of development work at the 

Cambridge Inkjet Research Centre, in some cases replicating 

previous studies in order to establish techniques and to underpin 

more novel analyses. In particular, we present some 

work used to provide input to current theoretical models 

developed by our research partners. 4 

APPARATUS AND EXPERIMENTAL TECHNIQUES 

Figure 1 illustrates the equipment used in this work. The key 

elements are as follows: 

• a high resolution digital camera with a lens system capable 

of imaging a field of view of a few millimeters, 

connected to a PC for control and image storage 

• an ink jet print head with drive electronics and data 

source 

• a very short duration 20 ns flash light source with 

delivery optics 

• means to delay the flash relative to the initiation of the 

printing event and to measure that delay time accurately 

The experiments for the measurements described below 

used a Xaar 126-200 print head with a linear array of 

50 m diam nozzles. The fluid was a simple semitransparent 

UV-curable ink with a Newtonian viscosity of 

20 mPa s at 25°C. 

Jets and drops emerging from ink jet printers are commonly 

visualized stroboscopically by creating composite images, 

superimposing tens or hundreds of individual events in 

a single frame. 5 This procedure relies on the repeatability of 

the drop formation process to give the impression of observing 

individual jets and drops. In some cases, as shown in 

Figure 2(a), the events are not completely reproducible, 

which results in some blurring of the image, particularly at 

long times after drop ejection. For stroboscopic images of 

this kind the flash duration is typically 1 s, which, given 

the high velocity and small size of the objects being ob- 

Received Mar. 1, 2007; accepted for publication Jun. 20, 2007. 

1062-3701/2007/515/438/7/$20.00. 

Figure 1. Schematic of apparatus. 

438

Hutchings, Martin, and Hoath: High speed imaging and analysis of jet and drop formation 

Figure 2. Comparison of images from different techniques: a composite 

image strobed illumination and b single event 20 ns flash 

illumination. 

served, can result in significant movement blur. However, 

some information about the development of the jet and 

drops can be obtained from strobed images by changing the 

timing or phase of the illuminating flashes relative to the 

printing event. 

Alternatively, a high-speed camera can be used to observe 

single events as they occur. 6 This requires the use of a 

camera with a very high framing rate (in megahertz), but 

typically the pixel resolution of such equipment is poor and 

only a small number of frames can be captured. 

All the images used in this work were obtained by using 

a short duration 20 ns flash and a high resolution still 

camera. Hence, each captured image was of a unique event. 

In cases where the events are reproducible, a sequence can be 

built up by taking pictures of successive events at increasingly 

greater time intervals from an appropriate event trigger 

signal. The time delay before firing the flash was set by using 

signal delay electronics and was measured with an oscilloscope 

detecting the trigger signal and an output from the 

flash. 

The apparatus shown in Fig. 1 was used to capture individual 

printing events. A typical image is shown in 

Fig. 2(b). Because of the very short duration of the flash 

illumination, there is no significant motion blur. In cases 

where the events are reproducible, the time development of 

the event can be investigated by capturing a series of images 

with different delays. Less reproducible events can also be 

captured repeatedly and information on their variability determined. 

Because of the high quality of the imaging, it is 

possible to determine drop and ligament sizes and shapes 

and then to use this information to compute drop and ligament 

volumes. By comparing images within timed sequences, 

the velocities of the various components of the ink 

stream can also be determined. 

The Xaar XJ126-200 drop-on-demand print head uses a 

shared-wall piezoelectric design to reduce the head drive 

voltage needed to form jets of a given velocity. As a result, 

neighboring nozzles are influenced by the need to actuate 

the shared walls. The print head can be driven in a singleshot 

mode that fires a particular group of nozzles at one 

time, cycling around the three groups A-B-C-A-B-C- and so 

on, but in our application, a single trigger event caused the 

print head to print in the sequence A-B-C with fixed short 

time delays between the three groups. The actual nozzles 

fired depend on the image presented to the print head. In 

this work, the images were solid blocks arranged either to 

print all the nozzles, a designated block of 16 nozzles, a 

single A-B-C set, or a single nozzle from group A, B, or C, 

according to the phenomenon under study. 

A logic circuit was devised to handle the asynchronous 

nature of the print command and the print head cycle time 

clock, in order to ensure reproducible firing of individual 

nozzles on a timescale of 1 s. The actual ink jet printing 

starts at a fixed time in relation to the clock edge, so that 

without the use of this logic circuit, an additional undesirable 

timing jitter, equal to the clock cycle time of 1.75 s, 

would be introduced, thereby disrupting pseudosequences 

obtained with shorter time steps. Residual timing uncertainties 

were 0.70 s, due to randomness in the triggering of 

some flashes, which were also associated with weaker light 

output and, thus, darker images. An oscilloscope was used to 

determine the precise relative timing of the nozzle firing and 

image capture to a precision of 20 ns. The timings of all 

images presented here are accurate to 0.1 s. 

IMAGE PROCESSING AND DATA EXTRACTION 

To extract quantitative information from the images, it is 

necessary to analyze them and decide which parts of the 

image belong to the background and which to the ink drops 

and ligaments. Standard techniques and various proprietary 

image processing tools are available to find edges and objects 

within images. 7 However, there are particular features of 

these backlit images, which make the use of such tools laborious 

and sometimes inaccurate. In particular, the background 

intensity often varies both within each image and 

from image to image. Within a single image, there is often a 

thin extended ligament that starts at or near the nozzle. 

Shading by the nozzle means that the background intensity 

varies significantly along the length of the ligament. Although 

the light source provides a very short duration pulse, 

the nature of the source means that the intensity can vary by 

10% or more from image to image. Drops and ligaments 

from a transparent fluid often show light central regions 

because light from the bright-field source passes straight 

through these areas, rather than being refracted away from 

the optical path as it is at the edges of these features. 

The example in Figure 2(b) illustrates some of these 

issues. An image processing method was developed, which 

would cope with these variations and artifacts and allow 

hundreds of images at a time to be analyzed automatically. 

Physically reasonable assumptions can be made about the 

objects being imaged to simplify this task. For example, liquid 

drops and jets will not have holes within them and the 

edges should be represented by smooth curves at a pixel 

level. The area within the image in which the drops and 

ligaments are expected to appear is often known. Fixed features 

within all the images in a timed sequence can be used 

to compensate for any small, inadvertent positional movement 

between the object and the imaging system during the 

process of image capture. 



Figure 3. Image processing sequence. 

Figure 4. Raw image above and processed version below, showing 

the results of automated feature selection and edge detection. The largest 

drop in this image has volume of 80 pl and the smallest a volume of 

0.2 pl. 

The image processing is carried out in several steps, as 

shown in Figure 3. First, from a selected area of interest (a) 

a relatively fast but inaccurate technique is used to find the 

approximate edges of the features by looking for regions of 

rapid brightness change (b). In the next step (c), the edges 

are examined in detail to determine which parts of the edges 

are inside and which parts are outside the feature, based on 

a threshold level determined by the local ranges of brightness 

levels. Finally, any “holes” within the features are filled 

(d). 

A computer program, PEJET, was written, which incorporates 

these processing techniques, together with a way to 

select a region of interest within a set of images. The program 

allows a fixed datum to be defined within the image, 

which is then used to correct for any slight shift of the camera 

relative to the object over the course of the experiment. 

The program includes a way to detect and label each feature 

and to output images indicating the features selected. It also 

outputs the size data associated with each feature into a text 

file or spreadsheet. The program can be set up to process a 

complete set of images (for example, a timed sequence) 

without operator intervention. Figure 4 is an example of the 

image output from this program in which the various drops 

and ligament have been recognized and their edges indicated. 

The program has not been optimized for speed because 

it is not used to process images as they are captured but 

rather to batch process them subsequently. The sensitivity of 

the edge detection algorithm can be adjusted to suit the 

contrast and noise level within the image. It can also be set 

to reject unwanted or spurious images by defining the region 

of interest and rejecting spots below a preset level. However, 

our experience with the images we have captured has been 

that the program will reliably detect features four or more 

pixels across. A drop with a diameter of four pixels, with the 

magnifications typically used, is equivalent to a drop volume 

of approximately 0.01 pl. The majority of the features observed 

are traveling at 10 m s −1 , and, hence they will move 

200 nm during the 20 ns flash duration, which is 

0.3 pixels at the typical magnifications used here. We 

would not expect the feature velocity to have any significant 

influence on the measurements of volume. 

As well as the calculated data the program also outputs 

images on which the detected edges have been marked. In 

this way, the proper functioning of the program was ensured 

by checking the edges detected; in some cases, the volume 

values were also calculated manually for comparison to those 

calculated by the program. By flashing a second time after a 

short delay, two images of exactly the same drop could be 

captured in the same frame but with the drop moved and 

with an evolved shape. The program correctly calculates the 

same volume for these second drop images. 

The threshold level selected to make these measurements 

can have some effect on the measured sizes of the 

objects. The correct threshold value can be determined in a 

number of ways. A calibration object of known size can be 

imaged in the system instead of the jets, and the correct 

threshold determined by experiment. Alternatively, an image 

feature of known size (such as the nozzles themselves) can 

be used to check the accuracy of the measured objects. It is 

also possible to consider the effect of optical blurring on a 

sharp edge and to compare edges in the images to those in 

computed images. 

Once the images have been processed, the dimensions 

and volume of each component of the ink stream can be 

computed. The camera and print head are usually arranged 

so that the drop and ligaments travel along the vertical axis 

of the image. If it is assumed that the ink stream has a 

circular cross section at all points, then each horizontal line 

of pixels in an object represents a circular slice through that 

object. By summing these slices, the volume of the whole 

stream, or of any horizontal slice through the stream, can be 

estimated. By comparing such measurements at various 

stages of drop development, a picture can be built up of how 

the volume of the object, or any part of it, changes over time. 

Particular features, such as the tip of the jet or drop or its 

center of mass, can be tracked, and the method thus provides 

a powerful tool to generate velocity and flow information. 

The evolution of drop and satellite formation can be 

tracked by dividing the ligament and drop into volume elements 

and processing successive images to calculate how the 

volume elements move over time. In the initial image of the 

sequence, each volume element is chosen to contain a fixed 

number of pixel slices (Figure 5) and the volume of each 

element is computed. In subsequent images, the boundaries 

of each fixed volume element can be determined by summing 

the volume of ink from each pixel slice until the volume 

of the element has been reached. In cases where, after a 



Figure 5. Tracking jet development. 

Figure 7. Raw and processed images showing satellite separation and 

lateral fluctuations. 

Figure 8. Image showing tail deflection while the ligament is still attached 

to the nozzle; the orifices of the neighboring nozzles are visible 

and no other nozzles were fired. 

Figure 6. a Group of jets emerging from a Xaar XJ-126 print head. The 

distance between neighboring nozzles is 137 m. b Image at a later 

time, for the nozzles shown in a, showing the evolution of the jets to form 

ligaments, satellites and main drops. 

short initial ink ejection period, the volume of ink external 

to the nozzle remains substantially constant, the accuracy of 

this determination can be improved by normalizing the total 

volumes measured from individual jets in successive images. 


In the present work, the print heads were operated with 

model UV-curable inks and the main ink drops had velocities 

of 6 ms −1 after 1mmflight. Ink jets emerging from 

an array of nozzles (spaced at 137 m), together with more 

fully developed ligaments, are shown in Figure 6(a). These 

sharp images resulting from a single 20 ns flash may be 

compared to those reported elsewhere (e.g., Ref. 6). 

During the early stages of the ejection of ink through 

the nozzle, the jet ligament becomes rapidly narrower, down 

to a diameter smaller than that of the orifice at the nozzle 

plate. This implies that the meniscus must lie within the 

nozzle, thereby allowing some air to enter the nozzle. 

Figure 6(b) displays jets at various later stages, showing 

ligament stretching, the snapping of ligaments, the formation 

of satellites, and almost spherical main drops followed 

by trails of satellites. 

Example images are shown in Figures 7 and 8. Figure 7 

shows a raw image and corresponding processed image with 

four jets, one after separation of a tail satellite, with lateral 

fluctuations, tail ligament thinning, and beading visible. It 

was possible, in principle, that these tail deflections could 

occur through the influence of adjacent jets via aerodynamic, 

acoustic, or other means. Figure 8 was obtained by 

firing only a single nozzle and demonstrates that deflection 

of the tail can occur without requiring influences from jets 

fired from neighboring nozzles. The time after firing at 



Figure 9. Edge profile for a jet that has broken away from the nozzle 

vertical bar, showing a satellite and the center projection dotteddashed 

broken line of the mass flow from the nozzle. A dotted line 

represents the lateral deflection of the end of the tail. The lateral scale is 

exaggerated. 

which deflection occurred was similar in all images, whether 

for single jets or multiple groups, or after cleaning the nozzle 

plate. Figure 8 also shows the ink patches associated with the 

orifices of the two unfired neighboring nozzles. 

Break off occurs when the stretched ligament thins 

down and snaps. The two sections then follow independent 

histories: one will probably move backward into the nozzle 

orifice and the other will retract toward the main head. Either 

or both sections may then produce satellites, but usually 

the initial rupture is close to the nozzle and all satellites 

originate from the long ligament. 

Figure 9 shows the edge profile, derived by the methods 

described above, from a jet that had broken. The time after 

the first rupture was enough to allow a satellite to form, and 

thus, at least one more ligament break had occurred between 

the nozzle plane and the main ligament. The ligament width 

fluctuations suggest that other breaks may have been imminent. 

In this image, the ligament does not appear to point 

away from the same apparent origin (dotted-dashed line) 

throughout its length but shows a distinct angular deviation 

by 1° at 520 m. The later tail section and the satellite 

appear to point (dotted line) from another location 

10 m off center. We believe that this implies that the thin 

ink jet ligament became attached to the edge of the nozzle. It 

is possible that at some point as it thins, the ligament becomes 

unstable in a central position and that some small 

disturbance or asymmetry will then cause it to move to the 

edge of the nozzle. 

Ligament development proceeds, as described earlier, 

with the stretching of the material between the nozzle plane 

and the ink jet head. The stretched ligament has a small but 

finite minimum radius during the process; we have found 

that the ink jet ligament can be represented by a truncated 

cone with a specific half angle at the typical distance at 

which ligament rupture occurs near the nozzle, even if the 

ligament happens to rupture elsewhere before this, due to 

the width fluctuations in the thinning ligament. 

The tail width fluctuations observed in still-attached 

ligaments are consistent with net mass movements along a 

truncated cone shape fitted to the downstream ligament behind 

the jet head; this cone had a diameter of 6 m at the 

Figure 10. Independence of tail deflection and tail width fluctuations. 

The center of the jet head was at a position of 888 m. b Tail deflection 

without width fluctuations at an earlier time for the nozzle of a. The 

center of the head was at a distance of 826 m. 

position of the nozzle plane. This is illustrated in 

Figure 10(a), which shows part of the 0.9 mm long ligament 

attached to the nozzle, together with a straight line 

representing an equivalent undisturbed conical profile 

matching the ink material volume. The ink jet head is not 

shown because it was 50 m wide. The ligament width is 

clearly not constant along the length, nor is it zero near the 

nozzle plane: the minimum ligament width was 6.4 m. 

These data relate to the conditions 120 s after printing: 

strong width fluctuations appear at positions up to 

250 m from the nozzle. The trajectories of the main 

drop and its ligament were accurately directed away from the 

center of the nozzle orifice for a relatively long time, until 

the tail was apparently disturbed in some way. This suggests 

that it would be wrong to model the system as axisymmetric. 

At later times, the ligament tail appeared to be straight but 

offset, and originating from the edge of the nozzle opening. 

These effects were not random; they were consistent and 

different for each nozzle, perhaps reflecting imperfections at 

the nozzle edges or some other variability, such as nozzle 

wetting or contamination. 

Figure 10(a) also shows the lateral position of a ligament 

tail plotted on the same scale as the width fluctuations. 

The center of the ligament lies accurately on the line joining 

the center of the head to the center of the nozzle back to a 

distance of 300 m from the nozzle, then shows a lateral 

displacement of up to 2 m before returning to the nozzle 



Figure 12. Ink volume passing beyond various downstream planes vs 

time. 

Figure 13. Illustration of the shapes of three long jets that have been fitted 

by simple empirical functions. 

Figure 11. a Jet tip profiles for nominally 6 m/s ink drops emerging 

from a 50 m diameter nozzle, at various times in microseconds following 

emergence. b Evolution of the diameters of the jet tip maximum 

width and ligament minimum width. c Percentage of total ink volume 

along a jet beyond the nozzle plane, for various ligament lengths and 

ratios of length to main head width. 

center; the angles between the original ligament axis direction 

and the tail lay between −0.5° at a distance of 300 m 

and +1° at the nozzle plane. Figure 10(b) shows the geometry 

of a jet from the same nozzle as that in Fig. 10(a), but 

16.1 s earlier. At this earlier time, there are no appreciable 

width fluctuations but still a significant lateral shift. The 

phenomena of width fluctuations and lateral shift therefore 

appear to be independent. Furthermore, the timing and distance 

information implies that the lateral shift has moved 

consistently with the axial stretching rate, while width fluctuations 

have appeared within 20 s of Fig. 10(b). The 

graphs in Fig. 10 were generated by averaging the images 

from three events captured with the same time delay. The 

small fluctuations derive from the digital nature of the images, 

which allow only certain values to be obtained for the 

width and position of the ligament and have an intrinsic 

error of ±1 pixel in the width measurement at each point. 

Ligament widths and tip diameters vary with time in a 

systematic fashion. Figure 11(a), for example, shows tip profiles 

of jets at several short times after emerging from the 

nozzle. The precision of the technique is illustrated by the 

linear uncertainty of only 1 pixel =0.61 m in this example, 

similar to the wavelength of the imaging light. Figure 

11(b) shows the rapid growth in the tip width of the emerging 

jet, followed by a rapid fall and then a rise toward the 

final drop size (corresponding to 100 pL printed drop volume 

in this example). The nozzle width is shown for comparison. 

The minimum ligament width falls quickly after the 

emergent tip starts necking, and continues to shrink as the 

ligament stretches. 

When the ligament is long and unbroken, the volume 

lying beyond the nozzle plane stretches as the ligament extends, 

while the volume fraction in the head increases. For 

the example shown in Fig. 11(c), the ratio of the length of 

the ligament to the diameter of the head reaches 20 before 

the ligament snaps, at which point there is still 30% of the 

total volume in the tail and 70% in the head section. The 

total downstream ink volume can be calculated (assuming 

reproducible ligaments with circular cross section) at different 

locations. Figure 12 shows that the ink volume beyond 

the nozzle plane initially overshoots the final drop volume, 

whereas it undershoots at downstream locations even as 



stretching, with additional exponential and cubic terms extending 

to the main drop hemisphere (see Figure 13). As 

noted above, the ligament behind the main drop tapers 

smoothly. 

Figure 14(a) shows a sequence of images of jets from 

the same nozzle as they develop over time. Two volume 

elements have been selected and tracked. The element nearer 

the nozzle exhibits considerable extension during jet development. 

In contrast, the element within the head shows a 

slight contraction, as shown in Fig. 14(b). 

CONCLUSIONS 

By using flash illumination of very short duration (ca. 

20 ns), high quality, single-event digital images of jets and 

drops can be captured. A computer program, PEJET, has been 

written to automate the processing of such images and to 

generate quantitative data about the whole ink stream. From 

this data, it is then possible to compute the variation in fluid 

volume, volume flow, and velocity as a function of both 

position and time. The method has been shown to have high 

accuracy. The results can be used to study the influence of 

nozzle design, drive waveform and fluid properties on jet 

and drop formation, as well as to provide accurate data for 

comparison to the results of computational modeling. 

The level of quantitative information that can be extracted 

from high speed flash images allows jet profiles and 

satellite formation to be studied over time. As examples, 

tail-width fluctuations, lateral deflections, and satellite velocities 

have been quantitatively analyzed. 

It has been shown that, at a late stage, the jet ligament is 

unstable and tends to move away from the center of the 

nozzle. At similar times the ligament rapidly forms fluctuations 

leading to break off and satellites. This late-stage lateral 

displacement of the ink jet ligament occurs in the absence of 

aerodynamic, acoustic, or other influences from adjacent 

jets. 


This work was supported by the UK Engineering and Physical 

Sciences Research Council (EPSRC) and by a consortium 

of industrial partners within the Cambridge Inkjet Research 

Centre. Rhys Morgan is thanked for his help in constructing 

the experimental equipment. 

Figure 14. a Image sequence and b element length evolution. 

close as 17 m (one-third of the nozzle diameter). This indicates 

that some ink flows backward into the nozzle from 

very close range, while the rest moves forward as a ligament. 

In these experiments, the ligament stretched for tens of microseconds 

before it broke off. 

The leading surfaces of the jets studied in this work 

have been found to be very closely hemispherical, once the 

emergent phase has passed, as would be expected if the 

shape is controlled by surface tension and negligible aerodynamic 

flattening occurs. The rear surface of the head exhibits 

a shape that can be approximated by constant, linear, and 

quadratic terms linked to the nozzle plane via ligament 

REFERENCES 

1 P. Pierron, S. Allaman, and A. Soucemarianadin, “Dynamics of jetted 

liquid filaments”, Proc. IS&T’s, NIP17 (IS&T, Springfield, VA, 2001) p. 

308. 

2 S. Allaman and G. Desie, “Improved ink jet in situ visualization 

strategies”, Proc. IS&T’s, NIP20 (IS&T, Springfield, VA, 2004) p. 383. 

3 H. Dong, W. W. Carr, and J. F. Morris, “An experimental study of 

drop-on-demand drop formation”, Phys. Fluids 18, 072102 (2006). 

4 J. Etienne, E. J. Hinch, and J. Li, “A Lagrangian–Eulerian approach for 

the numerical simulation of free-surface flow of a viscoelastic material”, 

J. Non-Newtonian Fluid Mech. 136, 157 (2006). 

5 W. T. Pimbley and H. C. Lee, “Satellite droplet formation in a liquid jet”, 

IBM J. Res. Dev. 21, 21 (1977). 

6 C. Rembe, J. Patzer, E. P. Hofer, and P. Krehl, “Realcinematographic 

visualization of droplet ejection in thermal ink jets”, Recent Progress in 

Ink Jet Technologies II (IS&T, Springfield, VA, 1999) p. 103. 

7 J. C. Russ, The Image Processing Handbook, 4th ed. (CRC Press, Boca 

Raton, FL, 2002). 




Effects of Thin Film Layers on Actuating Performance 

of Microheaters 

Min Soo Kim, Bang Weon Lee, Yong Soo Lee, Dong Sik Shim and Keon Kuk 

Micro Systems Lab, Samsung Advanced Institute of Technology (SAIT), Yongin, Gyeonggi 446-712, Korea 

E-mail: minskim@samsung.com 

Abstract. We investigated effects of thin film layers on actuating 

performance of microheaters. Bubble behaviors on microheaters 

were observed experimentally, and heat conduction characteristics 

in thin film layers were analyzed numerically. Nine kinds of tantalum 

nitride (TaN) microheaters were prepared. Step-stress test showed 

that voltage limits of nonpassivated heaters were 50% of those of 

passivated heaters. Open pool bubble test was carried out using 

deionized water as a working fluid. Nonpassivated heaters produced 

comparable bubbles with only 20–50% of input energy required for 

passivated heaters. However, nonpassivated heaters could be operated 

only in a narrow range of driving voltage. We constructed a 

hybrid model for bubble nucleation prediction correlating nucleation 

times with driving powers. Based on work of bubble formation estimated 

from bubble volume evolution, actuation efficiencies of 

microheaters were calculated and compared. Efficiencies of 

nonpassivated heaters were much higher than those of passivated 

heaters. However, nonpassivated heaters failed to show robust actuating 

characteristics over a wide range of power density. As promising 

microactuators, nonpassivated heaters need further investigation 

from a viewpoint of reliability. © 2007 Society for Imaging 

Science and Technology. 


INTRODUCTION 

Microheaters are one of the actuators widely adopted in 

various microsystems. Their materials and simple structures 

are compatible with standard silicon processes. Also, large 

actuating force enables the size small compared to the other 

actuators, such as piezoelectric, electrostatic, electromagnetic, 

and acoustic. This is favorable for an array configuration 

of high spatial density, such as an ink jet print head. 

Characteristics of various types of ink jet heads have been 

summarized. 1 

Many studies of microheaters contributed understanding 

of bubble generation mechanism. Asai 2 proposed a 

bubble nucleation theory to guide the design of thermal 

inkjet heads, and Andrews 3 showed complete bubble cycles 

from nucleation to collapse by visualization techniques. 

Rembe et al. 4 visualized nonreproducible phenomena on 

microheaters with real high speed cine photomicrography. 

Kuketal. 5 studied thermal efficiencies of microheaters, especially 

focusing on the effects of heater size and aspect 

ratio. 

In this study, effects of thin film layers have been investigated 

on actuating performance of microheaters. Bubble 

Received Jan. 8, 2007; accepted for publication May 7, 2007. 

1062-3701/2007/515/445/7/$20.00. 

behaviors on microheaters were observed experimentally, 

and heat conduction characteristics in thin film layers were 

analyzed numerically. Step-stress tests (SSTs) and open pool 

bubble visualization were carried out. Bubble volumes were 

estimated from the images taken experimentally. A hybrid 

model was constructed for prediction of bubble nucleation. 

Work of bubble formation was calculated from bubble volumes 

using both nucleation pressure (i.e., maximum bubble 

pressure) from heat conduction simulation and pressure 

profile following Asai’s model. 6 Finally, actuation efficiencies 

of microheaters with different passivation layers were obtained. 

Their actuating performances were compared, and 

the effects of thin film layers were discussed. 

METHODS 

Preparation of Microheaters 

Microheaters were fabricated using conventional thin film 

processes (Figure 1). Thermal barrier layer of silicon dioxide 

SiO 2 was thermally grown up to 3 m on silicon wafer. 

Tantalum nitride TaN heater and aluminum (Al) electrode 

were consecutively deposited by sputtering and patterning. 

The sheet resistance of TaN film was 50 /. Silicon 

nitride SiN x film was deposited using the PECVD technique 

as a heat transfer layer. Finally, tantalum (Ta) film of 

3000 Å was deposited as an anticavitation layer. 

Nine kinds of microheaters were prepared with different 

thin film layers. Microheaters had a square shape and the 

same planar size of 22 m22 m. Three different thicknesses 

of SiO 2 were 1 m, 2 m, and 3 m. Three different 

types of passivation layers included nonpassivation, SiN x of 

4000 Å with Ta of 3000 Å, and SiN x of 6000 Å with Ta of 

3000 Å. Heaterswerelocated300 m from a chip edge 

for side view images of bubble. Table I lists mean values of 

electrical resistance of the fabricated heaters along with their 

standard deviations. The number of data points was 60 for 

each type. 

A step-stress test has been performed to compare voltage 

limits of different microheaters. Voltage limit means a 

maximum endurable voltage, defined as maximum voltage 

before causing more than 1% variation from the initial value 

of resistance. At a fixed pulses width, electrical pulses of a 

prescribed voltage level were repeatedly given to a 

microheater. After 100,000 cycles at a frequency of 1000 Hz, 

electrical resistance was measured and then the voltage level 

was increased with an increment of 1V. Maximum endur- 

445

Kim et al.: Effects of thin film layers on actuating performance of microheaters 

Table I. Values of electrical resistance of the fabricated microheaters. 

SiO 2 

Passivation Type 1 µm 2 µm 3 µm 

No passivation 

Mean a Ω 60.5 64.0 59.2 

Standard deviation Ω 2.2 1.3 1.0 

SiN x 4000 Å and Mean Ω 60.6 66.1 65.4 

Ta 3000 Å 


SiN x 6000 Å and Mean Ω 61.0 67.6 68.6 

Ta 3000 Å 


a The number of samples is 60 for each heater. 

Figure 2. Experimental setup for open pool bubble visualization. 

Figure 1. Optical microscope images of the fabricated microheaters: a 

nonpassivated heater, 22 m22 m microheater with SiO 2 of 3 m 

and b passivated heater, 22 m22 m microheater with SiO 2 of 

1 m, SiN x of 6000 Å and Ta of 3000 Å. 

able voltage was recorded as the one causing variation more 

than 1% from the initial resistance. 

Open Pool Bubble Visualization 

Bubbles on the heaters were visualized with a microscope in 

an open pool setup shown in Figure 2. A xenon stroboscope 

provided sample illumination through the microscope. The 

bubble was synchronized with the input pulse to the heater, 

and bubble images were captured by high speed CCD 

cameras. 7 Exposure time of the CCD was 0.3 s, and frame 

time of consecutive images was 0.1 s. Estimation of bubble 

volume at a fixed time requires two synchronized bubble 

images, plane and side views with two CCD arrangements as 

shown in Figure 3. CCD1 captured bubble width, length, 

and height could be obtained from CCD2 image. 5 

Square wave electrical heating pulses were applied to 

microheaters with 8Hzrepetition frequency. Heating current 

was measured by a current probe. During pulse heating, 

voltage difference between electrodes was recorded using a 

Figure 3. Two CCDs for capturing both plane and side views of bubble. 

digital oscilloscope. Electrical power multiplied by pulse 

width represents input energy to the heater. 8 

Bubble volume was obtained as a function of time by 

image processing, using both experimental plane and side 

views of the bubble. 5 Work of bubble formation was estimated 

following the same procedure as in our previous 

article. 5 

Hybrid Model for Bubble Nucleation Prediction 

In the present study, three models for bubble nucleation 

prediction were compared to experimental data. In the temperature 

model, a bubble starts to nucleate when interfacial 



Table II. Properties of thin film materials. 

Material 

Density 

kg/m 3 

Heat Capacity 

kJ/kg 

Thermal 

Conductivity 

W/m K 

Si 2340.0 760.0 165.0 

SiO 2 2070.0 840.0 1.4 

TaN 16,600.0 150.0 57.0 

Al 2690.0 898.7 221.5 

SiN x 3,200.0 790.0 1.67 

Ta 16,600.0 150.0 57.0 

temperature T int , which is defined as maximum temperature 

at the interface between heater and fluid, reaches a critical 

temperature T cri . In the heat flux model, 9 instead of a 

single fixed interfacial temperature T int , the critical temperature 

is determined by considering heat flux through the interface 

as follows: 

T heat =230°C+L char 

T 

z , 

where L char is a thermal characteristic thickness of the heat 

flux model. 

However, these two models showed agreement with experimental 

data only over a partial range of power density 

(see Bubble Nucleation). In the present study, a hybrid 

model was constructed from these two models to correlate 

bubble nucleation time with power density. Based on the 

results from the two models, the shorter nucleation time 

between two nucleation times predicted by temperature 

model and heat flux model in the hybrid model was adopted 

as follows: 

t hybrid = mint temp ,t heat , 

where t temp represents a nucleation time determined by temperature 

model, and t heat by heat flux model. 

Numerical analyses of heat transfer in thin film layers 

have been performed using CFD-ACE v2004 for seven different 

heaters: three nonpassivated heaters with thermal barrier 

SiO 2 thickness of 1 m, 2 m, and 3 m, and four 

passivated heaters with passivation layer SiN x thickness of 

4000 Å and 6000 Å, each having two different thermal barrier 

thicknesses of 1 m and 3 m. 

In numerical simulation, thermal properties of thin film 

layers, such as diffusivity, have a great influence on prediction 

of interfacial temperature. However, properties of thin 

film layers have been reported to vary largely, depending on 

measurement techniques as well as fabrication methods. In 

our numerical analyses, bulk properties were assumed for 

thin film layers (Table II). Inevitably, this provided higher 

interfacial temperature than expected. In the present study, 

however, our concern was to obtain relative nucleation time 

as a function of power density rather than absolute value of 

interfacial temperature. As a critical temperature, maximum 

1 

2 

Figure 4. Results from step-stress test of microheaters, presenting voltage 

limits as a function of pulse width for three different passivation types. 

temperature at the liquid interface was used instead of average 

temperature to obtain better agreement with experimental 

data. In the present simulation, maximum temperature 

was 40°C higher than average temperature. Critical temperature 

and thermal characteristic thickness for the heat 

flux model were determined by matching numerical results 

with experimental data for the specific case of 4000 Å SiN x 

and 3 m SiO 2 . In the present study, the values were chosen 

as T cri =360°C and L char =0.2 m. 

RESULTS AND DISCUSSIONS 

Step-Stress Test of Microheaters 

Results from step-stress test are quite straightforward and 

intuitive. For different passivation types, Figure 4 depicts 

voltage limits at several selected pulse widths. Thermal barrier 

thickness SiO 2 was fixed with 1 m. For a specific 

passivation type, higher voltage could be used along with 

shorter pulse width. Voltage limits of passivated heaters were 

about two times higher than those of nonpassivated heaters. 

For example, at a pulse width of 1 s, voltage limit was 5V 

for nonpassivated heater, whereas it was 13 V for passivated 

heater with SiN x 6000 Å and Ta 3000 Å. This implies that, 

in terms of electrical power, existence of passivation layers 

helped the heater endure about six times higher voltage. 

Also, the heater with thicker passivation layers could stand 

higher voltage. On the other hand, the nonpassivated heater 

had a narrow range of driving voltage as compared to passivated 

heaters, which means that nonpassivated heaters have 

a limited operating window. 

Table III shows voltage limits of nine microheaters at 

1 s pulse width. Voltage limits were higher for thicker passivation 

layers. On the other hand, effects of thermal barrier 

thickness seem to appear with increasing thickness of passivation 

layer. Our explanation is that thinner thermal barrier 

layer facilitates relaxation of thermal stresses due to more 

thermal dissipation through itself for the same thickness of 

passivation layer. In the present study, the highest voltage 

limit was obtained for the microheater of the thickest passivation 

layer with the thinnest thermal barrier layer. 



Table III. Maximum voltages at 1 µs pulse width for three different passivation types, 

each with three different thicknesses of thermal barrier layer SiO 2 . 

SiO 2 

1µm 2µm 3µm 

No Passivation 5 5 5 

SiN x 4000 Å and Ta 3000 Å 10 9 9 

SiN x 6000 Å and Ta 3000 Å 13 9 10 

Figure 6. Input energy to microheaters of different passivation types at 

normal bubble creation. 

Figure 5. Driving conditions of microheaters of different passivation types 

for normal bubble creation, along with corresponding SST results in Fig. 

4. 

Driving Conditions for Bubble Actuation 

Figure 5 depicts relations between driving voltage and pulse 

width for bubble creation on microheaters with different 

passivation layers. Again, thermal barrier thickness SiO 2 

was fixed at 1 m. In our open pool visualization a bubble 

that is stable and repeatable in shape as well as saturated in 

size was chosen as a normal bubble. A normal bubble has a 

consistent shape with time (dV/dt0, whereV is bubble 

volume and t is time) as well as a maximal size with driving 

energy (dV/dE0, where E is driving energy). For a 

nonpassivated heater, the available driving voltage margin 

was relatively small. Also, at a fixed pulse width, a normal 

bubble was observed near the voltage limit obtained from 

the SST. However, passivated heaters could produce normal 

bubbles far below the voltage limit. On the other hand, at a 

fixed driving voltage, longer pulse width was required for the 

heater with thicker passivation layer, which means that more 

input energy is needed for heaters with thicker passivation 

layers due to increased energy consumption in the passivation 

layer during energy transfer to fluid. Therefore, more 

input energy should be provided so that the prescribed thermal 

energy can be transferred to fluid for normal bubble 

creation. 

Input energy for normal bubble creation is plotted in 

Figure 6. Nonpassivated heaters consumed input energy 

about 20–50% compared to passivated heaters. More energy 

was required for heaters with thicker passivation layers as 

stated earlier. Experiments showed that input energy decreased 

with increasing driving voltage, i.e., with increasing 

driving power. At high driving power, relatively rapid temperature 

elevation seemed to make nucleation occur earlier 

and thus reduce input energy. On the other hand, it is well 

known that too high a driving power causes nucleation to 

occur too rapidly and too small a bubble forms. 5 Therefore, 

this implies existence of a lower limit in driving power for 

minimizing input energy. For passivated heaters, input energy 

seemed to saturate after around 9V, whereas nonpassivated 

heaters showed no saturation region, possibly due to 

a narrow range of driving voltage from the SST. 

Characteristics of bubble creation are presented in Figure 

7. Bubble nucleation time is depicted in Fig. 7(a) and 

bubble life in Fig. 7(b). The open pool bubble test may have 

asystemdelayof0.1 s between trigger signal and application 

of the actual heating pulse to a microheater. Therefore, 

the obtained nucleation time may have a corresponding 

uncertainty. From Fig. 7(a), nucleation time was largely dependent 

on driving power. Higher driving power resulted in 

earlier bubble nucleation. At the same driving power, faster 

bubble nucleation was observed on nonpassivated heaters 

due to rapid temperature elevation with minimal energy 

loss. On the other hand, from Fig. 7(b), the created bubbles 

lasted longer on passivated heaters. Bubble life decreased 

with increasing driving power. All these observations support 

the inference that bubble life is affected mainly by heat 

transfer time before nucleation. However, effects of thickness 

of passivation layers seemed to be different depending on the 

power density level as seen from Fig. 7(b). Above a certain 

driving voltage (here, 8 V), bubble life became longer on a 

thinner passivation layer. This reversal might be explained by 

more energy transfer to fluid on thinner passivation layers at 

high power density resulting in relatively wider heated region 

adjacent to the heater surface. This distinct behavior, 

depending on power density, needs further analysis. 

Bubble Nucleation, Work of Bubble Formation, and 

Actuation Efficiency 

With values of T cri =360°C and L char =0.2 m obtained 

above, our hybrid model predicted nucleation times in good 

agreement with experimental data as shown in Figure 8(a). 

The temperature model predicted later nucleation at low 



Figure 7. Bubble characteristics: a bubble nucleation time and b 

bubble life for three different passivation types. 

power density than experimental observations [Fig. 8(b)], as 

did the heat flux model at high power density [Fig. 8(c)]. 

This means that a single critical temperature cannot cover 

the whole range of power density and thus it should vary, 

depending on the power density level. From above observation, 

we constructed a hybrid model where the critical temperature 

was determined as the one giving shorter nucleation 

time between the two models [Eq. (2)]. As a result, the 

hybrid model could fit experimental data excellently in the 

whole range of power density. 

Based on the hybrid model, input energy before nucleation 

was calculated as shown in Figure 9(a). For nonpassivated 

heaters, input energy decreased 50% when power 

density changed from 0.8 GW/m 2 to 2GW/m 2 .However, 

for passivated heaters, input energy was little affected by 

change of power density. As shown in Fig. 9(b), the energy 

transferred to fluid before nucleation was inversely proportional 

to the power density for both nonpassivated and passivated 

heaters. 

Estimated work of bubble formation and efficiencies are 

presented in Figure 10. Bubble work decreased with increasing 

power density for all passivation types, as shown in Fig. 

10(a). This might be explained by smaller bubble size and 

shorter bubble life, along with less heat transfer to fluid owing 

to earlier nucleation for higher power density. Results 

also showed that comparable work of bubble formation 

could be obtained from nonpassivated heaters. For passivated 

heaters, thickness of passivation layer SiN x affected 

Figure 8. Prediction of nucleation time from the hybrid model along with 

the corresponding experimental data: a hybrid model, b temperature 

model, and c heat flux model. 

both work of bubble formation and efficiency. Heaters with 

4000 Å SiN x produced more work showing higher efficiency 

than those with 6000 Å SiN x at the same power density. 

On the other hand, as seen in Fig. 10(b), in the power 

density range of 1–2 GW/m 2 , nonpassivated heaters 

showed much higher efficiencies although they produced 

less work. From Fig. 9, nonpassivated heaters could deliver 

almost half the input energy to the fluid. In the power den- 



Figure 9. Simulation results based on the hybrid model: a input energy 

up until the nucleation time and b effective energy transferred to fluid for 

different thin film layers. 

sity of 1 GW/m 2 , the nonpassivated heater could transport 

12% more energy to fluid as compared to the passivated 

heaters, even with less input energy. Therefore, 

nonpassivated heaters could produce comparable work [Fig. 

10(a)] with less input energy [Fig. 9(a)] at lower power density. 

This strongly supports higher efficiency of nonpassivated 

heaters as depicted in Fig. 10(b). 

Furthermore, in Fig. 10(b), efficiency of nonpassivated 

heaters showed an opposite trend with power density. This 

could be explained by a nonlinear effect of passivation layers 

on the heat transfer from heater to fluid. In passivated heaters, 

as power density increases, the temperature gradient at 

fluid interface is kept almost constant due to existence of 

passivation layers, while it increases correspondingly for 

nonpassivated heaters. From Fig. 8, higher power density 

resulted in earlier bubble nucleation, showing a much larger 

rate of decrease for nonpassivated heaters. This implies that 

a relatively large decrease in input energy occurs for nonpassivated 

heaters, as is confirmed again in Fig. 9(a). Earlier 

nucleation means a decrease in heat transfer time for transporting 

thermal energy to fluid. However, for nonpassivated 

Figure 10. a Estimated work of bubble formation and b actuation 

efficiency defined as the ratio of work to input energy. 

heaters, effective energy transferred to fluid is little affected 

due to absence of energy consumption in passivation layers 

in the passivated heaters. Thus, prescribed thermal energy 

for bubble creation can be delivered to the fluid with much 

less input energy (Fig. 9). Also, work of bubble formation is 

roughly proportional to the energy input into the fluid 

[Figs. 10(a) and 9(b)]. Therefore, the actuation efficiency, 

defined as a ratio of work to input energy, increases with 

increasing power density for nonpassivated heaters. 

Ink jet heads have been fabricated adopting some of 

heaters considered in the present study. Two kinds of passivation 

layers (i.e., SiN x =4000 Å or 6000 Å) and two kinds 

of thermal barrier layers (i.e., SiO 2 =2 m or 3 m) were 

tested. Heater sizes were 400 m 2 =20 m20 m or 

576 m 2 =24 m24 m. The fabricated ink jet heads 

adopted a traditional top shooting structure, and one chip 

had 50 heaters. Chamber sizes were 24 m24 m 

17 m =WLH or 28 m28 m17 m for 

each heater. Nozzle plate thickness was 12 m, and orifice 

exit diameter was 12 m. Inlet flow passage was 14 m 

20 m17 m =WLH. Ejected droplet volume 

was two to three picoliters. Table IV provides droplet ejec- 



Table IV. Droplet ejection performance of the inkjet heads adopting some of the 

passivated heaters considered in the present study. 

SiN x 

Å 

SiO 2 

µm 

Heater 

µmµm 

Voltage 

V 

Pulse 

µs 

Energy 

µJ 

Speed 

m/s 

Frequency 

kHz 

4000 2 2020 8 1.3 1.35 16.9 18 

2424 8 1.5 1.51 17.2 15 

3 2020 8 1.0 1.12 15.4 15 

2424 8 1.1 1.21 17.2 18 

6000 2 2020 8 1.4 1.52 15.8 18 

2424 8 1.4 1.44 16.2 18 

3 2020 8 1.4 1.49 16.3 15 

2424 9 1.1 1.52 15.3 15 

tion performance of the fabricated ink jet heads under typical 

driving conditions. Ink jet heads with thicker passivation 

layers consumed more energy for stable ejection. Obviously, 

this is consistent with the result that more energy is required 

for normal bubble creation with a thicker passivation layer 

[Figs. 6 and 9(a)]. In Table IV a thermal barrier layer SiO 2 

seemed to become significant with respect to input energy 

for thinner passivation layers. When the thermal barrier 

layer becomes too thin, undesirable heat loss may increase 

during the heating period. 

CONCLUSIONS 

Results from step-stress test emphasize that maximum 

power density level for the nonpassivated heater should be 

kept low compared to the passivated heaters. Open pool 

bubble actuation indicates that the energy margin is very 

small for nonpassivated heaters. On the contrary, nonpassivated 

heaters showed much higher actuation efficiency 

even with lower input energy, but only over a limited range 

of power density. The hybrid model could provide nucleation 

times in excellent agreement with experimental data. 

As promising microactuators, applicability of nonpassivated 

heaters needs further investigation from the viewpoint of 

reliability. Our investigation will be helpful for development 

of the thermal ink jet heads for more reliable and higher 

performance. 

REFERENCES 

1 K. Silverbrook, US Patent 6,338,547B1 (2002). 

2 A. Asai, “Application of the nucleation theory to the design of bubble jet 

printers”, Jpn. J. Appl. Phys., Part 1 28(5), 909 (1989). 

3 J. R. Andrews, “Micro-boiling and pico-jetting”, IMA Workshop: Analysis 

and Modeling of Industrial Jetting Processes (IMA, Cambridge, UK, 

2001). 

4 C. Rembe, S. Wiesche, M. Beuten, and E. P. Hofer, “Nonreproducible 

phenomena in thermal ink jets with real high-speed cine 

photomicrography”, Proc. SPIE 3409, 316 (1998). 

5 K. Kuk, J. H. Lim, M. S. Kim, M. C. Choi, C. H. Cho, and Y. S. Oh, 

“Research on micro heater efficiency for thermal inkjet head”, J. Imaging 

Sci. Technol. 49(5), 545 (2005). 

6 A. Asai, “Bubble dynamics in boiling under high heat flux pulse 

heating”, J. Heat Transfer 113, 973 (1991). 

7 C. Rembe, J. Patzer, E. Hofer, and P. Kreh, “Real cinematographic 

visualization of droplet ejection in thermal ink jets”, J. Imaging Sci. 

Technol. 40(5), 401 (1996). 

8 J. H. Lim, Y. S. Lee, H. T. Lim, S. S. Baek, K. Kuk, and Y. S. Oh, 

“Visualization of bubbles with various current density distributions in 

thermal inkjet head”, Proc. IEEE MEMS Conference (IEEE, Piscataway, 

NJ, 2003) pp. 197–200. 

9 W. Runge, “Nucleation in thermal ink-jet printers”, Proc. IS&T’s NIP8 





Hybrid Stacked RFID Antenna Coil Fabricated by Ink Jet 

Printing of Catalyst with Self-assembled Polyelectrolytes 

and Electroless Plating 

Chung-Wei Wang, Ming-Huan Yang, Yuh-Zheng Lee and Kevin Cheng 

DTC/Industrial Technology Research Institute, Hsinchu, Taiwan, Republic of China 

E-mail: wangcw@itri.org.tw 

Abstract. This article describes a method of forming a stacked hybrid 

metal structure and pattern to enhance radio-frequency identification 

antenna coil inductance. The essential strategies included 

the use of multilayer self-assembled polyelectrolytes to modify the 

surface property of substrates, an ink jet printing process for a Pd 

containing catalyst, and a stacked hybrid metal layer formed by 

electroless plating in subsequent processes. The results demonstrate 

that the minimum line width and line spacing can reach 

100 m/100 m, and electrical performance is compared to prior 

research in employing different approaches. The method presented 

in this article enhances the capability by adapting to any substrate 

surface using the self-assembled polyelectrolyte technique. Therefore, 

results were verified on different substrates, such as PI, PET, 

and FR-4; satisfactory electric performance for application was obtained. 

In detail, the inductance of the antenna improved from 300 

nH to 20 H for a monolayer coil, and 600 nH to 50 H for a double 

layer coil, depending on the metal thickness. © 2007 Society for 

Imaging Science and Technology. 


INTRODUCTION 

In recent years, printed organic electronics have been intensively 

developed because of their advantages of low cost, easy 

and flexible processing, efficient usage of material, and large 

scale manufacturing compared to conventional lithography 

and vacuum processes. These applications 1 include polymer 

light-emitting diodes (PLED), organic thin-film transistors 

(OTFT), organic bistable memory (OBD), metal circuits, etc. 

According to the forecast data from marketing research institutes, 

printed logic integrated circuits (ICs) will gradually 

attract more attention and will dominate in the future. 1 

Among the logic ICs, radio-frequency identification (RFID) 

tags seem to be one of the most promising vehicles. These 

low cost printed products can replace not only bar codes but 

also expensive inorganic RFID silicon chips used for supply 

chain and automated retail checking. Generally, a simple 

RFID system (Figure 1) consists of a reader and a tag. There 

are many active and passive devices in a tag, including, at the 

least, an antenna, a capacitor, a diode, a transistor, and some 

memory. The system can be operated at lower frequencies, 

such as 125 kHz, 13.56 MHz, or860−930 MHz, and at a 

higher speed of 2.45 GHz for different applications. Many 

companies have focused on printed RFID development with 

organic semiconductors; however, the operation frequency 

of these RFIDs is limited due to the low carrier mobility 

characteristic of organic semiconductors. Hence, a highly 

conductive coil for the tag antenna is necessary. 2,3 

At present, most tag antennas with good conductivity 

are fabricated by copper or aluminum etching. The line 

width and line space are 600−800 m and 200−400 m, 

respectively, in each common etching case. Metal paste consisting 

of silver can be used in screen printing or ink jet 

printing processes but suffers from low resolution, poor adhesion, 

and high cost. In addition, the resulting metal paste 

patterns always need a thermal curing process to sinter the 

metal particles in order to achieve acceptable conductivity. 

The curing temperature increases with silver particle size, 

and it represents a challenge for application to flexible 

substrates, 3,4 especially with the screen process. 

Besides the issue of conductivity, high Q value inductors 

and well behaved capacitors are also required for power coupling 

and communication. High Q inductors can be 

achieved via several methods, which include adding magnetic 

material to the coil, increasing the number of circles, 

and reducing the resistance of the antenna coil. Among 

those measures, the addition of magnetic material is the best 

method for enhancing Q value. 4,5 


1062-3701/2007/515/452/4/$20.00. 

Figure 1. An RFID system. 

452

Wang et al.: Hybrid stacked RFID antenna coil fabricated by ink jet printing of catalyst with self-assembled polyelectrolytes and electroless plating 

Because of the need for high inductance value, in this 

article, a hybrid circuit fabrication method combined with 

magnetic nickel layering onto an ink jet printed copper 

RFID antenna coil is presented. In this approach, the inductance 

of the antenna coil can be enhanced from 300 nH to 

20 H for monolayer cases, and from 600 nH to 55 H for 

double-layer cases. We will discuss the details below. 

EXPERIMENTAL 

Chemical Preparation 

Poly(allyamine hydrochloride) (PAH) (MW55,000– 

65,000) and poly(acrylic acid) (PAA) (MW70,000, 25% 

aqueous solution) were purchased from Aldrich. All chemicals 

were used without further purification. Deionized water 

(DI) was used in all aqueous solutions and rinsing procedures. 

The concentration of the polyelectrolyte-dipping solutions 

was 10 mM, based on the molecular weight (MW) of 

the repeat unit. The pH values of the PAH solution and PAA 

solution were adjusted to 7.5 and 3.5 by the addition of 1M 

HCl and 1 M NaOH, respectively. Layer-by-layer assembly 

was carried out using an autodipping system. Before the 

layer-by-layer assembly process, double layer substrates were 

drilled to afford via holes to connect the two separated antenna 

circuits on each substrate side. 

Ink Formulation and Pd-Catalyst Patterning 

The catalyst ink was prepared by dissolving a Pd complex 

compound in DI water. The ink viscosity and surface tension 

were about 8−20 cps and 40 dyne/cm, respectively. 

The printing process was conducted on a third generation 

print platform, designed and integrated for 550 mm 

650 mm substrates by the Industrial Technology Research 

Institute, Taiwan (ITRI). Polyimide (PI), polyethylene 

terephthalate (PET), and FR-4 substrates with different selfassembly 

layers were printed with the catalyst solution. 

While the solution was drying, the catalyst would be adsorbed 

on the substrate surface and would gradually diffuse 

into the layer of poly(allylamine hydrochloride) beneath; 

therefore, an ink jet defined area was afforded. 6–8 

Electroless Plating of Nickel 

After the Pd-containing catalyst was patterned on the selfassembled 

polyelectrolyte layer, a commercial formulation, 

consisting of 40 g/L nickel sulfate, 20 g/L sodium citrate, 

10 g/L lactic acid, and 1 g/L dimethylaminobenzaldehyde 

(DMAB) in DI water, for the electroless nickel plating, was 

used to further exchange the Ni with the Pd, and leaves the 

Ni at the location of the predetermined pattern. The optimum 

condition to obtain complete transport between Ni 

and Pd corresponds to pH 6.8, as determined by titration 

with 0.3 M ammonium hydroxide solution. Before deposition 

of nickel metal, the substrate needed to be dipped into 

the accelerator solution for about 3−10 s and then washed 

with DI water and dried. In order to keep the solution homogeneous, 

an air bubble generator was used in the plating 

bath. 6–8 

Electroless Plating of Copper 

A pure Ni circuit has good magnetic characteristics but is 

weak in terms of electronic performance, i.e., resistivity. We 

Figure 2. Schematic structures of the a single side antenna coil and b 

double side antenna coil. 

therefore explore how a combined layer of Cu with Ni, in a 

certain thickness ratio, may incorporate both the desired 

electrical and magnetic performance. Accordingly, we tried 

growing a copper layer on the Ni layer by the solution process. 

Theoretically, nickel and nickel plating solutions would 

induce copper ion reduction; the substrate with a nickel antenna 

coil, obtained from the previous plating process, then 

needs to be washed with DI water for 10 min to remove 

nonadsorbed nickel particles and nickel plating solution 

from the surface, and then dried prior to electroless plating 

of copper. All of the reagents for copper electroless plating 

were dissolved in DI water at room temperature with stirring. 

To keep the solution homogeneous, an air bubble generator 

was also used in the electroless plating bath; the pH 

value of which was controlled. The copper deposition was 

carried out at 40 ° C. Immersion time and temperature are 

two key factors with which to control the metal thickness. 

After being removed from the electroless plating solution, 

the samples were rinsed with DI water to remove residual 

plating solution and then set aside to dry. The structures 

of monolayer and double layer antenna coil are shown in 

Figure 2. 7,8 


Table I lists the surface treatment processes to form selfassembled 

polyelectrolyte layers on a substrate. The repeating 

step for PAH/PAA bilayer deposition depends on the 

substrate. From our experience, for example, the PET substrate 

needs three repeats. Similarly, if this surface treatment 

is applied to the polyimide substrate, the number of repeating 

steps increases to seven. An antenna coil pattern with a 

line width of 75−100 m and line space of 75−100 m, 

was accordingly obtained. The resulting conductance can be 

adjusted by varying the nickel and copper electroless plating 

recipes. The edge of a nickel line of 2 m thickness was 

uniform, as shown in Figures 3(a)–3(c). The line edge after 

electroless copper plating was still smooth, as shown in Figs. 

3(d) and 3(e), where the copper metal thickness was 5 m. 

The ratio of Ni film to Cu film is 0.4 2 m/5 m, 

which is determined by the resistivity and inductance. How- 



Table I. Process flow of self-assembled polyelectrolyte layers before ink jet printing of 

catalyst. 

Step 

Substrate cleaning 

Immerse into PAH aq 

Substrate flushing 

Immerse into PAA aq 

Repeating 

PAH/PAA bilayer 

deposition 

Immerse into PAH aq 

10 mM 

Condition 

Washing with toluene 

and DI-water in sequence 

PAH aq 10 mM 

DI-water 

PAA aq 10 mM 

Up to three pairs of PAH 

/PAA layers 

PAH aq 10 mM 

Figure 3. a and d are 100 m wide printed single side antenna coil 

structure 27 circles layout; b, c, d, andf are optical micrographs 

of printed inductor formed on commercial flexible substrates. 

ever, many unexpected particles were present within the line 

spaces [Fig. 3(f)], possibly from unwanted residual nickel 

particles and the nickel electroless solution. To measure the 

electrical characteristics, an RLC meter (Agilent 4294A) was 

employed, and the data thus recorded are shown in Table II. 

Table II indicated that adding Ni will dramatically increase 

the inductances, the L- and the Q-factor, but not reduce the 

resistivity. The inductances of monolayer hybrid RFID antenna 

coils (samples A and B) were about 50 times higher 

than that of the copper antenna coil (sample C), although 

the resistances were similar in both cases. The inductances of 

the double layer hybrid RFID antenna coils (samples D and 

E) were also 90 times higher than for the double layer 

copper coil, in spite of the resistance of a double layer hybrid 

metal antenna being higher than that of the double layer 

copper by 12 . 

The measured deposition rates for electroless nickel and 

copper were 12 m for 60 min and 5 m for 60 min, respectively. 

In practice, the deposition rates were dependent 

on temperature, metal ion concentration, pH value of the 

bath solution, catalyst concentration, activator, the bath stability, 

and finally by electroless deposition time. In this study, 

the electroless bath stability was very important, because it 

influenced both the deposition speed and quality. The maximum 

thickness of nickel, which can be formed is up to 

35 m for over 3h, and copper can be deposited up to 

20 m over 4h. Compared to the results of Shan et al., 8 

who were able to deposit 5 m over 1hat saturation, this 

study made an obvious improvement in both growing rate 

and circuit performance. Accordingly, we changed the formulation 

for electroless plating, and have an activator process 

prior to electroless plating. We adopt the mist activator 

for 5−10 sec to coat a thin layer of activator on catalyst in 

order to speed up the reaction with electroless plating. 

The incorporation of magnetic metal into the coil structure 

can dramatically enhance the inductance and the Q 

factor. The maximum Q factor, near 0.62, was achieved from 

a hybrid metal structure of 2 m nickel and 5 m copper 

in a single side antenna coil structure. A Q factor near 1.35 

was achieved for the double side antenna, which is about 

double than that obtained in the single side case. The conductivity 

of the printed antennas was 90% and 80% that of 

bulk copper for the single side antenna and double side antenna, 

respectively. The higher resistivity was due to the impurities 

imbedded in the antenna coil by the electroless plating 

process. Hence, the Q factor could be further improved 

by resistivity reduction of the plating metal. 

In addition to the Q factor, the impedance between antenna 

coil and interconnected coaxial cable should be about 

50 , on the basis of the layout of Fig. 1, in order to get the 

desired communication response of the single side circuit 

and double side circuit RFID antenna coils, as presented in 

Table II. The resulting impedance values were 29 and 

49 for the single side antenna and double side RFID an- 



Table II. Characteristics of the printed antenna coils in single side and double side circuits. 

Sample 

L 

µH 

R ac 

Ω 

Q 

Impedance 

Ω 

Impedance phase 

° 

R dc 

Ω 

Conductivity 

s/m Cu 

Single side circuit A 

2 µmNi+5 µmCu 

Single side circuit B 


Single side circuit C 

only 5 µm Cu 

Double sides circuit D 


Double sides circuit E 

2 µmNi+5 µm Cu 

Double sides circuit F 

Only 5 µm Cu 

16 21.1081 0.62 24.5845 32.2368 20.45 0.904 

17 25.8908 0.54 29.2636 27.3605 24.33 0.767 

0.3 23.3501 0.0093 N/A N/A N/A N/A 

55 40.1211 1.1 48.1572 40.2387 40.78 0.828 

48 44.1578 0.96 50.3541 38.2173 42.35 0.758 

0.6 28.5872 0.02 N/A N/A N/A N/A 

for different layout designs. The resulting antenna coil is 

flexible and passes the 3M tape peeling test for PI, PET, and 

FR-4 substrates. The inductance of this antenna is improved 

from 300 nH to 20 H for the single side coil, and 600 nH 

to 50 H for double side coil, compared to the pure copper 

coil under the same conditions. Also, we found that the electroless 

plating recipes used for nickel and copper leads to 

side reactions, which resulted in impurities left in the coil, 

and led to a slight elevation of resistivity, which we expect to 

be able to change in the near future. 

Figure 4. Photograph of the peeling test performed on the single side 

antenna with 3M TM-650 tape. 

tennas, respectively. Obviously, a double side antenna with 

this design is more suitable than the single side antenna for 

use in RFID application. Furthermore, the double-side design 

can reduce the substrate size and signal interference 

attributed to the dense single side coil architecture. 

The adhesion of this printed flexible RFID hybrid antenna 

coil was tested using 3M tape (TM–650). As shown in 

Figure 4, no observed peels proved that excellent adhesion 

between antenna coil and flexible substrate is achieved. 6 

CONCLUSIONS 

We have successfully demonstrated that monolayer and 

double-layer RFID antenna coils can be manufactured by the 

use of multilayer SAMs, ink jet printing, and electroless plating 

processes. Incorporation of the magnetic metal nickel 

could dramatically enhance the inductance value and therefore 

the Q value. We can employ a multilayer antenna to 

obtain higher inductance with respect to the requirements 

REFERENCES 

1 K. Cheng, M.-H. Yang, W. W. W. Chiu, C.-Y. Huang, J. Chang, and T.-F. 

Yin, “Ink-jet printing, self-assembled polyelectrolytes, and electroless 

plating: Low cost fabrication of circuits on flexible substrates at room 

temperature”, Mater. Res. Soc. Symp. Proc. 26, 247–264 (2005). 

2 S. Molesa, D. R. Redinger, D. C. Huang, and V. Subramanian, “Highquality 

ink jet-printed multilevel interconnects and inductive 

components on plastic for ultra-low-cost RFID applications”, Mater. 

Res. Soc. Symp. Proc. 769, 831-836 (2003). 

3 S. E. Molesa, A. de la F. Vornbrock, P. C. Chang, and V. Subramanian, 

“Low-voltage ink jetted organic transistor for printed RFID and display 

applications”, IEEE Device Research Conference 26, 742–747 (2005). 

4 V. Subramanian, “Towards printed low-cost RFID Tags: Device, materials 

and circuit technologies”, MA, (March 16–19, 2003). 

5 D. Redinger, R. Farshchi, and V. Subramanian, “Ink jetted passive 

components on plastic substrate for RFID”, IEEE Trans. Electron 

Devices 51, 1978 (2004). 

6 C.-W. Wang, M.-H. Yang, J. Chang, K. Cheng, C.-P. Yu, C. H. Yu, C.-M. 

Chang, and C.-M. Chang, “Surface characteristics of ink-jet printed 

circuits by polyelectrolyte multilayers with zone model analysis and salt 

solution improvement”, IS&T’s 1st Digital Fabrication Conference (IS&T, 


7 T.-F. Guo, S.-C. Chang, S. Pyo, and Y. Yang, “Vertical integrated 

electronic circuits via a combination of self-assembled polyelectrolytes 

ink-jet printing and electrodeless metal plating processes”, Langmuir 

18(21), 8142–8147 (2002). 

8 P. Shan, Y. Kevrekidis, and J. Benziger, “Ink-jet printing of catalyst 

patterns for electroless metal deposition”, Langmuir 15, 1584–1587 

(1999). 




Top Contact Organic Thin Film Transistors with Ink Jet 

Printed Metal Electrodes 

Kuo-Tong Lin, Chia-Hsun Chen, Ming-Huan Yang, Yuh-Zheng Lee and Kevin Cheng 

Display Technology Center, Industrial Technology Research Institute Hsinchu, Taiwan 310, 

Republic of China 

E-mail: BillyLin@itri.org.tw 

Abstract. Ink jet printing conductive metal nanopastes have been 

studied in research on printed circuit boards and passive components. 

This technique provides a manufacturing method that can 

replace more expensive processes, such as lithography or metal 

evaporation. In this paper, we demonstrated a printed silver topcontact 

source-drain electrodes on bottom-gate organic thin film 

transistors (OTFTs). The soluble conjugated polymer, regioregular 

poly(3-hexylthiophene) (RR-P3HT), was used as the active channel 

material. To increase electric transport efficiency by tuning the surface 

hydrophibic characteristics, a buffer layer, V 2 O 5 , between organic 

semiconductor and printed electrodes, was introduced to resolve 

the incompatibility of these two layers. The results indicated 

the printed Ag on V 2 O 5 film is hydrophilic (contact angle 10°), 

different with that on P3HT only (hydrophibic, contact angle about 

65°). High hydrophilic surface of V 2 O 5 led to self-aligning behavior of 

patterned Ag nanopastes and is helpful to get a flat film. This study 

also compares the characteristics of OTFTs to traditionally evaporated 

electrodes and to ink jet printed source-drain electrodes, in 

order to analyze the mechanism action of the buffer layer and its 

related process differences. © 2007 Society for Imaging Science 



INTRODUCTION 

In recent years, a trend in developing new electronic devices 

is to minimize the element size and to increase the density of 

devices, which is a solution to reducing cost and simplifying 

processing. 1 The major cost of fabricating an electronic device 

comes from the processing procedures rather than from 

the raw materials. Traditional photolithography and the high 

vacuum process can produce devices that scale down to several 

micrometers, but the infrastructures are expensive. Low 

cost process technologies for fabricating electronic devices 

include screen printing, 2,3 micromolding in capillaries, 4 soft 

lithographic stamping, 5 and ink jet printing. 6 The reduction 

of processing procedures through direct deposition of a material 

could reduce the cost and complexity of the fabrication 

process further. Ink jet printing is a potential alternative to 

the existing deposition approaches. As a maskless process, it 

can reduce the cost of physical molding and chemical waste 

of material. Moreover, ink jet printing technology is much 

easier extended to process on a large area substrate than are 

other processes. 

Overarching inks can be used for the process of ink jet 


1062-3701/2007/515/456/5/$20.00. 

printing. These inks comprise inorganic and organic electronic 

materials and are expected to form wiring, electrodes, 

and many other kinds of components of an electric device. 

The application of ink jet printed electronic devices includes 

thin film transistors (TFTs), radio frequency identification 

(RFID) circuits, organic light-emitting diodes (OLED), and 

printed circuit boards (PCBs). 7 Among various inks used in 

the past few years, conducting inks have been newly developed 

for forming the electrodes of the devices. Poly(ethylene 

dioxythiophene) (PEDOT) is a good candidate for forming 

ink jet printed electrodes, but the conductivity is several orders 

lower than that of metals. The conductivity of the metal 

nanopastes is three to four orders higher than that of 

PEDOT. After low temperature sintering 150°C, the 

nanoparticles form aggregates, and film thickness can be 

controlled by tuning the drop size and the solid content of 

the inks. The fabrication processes for flexible display devices 

require low processing temperatures, however. 

Burgietal. 8 discussed the influence of the work function 

for the metal on the device performance with poly(3– 

hexylthiophene) (PHT) as an active layer. The source/drain 

metal electrode is chosen to be Cr/Au, Ag, or Au only. Other 

metals form Schottky barriers at the interface so that the 

resistance is high and the current is blocked therein. 

This study explored the effects of various printing conditions 

with a Dimatix SE-128 piezoelectric printhead on the 

quality of the printed film, and we discuss the influence of 

the thin transition metal buffer layer on TFT device performance. 

Using a transition metal oxide as the hole injection 

layer is an effective way to improve the characteristics of 

organic thin film transistors (OTFTs). 9 At the same time, the 

contact angle of the nanopaste on the transition metal oxide 

is smaller than that on P3HT; thus, the printing quality of 

the metal line is better. Furthermore, the thickness of this 

buffer layer is so thin that device performance is not significantly 

influenced. 

TOP CONTACT–BOTTOM GATE OTFT STRUCTURE 

AND PROCESSES 

The structure of the organic thin film transistor with ink jet 

printed silver electrodes is shown in Figure 1. The bottom 

gate with heavily doped Si is selected, and SiO 2 with thickness 

of 300 nm is used as the gate dielectric. The Si/SiO 2 

substrate was cleaned sequentially with acetone and isopro- 

456

Lin et al.: Top contact organic thin film transistors with ink jet printed metal electrodes 

Figure 1. The structure of the polymer thin film transistor device PTFT 

with a silver top electrode. 

Figure 3. Profiles of the sintered silver nanopaste dot formed by ink jet 

printing on the different substrates: a P3HT and b V 2 O 5 . 

Figure 2. Observation of the break-off behavior of the printed drop beneath 

an arbitrary nozzle. 

Table I. Contact angles for silver nanopaste on the P3HT and V 2 O 5 . 

Substrate 

Contact Angle 

P3HT 65° 

V 2 O 5 10° 

panol in an ultrasonic bath. After 150 W, 450 mtorr, and 

5 min, oxygen plasma treatment, a monolayer of OTS was 

self-assembled onto the SiO 2 layer by dipping the substrate 

(with SiO 2 layer) into the 60°C OTS dilute toluene solution 

for 20 min. Then, regioregular poly(3-hexylthiophene) 

(P3HT) (purchased from Rieke) was used as the organic 

active layer. P3HT dissolved in xylene 0.5 wt. % was spin 

coated on the substrate and baked at 90°C for 90 mins. The 

thickness of this P3HT film was about 50 nm. The transition 

metal buffer layer comprised of dielectric material, vanadium 

oxide V 2 O 5 ,was then thermally evaporated by 

shadow mask on the P3HT; its thickness was 3nm. The 

shape of V 2 O 5 layer is defined by a metal mask, and Ag is 

ink jet printed approximately within the area of V 2 O 5 . The 

printing position need not to be correctly defined, since the 

surface energy behavior self-aligns this Ag ink on the low 

contact angle surface, i.e., V 2 O 5 . To compare the performance 

of this device with ink jet printed silver nanopaste top 

electrodes, we made a device with evaporated silver film top 

electrodes. 

The ink jet printing platform, developed by the Display 

Technology Center (DTC) in the Industrial Technology 

Research Institutes (ITRI), equipped with Dimatix SE-128 

piezoelectric printhead, was used to discharge the silver 

nanoparticle inks (AG-IJ-G-100-S1), purchased from Cabot. 

The low resistivity patterned silver film was obtained after 

postbaking at 150°C for 30 min in a glove box to sinter the 

Figure 4. Morphology of the sintered silver nanopaste dot formed by ink 

jet printing on the P3HT substrate with different drop densities. 

Figure 5. Morphology of the sintered silver nanopaste dot formed by ink 

jet printing on the V 2 O 5 /P3HT substrate with different drop densities: a 

100 dpi, b 200 dpi, c 300 dpi, and d 400 dpi. 

predeposited precursor material. The thickness of the metal 

layer was 200 nm. In contrast, the pressure of the chamber 

was maintained at about 510 −6 torr during evaporation of 

silver electrodes. Finally, measurements of V D versus I D and 

V G versus I D for devices were conducted using a Keithley 

4200 semiconductor parameter analyzer. 

The field effect mobility was calculated in the saturation 

regions from the following equation: 



Figure 6. I d -V ds characteristics of the PTFT, wherein the source-drain electrodes are: a evaporated Ag, b 

evaporated V 2 O 5 /Ag, c evaporated V 2 O 5 /Ag with postannealing, and d evaporated V 2 O 5 /IJP Ag with 

postannealing. 

I DS = WC i /2LV G − V T 2 , 

where W is the channel width, L is the channel length, C i is 

the capacitance per unit area of the insulator, and V T is the 

threshold voltage. 


Stability of the Ink Jet Printed Ink 

This study improved the jetting behavior of piezoelectric 

print heads by controlling three parameters, i.e., pulse waveform, 

pulse frequency, and driving voltage. Suitable range of 

waveform modulation and pulse frequency are defined corresponding 

to given driving voltages for appropriate ink discharging. 

The jetting behavior can be observed by using a 

strobe image capturing system integrated in the printing 

platform. The break-off behavior of the jettable silver 

nanoparticles ink was captured within 60 s and shown in 

Figure 2. The velocity of the jetted drop was about 6.4 m/s 

and stable. With a stand-off of 1mm, we successfully modulated 

the ink jet printing parameters of Dimatix SE-128 for 

silver nanoparticles of ink, and obtained uniform and continuous 

silver paste lines. 

1 

S-D 

Electrodes 

Table II. Electrical characteristics of the various PTFTs. 

Mobility 

cm 2 / V.s 

Threshold 

Voltage 

V 

On/Off 

Ratio 

Evaporate Au 3.48 10 −3 −9.49 14268 

Evaporate Ag 1.88 10 −3 −11.17 629 

Evaporate V 2 O 5 / Ag 2.23 10 −3 −10.63 1398 

Evaporate V 2 O 5 /Ag 

1.19 10 −3 −9.95 3107 

with postannealing 

Evaporate V 2 O 5 / IJP Ag 

with postannealing 

2.98 10 −4 −15.27 138 

Morphology of the Ink Jet Printed Nanopaste 

The contact angles of the silver nanopaste between the P3HT 

film and the V 2 O 5 buffer film are shown in Table I. The 

contact angle on P3HT film is 65°, which is larger than that 

on the V 2 O 5 film. The morphology of printed nanopaste is 

related to the contact angle on the surface. The morphologies 

of the nanopaste on these two substrates are shown in 

Figure 3. As the nanopaste ink jet printed on the P3HT film 

was postannealed, the dot diameter shrunk to about 30 m 



Figure 7. I d -V gs characteristics of the PTFT, wherein the source-drain electrodes are: a evaporated Ag, b 

evaporated V 2 O 5 /Ag, c evaporated V 2 O 5 /Ag with postannealing, and d evaporated V 2 O 5 /IJP Ag with 

postannealing. 

and formed a coffee-ring structure. The thickness of the ring 

edge was 2.5 m. The drop morphology on P3HT alone 

was so rough that it was difficult to form a flat film and 

pattern uniform metal electrodes. On the other hand, it was 

a good choice to print nanopaste on V 2 O 5 /P3HT film. The 

dot size was 73 m and did not shrink. A coffee ring still 

formed, but the morphology of the dot was flatter, and it 

was easier to construct lines and patterns of metal electrodes. 

The ring thickness was 0.5 m. In addition, the 

more hydrophilic surface of V 2 O 5 led to self-aligning behavior 

of patterned Ag nanopastes. After sintering the patterned 

Ag nanopaste at 150°C for 30 min, the resultant resistivity of 

the ink jet printed electrode was about 310 −8 m, which 

was two to three times that for the normal bulk silver. 

Currently, no OTFT-related article has reported how to 

ink jet print Ag ink onto the semiconductor layer and discussed 

the resulting interface. All of the previously reported 

devices were bottom-contact devices, and Ag ink was printed 

onto the dielectric layer. However, the contact angle of Ag 

ink to hydrophilic dielectric layer was small, and the printed 

pattern was easy to fix in position. On the other hand, the 

contact angle of Ag ink to the hydrophibic organic semiconductor 

was large, which makes the printed ink deposits 

shrink relative to each other, and it is accordingly hard to 

form a uniform film, as shown in Figure 4. In this article an 

approach is proposed to improve the morphology of the 

electrode ink on the semiconductor by inserting an inorganic 

buffer layer. The morphology of the nanopaste on 

V 2 O 5 /P3HT film is shown in Figure 5. As the printing density 

increased to 400 dpi, the drops began to aggregate and 

formed uniform lines or patterns. The best printing density 

was found to be 400 dpi, with the width of the line 

62 m. Further increase in the printing density led to 

thicker and wider conductive Ag lines. If we discharged 

denser inks, it would be harder to control the morphology 

because more ink would lead to unstable boundaries of the 

Ag lines. 

DEVICE PERFORMANCE 

The source-drain (S-D) electrodes are bilayer electrodes 

formed on the organic active layer. The channel width and 

length of each device were 2000 m and 140 m, respectively. 

The device performance is stable, on the basis of the 

averaged results obtained with three devices. Basically, no 

obvious difference was found in device performance between 

evaporated silver and bilayer V 2 O 5 /Ag electrodes. The 



results are shown in Figures 6(a) and 6(b). The device with 

the Ag S-D electrode had an on-off ratio I on /I off and mobility 

similar to the device with V 2 O 5 /Ag S-D electrode. The 

on-off ratio was near 10 3 , and the mobility was 

10 −3 cm 2 /V s. However, the current of the V 2 O 5 /Ag S-D 

electrode was about twofold larger than that of the device 

with Ag S-D electrode, indicating better hole injection after 

inserting the V 2 O 5 , which can also be a hole injection layer 

as well as providing surface energy modification. In the case 

of ink jet printed silver electrode, the patterned modification 

of the V 2 O 5 buffer layer was necessary for the correct definition 

of the S-D boundary. The Ag nanopaste ink was selfaligned 

onto the patterned S-D V 2 O 5 buffer layer and did 

not deposit onto the P3HT due to the differences in contact 

angle, i.e., the surface energy. The device performance after 

postannealing is shown in Figs. 6(c) and 6(d). We observed 

that there is no decrease in the performance of the device 

with evaporated S-D electrodes after postannealing at 150°C 

for 30 min. The performance of the device with ink jet 

printed Ag electrode was not as good. The mobility and 

current dropped one order of magnitude, which may be due 

to the conductivity decrease of the ink jet printed Ag electrode 

and organic residuals from the nanopastes being 

present at the metal/semiconductor interface. These results 

are summarized in Table II. It is worth noting that the choice 

of S-D electrode and interface engineering are important. 

The I d -V gs characteristics of these four devices are shown in 

Fig. 7 for comparison. There are some leakage currents in 

our newly developed device, which we will discuss and try to 

improve in the future. 

CONCLUSION 

In this paper, we successfully demonstrated a top contact– 

bottom gate OTFT device fabricated by ink jet printing silver 

nanopaste electrode, along with thermally evaporation and 

spin coating. By modulating the ink jet printing parameters, 

a well-sintered silver electrode was formed to fabricate a TFT 

device with a channel width of 2000 m and channel 

length of 140 m. The morphology of the nanopaste drop 

on the P3HT was rough, and it was hard to form flat lines 

and patterns of the metal electrodes; thus, we introduced a 

thin transition metal oxide as the buffer layer. The more 

hydrophilic surface of V 2 O 5 led to self-aligning behavior of 

patterned Ag nanopaste, and a continuous conductive line 

could be formed. The current of the V 2 O 5 /Ag S-D electrode 

was about twofold larger than that of a device with an 

evaporated Ag S-D electrode, indicating better hole injection 

after inserting the V 2 O 5 , which can also be a hole injection 

layer, as well as providing surface energy modification. The 

device performance of the ink jet printed Ag electrode was 

inferior to that of the evaporated one after annealing. The 

mobility and current dropped one order of magnitude, 

which may be due to the conductivity decrease of the ink jet 

printed Ag electrode post annealing and the organic residuals 

from the nanopaste present in the metal/semiconductor 

interface. In the future, we expect to improve the interface 

between the S-D electrodes and the semiconductor layer of 

the printed OTFT device to achieve comparable performance 

to an evaporated or spin-coated counterpart. 

REFERENCES 

1 P. Calvert, Chem. Mater. 13, 3299 (2001). 

2 C. D. Sheraw, L. Zhou, J. R. Huang, D. J. Gundlach, T. N. Jackson, M. G. 

Kane, I. G. Hill, M. S. Hammond, J. Campi, B. K. Greening, J. Francl, 

and J. West, Appl. Phys. Lett. 80, 1088 (2002). 

3 H. Klauk, D. J. Gundlach, and T. N. Jackson, IEEE Trans. Electron 

Devices 20, 289 (1999). 

4 J. A. Rogers, Z. Bao, and V. R. Raju, Appl. Phys. Lett. 72, 2716 (1998). 

5 J. A. Rogers, Z. Bao, A. Makhija, P. Braun, Adv. Mater. (Weinheim, Ger.) 

11, 741 (1999). 

6 H. Sirringhaus, T. Kawase, R. H. Friend, T. Shimoda, M. Inbasekaran, W. 

Wu, and E. P. Woo, Science 290, 2123 (2000). 

7 K. Cheng, M. Yang, W. Chiu, C. Huang, J. Chang, T. Ying, and Y. Yang, 

Macromol. Rapid Commun. 26(4), 247 (2005). 

8 L. Burgi, T. J. Richards, R. H. Friend, and H. Sirringhaus, Appl. Phys. 

Lett. 94, 6129 (2003). 

9 C. W. Chu, S. H. Li, C. W. Chen, V. Shrotriya, and Y. Yang, Appl. Phys. 

Lett. 87, 193508 (2005). 




Ring Edge in Film Morphology: Benefit or Obstacle 

for Ink Jet Fabrication of Organic Thin Film Transistors 

Jhih-Ping Lu 

Dept. of Photonics and Display Institute, National Chiao Tung University, Hsinchu, Taiwan, 30010, 

Republic of China and Display Technology Center, Industrial Technology Research Institute, Hsinchu, Taiwan 

310, Republic of China 

E-mail: Anthony_lu@itri.org.tw 

Ying-pin Chen 



Yuh-Zheng Lee and Kevin Cheng 

Display Technology Center, Industrial Technology Research Institute, Hsinchu, Taiwan 310, 


Fang-Chung Chen 



Abstract. In this study, a novel process using ink jet printing with 

dilute PMMA [poly(methylmethacrylate)] solution to form fine separation 

banks as confined boundaries for the subsequent depositing 

poly(3,4–ethylene dioxythiophene) (PEDOT) was proposed. The 

PEDOT comprised the source and drained electrodes with a gap of 

several micrometers due to the innate characteristic of the ring edge 

of ink jetted PMMA. Using this technique, the organic device was 

designed with a bottom gate and evaporated pentacene as the 

channel material. The device mobility and on/off ratio were about 

3.610 −4 and 210 2 cm 2 /V sec, respectively. © 2007 Society for 

Imaging Science and Technology. 


INTRODUCTION 

Inorganic semiconductor technology and related applications 

have already attained a mature stage in the past decades. 

Because of this revolutionary advancement, science 

and technology have grown vigorously. The demand for 

lightweight and flexible electronic products is gradually increasing. 

However, high temperature processing of inorganic 

semiconductors is difficult to apply to this demand. Low 

temperature direct patterning technology and organic electronics 

have accordingly been intensively developed. 1 With 

respect to organic electronics, pentacene and regioregular 

poly(3-hexylthiophene) (RR-P3HT) are the most popular 

materials used in fabrication of organic thin film transistors 

(OTFT). The electrical performance of pentacene is superior 

among the present organic materials, but it is processed by 

vacuum evaporation, e.g., thermal evaporation 2 or organic 


1062-3701/2007/515/461/4/$20.00. 

vapor phase deposition. 3 Despite of the lower mobility of the 

polymeric material, the solubility of P3HT, compared to 

pentacene, makes it compatible with the printing process, 

which has the advantages of low temperature processing, 

large area manufacturing capability, etc. 

Among patterning technologies, ink jet printing is considered 

the most promising potential candidate for industrial 

mass production. In recent years, ink jet printing is used not 

only on paper and wide format media, but also widely applied 

to patterning deposition of organic and inorganic electronics. 

There are many investigations of ink jet printing 

patterning, such as deposition of insulators, organic 

semiconductors, 4 or formation of conductive paste circuits, 5 

and microlens array. 6 Fabrication of organic TFT (OTFT), a 

stacked device, needs to integrate several printing steps; thus, 

it is more difficult and complicated than a single material 

deposition process. In spite of these challenges, the use of 

ink jet printing to make good polymer TFT devices has already 

been discussed in many papers. 7,8 

Most research on OTFT has been concerned with new 

material development, surface modification or structural design 

for improving the electric characteristics. 9 As is well 

known, the current from drain to source is inversely proportional 

to the channel length. It is favorable to shorten channel 

length instead of increasing channel width W for better 

TFT array resolution. However, the printing feature is limited 

by the drop size and directionality of the ink jet droplet. 

The printed line width is usually greater than 80 m with 

the use of a print head with 35 pl droplet, such as a Spectra 

SE-128. The variation of ink jet printing directionality also 

makes it difficult to obtain small and uniform gate channel 

461

Lu et al.: Ring edge in film morphology: Benefit or obstacle for ink jet fabrication of organic TFTs 

Figure 1. Structure of the organic thin film transistor OTFT device with a 

ridge of PMMA ring edge. 

length over the whole printing area. In this study, we separated 

the organic electrodes by a poly(methyl methacrylate) 

(PMMA) ridge, formed by the natural phenomenon of capillary 

flow to obtain gate lengths of 10 m using an ink jet 

printing method. 

PMMA PARTITION TO SEPARATE CHANNELS 

The phenomenon of ring shaped patterns from dried solution 

is common in our daily life. When a liquid droplet lands 

on a substrate, there are three interface interactions governing 

the droplet drying behavior. These interactions occur 

between the individual interface of solid to liquid, liquid to 

gas, or gas to solid. When the solvent gradually evaporates, 

liquid droplets are dried to the solid state. Finally, the solute 

accumulates around the peripheral boundary as ridges, 

much thicker than that of the center area and is called the 

“coffee ring shape.” 

Much effort has been devoted to explain the coffee ring 

phenomenon. Deegan et al.’s hypothesis 10 is the most acceptable 

to describe the ring shaped formation behavior. He 

assumed that the liquid evaporates faster around the periphery 

than at the center part. This makes the peripheral 

boundary easier to dry than the center area and, therefore, to 

form a contact line around the liquid peripheral. This results 

in a concentration gradient of dissolving solute, and the solution 

accompanying the solute begins to diffuse from the 

center to the edge. When the drying process is completed, 

most solute in the liquid has been carried to the edge and 

formed into a ring shaped profile. The natural behavior of 

ridge formation is an obstacle for obtaining smooth films for 

organic electronics, such as polymer light emitting diodes 

(PLED) and organic thin film transistors. As an alternative, 

this study exploited this behavior as a beneficial micropatterning 

method to define the gate length of OTFT. Through 

this method, OTFTs with a several micrometer channel 

length are fabricated. 

EXPERIMENTAL 

The structure of the organic thin film transistor (OTFT) 

used in this study is shown in Figure 1. The bottom gate is 

heavily doped Si on which thermal oxide, SiO 2 , 300 nm 

thick is formed. Sputtered Au 70 nm thick and Cr 30 nm,as 

an adhesive layer, was used to form the interconnection 

tracks with spacing of 100 m, as defined by a stainless 

shadow mask. Poly(methylmethacrylate) (PMMA) ridge several 

micrometers wide was formed by the coffee-ring edge 

effect. The statistical data for the ring width distribution 

Figure 2. Side and top views of the process flow schematic illustration of 

ring ridge patterned organic thin film transistors. 

yield a mean value of 7.88 m and a standard deviation of 

1.68 m for 25 devices. After suitable plasma treatment to 

remove the thinner PMMA film from the surface, the diluted 

poly(ethylene dioxythiophene) (PEDOT 4071, purchased 

from Bayer) was ink jet printed on both sides of the ridge as 

source and drain electrodes, using a thermal bubble print 

head capable of 80 pl droplets as developed by our institute 

(ITRI). 

The fabricating procedures for ring ridges, electrodes, 

and the pentacene active layer are shown in Figure 2. As 

shown in Fig. 2(a), the Si/SiO 2 substrate with Au/Cr interconnection 

pads were treated with O 2 plasma at 200 W, 800 

SCCM (standard cubic centimeter per minute) for 1 min to 

improve the surface wettability. In Fig. 2(b), 1 wt. % PMMA 

solution dissolved in anisole was printed and there was one 

ridge located between the two Au/Cr pads. An ion-coupled 

plasma was used to etch the thinner part inside the ring 

ridge by treating 50 sec O 2 plasma with etching rate 

6.78 Å/sec, and therefore, two thicker ridges were left on 

both sides. Sequentially, carbon tetrafluoride plasma was applied 

to make the PMMA surface liquid repellent as shown 

in Fig. 2(c). The residual ridge was treated with CF 4 plasma 

for 100 sec under a pressure of 1 torr with a small etching 

rate of 0.3 Å/sec. The treated PMMA surface then had contact 

angles of 106° and 60° for water and anisole, respectively. 

Therefore, after the dry etching and repellent treatment, 

a ridge with width of 5.37 m was made as shown in 

Figure 3. Then PEDOT solution was ink jet printed on both 

sides of the ridge to connect to each of the Au/Cr pads, 



Figure 3. Microscopy of a PMMA residual ridge after plasma etching. 

Figure 5. Organic thin film transistor characteristics of a ring ridge patterned 

with pentacene as the active layer. 

Figure 4. Profile of printed PEDOT by ink jet printing on both sides of the 

etched repellant PMMA ridge. 

which were used as a source or drain electrode as shown in 

Fig. 2(d). Figure 4 shows the profile of PEDOT, which was 

printed on both sides of the etched PMMA ridge. This profile 

was obtained by XP-1 profilometer (AMBIOS technology). 

From the profile, we can see that the printed PEDOT 

can be adequately confined between ridges, and the height of 

PEDOT domains and ridges are around 70 nm and 110 nm, 

respectively. Hexamethyldisilazane (HMDS) and pentacene 

were thermally evaporated in sequence onto the above substrate 

to complete the OTFT devices, as shown in Fig. 2(e). 

In order to prevent source/drain electrodes from interconnecting, 

it is better to make the PMMA ridge as repellent to 

the PEDOT solution as possible. 


According to the above procedures, utilizing the ink jet 

printing technique incorporated with ring edge effect we 

were able to produce a narrow line of 10 m width, for 

which the mean and standard deviation are 5.16 m and 

0.31 m for 25 samples. Applying this patterning method to 

fabricate an OTFT device, better MOS (metal-oxide semiconductor) 

characteristics could be obtained. The I d versus 

V d and I d versus V g measured using a Keithley 4200 semiconductor 

parameter analyzer are shown in Figure 5. The 

field effect mobility can be calculated at the saturation region 

from the following equation: 

IDS = WC i 

2L 

V G − V T 2 , 

where C i is the capacitance per unit area of the insulator, and 

V T is the threshold voltage. 

1 

Figure 6. Capacitance comparison of plasma treatment on the oxide 

layer. 

The mobility, on/off ratio and threshold voltage are 

about 3.610 −4 cm 2 /V sec, 210 2 , and 22 V, respectively. 

It is found that the off-current and threshold voltage are 

relatively high in this device, which may be attributed to the 

negative charging into the PMMA ridge after the plasma 

treatment. In addition, pentacene cannot form a better 

alignment on the HMDS/PMMA ring ridge owing to the 

poor mobility of these devices. Further study is needed to 

clarify the above problems. 

Besides the patterning process, the plasma charging effect 

on an oxide layer should be considered because the insulator 

layer was also treated during the etching and repellent 

treatment. This issue was addressed in terms of the 

capacitance variation of devices with MIM structure, doped 

Si/oxide/Al, with or without plasma treatment on the oxide 

layer. After examination with an HP 4194 impedance ana- 



lyzer, we found the capacitance to be without dramatic 

change after plasma treatment under the conditions used in 

the patterning process, as shown in Figure 6. Most devices 

have similar capacitance, even though some have a maximum 

variation of 5%. 11 

CONCLUSION 

In this paper, we overcame the restriction of line width by 

using the direct ink jet printing technique (with droplets of 

35 pl the line width is generally around 80–150 m) to 

obtain fine lines of 5 m using the ring-edge effect. We 

successfully applied this novel method to fabricate OTFT 

devices with small channel lengths of several micrometers. 

This patterning method was proven feasible. Devices with 

common characteristics of transistors were proposed, 

though the results have yet to be optimized, where the mobility 

and on/off ratio were about 3.610 −4 cm 2 /V sec and 

210 2 cm 2 /V sec, respectively. It is evident that this patterning 

method of ink jet printing can be widely applied 

after additional studies in the future. 

REFERENCES 

1 S. R. Forrest, Nature (London) 428, 911-918 (2004). 

2 Y.-Y. Lin, D. J. Gundlach, S. F. Nelson, and T. N. Jackson, IEEE Trans. 

Electron Devices 44, 1325–1331 (1997). 

3 M. Baldo, M. Deutsch, P. Burrows, H. Gossenberger, M. Gerstenberg, V. 

Ban, and S. Forrest, Adv. Mater. (Weinheim, Ger.) 10, 1505-1514 (1998). 

4 K. F. Teng and R. W. Vest, IEEE Electron Device Lett. 9, 591 (1988). 

5 B.-J. de Gans, P. C. Duineveld, U. S. Schubert, Adv. Mater. (Weinheim, 

Ger.) 16, 203 (2004). 

6 K. Cheng, M. Yang, W. Chiu, C. Huang, J. Chang, T. Ying, and Y. Yang, 

Macromol. Rapid Commun. 26, 247 (2005). 

7 W. R. Cox and T. Chen, Opt. Photonics News 12, 32–35 (2005). 

8 H. Sirringhaus, T. Kawase, R. H. Friend, T. Shimoda, M. Inbasekaran, W. 

Wu, and E. P. Woo, Science 290, 2123 (2000). 

9 C. W. Sele, T. von Werne, R. H. Friend, and H. Sirringhaus, Adv. Mater. 

(Weinheim, Ger.) 17(8), 997 (2005). 

10 R. D. Deegan, O. Bakajin, T. F. Dupont, G. Huber, S. R. Nagel, and T. A, 

Nature (London) 389, 827-829 (1997). 

11 J. P. Lu and F. C. Chen (unpublished). 




Digital Fabrication Using High-resolution Liquid Toner 

Electrophotography 

Atsuko Iida, Koichi Ishii, Yasushi Shinjo, Hitoshi Yagi and Masahiro Hosoya 

Advanced Electron Devices Laboratory, Corporate Research & Development Center, Toshiba Corporation, 1, 

Komukai-Toshiba-cho, Saiwai-ku, Kawasaki, 212-8582, Japan 

E-mail: atsuko.iida@toshiba.co.jp 

Abstract. We have developed a digital fabrication process using 

high-resolution liquid toner electrophotography, consisting of fine liquid 

toner, a high-resolution exposure system, and nonelectrical 

transfer. Fine pitch multiline patterns of Cu wiring can be obtained 

by printing fine lines with seed toners and by electroless plating 

deposited on lines. Submicron-diameter seed toners have superfine 

conductive particles on their surfaces. Adhesion between the seed 

toner layer and Cu layer was increased by applying surface modification. 

Multiline patterns of 1 pixel line width (21.6 m) with the 

volume resistivity of 2.110 −6 Ωcm were realized by using a 1200 

dpi resolution light-emitting diode. Furthermore, the development 

process of multiline patterns with 2540 dpi resolution was examined 

by numerical simulations based on the electrophoretic characteristics 

of liquid toner and on the electrostatic forces. The capability of 

multiline-pattern formation of line and space (L/S)10/10 m was 

confirmed. The actual toner images of L/S10/10 m multiline pattern 

were obtained by using a 2540 dpi resolution luster scanning 

unit (LSU). Theoretical and experimental results confirm that the 

fabrication process using liquid toner electrophotography is available 

for realizing high-resolution multiline patterns. © 2007 Society 

for Imaging Science and Technology. 


INTRODUCTION 

Fine-pitch-pattern formation of wiring circuit boards and 

reduction in the cost of manufacturing processes are critically 

important requirements for electronic components and 

semiconductor packages. Recently, digital fabrication of electronic 

components using printing technology has been 

reported. 1–4 Digital fabrication of circuit boards offers several 

advantages, notably simple mask-less process, and reduced 

costs. 

We have been developing a high resolution liquid toner 

electrophotographic imaging technology with the potential 

to produce fine images equal in quality to those produced by 

offset printing. 5–13 Extremely high resolution images have 

been realized by our technology that includes a full color 

Image-On-Image (IOI) development process, highresolution 

LSU with 2540 dpi, and the non-electric transfer 

process called “shearing transfer”. 

In this paper, we applied our imaging technology to a 

new digital fabrication process for an electronic circuit 

board. 14 The capability of fine line-pattern formation by using 

liquid toner electrophotographic technology was verified 

theoretically and experimentally. 

HIGH-RESOLUTION LIQUID TONER IMAGING 

PROCESS 

Figure 1 shows a typical configuration of our liquid toner 

electrophotographic imaging system. It includes a photoreceptor 

drum, a scorotron charger, an exposure unit, a development 

unit, a dryer, and a transfer unit at the periphery of 

the photoreceptor drum. The toner image is dried properly 

by the dryer on the photoreceptor and is transferred to the 

intermediate transfer roller by the non-electric transfer process, 

and then secondary transferred to and thermally fixed 

on a substrate. The highly efficient performance of the 

shearing transfer is realized for the drying states in which the 

toner layer includes the solvent in the range of 

5–15 wt %. 7,8 

The searing transfer method is the nonelectrostatic process, 

which is used in the newly developed liquid toner 

electrophotograpy. 5–10 The intermediate transfer roller consists 

of a metal roller with 0.2 mm thick elastic layer of 

silicone rubber on the surface. The shearing transfer does 

not depend on an adhesive force of the elastic layer of the 

intermediate transfer roller. Figure 2 shows a schematic description 

of the mechanism of this method. In this method, 

the velocity of the surface of the intermediate transfer roller 

is slower than that of the surface of the photoreceptor drum 

by several percent. A distortion occurs in the nip of the 

elastic layer due to the difference of the velocities, and a 

 

IS&T Member 

Received Mar. 5, 2007; accepted for publication Jun. 7, 2007. 

1062-3701/2007/515/465/8/$20.00. 

Figure 1. Configuration of liquid toner electrophotographic technology. 

465

Iida et al.: Digital fabrication using high-resolution liquid toner electrophotography 

Figure 2. A schematic of the mechanism of shearing transfer: a Distortion 

in the nip area and restoration in the backward of the nip occur by 

the difference of the velocities. b The restoration and friction cause a 

shearing stress. Shearing stress slides the toner layer on the photoreceptor 

and transfers it to the transfer roller. 

restoration occurs behind the nip [Fig. 2(a)]. The restoration 

and friction cause a shearing stress between the toner layer 

and the photoreceptor drum, which slides the toner layer on 

the photoreceptor and transfers it to the intermediate roller 

[Fig. 2(b)]. The toner layer moves in unity by the capillary 

condensation because the toner particles are covered with a 

thin film of the solvent during the transfer performance. 7,8,15 

The method is effective and well directed for a highresolution 

image because there is no degradation in quality 

without toner scattering in an electric transfer method. 

Figure 3 shows the comparison of the toner images before 

and after the transfer process observed by a stereoscopic 

microscope. The toner image on the paper is in good agreement 

with that on the photoreceptor surface. The difference 

of their line widths was observed, but image destruction did 

not occur due to the transfer. The observation confirms that 

the shearing transfer method is an excellent method. 

Figure 3. Comparison of the images before and after the transfer process 

2540 dpi, 1 point Chinese character: a Toner image before the transfer 

on the photoreceptor drum and b toner image after the transfer on 

the paper. 

Figure 4. Printed image obtained by liquid toner electrophotography: a 

Comparison of the printed images on the paper and b toner particle. 

A printed image realized by our technology is shown in 

Figure 4(a). Comparison to a printed image realized by dry 

toner reveals that a finer pitch line was obtained by our 

technology. Figure 4(b) shows an image of a toner particle 

observed by a field emission scanning electron microscope 

(FE-SEM). The average diameter of toner particles was 

200 nm. 

DIGITAL FABRICATION PROCESS 

The digital fabrication process shown in Figure 5 proceeds as 

follows: first, fine-pitch patterns are printed on the substrate 

by using seed toners [Fig. 5(a)], and then a surface modification 

is added on the printed pattern [Fig. 5(b)]. Finally, 

conductive layers are deposited on the substrate by electroless 

plating [Fig. 5(c)]. The same electrophotographic system 

for imaging technology can be used for digital fabrication. 

Patterns formed on the photoreceptor can be transferred to 

the substrate, which is almost or partly covered with metallic 

layers, by using shearing transfer. Silicon wafers, glass, and 

resin films (such as polyimide, glass epoxy, or polyethylene 

terephthalate) are used for the substrate. 

Image Formation of Fine-pitch Patterns 

The basic particle of the seed toner was composed of acrylic 

thermoplastic resin. The superfine conductive particles were 

chosen from the metal particles of Ag, Cu, Pd, or any other 

materials with the catalytic ability to perform as seed for 

electroless plating. Their average diameters were 20–80 nm. 

Toner particles prepared by adding a charge controlling 

agent were dispersed in an insulative solvent, Isopar ® manufactured 

by ExxonMobil Chemical Japan Pte. Ltd. 

A positively charged organic photoreceptor with a single 

coated layer of the phthalocyanine pigment was used. The 



Figure 6. Line/space=1/1 pixel multiline pattern with 1200 dpi exposure 

system. 

Figure 5. Digital fabrication process: a Printing seed toner, b surface 

modification, c electroless plating. 

photoreceptor drum rotating at a speed of 100 mm/sec was 

charged uniformly at +700 V by a charger, and a latent image 

was formed on the photoreceptor by a 1200 dpi LED 

exposure system. The exposure time was 1.6 s. Then, the 

latent image was visualized by a development unit. The voltage 

applied to the development roller was +550 V. Excess 

liquid included in the image was removed by a squeeze 

roller, and the image was dried by an air dryer. Next, the 

image was transferred to an intermediate transfer roller under 

pressure of 0.74 MPa at the 100°C without electrostatic 

forces. The velocity of the surface of the intermediate transfer 

roller was slower than that of the surface of the photoreceptor 

drum by 2.5%. Finally, the image was transferred to a 

substrate under pressure of 0.74 MPa by a backup roller at 

100°C. 

Fine-pitch pattern of L/S=1 pixel/1 pixel 

=21.6 m/21.6 m with 1200 dpi resolution was printed 

on polyimide film as shown in Figure 6. The fine-pitch pattern 

formed on the photoreceptor surface was transferred to 

the substrate without deterioration during our transfer 

process. 

In the case of the dry toner electrophotographic system, 

the toner particle size is 4−8 m in diameter. 3,4 The limit of 

the L/S using dry toner is about 80/80 m and the line 

edge shows a ragged shape in the range of several tens of 

m. Compared with the dry toner electrophotography, it is 

possible to achieve higher resolution patterns owing to edge 

sharpness and to provide higher performance in electroless 

plating owing to uniform dispersion of conductive particles 

over the printed pattern. Additionaly, the amount of scattered 

particles in the vicinity of the line, in the case of using 

the shearing transfer process, is considerably less than that in 

the case of using the electric transfer process. 

Surface Modification after Printing Pattern 

Liquid toner imaging technology is advantageous for realizing 

a flat and smooth surface for printed patterns. However, 

a surface modification is required in order to successfully use 

the electroless plating process. The dry etching process was 

applied to the patterns after printing to form a surface with 

irregularities. Plasma etch system TE-7500M of Tokyo 

Electron Ltd. was used for plasma etching with fluorocarbon 

and oxygen mixture gases. The rate of mixture gases was 

1 sccm for C 4 F 8 and 10 sccm for O 2 , and the total gas pressure 

was 7.3 Pa. The applied power was 50 W, and etching 

time was 10 sec. The fluoride deposition was removed by 

cleaning with HCl solution after plasma etching. 

Figure 7 shows the SEM images of the pattern surface, 

both of as printed [Fig. 7(a)] and after adding surface modification 

[Fig. 7(b)]. The pattern was printed on the silicon 

wafer with an insulating coating of epoxy resin film. A 

slightly rough surface after surface modification is observed, 

as shown in Fig. 7(b). The resin of the surface was selectively 

decomposed, and the superfine conductive particles remained 

after the dry eching process. The weight of the resin 

decreased by 5%, and the amount of exposed superfine conductive 

particles increased at the surface. 

In the case of the as-printed pattern without any surface 

modification, however, the Cu film was deposited on it at the 

initial state of the electroless plating process and the adhesion 

of the Cu film was insufficient. The deposited Cu film 

was removed during the electroless plating process. It was 

not necessary to add the special etching process whenever 

the line width was more than 50 m using conductive particles 

of several hundred nanometers in diameter, the Cu 

film was deposited successfully without the plasma etching 

in that case. 

The failure of the electroless plating was considered to 

be attributable to two factors, namely, only a small area of 

the superfine conductive particles was exposed for electroless 

plating solution and the “anchor effect” was not obtained 

sufficiently for the flat surface of the as-printed pattern. 

Therefore, surface modification as described above was 

needed before electroless plating. 



Figure 7. SEM images of pattern surface of seed toner layer: a Printed 

pattern and b after surface modification. 

Figure 9. SEM cross-sectional view of Cu conductive line. 

Figure 8. SEM images of L/S=1/1 pixel Cu conductive lines. Substrate: 

Si wafer with insulation film of epoxy resin. a Cu conductive 

pattern and b edge of Cu line. 

Electroless Plating 

Samples were prepared by cleaning in low-concentration 

acid before electroless plating. The Cu layer was deposited 

on the printed surface by using a plating solution based on 

ethylenediamine-tetraacetic acid. Samples were dipped in 

Thru-Cup ELC-SP ® ,providedbyC.Uyemura&Co.,Ltd.,at 

60°C. 

SEM images of Cu conductive pattern are shown in 

Figure 8. The thickness of the toner layer was 1 m and 

that of Cu layer was 3 m. Figure 8(b) shows an edge of 

the Cu line. The raggedness of the line edge was 2 m. Cu 

lines were clearly isolated from one another. The amount of 

scattered particles between the printed lines was decreased 

by applying the surface modification, as shown in Fig. 6. A 

high-resolution multiline pattern was obtained by using our 

fabrication process. 

Figure 9 shows the SEM cross-sectional view of the Cu 

line. The cross section of the seed toner layer has fine irregularities 

whose maximum height (Rz) per reference length c 

(where c=1 m) is300 nm. The very fine irregularities 

were obtained by the plasma etching. Adhesion between the 

seed toner layer and Cu layer was increased dramatically by 

applying surface modification. No peeling off of the layers 

was observed during the electroless plating. 

The volume resistivity of the Cu line was 

2.110 −6 cm, which is 1.2 times as high as that of bulk 

Figure 10. Reliability test: a Test pattern L/S=2/4 pixel, b reflow 

oven condition, and c reflow test. 

Cu. The volume resistivity was slightly higher than that of 

bulk Cu because the fabrication process was insufficiently 

optimized in the present study. That is an issue to be tackled 

in the next phase of this research. 

Reliability Test 

A reliability test was carried out. Change of the resistance of 

the wiring patterns for L/S=2 pixel/4 pixel shown in 

Figure 10(a) was measured after applying the reflow test. 

Samples were covered with Au t=50 nm/Ni t=4 m 

layers by additional electroless plating. The reflow condition 

was set up for a peak temperature of 260°C, as shown in 

Fig. 10(b), which is the usual condition for a lead-free solder 

bump. The reflow test is shown schematically in Fig. 10(c). 

Results of the reflow test are shown in Figure 11. The 

rates of resistance change were 10% for every sample after 

the reflow test had been performed eight times. The reflow 

test is severe in the case of the samples including the resin 

layer. Though the surface of the insulating film was partly 

decomposed after the reflow test had been performed a few 

times, the resistance values were stable. The circuits will accordingly 

be compatible with the conventional bump assembly 

process. 



Figure 11. Results of the resistance change after the reflow test. 

Figure 13. Surface potential distributions V p of 2540 dpi multiline pattern: 

a L/S=10/10 m andb L/S=10/20 m. 

Figure 12. Analysis model of the development area: V p x, surface potential 

distribution of photoreceptor; V d , development bias; E x , E y , electric 

field in the development area; and tx,y, potential distribution in the 

development area. 

Furthermore, for the purpose of examining the capability 

of multiline-pattern formation using liquid development 

with a 2450 dpi resolution exposure system, a theoretical 

analysis is discussed in the next section. 

THEORETICAL ANALYSIS OF MULTILINE-PATTERN 

FORMATION USING LIQUID TONER 

DEVELOPMENT 

In our previous study, 9–13 the high-resolution liquid toner 

development process of a very fine single dot with 2540 dpi 

resolution was analyzed theoretically and experimentally. In 

the present study, the development processes of 

L/S=1 pixel/1 pixel multiline patterns formed with the 

2540 dpi resolution exposure system are analyzed. 

Analysis Models 

An analyzed model is a multiline pattern including ten lines, 

whose line width is 1 pixel =10 m and space is 10 m 

and 20 m in each case. The development area illustrated in 

Figure 12 is defined as the rectangle of 300 m for 

L/S=10/10 m and 380 m for L/S=10/20 m, respectively, 

on the x-axis and 150 m on the y-axis. The surface 

potential distributions on the photoreceptor V p determined 

from the exposure energy distribution and the photoinduced 

discharge curve (PIDC) of the positively charged 

photoreceptor are shown in Figure 13 (V 0 =700 V, 

V d =400 V). 9–13,16 However, V p for L/S=10/20 m were 

sufficiently isolated from each other, but those for 

L/S=10/10 m were lower because of the narrow space. 

Two-Dimensional Liquid Development Model 

The toner particles migrate toward the latent image due to 

the effect of the electric field E originated from the surface 

potential on the photoreceptor V p and the development bias 

V d . The space and time distributions of the charge density 

are brought about by the continuity equations [Eqs. (1) and 

(2)] and Poisson’s equation [Eq. (3)]. 9–13,17–19 The calculation 

was performed using differential equations. The charge 

density of the toner particles p is positive, and the charge 

density of the counter ions n is negative. In the initial state 

of the development process, both the charge densities, p 

and n , are homogeneous in the area of analysis and their 

absolute values P 0 are the same. The values used in the 

analysis are listed in Table I 

2 t 

x 2 

p 

t =− p p E x 

x 

n 

t 

+ 2 t 

y 2 

=+ n n E x 

x 

− p p E y 

, 1 

y 

+ n n E y 

, 2 

y 

=− E x 

x + E y 

=− 

y p + n 

. 3 

t 

Table I. Parameters for numerical simulation. 

Parameters Symbol Value Unit 

Development time T d 48 msec 

Initial charge density of liquid toner P 0 1.54 C/m 3 

Relative dielectric constant of liquid toner t 2.03 - 

Development gap dt 150 µm 

Mobility of toner particle µ p 4.00 10 −10 m 2 / Vsec 

Mobility of counter ion µ n 2.00 10 −10 m 2 / Vsec 



Figure 15. Time dependence of toner distributions for L/S 

=10/10 m: a 1 msec, b 5 msec, and c 48 msec. 

toner particles. Toner particles deposited almost equally on 

each line in the early stage of the development process 

t=1 msec. However, toner density of the area outside the 

multiline pattern decreased rapidly and the depletion area 

became wider as the development process proceeded. The 

amounts of deposited toner on both outer lines were less 

than those of deposited toner on inner lines in the last stage 

of the development process t=48 msec. 

Multiline Formation 

Figure 16 shows the L/S dependence of toner distribution in 

the case of L/S=10/10 m and 10/20 m, respectively. Although 

the surface potential V p for L/S=10/10 m becomes 

lower, the lines formed on the photoreceptor are sufficiently 

isolated from one another. 

In this result, the above mentioned “edge effects” were 

also observed. In particular, the edge effects are rather significant 

for the fine-pitch mode. 

Figure 14. Calculated potential distributions in the development area: 

a L/S=10/10 m andb L/S=10/20 m. 

Potential and Charge Density Distribution of Multiline 

Pattern 

The calculated potential distributions in the development 

area for L/S=10/10 m and 10/20 m, respectively, are 

shown three-dimensionally in Figure 14. The x-axis indicates 

the position in subscanning direction on the photoreceptor 

surface, and the y-axis indicates the development gap between 

the photoreceptor and the development roller. It indicates 

that the potential gradient in the positive direction of 

y-axis is low except in the vicinity of the photoreceptor surface. 

Figure 15 shows the time dependence of toner distribution 

for L/S=10/10 m. Clear contrast of the charge density 

was observed in the vicinity of the photoreceptor surface. 

The multiline pattern formed gradually by migration of 

Figure 16. L/S dependence of toner distributions for multiline pattern: 

a L/S=10/10 m and b L/S=10/20 m. 



In order to obtain fine multiline patterns with a uniform line 

width, image processing will be required. 

The numerical analysis of fine-pitch multiline patterns 

agrees well with that of the actual toner images obtained 

experimentally. We infer that the simulation results precisely 

describe the actual toner development system. 

Figure 17. L/S=10/10 m multiline pattern on the Si substrate. 

EXPERIMENTAL RESULTS 

The seed toner images of L/S=10/10 m multiline pattern 

were realized by using the real 2540 dpi resolution LSU. 

Figure 17 shows the lines printed on Si wafer with the insulating 

coating. Figure 18(a) shows the lines on the polyimide 

film. Width and height of the lines were measured by laser 

microscope [Fig. 18(b)]. The edge effect is observed; the 

amount of deposited toner for the outer line (line No. 1) was 

less than that of deposited toner for inner lines (Nos. 2–4). 

Figure 18. L/S=10/10 m multiline pattern on the polyimide film: a 

Printed pattern on the polyimide film and b line shape data measured by 

laser microscope. 

CONCLUSION 

We have developed a digital fabrication process using highresolution 

liquid toner electrophotography. The multiline 

pattern of L/S=21.6/21.6 m was formed by printing the 

seed toner with a 1200 dpi resolution LED exposure system, 

and the Cu layer was deposited on the printed lines by electroless 

plating after surface modification. The liquid toner 

development of multiline patterns formed with 2540 dpi 

resolution LSU was analyzed theoretically. The numerical 

analysis agrees well with the experimental results. These results 

confirm that the fabrication process using liquid toner 

electrophotography is available for forming high-resolution 

multiline patterns. 


The authors wish to thank H. Aoki and N. Yamaguchi of 

Semiconductor Company, Toshiba Corporation, for advice 

on the assembly technology. They also wish to thank T. 

Yasumoto and N. Yanase of Semiconductor Company, 

Toshiba Corporation, for advice on the surface modification. 

They are grateful to S. Sakamoto and Y. Yamamoto of C. 

Uyemura & Co., Ltd., for supporting the electroless plating 

process. 

REFERENCES 

1 P. H. Kydd and D. L. Richard, “Electrostatic printing of Parmod 

electrical conductors”, Proc. IS&T’s NIP14 (IS&T, Springfield, VA, 1998) 

pp. 222–225. 

2 P. Calvert, “Inkjet printing for materials and devices”, Chem. Mater. 13, 

3299 (2001). 

3 N. Yamaguchi, H. Aoki, C. Takubo, K. Imamiya, T. Yamauchi, and H. 

Hashizume, “New process using electrophotographic technology for 

manufacturing printed circuit board”, Proc. ISJ’s Japan Hardcopy 2004 

(The Imaging Society of Japan, Tokyo, 2004) pp. 121–124. 

4 H. Aoki, N. Yamaguchi, and C. Takubo, “A study of electrophotography 

process for manufacturing printed circuit board”, Proc. IS&T’s NIP20 


5 H. Yagi, Y. Shinjo, H. Oh-oka, M. Saito, K. Ishii, I. Takasu, and M. 

Hosoya, “Image-on-image color process using liquid toner”, Proc. 

IS&T’s NIP16 (IS&T, Springfield, VA, 2000) pp. 246–250. 

6 S. Hirahara, T. Watanabe, M. Saito, A. Iida, K. Ishii, and M. Hosoya, 

“Liquid toner image transfer using shearing stress [I]”, J. Imag. Soc. Jpn. 

42, 17 (2003). 

7 A. Iida, Y. Shinjo, H. Nukada, S. Hirahara, N. Yoshikawa, and M. 

Hosoya, “A study of the relationship between drying state of toner and 

transfer performance in the IOI color process using liquid toner”, Proc. 

International Congress of Imaging Science 2002 (The Imaging Society of 

Japan, Tokyo, 2002) pp. 596–597. 

8 A. Iida, Y. Shinjo, H. Nukada, N. Yoshikawa, S. Hirahara, and M. 

Hosoya, “Relationship between drying state of toner image and transfer 

performance in the IOI color process using liquid toner”, J. Imag. Soc. 

Jpn. 42, 24 (2003). 

9 K. Ishii, M. Takahashi, H. Nagato, K. Higuchi, M. Hosoya, and K. 

Komata, “2540 dpi full color imaging creation with a liquid 

electrophotography system”, Proc. IS&T’s NIP19 (IS&T, Springfield, VA, 

2003) pp. 9–12. 

10 K. Ishii, M. Takahashi, H. Nagato, K. Higuchi, M. Hosoya, and K. 

Komata, “Liquid toner electrophotography for 2540 dpi high resolution 

image”, J. Imag. Soc. Jpn. 44, 437 (2005). 



11 H. Yagi, Y. Shinjo, H. Oh-oka, M. Saito, K. Ishii, H. Nukada, and M. 

Hosoya, “Image-on-image color process using liquid toner”, J. Imag. 

Soc. Jpn. 44, 416 (2005). 

12 Y. Shinjo, H. Yagi, M. Takahashi, K. Ishii, I. Takasu, and M. Hosoya, 

“Theoretical analysis and experiment of high-resolution development 

using liquid toner”, J. Imag. Soc. Jpn. 44, 429 (2005). 

13 I. Takasu, H. Yagi, Y. Shinjo, M. Takahashi, and M. Hosoya, “Analysis of 

2540 dpi dot reproduction by liquid development”, Proc. IS&T’s NIP20 


14 A. Iida, K. Ishii, Y. Shinjo, H. Yagi, M. Saito, H. Nukada, M. Takahashi, 

and M. Hosoya, “A study of digital fabrication using high-resolution 

liquid toner electrophotography”, Proc. IS&T’s Digital Fabrication 2005 


15 J. N. Israelachivili, Intermolecular and Surface Forces (Academic Press, 

New York, 1992). 

16 Y. Kishi, M. Kadonaga, and Y. Watanabe, “Numerical simulation of toner 

motion in magnetic brush development system”, Proc. ISJ’s Japan 

Hardcopy ’99 (The Imaging Society of Japan, Tokyo, 1999) pp. 177–180. 

17 I. Chen, “A charge transport model of liquid development for 

electrography”, J. Imaging Sci. Technol. 39(6), 473 (1995). 

18 F. J. Wang, G. A. Demoto, H. R. Till, and J. F. Knapp, “Electrophoretic 

deposition of liquid toners in a plate-out cell—A numerical analysis”, 

Proc. IS&T’s NIP15 (IS&T, Springfield, VA, 1999) pp. 623–626. 

19 S. Aoki, H. Ishida, and G. Takano, “3D toner image simulation 

considering the latent image formation with laser exposure”, Proc. ISJ’s 

Japan Hardcopy 2003 (The Imaging Society of Japan, Tokyo, 2003) pp. 

273–276. 


COMING SOON!... to the IS&T Wiley Series 

COLOR CONSTANCY 

Marc Ebner 

Julius-Maximilians-Universitat Wurzburg, Germany 

ISBN: 978-0-470-05829-9 

Hardcover 

416 pages 

June 2007 

List price: $135.00 

IS&T price 

Member $115.00 Non-member $125.00 

This book provides a comprehensive introduction to the field of color constancy, describing all the 

major color constancy algorithms, as well as presenting cutting edge research in the area of color 

image processing. Beginning with an in-depth look at the human visual system, Ebner goes on to: 

• examine the theory of color image formation, color reproduction and different 

color spaces; 

• discuss algorithms for color constancy under both uniform and non-uniforms 

illuminants; 

• describe methods for shadow removal and shadow attenuation in digital images; 

• evaluate the various algorithms for object recognition and color constancy and 

compare this to data obtained from experimental psychology; 

• set out the different algorithms as pseudo code in an appendix at the end of the book. 

Color Constancy is an ideal reference for practising engineers, computer scientists and researchers 

working in the area of digital color image processing. It may also be useful for biologists or scientists 

in general who are interested in computational theories of the visual brain and bio-inspired 

engineering systems. 

For more information please visit: 

http://www.wiley.com/WileyCDA/WileyTitle/productCd-0470058293.html

Journal of the Imaging Society of Japan VOL.46 NO.4 

2007 

CONTENTS 

Original Papers 

Analysis of the Penetration Action to the Ink-Jet Media of Pigment Ink-Jet Ink 

K. MURAYAMA, K. INOUE, M. YATAKE and K. KOROGI ...2362 

An Action Model of Silica Additives on the Electrification and the Adhesion of Toner 

H. OKADA, M. TAKEUCHI and J.M. EUN ...2417 

Investigation of Color Electrophoretic Display Utilizing Electrophoretic Colored Particles 

S. SUNOHARA and T. KITAMURA ...24713 

Imaging Today 

Latest Toner Technologies 

IntroductionH. YAMAZAKI, K. NAGATO, N. KOBAYASHI and H. NAITO ...25420 

Current Technologies of Suspension Polymerization Toner M. UCHIYAMA ...25521 

Technology Development of EA-HG Toner F. KIYONO ...26127 

Development of KONICA MINOLTA HD(High Definition) Digital Toner 

Produced by Emulsion Polymerization and Coagulation Method T. YAMANOUCHI ...26632 

Ester Elongation Polymerized Toner 

H. YAMADA, J. AWAMURA, H. NAKAJIMA, A. KOTSUGAI, F. SASAKI and T. NANYA ...27137 

Toner Manufacturing Using a Chemical Milling Method E.J. CHOI, C.H. KIM and H.N. YOON ...27743 

The Powder Processing Machines to Do with Grinding Toner Production and a New Technology 

for Toner Particle Production M. YOSHIKAWA ...28349 

The Manufacture of the Small Particle Size Toner in a Dry Process for High Speed Printing, 

Fine Quality and Oil-less Fusing S. OMATSU ...28955 

Lectures in Science 

Introduction of OpticsII 

The Behavior of a Beam of Light upon Reflection or Refraction at a Single Spherical Surface 

H. MUROTANI ....29561 

ISJ’s Awards 2006 30268 

Information Materials for the 50th ISJ General Meeting 31278 

Meeting Reports 340106 

Announcements 342108 

Guide for Authors 346112 

Contents of J. Photographic Society of Japan347113 

Contents of J. Printing Science and Technology of Japan 348114 

Contents of J. Inst. Image Electronics Engineers of Japan 349115 

Contents of Journal of Imaging Science and Technology 350116 

Essays on Imaging 

The Imaging Society of Japan 

c/o Tokyo Polytechnic University, 2-9-5, Honcho, Nakano-ku, Tokyo, 1648768 Japan 

Phone :033373-9576 Fax :033372-4414 E-mail :

imaging.org 

IS&T 

your source for imaging science information 

CHECK OUT THESE NEW OFFERINGS FROM THE IS&T BOOKSTORE 

CLOSE RANGE PHOTOGRAMMETRY 

Principles, Methods and Applications 

Thomas Luhmann, Stuart Robson, Stephen Kyle, and Ian Harley, published by John Wiley & Sons, 2007, 528 pp, 

ISBN: 978-0-470-10633-4, 2 lbs. 

English-language edition of a book originally published in 2003 in German. The translation was undertaken with cooperation of 

the original author and a team of three researchers in the UK. The book describes the fundamentals of photogrammetry, the 

equipment used in both hardware and software, analytical methods and various real-world applications where the technique has 

found broad use over the past 20 years. Photogrammetry is a very strong field in Germany, and many of the best researchers, 

leading companies and equipment manufacturers are based there. Because of this, the book is authoritative, and its translation 

into English will be appreciated by users in North America who often don’t have access to the latest writing on the topic. 

Members $100 / Non-Members $110 (list price $120) 

DIGITAL VIDEO AND HDTV 

Algorithms and Interfaces, First Edition 

Charles Poynton, published by Morgan Kauffman, 2002, 736 pp., ISBN: 1-55860-792-7, 4 lbs. 

Rapidly evolving computer and communications technologies have achieved data transmission rates and data storage 

capacities high enough for digital video. But video involves much more than just pushing bits! Achieving the best possible 

image quality, accurate color, and smooth motion requires understanding many aspects of image acquisition, coding, 

processing, and display that are beyond the usual realm of computer graphics. At the same time, video system designers are 

facing new demands to interface with film and computer system that require techniques outside conventional video engineering. 

Member: $58.00 / Non-Member: $63.00 (list price $67.95) 

DIFFRACTION, FOURIER OPTICS AND IMAGING 

Okan K. Erosy, published by John Wiley & Sons, 2006, 413 pp, ISBN: 978-0-23816-4, 2 lbs. 

This book presents current theories of diffraction, imaging, and related topics based on Fourier analysis and synthesis 

techniques, which are essential for understanding, analyzing, and synthesizing modern imaging, optical communications 

and networking, as well as micro/nano systems. Applications covered include tomography; magnetic resonance imaging; 

synthetic aperture radar (SAR) and interferometric SAR; opticalcommunications and networking devices; computer-generated 

analog holograms; and wireless systems using EM waves. 

Members $95.00 / Non-Members $105.00 (list price $110.00) 

DIGITAL IMAGE PROCESSING 

PIKS Scientific Inside, 4th Edition (includes CD Rom) 

William K. Pratt, published by John Wiley & Sons, 2007, 782 pp., ISBN: 0-471-76777-0, 3 lbs. 

The Fourth Edition of Digital Image Processing provides a complete introduction to the field and includes new information that 

updates the state of the art. The text offers coverage of new topics and includes interactive computer display imaging examples 

and computer programming exercises that illustrate the theoretical content of the book. These exercises can be implemented using 

the Programmer’s Imaging Kernel System (PIKS) application program interface included on the accompanying CD. 

Members $105.00 / Non-Members $115.00 (list price $125.00)

IS&T Corporate Members 

IS&T Corporate Members provide significant financial support, thereby assisting the Society in achieving its goals of 

disseminating information and providing professional services to imaging scientists and engineers. In turn, the Society 

provides a number of material benefits to its Corporate Members. For complete information on the Corporate 

Membership program, contact IS&T at info@imaging.org. 

Sustaining Corporate Members 

Adobe Systems Inc. 

345 Park Avenue 

San Jose, CA 95110-2704 

Eastman Kodak Company 

343 State Street 

Rochester, NY 14650 

Hewlett-Packard Company 

1501 Page Mill Road 

Palo Alto, CA 94304 

Lexmark International, Inc. 

740 New Circle Road NW 

Lexington, KY 40511 

Xerox Corporation 

Wilson Center for Research and 

Technology 

800 Phillips Road 

Webster, NY 14580 

Supporting Corporate Members 

Fuji Photo Film Company, Ltd. 

210 Nakanuma, Minami-ashigara 

Kanagawa 250-0193 Japan 

Konica Minolta Holdings Inc. 

No. 1 Sakura-machi 

Hino-shi, Tokyo 191-8511 Japan 

TREK, Inc./TREK Japan KK 

11601 Maple Ridge Road 

Medina, NY 14103-0728 

Pitney Bowes 

35 Waterview Drive 

Shelton, CT 06484 

Donor Corporate Members 

ABBY USA Software House, Inc. 

47221 Fremont Blvd. 

Fremont, CA 94538 

Axis Communications AB 

Embdalavägen 14 

SE-223 69 Lund, Sweden 

Ball Packaging Europe GmbH 

Technical Center Bonn 

Friedrich-Woehler-Strasse 51 

D-53117 Bonn, Germany 

Cheran Digital Imaging & Consulting, Inc. 

798 Burnt Gin Road 

Gaffney, SC 29340 

Clariant Produkte GmbH 

Division Pigments & Additives 

65926 Frankfurt am Main Germany 

Felix Schoeller Jr. GmbH & Co. KG 

Postfach 3667 

D-49026 Osnabruck, Germany 

Hallmark Cards, Inc. 

Chemistry R & D 

2501 McGee, #359 

Kansas City, MO 64141-6580 

ILFORD Imaging Switzerland GmbH 

Route de l’Ancienne Papeterie 1 

CH-1723 Marly, Switzerland 

MediaTek Inc. 

No. 1 Dusing Rd., 1 

Hsinchu 300 R.O.C, Taiwan 

Pantone, Inc. 

590 Commerce Blvd. 

Carlstadt, NJ 07072-3098 

Quality Engineering Associates (QEA), Inc. 

99 South Bedford Street, #4 

Burlington, MA 01803 

The Ricoh Company, Ltd. 

16-1 Shinei-cho, Tsuzuki-ku 

Yokohama 224-0035 Japan 

Sharp Corporation 

492 Minosho-cho, Yamatokoriyama 

Nara 639-1186 Japan 

Sony Corporation/ 

Sony Research Center 

6-7-35 Kita-shinagawa 

Shinagawa, Tokyo 141 Japan 

Torrey Pines 

5973 Avenida Encinas, Suite 140 

Carlsbad, CA 92008 

*as 9/1/07

Advanced Measurement Systems for All R&D and 

Quality Control Needs in Electrophotography, 

Inkjet and Other Printing Technologies 

PDT ® -2000 series 

Electrophotographic characterization, 

uniformity mapping, and defect 

detection for large and small format 

OPC drums 

PDT ® -1000L 

PDT ® -1000 

ECT-100 TM 

OPC drum coating thickness 

gauge 

Electrophotographic 

Component Testing 

MFA-2000 TM 

Magnetic field distribution 

analysis in mag roller magnets 

DRA-2000 TM 

Semi-insulating components testing 

including charge rollers, mag rollers, 

transfer rollers, transfer belts, and 

print media 

TFS-1000 TM 

Toner fusing latitude testing 

Objective Print Quality 

Analysis for All Digital 

Printing Technologies 

IAS ® -1000 

Fully-automated high volume print 

quality testing 

Scanner-based high speed print 

quality analysis 

Scanner IAS ® 

Personal IAS ® 

Handheld series for print quality, 

distinctness of image (DOI), and 

color measurements. Truly portable; 

no PC connection required 

PocketSpec TM 

DIAS TM 

Quality Engineering Associates, Inc. 

99 South Bedford Street #4, Burlington, MA 01803 USA 

Tel: +1 (781) 221-0080 • Fax: +1 (781) 221-7107 • info@qea.com • www.qea.com

imaging.org 

your source for imaging technology conferences 

2008 MEETING CALENDAR 

4th European 

Conference 

on Colour in 

Graphics, Imaging, 

and Vision 

CGIV 2008 

Archiving 2008 

June 24-27, 2008 

Bern, Switzerland 

June 9-13, 2008 

Terassa, Spain 

CALL FOR PAPERS 

Abstract deadline: November 15, 2007 

www.imaging.org/conferences/cgiv2008/ 


Abstract deadline: November 26, 2007 

www.imaging.org/conferences/archiving2008/ 

www.imaging.org/conferences/NIP24/ 

NIP24 

24th International Conference on 

Digital Printing Technologies 

Abstract deadline: February 11, 2008 

September 7-12, 2008 

Pittsburgh, Pennsylvania 

Digital 

Fabrication 

2008 

www.imaging.org/conferences/DF2008/ 

CIC16 

16th Color Imaging Conference 

November 10-14, 2008 

Portland, Oregon 


Abstract deadline: April 15, 2008 

www.imaging.org/conferences/CIC16/ 

Visit www.imaging.org to download programs and general information on these and other IS&T-sponsored conferences.

JIST - Society for Imaging Science and Technology

Create successful ePaper yourself

Delete template?

Save as template?