Engineering 02 08 2010092609263635 PDF

ISSN: 1947-3931 Volume 2, Number 8, August 2010
Engineering
Prof. David L. Carroll
Wake Forest University, USA
ISSN: 1947-3931
www.scirp.org/journal/eng/
9 771947 393005
08
Journal Editorial Board

ISSN: 1947-3931 (Print), 1947-394X (Online)
H http://www.scirp.org/journal/ eng
Editor-in-Chief
Editorial Board
Dr. Rehan Ahmed
H e r i o t-Wa t t U n i v e r s i t y, U K
Prof. MohD A. M. Al-Nimr
J o r d a n U n i v e r s i t y o f S c i e n c e a n d Te c h n o l o g y, J o r d a n
Prof. Jan Awrejcewicz
Te c h n i c a l U n i v e r s i t y o f L o d z , Po l a n d
Prof. Shahnor Basri
P u t ra U n i v e r s i t y, M a l a y s i a
Prof. Alain. Bernard
E c o l e C e n t ra l e d e N a n t e s , F ra n c e
Dr. P. Chandramohan
A n n a U n i v e r s i t y, I n d i a
Dr. Hongyang Chen
T h e U n i v e r s i t y o f To k y o , J a p a n
Prof. Ming Chen
S o u t h e a s t U n i v e r s i t y, C h i n a
Dr. Rahat Iqbal
C o v e n t r y U n i v e r s i t y, U K
Prof. Shouetsu Itou
Ka n a g a w a U n i v e r s i t y, J a p a n
Prof. Jae Moung Kim
I N H A U n i v e r s i t y, Ko r e a ( S o u t h )
Prof. Chui-Chi Lee
S H U -T E U n i v e r s i t y, Ta i w a n ( C h i n a )
Prof. Yong-Gang Lv
C h o n g q i n g U n i v e r s i t y, C h i n a
Prof. John Marsh
I E E E P h o t o n i c s S o c i e t y, U K
Dr. Mehrdad Massoudi
U . S . D e p a r t m e n t o f E n e r g y / N E T L- P i t t s b u r g h , U S A
Dr. Koduri Venkata Surya Ramam
Universidad of Concepcion, Chile
Prof. Sergey M. Smolskiy
M o s c o w Po w e r E n g i n e e r i n g U n i v e r s i t y, R u s s i a
Prof. Hongbin Sun
Ts i n g h u a U n i v e r s i t y, C h i n a
Prof. Baolin Wang
H a r b i n I n s t i t u t e o f Te c h n o l o g y, C h i n a
Prof. Chengshan Wang
T i a n j i n U n i v e r s i t y, C h i n a
Prof. Gaofeng Wang
Wu h a n U n i v e r s i t y, C h i n a
Prof. Zhao Xu
Te c h n i c a l U n i v e r s i t y o f D e n m a r k , D e n m a r k
Dr. Wei Yan
Tr e n d M i c r o , U S A
Dr. V.P. Yashnikov
Russian Academy of Sciences, Russia
Prof. Tsutomu Yoshihara
Wa s e d a U n i v e r s i t y, J a p a n
Prof. Xiangjun Zeng
C h a n g s h a U n i v e r s i t y o f S c i e n c e a n d Te c h n o l o g y, C h i n a
Dr. Hongyu Zhang
C e r e s I n c ., T h o u s a n d O a k s , U S A
Prof. Luowei Zhou
Editorial Assistant
Yanna Li
Scientific Research Publishing, USA
eng@scirp.org
Engineering, 2010, 2, 559-672

Published Online August 2010 in SciRes (http://www.SciRP.org/journal/eng/)
TABLE OF CONTENTS
Volume 2
Number 8
August 2010
Experimental Comparative and Numerical Predictive Studies on Strength Evaluation of Cement

Types: Effect of Specimen Shape and Type of Sand
H. Hodhod, M. A. M. Abdeen559
Hydrogen Pick up in Zircaloy-4: Effects in the Dimensional Stability of Structural Components

under Nuclear Reactor Operating Conditions
P. Vizcano, C. P. Fagundez, A. D. Banchik573
Electrochemical Generation of Zn-Chitosan Composite Coating on Mild Steel and its Corrosion Studies
K. Vathsala, T. V. Venkatesha, B. M. Praveen, K. O. Nayana580
Tunable Erbium-Doped Fiber Lasers Using Various Inline Fiber Filters

S.-K. Liaw, K.-C. Hsu, N.-K. Chen585
Behaviour of a Composite Concrete-Trapezoidal Steel Plate Slab in Fire

T. Hozjan, M. Saje, I. Planinc, S. Srpi, S. Bratina594
The Effect of Initial Oxidation on Long-Term Oxidation of NiCoCrAlY Alloy

C. Zhu, X. Y. Wu, Y. Wu, G. Y. Liang602
Highly Nonlinear Bending-Insensitive Birefringent Photonic Crystal Fibres

H. Ademgil, S. Haxha, F. AbdelMalek608
Progress in Antimonide Based III-V Compound Semiconductors and Devices

C. Liu, Y. B. Li, Y. P. Zeng617
Lie Group Analysis for the Effects of Variable Fluid Viscosity and Thermal
Radiation on Free Convective Heat and Mass Transfer with Variable Stream Condition
P. Loganathan, P. P. Arasu625
Statistical Modeling of Pin Gauge Dimensions of Root of Gas Turbine Blade in Creep Feed Grinding
Process
A. R. Fazeli635
Wind Turbine Tower Optimization under Various Requirements by Using Genetic Algorithm
S. Yldrm, . zkol641
A Device that can Produce Net Impulse Using Rotating Masses

C. G. Provatidis648
Computer-Aided Solution to the Vibrational Effect of Instabilities in Gas Turbine Compressors

E. A. Ogbonnaya, H. U. Ugwu, C. A. N. Johnson658
Flipped Voltage Follower Design Technique for Maximised Linear Operation

C.-M. Chen, K. Hayatleh, B. L. Hart, F. J. Lidgey665
Wire Bonding Using Offline Programming Method

Y. L. Foo, A. H. You, C. W. Chin668
Copyright 2010 SciRes.
ENG
Engineering
Journal Information
SUBSCRIPTIONS
Engineering (Online at Scientific Research Publishing, www.SciRP.org) is published monthly by Scientific Research Publishing, Inc.,
USA.
Subscription rates:
Print: $50 per issue.
To subscribe, please contact Journals Subscriptions Department, E-mail: sub@scirp.org
SERVICES
Advertisements
Advertisement Sales Department, E-mail: service@scirp.org
Reprints (minimum quantity 100 copies)
Reprints Co-ordinator, Scientific Research Publishing, Inc., USA.
E-mail: sub@scirp.org
COPYRIGHT
Copyright2010 Scientific Research Publishing, Inc.
All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by
any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except as described below, without the
permission in writing of the Publisher.
Copying of articles is not permitted except for personal and internal use, to the extent permitted by national copyright law, or under
the terms of a license issued by the national Reproduction Rights Organization.
Requests for permission for other kinds of copying, such as copying for general distribution, for advertising or promotional purposes,
for creating new collective works or for resale, and other enquiries should be addressed to the Publisher.
Statements and opinions expressed in the articles and communications are those of the individual contributors and not the statements
and opinion of Scientific Research Publishing, Inc. We assumes no responsibility or liability for any damage or injury to persons or
property arising out of the use of any materials, instructions, methods or ideas contained herein. We expressly disclaim any implied
warranties of merchantability or fitness for a particular purpose. If expert assistance is required, the services of a competent
professional person should be sought.
PRODUCTION INFORMATION
For manuscripts that have been accepted for publication, please contact:
E-mail: eng@scirp.org
Engineering, 2010, 2, 559-572

doi:10.4236/eng.2010.28072 Published Online August 2010 (http://www.SciRP.org/journal/eng).
Experimental Comparative and Numerical Predictive

Studies on Strength Evaluation of Cement Types: Effect of
Specimen Shape and Type of Sand
1
Hossam Hodhod1, Mostafa A. M. Abdeen2
Department of Structural Engineering, Faculty of Engineering, Cairo University, Giza, Egypt

Department of Engineering Mathematics & Physics, Faculty of Engineering, Cairo University, Giza, Egypt
E-mail: {hossamhodhod, mostafa_a_m_abdeen}@hotmail.com
Received May 6, 2010; revised July 23, 2010; accepted July 25, 2010
Abstract
Quality of cement is evaluated via group of tests. The most important, and close to understanding, is the
compressive strength test. Recently, Egyptian standards adopted the European standards EN-196 and
EN-197 for specifying and evaluating quality of cements. This was motivated by the large European investments in the local production of cement. The current study represents a comparative investigation, experimental and numerical, of the effect of different parameters on evaluation of compressive strength. Main parameters are shape of specimens and type of sand used for producing tested mortars. Three sets of specimens
were made for ten types of cements. First set were 70.6 mm cubes molded according to old standards using
single sized sand. Second group were prisms molded from standard sand (CEN sand) according to the recent
standards. Third group were prisms molded from local sand sieved and regenerated to simulate same grading
of CEN sand. All specimens were cured according to relevant standards and tested at different ages (2,3,7,10
and 28 days). Results show that CEM-I Type of cement does not fulfill, in all of its grades, the strength requirements of Ordinary Portland cement OPC specified in old standards. Also, the use of simulated CEN
sand from local source gives strengths lower than those obtained using standard certified CEN sand. A limited number of tests were made on concrete specimens from two most common CEM-I types to investigate
effect on concrete strength and results were also reported. Numerical investigation of the effect of specimen
shape and type of sand on evaluation of compressive strength of mortar specimens, presented in the current
study, applies one of the artificial intelligence techniques to simulate and predict the strength behavior at
different ages. The Artificial Neural Network (ANN) technique is introduced in the current study to simulate
the strength behavior using the available experimental data and predict the strength value at any age in the
range of the experiments or in the future. The results of the numerical study showed that the ANN method
with less effort was very efficiently capable of simulating the effect of specimen shape and type of sand on
the strength behavior of tested mortar with different cement types.
Keywords: Cement Type, Sand Type, Mortar Specimen, Strength, Modeling, Artificial Neural Network
1. Introduction
For decades, engineers used to apply cement based on
certain classification [1-3]. This classification refers to
its composition and consequently relevant properties.
Among these properties, strength was the main target of
using a specific type of cement. Ordinary Portland cement (OPC), sulphate resisting cement (SRC) and white
cement share almost same values for compressive
strength at different ages. One type: namely rapid hardCopyright 2010 SciRes.
ening cement had the higher early strength than others.

Recently, end of the year 2006, the Egyptian standards [4]
decided to adopt the European standards EN196 & EN
197 [5] for producing, specifying and testing almost all
types of cements. The new standard took the designation
ES4756 and included all types of cements but SRC. The
new standard included a drastic change in specifying
cement types, and appeared ambiguous in many aspects
since it added ranks and rate of hardening for the same
composition of cement. This raised many questions
ENG
560
H. HODHOD ET AL.
about the actual composition of cement and its properties.

Also, questions were raised about the properties of cemtitious mixes for which cement is used and whether the
correlations between properties and type of cement will
remain the same or not. Besides, methods of testing cement to evaluate its compressive strength were changed
from using cubic specimens (70.6 mm side length) [6] to
part of prism (with 40 mm square cross section). Moreover, the standard dictated the use of specific type of
sand, which is not available locally, for making mortar
specimens. This sand must be procured from certified
suppliers and is called CEN sand. Such a condition
raised a question about the role of this sand in hydration
process and strength development too. All these questions motivate the need for research to clarify nature of
new cement types and declare their properties and effect
on properties of cementitious mixes. One local attempt [7]
was made and yielded that new standards are efficient.
However, the contradictions between test results of the
study and the known size effect rule urge the need for
more investigation.
Since the experimental work needs a lot of effort, time
and money, which is quite clear from the literature mentioned previously, the need for utilizing new methodologies and techniques to reduce this effort, save time and
money (and at the same time preserving high accuracy)
is urged. Artificial intelligence has proven its capability
in simulating and predicting the behavior of the different
physical phenomena in most of the engineering fields.
Artificial Neural Network (ANN) is one of the artificial
intelligence techniques that have been incorporated in
various scientific disciplines. Solomatine and Toorres [8]
presented a study of using ANN in the optimization loop
for the hydrodynamic modeling of reservoir operation in
Venezuela. Kheireldin [9] presented a study to model the
hydraulic characteristics of severe contractions in open
channels using ANN technique. The successful results of
his study showed the applicability of using the ANN approach in determining relationship between different
parameters with multiple input/output problems. Abdeen
[10] developed neural network model for predicting flow
characteristics in irregular open channels. The developed
model proved that ANN technique was capable with
small computational effort and high accuracy of predicting flow depths and average flow velocities along the
channel reach when the geometrical properties of the
channel cross sections were measured or vice versa. Allam [11] used the artificial intelligence technique to predict the effect of tunnel construction on nearby buildings
which is the main factor in choosing the tunnel route.
Allam, in her thesis, predicted the maximum and minimum differential settlement necessary precautionary
measures. Park and Azmathullah et al. [12] presented a
study for estimating the scour characteristics downstream
of a ski-jump bucket using Neural Networks (NN). Abdeen [13] presented a study for the development of ANN
models to simulate flow behavior in open channel infested by submerged aquatic weeds. Mohamed [14] proposed an artificial neural network for the selection of
optimal lateral load-resisting system for multi-story steel
frames. Mohamed, in her master thesis, proposed the
neural network to reduce the computing time consumed
in the design iterations. Abdeen [15] utilized ANN technique for the development of various models to simulate
the impacts of different submerged weeds' densities, different flow discharges, and different distributaries operation scheduling on the water surface profile in an experimental main open channel that supplies water to different distributaries.
2. Problem Description
To study the effect of specimens shape and types of sand
used for producing tested mortars on evaluation of compressive as well as flexural tensile strengths, experimental and numerical techniques will be presented in this
study. The experimental program and its results will be
described in detail in the following sections. After that,
numerical approach will be discussed to show the efficiency of numerical techniques. The numerical models
presented in this study utilized Artificial Neural Network
technique (ANN) using the data of experiments and then
can predict the strength value in the range of the experiment or in the future.
3. Experimental Program
The experimental program focuses on evaluating compressive strength of mortar made from new cement types.
Ten types of cements with different grades and rate of
hardening were procured from local market in Egypt.
Compressive strength was evaluated for each type using
the cubic specimens (70.6 mm side length) and using the
testing of part of prism (40*40*160 mm). The last method was employed twice. First with local sand following
the same grade specified in ES4756 (and EN 196), and
second with certified CEN sand according to same standards. Specimens were tested at ages of 2, 3, 7, 10 and
28 days.
Concrete mixes with same proportions were cast from
different types of cements. Slump and compressive
strength were measured for each mix to investigate the
effect of type of cement on concrete properties. Compressive strength was measured at 3, 7 and 28 days.
4. Materials and Specimens

Constituents for mortar and concrete mixes were:
4.1. Water
Tap water was used for mixing and curing of all speciENG
H. HODHOD
mens.
4.2. Cement
Ten types of cement were used. All were supplied in
bags carrying the symbols of both ES4756 and EN-197.
They were all produced locally in Egypt by different
Cement Companies. The ten types covered CEM I (ordinary Portland cement) with different grades and rates of
hardening. The types also included white cement, sulphate resisting cement (SRC) and CEM II type cements.
Table 1 shows the investigated types of cement.
ET AL.
561
ing compressive strength of mortar with mix proportions

of water: cement: sand = 0.4:1:3 by weight. Sand was 0.6
0.85 mm local sand. Constituents were mixed manually.
Steel molds were used for casting. The other two sets of
specimens were prisms (40*40*160 mm) cast from mixtures with proportions of water: cement: sand = 0.5:1:3
by weight.
For one set, standard CEN sand was used for casting.
For the other set, regenerated local sand with grading
4.3. Sand
Two types of sand were used for the current study: CEN
sand that was imported from France in sealed transparent
bags (Figure 1), and local siliceous sand. Local siliceous
sand was used in its natural grading (Figure 2) for casting concrete. This sand was sieved to get the single size
sand required for testing mortar cubes according to old
ES (still effective as part of local code of practice ECP
203/2001 app.3. The local sand was also used to regenerate the CEN sand by collecting different sizes in the percentages specified in EN-196. The grading of this regenerated sand, and limits of CEN sand, are shown in Figure 3.
Percentage Passing
Figure 1. Bags of CEN sand.
4.4. Gravel
Local siliceous gravel was used for casting concrete specimens. Gravel has a maximum nominal size of 20 mm.
100
80
60
40
20
0
0.01
0.1
10
Sieve Opening Size (mm)
4.5. Specimens (Cubes and Prisms)

Standard cubes with 70.6 mm side were used for evaluat-
Figure 2. Grading of local sand used in concrete mixes.
Table 1. Investigated types of cement.

No
1
Strength Evaluation (on Mortar)

Flexure
Compn
Flexure
Compn
Cube Compn.
Local sand(*)
CEN Sand
Type of Cement
Manufacturer
CEM I-52.5N
SINAI
El-MASRIYA
HELWAN
CEM I-42.5N
(SRC)(**)
CEM I-42.5N (SRC)
CEM I-42.5N
El-MASRIYA
CEM I-42.5R
HELWAN
CEM I-42.5N
HELWAN
(White)
7
CEM I-32.5R
ELKAUMIYA (NCC)
8
CEM I-32.5R (SRC)
ASSIUT (CEMEX)
9
CEM II-B-S 32.5N
HELWAN
10
CEM II-B-L -32.5N
ELKAUMIYA (NCC)
(*) Sand having a grading similar to CEN sand.
(**) This cement will be denoted in figures as SRC-1.
6
ENG
H. HODHOD
562
ET AL.
Percentage Passing
100
80
CEN Upper Limit

CEN Lower Limit
Regenerated Sand
60
40
20
0
0.01
0.1
10
Sieve Opening Size - mm
Figure 3. Grading limits of CEN sand and grading of locally

regenerated sand.
similar to CEN was used. Constituents were mixed mechanically using 5 liter mixer. Steel molds were used for
casting. For all sets, specimens were compacted using
vibrator and left covered with impervious sheet for 24
hours. Then specimens were demolded and immersed in
water till day of testing. Concrete cubes (with 150 mm
side length) were cast to evaluate concrete strength. Mix
proportions were water: cement: sand: gravel = 0.6:1:
1.5:3.0. Constituents were mixed mechanically using 140
liter tilting type mixer. Dry materials were mixed first for
about one minute. Then, water was added gradually and
mixing continued till uniform mix was obtained. Concrete was cast in steel molds and compacted using a vibrating table. Specimens were covered with plastic
sheets for 24 hours. Then molds were removed and specimens were wet cured till age of testing.
5. Test Results
The test results are explained in the following sections.
5.1. Cement Setting Time

Initial and final setting times measured for different
types of cement are shown in Figure 4. One can see that
the initial setting time ranges from 70 to 120 min. Final
setting time ranges from 140 to 240 min. Generally, final
setting time is almost double the initial setting time. The
least setting time was recorded for CEM I 52.5 N and the
longest setting time was recorded for CEM II B-S-32.5 N.
Setting time increases as cement grade decreases, and
SRC shows less setting time for same grade. Rate of setting (expressed by N or R after grade) does not seem to
affect setting time results. Recorded values of initial and
final setting times comply with limits of ES 4756 and
EN-197.
5.2. Mortar Compressive Strength

Compressive strength measured for all specimens and
Figure 4. Setting time of different types of cement.
types of cements are plotted versus time in Figure 5.

One can see that cube specimens specified in old standards give strength lower than part of prisms specified in
the current standards. Large size of cubes helps reducing
its strength as the grading of the single sized sand does.
However, the low w/c ratio is supposed to help increasing the strength of cubes. This indicates that the effect of
size and confinement of prism specimens and the grading
and type of CEN sand could compensate for the increase
of w/c ratio of the specimens.
There is a difference between results obtained for
CEN sand and regenerated sand composed by adding the
proper percentage of each size from local sand. CEN
sand always gives higher strength. This implies that it is
not only sand grading that contributes to the strength.
Shape of particles and probably some chemical characteristics of sand may also contribute to this increase of
strength. These last two points need more research for
clarification.
It must be said that the strength of prisms does not fulfill the requirements of cement grade for all types. The
strength of prisms at 28 days reaches a percentage from
43 to 70% of corresponding grade of cement. Although
the compaction was not made using a jolting table (as
specified by current standard test method), this is not
expected to yield such big difference for all types of cement.
It is note-worthy that similar strength values were obtained for all rapid setting cements regardless of their
grade. However, the normal setting CEM I 52.5 N gave
the highest strength among all other cement mortars.
Strength factor, which is the ratio of 28 day strength to
the strength at a specific age, is plotted in Figures 6-8
for different specimens and different types of cement.
The small strength values for CEN sand prisms denote
the rapid strength development of strength with this sand.
However, for same sand, it seems that strength development does not follow the indication R & N for Type of
cement, where smaller factors are observed for normal
setting cements. Similar trend was found for other types
of sand.
ENG
H. HODHOD
ET AL.
563
Figure 5. Compressive strength of mortar specimens produced under different conditions, for different types of cement.
ENG
H. HODHOD ET AL.
564
Figure 6. Strength factor for mortar prisms produced from CEN sand (left: rapid setting cements, right: normal setting cements).
Figure 7. Strength factor for mortar prisms produced from regenerated sand (left: rapid setting cements, right: normal setting cements).
Figure 8. Strength factor for mortar cubes (left: rapid setting cements, right: normal setting cements).
5.3. Mortar Tensile Strength

Flexural tensile strength was measured for prism specimens since it is the first step in producing compression
specimens. Measured flexural strength for all types of
cements and for different sands are plotted in Figure 9.
One can observe the effect of CEN sand in increasing
strength of mortar. This effect confirms the above mentioned need for investigation of particle shape and chemical reactivity of CEN sand.
One more finding can be found when tensile flexural
strength is plotted versus compressive strength at different ages, as in Figure 10. It can be seen that there exists a
significant increase of tensile strength between 7 and 28

days. This could be observed for both types of sand. This
implies that the correlation between flexural strength and
compressive strength is significantly different at early
and later ages.
5.4. Concrete Slump

Concrete slump measuring results are shown in Figure
11 for all cement types. One can identify 3 main ranges
of slump: 0-40 mm, 40-80 mm, and 80-120 mm. First
low range of slump was recorded for rapid setting cements and 52.5 grade cement. Highest slump range was
ENG
H. HODHOD
ET AL.
565
Figure 9. Flexural strength of mortar prisms produced different sand types, for different types of cement.
ENG
H. HODHOD ET AL.
566
Concrete Compressive Strength (MPa)
Figure 10. Flexural strength vs. compressive strength (left: cen sand, right: regenerated sand).
30
25
CEM I 52.5N
CEM I 42.5N-SRC
20
CEM I 42.5N
CEM I 42.5R
CEM I 42.5N white
15
CEM I 32.5R
CEM I 32.5R (SRC)
10
CEM II-B-S32.5N
CEM II B-L32.5N
5
0
0
10
20
30
Age - days
Figure 11. Slump values for different types of cement.
observed for CEM II cements. The medium grade was

observed for the rest normal setting CEM I type of cement. Since the water consumption is related to rates of
hydration and heat evolution. One can conclude that
grade 52.5 has high rate of hydration. It is noteworthy
that CEM I 52.5 R does not exist in local Egyptian market. The high slump of CEM II cement mixes can be
correlated to low clinker content.
Figure 12. Measured concrete compressive strength for

different types of cement.
strength of different grades of cement. At 28 days, SRC

of 32.5 grade yields same strength as 42.5 grade. For
normal setting cements, there is a clear distinction between strength of different grades at all ages. Strength
ratio at 28 days is almost proportional to grade of cement.
One can still observe that SRC show higher values of
strength than other CEM I cements of same grade.
5.5. Concrete Compressive Strength
6. Numerical Model Structure
Measured values of concrete compressive strength are

plotted versus age, for all types of cement, in Figure 12.
Generally, one can see in Figure 12 that the effect of
cement grade can be distinguished in the limits where top
curve belongs to grade 52. N and bottom curve belongs
to grade 32.5 N. Curves for higher grades of cement are
shown in Figure 13 Left. One can see that, up to 7 days all
42.5 grade cements show almost same strength. However,
at later ages (28 days) the rapid setting type shows higher
strength than the normal setting ones. One can also see that
SRC cement show slightly higher strength than similar
42.5 N cements. Curves for low grade cements are shown
in Figure 13 Right. The effect of setting rate can be identified between 32.5 N and 32.5 R cements.
Still SRC cement shows higher strength at 28 days.
Figure 14 shows strength development for rapid setting
and normal setting cements, respectively. For rapid setting cements, there is no difference between early age
Neural networks are models of biological neural structures. Briefly, the starting point for most networks is a
model neuron as shown in Figure 15. This neuron is connected to multiple inputs and produces a single output.
Each input is modified by a weighting value (w). The
neuron will combine these weighted inputs with reference
to a threshold value and an activation function, will determine its output. This behavior follows closely the real
neurons work of the humans brain. In the network structure, the input layer is considered a distributor of the signals from the external world while hidden layers are considered to be feature detectors of such signals. On the
other hand, the output layer is considered as a collector of
the features detected and the producer of the response.
6.1. Neural Network Operation

It is quite important for the reader to understand how the
ENG
30
25
20
CEM I 52.5N
CEM I 42.5N-SRC
15
CEM I 42.5N
CEM I 42.5R
10
CEM I 42.5N white
5
0
0
10
20
ET AL.
H. HODHOD
567
30
25
20
CEM I 32.5R
CEM I 32.5R-SRC
15
CEM II-B-S32.5N
CEM II B-L32.5N
10
5
0
0
30
10
20
30
Age - days
Age - days
30
25
20
CEM I 42.5R
15
CEM I 32.5R
CEM I 32.5R-SRC
10
5
0
0
10
20
30
Figure 13. Measured concrete compressive strength (left: high grades of cement, right: low grades of cement).
30
25
CEM I 52.5N
20
CEM I 42.5N-SRC
CEM I 42.5N
15
CEM I 42.5N white

CEM II-B-S32.5N
10
CEM II B-L32.5N
5
0
0
Age - days
10
20
30
Age - days
Figure 14. Measured concrete compressive strength (left: rapid setting cements, right: normal setting cements).
the internal value of this operation, Uj. This value is then

biased by a previously established threshold value, tj, and
sent through an activation function, Fth. This activation
function can take several forms such as Step, Linear,
Sigmoid, Hyperbolic, and Gaussian functions. The Hyperbolic function, used in this study, is shaped exactly as
the Sigmoid one with the same mathematical representation, as in Equation (3), but it ranges from 1 to +1 rather
than from 0 to 1 as in the Sigmoid one.
f x
Figure 15. Typical picture of a model neuron that exists in
every neural network.
neural network operates to simulate different physical

problems. The output of ach neuron is a function of its
inputs (Xi). In more details, the output (Yj) of the jth neuron in any layer is described by two sets of equations as
follows:
U j X i wij
(1)
Y j Fth U j t j
(2)
and
For every neuron, j, in a layer, each of the i inputs, Xi,

to that layer is multiplied by a previously established
weight, wij. These are all summed together, resulting in
1
1 e x
(3)
The resulting output, Yj, is an input to the next layer or

it is a response of the neural network if it is the last layer.
In applying the Neural Network technique, in this study,
Neuralyst Software, Shin [16], was used.
6.2. Neural Network Training

The next step in neural network procedure is the training
operation. The main purpose of this operation is to tune
up the network to what it should produce as a response.
From the difference between the desired response and
the actual response, the error is determined and a portion
of it is back propagated through the network. At each
neuron in the network, the error is used to adjust the
weights and the threshold value of this neuron. Consequently, the error in the network will be less for the same
ENG
H. HODHOD ET AL.
568
considered in this study. Two models for sitting time

(initial and final), model for cube compression strength,
two models for prism compression strength (CEN and
regenerated sand) and two models for flexural strength
(CEN and regenerated sand).
inputs at the next iteration. This corrective procedure is

applied continuously and repetitively for each set of inputs and corresponding set of outputs. This procedure
will decrease the individual or total error in the responses
to reach a desired tolerance.
Once the network reduces the total error to the satisfactory limit, the training process may stop. The error
propagation in the network starts at the output layer with
the following equations:
wij wij' LR e j X i
and,
ej
7.1. Neural network Design

To develop a neural network models to simulate the effect of specimen shape and type of sand on the strength
behavior of tested mortar, first input and output variables
have to be determined. What we have in the current
study, to be considered as an input variable, is the types
of cement used in the mortar specimen. So from the
name of cement type we have to create a certain numeric
characteristic values could be used as input variables in
the present models as shown in Table 2.
Table 3 is designed to summarize all neural network
key input variables and output for all the seven models
presented in the current study. Some abbreviations used
in Table 3 due to space limitation as follows:
Str.: Strength
Compn.: Compression
Flex.: Flexural
(4)
Yj 1 Yj d j Yj
(5)
where, wij is the corrected weight, w ij is the previous

weight value, LR is the learning rate, ej is the error term,
Xi is the ith input value, Yj is the ouput, and dj is the desired output.
7. Simulation Models
To fully investigate numerically the effect of specimen
shape and type of sand on the strength behavior of tested
mortar with different cement types, seven models are
Table 2. Characteristic values for types of cement.

No
Type of Cement
CEM I-52.5N
CEM I-42.5N
(SRC) (**)
CEM I-42.5N
(SRC)
CEM I-42.5N
CEM I-42.5R
CEM I-42.5N
(White)
CEM I-32.5R
CEM I-32.5R
(SRC)
CEM II-B-S-32.5N
CEM II-B-L-32.5N
2
3
4
5
6
7
8
9
10
I or II
32.5 or 42.5
or 52.5
N or R
A or B
S or L
SRC or
White
Manufacturer
52.5
19
100
42.5
19
100
60
67
42.5
19
100
1
1
42.5
42.5
19
23
100
100
0
0
42.5
19
100
50
32.5
23
100
60
67
60
60
0
0
32.5
23
100
67
2
2
32.5
32.5
19
19
60
60
24
12
50
50
0
0
Table 3. Key input variables and output for all ANN models.
Input Variables
Model
I
or II
Initial Sitting Time

Final Sitting Time
Cube Str.
Prism Str.
(CEN sand)
Prism Str.
(Regenerated)
Prism Flex. Str.
(CEN)
Prism Flex. Str.
(Regenerated)
32.5 or
42.5 or
52.5
N or
R
A or
B
S or
L
SRC or
White
Manufacturer
Days
Initial Time
Final Time
Compn. Str.
Compn. Str.
Compn. Str.
Output
Flex. Ten.
Str.
Flex. Ten.
Str.
ENG
H. HODHOD
ET AL.
569
The parameters of the various network models developed in the current study for the different simulation
models are presented in Table 5. These parameters can
be described with their tasks as follows:
Learning Rate (LR): determines the magnitude of the
correction term applied to adjust each neurons weights
during training process = 1 in the current study.
Momentum (M): determines the life time of a correction term as the training process takes place = 0.9 in
the current study.
Training Tolerance (TRT): defines the percentage
error allowed in comparing the neural network output to
the target value to be scored as Right during the training process = 0.001 in the current study.
Testing Tolerance (TST): it is similar to Training
Tolerance, but it is applied to the neural network outputs
and the target values only for the test data = 0.003 in the
current study.
Input Noise (IN): provides a slight random variation
to each input value for every training epoch = 0 in the
current study.
Ten.: Tensile
Several neural network architectures are designed and
tested for all simulation models investigated in this study
to finally determine the best network models to simulate,
very accurately, the effect of specimen shape and type of
sand on the strength behavior of tested mortar with different cement types based on minimizing the Root Mean
Square Error (RMS-Error). Figure 16 shows a schematic
diagram for a generic neural network. The training procedure for the developed ANN models, in the current
study, uses the experimental data presented in the previous sections of the current study. After the ANN models
are settled for all cases, prediction procedure takes place
to predict the compression as well as tensile strengths at
different age-days rather than those days measured in the
experiment (internal and after 28 days).
Table 4 shows the final neural network models for the
seven simulation models and their associate number of
neurons. The input and output layers represent the key
input and output variables described previously for each
simulation model.
Input # 1
Output # 1
Input # 2
Output # 2
Hidden layer
3 neurons
Hidden layer
3 neurons
Figure 16. General schematic diagram of a simple generic neural network.

Table 4. The developed neural network models.
Simulation
Model
No. of Layers
Initial Sitting
Time
Final Sitting
Time
Cube Str.
Prism Str.
(CEN sand)
Prism Str.
(Regenerated)
Prism Flex. Str.
(CEN)
Prism Flex. Str.
(Regenerated)
Input Layer
No. of Neurons in each Layer

Second
First Hidden
Third Hidden
Hidden
Output Layer
ENG
H. HODHOD ET AL.
570
Table 5. Parameters used in the developed neural network models.
Training
Epochs
MPRE
RMS-Error
Initial
Sitting
Time
Final
Sitting
Time
Cube
Str.
Compn.
Str. (CEN)
Compn. Str. (Regenerated)
Flex. Str.
(CEN)
Flex. Str. (Regenerated)
1146
4985
672361
301098
179853
315475
505672
0.067
0.0005
0.034
0.0005
1.175
0.0008
0.174
0.0004
0.281
0.0003
1.512
0.0016
0.321
0.0002
Function Gain (FG): allows a change in the scaling

or width of the selected function = 1 in the current study.
Scaling Margin (SM): adds additional headroom, as a
percentage of range, to the rescaling computations used
by Neuralyst Software, Shin (1994), in preparing data for
the neural network or interpreting data from the neural
network = 0.1 in the current study.
Training Epochs: number of trails to achieve the present accuracy.
Percentage Relative Error (PRR): percentage relative error between the numerical results and actual measured value and is computed according to Equation (6) as
follows:
PRE = (Absolute Value (ANN_PR AMV)/AMV)*
100
(6)
where:
ANN_PR: Predicted results using the developed ANN
model
AMV: Actual Measured Value
MPRE: Maximum percentage relative error during the
model results for the training step.
8. Results and Discussions

Numerical results using ANN technique will be presented in this section for all the seven models. Due to
space limitation in the present paper the numerical results of one type of cement (CEM I 52.5 N) will be presented to show the simulation and prediction powers of
ANN technique for compressive as well as tensile
strengths.
8.1. Sitting Time

For the sitting time models (initial and final), Table 6
presents the ANN results with experimental ones. One
can see from this table that ANN technique can simulate
very efficient the experiment results for different types of
cements for mortar specimens.
8.2. Mortar Compressive Strength

Three ANN models are developed to simulate and predict the effect of specimen shape and type of sand on
evaluating the compressive strength of mortar specimens
for all the types of cement presented in the current study
at different ages. Figures 17 and 18 show the ANN

results and experimental ones for compressive strength
(cube and prism specimens) for one type of cement at the
ages of experiment (2,3,7,10,28 days) and then predict
the behavior at 14 days (internally) and after 28 days up
to 42 days (externally). From these figures, it is very
clear that ANN technique succeeded very well to simulate and predict the compressive strength behavior at
different ages for different specimens.
8.3. Mortar Tensile Strength

Another two ANN models are developed to simulate and
predict the flexural tensile strength for prism specimen
for two types of sand (CEN and Regenerated) at different
ages. Figure 19 presents the numerical results and experimental ones at the ages of experiments (2,3,7,10,28
days). One can observe that ANN technique can simulate
the tensile behavior and then predict the strength at ages
different than the ages of experiment (before and after 28
days) very successfully.
9. Conclusions
Based on the experimental investigation conducted in the
course of the current research, the following can be concluded:
1) There is an inverse proportion between setting time
and cement grade, and a direct proportion between grade
and water requirement for standard consistency.
2) Applying old cement standards, for testing and
evaluating mortar compressive strength of cement mortar, results in rejection of new cement types. Using of
C E MI52.5N
25.0
Compressive Strength- Mpa
Simulation
Parameter
20.0
15.0
10.0
Experiment
ANN Training
5.0
ANN Prediction
0.0
0
10
20
30
40
50
Age - days
Figure 17. Cube compressive strength.
ENG
H. HODHOD
ET AL.
571
Table 6. Sitting time models results.

Simulation Model
Initial
Final
No. 2
No. 3
No. 4
No. 5
No. 6
No. 7
No. 8
No. 9
No. 10
70.0
85.0
75.0
100.0
90.0
80.0
110.0
80.0
120.0
100.0
ANN
69.9
84.9
75.0
100.0
89.9
79.9
109.9
80.0
119.9
100.0
Exp.
140.0
140.0
180.0
210.0
200.0
170.0
225.0
185.0
240.0
200.0
ANN
140.0
140.0
180.0
210.0
200.0
170.0
225.0
185.0
240.0
200.0
C E MI52.5N
25
20
15
10
Experiment
ANN Training
C E MI52.5N
30.0
30
No. 1
Exp.
ANN Prediction
25.0
20.0
15.0
10.0
Experiment
ANN Training
5.0
ANN Prediction
0.0
0
10
20
30
40
50
10
Age - days
20
30
40
50
Age - days
Figure 18. Prism compressive strength (left: cen sand, right: regenerated sand).
C E MI52.5N
12
8
Experiment
ANN Training
C E MI52.5N
16
Flexural Strength - MP
Flexural Strength - MP
16
ANN Prediction
12
8
Experiment
ANN Training
ANN Prediction
0
0
10
20
30
40
Age - days
50
10
20
30
40
50
Age - days
Figure 19. Prism flexural strength (left: cen sand, right: regenerated sand).
jolting table for compaction is essential for obtaining

successful test results according to new standards (EN
196 and ES 4756).
3) Standard CEN sand cannot be regenerated locally
based only on its grading. Further investigation is required to get its other properties like particle shape and
chemical reactivity. There is some evidence on having
early strength development when CEN sand is used in
mortar.
4) Sulphate resisting cements show higher strength
than CEM I cements of same grade, for both mortar and
concrete mixtures.
5) There is some evidence that locally available cements do not follow the rate of strength development
denoted on packs.
6) For normal setting cements (N coded) there is a
clear distinction between concrete strength obtained for

specific cement grade. However, this could not be seen
for rapid setting types (R coded).
7) Correlation between flexural tensile and compressive strength of mortar differs significantly between early
and later ages.
Based on the results of implementing the ANN technique in this study, the following can be concluded:
1) The developed ANN models presented in this study
are very successful in simulating the effect of specimen
shape and type of sand on the behavior of mortar specimens (initial and sitting times, compressive strength,
flexural tensile strength) for different types of cement.
2) The presented ANN models are very efficiently capable of predicting the strength behavior at different ages
rather than the ages of the experimental results (in the
ENG
H. HODHOD ET AL.
572
201-206.
range of the experiment or in the future).
10. Acknowledgements
The Authors would like to express their gratitude towards Prof. Dr. Farouk El-Hakim of 15th May institute
for Civil and Arch. Engineering, and undergraduate students (4th yearcivil) for the help they provided during
the experimental part of this research.
11. References
[1]
A. M. Neville, Properties of Concrete, John Wiley &

Sons, New York, 1997.
[2]
S. H. Kosmatka, et al., Design and Control of Concrete

Mixtures, 5th Edition, Cement Association of Canada,
Ottawa, 2002.
[3]
Egyptian Standards (ES) 372, Standards of Ordinary and

Rapid Hardening Portland Cements, Egyptian General
Authority for Standards, Cairo, 1991.
[4]
Egyptian Standards (ES) 5476, Standards of Cements,

Egyptian General Authority for Standards, Cairo, 2007.
[5]
EN 197, CEMENT: Part 1: Composition, Specifications

and Conformity Criteria for Common Cements, 2004.
[6]
Egyptian Code of Practice 203, Basics of Design and

Regulations of Construction of Reinforced Concrete
Structures: Appendix III, Guide for Testing of Concrete
Materials, Egyptian Ministry of Housing, Egypt, 2001.
[7]
K. M. Yosri, Properties of New Cements Produced in

Egypt as per ES 4756/2005, HBRC Journal, Vol. 3, No.
3, 2007, pp. 23-33.
[8]
D. Solomatine and L. Toorres, Neural Network Approximation of a Hydrodynamic Model in Optimizing

Reservoir Operation, Proceedings of the 2nd International Conference on Hydroinformatics, Zurich, 1996, pp.
[9]
K. A. Kheireldin, Neural Network Application for Modeling Hydraulic Characteristics of Severe Contraction,
Proceedings of the 3rd Internetional Conference, Hydroinformatics, Copenhagen, 24-26 August 1998, pp. 41-48.
[10] M. A. M. Abdeen, Neural Network Model for Predicting

Flow Characteristics in Irregular Open Channel, Scientific Journal, Faculty of Engineering-Alexandria University, Vol. 40, No. 4, 2001, pp. 539-546.
[11] B. S. M. Allam, Artificial Intelligence Based Predictions
of Precautionary Measures for Building Adjacent to Tunnel Rout during Tunneling Process, Ph.D. Thesis, Faculty of Engineering, Cairo University, Giza, 2005.
[12] H. Azmathullah, M. C. Deo and P. B. Deolalikar, Neural
Networks for Estimation of Scour Downstream of a SkiJump Bucket, Journal of Hydrologic Engineering, ASCE,
Vol. 131, No. 10, 2005, pp. 898-908.
[13] M. A. M. Abdeen, Development of Artificial Neural
Network Model for Simulating the Flow Behavior in
Open Channel Infested by Submerged Aquatic Weeds,
Journal of Mechanical Science and Technology, KSME
International Journal, Vol. 20, No. 10, 2006, pp. 16911879.
[14] M. A. M. Mohamed, Selection of Optimum Lateral
Load-Resisting System Using Artificial Neural Networks, M.Sc. Thesis, Faculty of Engineering, Cairo
University, Giza, 2006.
[15] M. A. M. Abdeen, Predicting the Impact of Vegetations
in Open Channels with Different Distributaries Operations on Water Surface Profile Using Artificial Neural
Networks, Journal of Mechanical Science and Technology, KSME International Journal, Vol. 22, No. 9, 2008,
pp. 1830-1842.
[16] Y. Shin, NeuralystTM Users GuideNeural Network
Technology for Microsoft Excel, Cheshire Engineering
Corporation Publisher, Pasadena, 1994.
ENG
Engineering, 2010, 2, 573-579

Hydrogen Pick up in Zircaloy-4: Effects in the Dimensional

Stability of Structural Components under Nuclear Reactor
Operating Conditions
Pablo Vizcano, Cintia Paola Fagundez, Abraham David Banchik
Centro Atmico Ezeiza, Comisin Nacional de Energa Atmica, Presbtero J. Gonzlez y Aragn N 15,
Buenos Aires, Argentina
E-mail: vizcaino@cae.cnea.gov.ar
Received December 3, 2009; revised February 11, 2010; accepted February 15, 2010
Abstract
In the present work, the expansion coefficient due to hydrogen incorporation was measured for the axial direction of a Zircaloy-4 cooling channel, similar to that installed in the Atucha I PHWR, Argentina, trying to
simulate the nuclear power reactor operating conditions. As a first step, the solubility curve of hydrogen in
Zircloy-4 was determined by two techniques: differential scanning calorimetry and differential dilatometry.
The comparison with classical literature curves showed a good agreement with them, although the calorimetric technique proved to be more accurate for these determinations. Dilatometry was able to detect the end of
hydride dissolution from concentrations around 60 wppm-H up to 650 wppm-H, where the eutectoid reaction:
takes place (at 550oC). We assume that this ability is a good indicator of the aptitude of the
technique to measure dimensional changes in the given hydrogen concentration range. Then, the expansion
of Zircaloy-4 homogeneously hydrided samples was measured at 300oC, the typical operating temperature of
a nuclear power reactor, obtaining a relative expansion of 2.21 10-4% per wppm-H. Considering the relative expansion observed for Zircaloy-4 at room temperature due to hydriding, starting from a hydrogen free
sample, the total relative expansion rate is calculated to be 5.21 10-4% per wppm-H.
Keywords: Thermal Analysis, Dimensional Change, Hydrides, Zircaloy-4
1. Introduction
Most of the core structural components of the nuclear
power reactors are made of Zicaloy-4, a reference zirconium alloy in many structural nuclear applications. During reactor operation, the initial dimensions of the Zrbase components could increase due to three different
degradation processes: hydrogen pick up, irradiation growth and creep.
The hydrogen incorporated into the matrix is a fraction
of the total amount of hydrogen produced during the
corrosion reaction between the zirconium and the coolant,
according to the reaction:
Zr + 2H2O ZrO2 + 4H
The crystalline defects produced by the fast neutron
irradiation induce changes in the initial dimensions of the
components depending on the fabrication texture. On the
other hand, the creep contribution to these processes depends on the magnitude of the external stress applied to
the component.
The pick up of hydrogen atoms by the metal induces
an expansion of its initial length. This expansion continues after crossing the solubility limit at the reactor operating temperature, since the hydrogen in excess to that
limit precipitates as ZrH1,5+x after some supersaturation in
solid solution. Due to the higher specific volume of the
zirconium hydride with respect to the zirconium matrix,
the onset of precipitation induces an additional dimensional change. This change in length depends on both,
the orientation at which the hydrides precipitated in the
matrix and the crystalline texture of the component.
The material under study in the present work is Zircaloy-4 taken from cooling channels similar to those installed in the Atucha I PHWR. These tubes have a fully
recrystallized microstructure, which induces hydride
precipitation at the grain boundaries. In addition, these
components show a strong texture in a quasi-radial direction: the c axis of the -Zr hexagonal cell is oriented in a
cone surrounding the radial direction of the tube [1]. The
ENG
P. VIZCANO
574
aim of the present work is to determine the expansion

coefficient of Zircaloy-4 for the axial direction of the
channel at the reactor operating temperature (300C) [2].
2. Experimental Procedure
2.1. Material and Sample Preparation
The Zircaloy-4 samples were taken from an off cut of a
cooling channel similar to those installed in the Atucha I
reactor. Rectangular samples of dimensions 10 5 1.7
mm were cut from the tube, with the length of the sample
parallel to the axial direction of the tube, as it is shown in
Figure 1.
The tubes are cold-shaped and welded (by the tungsten
inert gas method) from fully recrystallized Zircaloy-4
sheets. The typical Kearns texture factors were measured
in a previous work for the [0002] pole (c axis of the
hexagonal cell). The range of values was: Faxial = 0.050.07, Ftangential = 0.22-0.26, Fradial = 0.67-0.73. Thus, about
6% of the c poles are aligned in the axial direction, 24%
in the tangential and 70% in the radial direction [1]. The
microstructure was fully recrystallized with a grain size
of 15-20 m. It can be observed in Figure 2.
10 mm
5 mm
Figure 1. Orientation and dimensions of the dilatometric
samples.
ET AL.
2.2. Hydriding
The hydrogen was incorporated by the cathodic charge
technique. The process was carried out in an electrolytic
cell at 80 2C. A diluted aqueous solution of sulfuric
acid was used as electrolyte, circulating a current density
of 5 mA/cm2 through the sample from periods of 18 to
96 h. As a result, hydride layers of different thickness
(from a few microns up to 50 microns) formed in the
samples.
The hydrogen was diffused into the bulk during the
dilatometric experiments. After the experiments, the
samples were polished to eliminate the oxide and any
remaining hydride layer on the surfaces. Finally, the hydrogen content was measured using a LECO RH-404
hydrogen meter. The error of these determinations is of
2%.
The hydrogen range of the samples hydrided in this
way varied from 50 to 650 wppm-H.
2.3. Differential Dilatometry

A Shimadzu TMA-60H vertical push rod differential
dilatometer, DD, was used to measure the difference in
expansion between a reference sample and a similar hydrided sample.
The experiments were carried out under inert gas atmosphere (high purity N2, 99.998%). As reference, an
uncharged Zircaloy-4 sample was used, containing about
20 wppm-H, which is incorporated during the fabrication
process of the channels. The minimum detection capacity
of DD is 0.25 m.
During the test, a constant load of 0.1 N was applied to
both samples. All the samples were subject to a nominal
thermal cycle made of a heating step at a rate of 5C/min.
After keeping the samples 30 minutes at the maximum
temperature they were cooled down at 5C/min. To avoid
the effect of the transformation, the maximum temperature was a few degrees below 550C.
2.4. Differential Scanning Calorimetry
20 m
Figure 2. Fully recrystallized microstructure of a Zircaloy-4

cooling channel. The typical size of the equiaxed grains is 20
6 m.
The calorimetric experiments were made using a thermal

flux differential scanning calorimeter Shimadzu, model
DSC-60. The dimensions of the samples were 4 4 1.7
mm, which were cut from the dilatometric samples after
the dilatometric experiments finished.
Two runs were performed for each sample at 5C/min,
in order to compare the results with the dilatometric data,
but the first one was discarded and TTSSd were determined in the second run. Figure 3 shows the calorimetric
heating curve of a hydrided sample where the points
usually associated with hydride dissolution are indicated:
the peak of the curve (pT, peak temperature), the maximum at the derivative of the DSC curve (msT, maximum
ENG
P. VIZCANO
0.20
Extrapolation lines
0.05
DSC curve
Derived curve
-4.0
0.10
-3.0
0.15
-2.0
d q/dt (mJ/sec )
msT
-1.0
dq/dt (mJ/sec)
0.25
0.00
pT
-5.0
25 C
100
200
300
400
o
500
-0.05
Temperature ( C)
Figure 3. Calorimetric curve of a sample containing 480

wppm-H, in the heating stage.
slope temperature) and the point where the baseline is

recovered or completion point (cT, completion temperature).
3. Results
3.1. Diffusion in the Bulk and TTSSd
Determinations
A typical diffusive dilatometric run is shown in Figure 4.
The differential apparatus needs a hydrogen-free reference sample (in fact it contains 20 wppm-H). Since the
reference is identical to the sample before hydrogen
charging, the expansion of the reference compensates
and cancels the thermal expansion of the -Zr phase in
the hydrided sample. Thus, the expansion measured with
the differential dilatometer only depends on the hydrogen
concentration of the hydrided sample. During the heating
stage, the hydride layer at the sample surface dissolves
and the hydrogen atoms diffuse into the bulk, raising the
concentration in solid solution. This process increases
continuously the dimensions of the sample as it is obser-
ved in Figure 4. Given enough time at the plateau temperature (550C), the hydride layer ends its dissolution
and hydrogen distributes homogeneously into the sample.
Depending on the thickness of the layer, it will dissolve
during the run or during the time at the plateau temperature. Yet, it is possible that a fraction remains undissolved. This will occur if at the plateau temperature the
solubility limit is reached without a complete dissolution
of the hydride layer. During the cooling stage, the sample
reduces its length but the differential expansion does not
return to zero: at room temperature, a difference in length between the sample and the reference still subsists
since the hydrogen diffused into the bulk is now precipitated as hydrides.
From the description given above, it is inferred that at
the first dilatometric run the hydrogen distribution is
controlled by the diffusion process. During this transient,
TTSSd cannot be determined. Thus, after an additional
mechanical polishing to eliminate any possible remaining hydride layer at the surface, TTSSd was measured in
the second run.
Figure 5 shows a dilatometric curve obtained after the
diffusive cycle, in the second run. During heating, the
sample increases its length again but when the dissolution finishes, the slope of the curve changes; at this point
TTSSd is determined. In Figure 5 this change in the
slope or knee is observed, for a sample containing 255
wppm-H, at 403C. This point is identified as the knee
temperature, keT. Another possible criterion, which is
not used in the present work, is to determine TTSSd at
the dilatometric derived curve: the change in the slope at
the knee produces a discontinuity, a step in the derived
curve, as it is shown in Figure 5 too. It is not an ideal
step; the discontinuity extends in a temperature range
of 40C to 50C. At the middle height of the step, the
step temperature, sT, can be determined. The step criterion proves to be more accurate than the knee criterion.
3.0
600
Expansion
14
Temperature
300
6
200
4
Lengh increase due to the
hydrogen incorporation
100
0
5000
10000
15000
Time (sec)
20000
2
0
25000
Figure 4. Dilatometric thermal cycle to diffuse the hydrogen

from the layers into the bulk.
0.0020
403 C
keT
0.0015
2.0
L(m)
10
400
Expansion curve
Derived curve
2.5
12
L (m)
Temperature(o(
C))
500
575
dL/dt (m/sec)
cT
0.0
ET AL.
1.5
0.0010
401 C
1.0
sT
0.0005
0.5
o
0.0
50 C
200
300
400
o
500
0.0000
Temperature ( C)
Figure 5. Dilatometric curve of the dissolution process and

derivative. The arrows indicate the change in the slope,
where the dissolution ends (keT) and the middle of the step
in the derived curve (sT). The sample contains 255 wppm-H.
ENG
P. VIZCANO
576
for other zirconium alloys [3], but no difference was observed between them for Zircaloy-4. Thus, TTSSd was
measured at the knee point (keT). In any case, it was observed that within an uncertainty interval of 2-4C, both
temperatures are virtually identical, Figure 5.
The dilatometric TTSSd data are plotted in Figure 6
as TTSSd vs the hydrogen concentration, [H]. The solubility equation obtained from these data is:
CkeT = 2.86 105 exp (-4730/keT)
(1)
On the other hand, with the DSC technique TTSSd

was determined following two criteria commonly used in
the literature: the peak, pT, and the completion temperature, cT, as it is shown in Figure 7. The fitting curves are
also included.
The solubility equations are:
CpT = 1.85 105 exp (-4362/pT)
(2)
(3)
CcT = 1.78 10 exp (-4546/cT)

500
Temperature(o(C
) )
450
400
350
300
250
200
150
100
Dissolution data
0
100
200
300
[H] (wppm)
400
500
600
Figure 6. Dissolution dilatometric data and fitting curve.
Temperature ((o
C))
500
400
300
200
100
pT data
cT data
0
100
200
300
400
500
600
[H] (wppm)
Figure 7. Dissolution calorimetric data measured at pT and
cT and fitting curves.
ET AL.
4. Discussion
4.1. Terminal Solid Solubility
The uncertainty of TTSSd determinations with the dilatometric technique can be estimated from Figure 5.
Where the expansion curve changes its slope (end of
dissolution), its derivative shows a step. This step extends over a temperature interval of about 40 to 50C, an
interval larger than the 25C between pT and cT in the
calorimetric curve (Figure 3). The knee temperature
virtually agrees with the temperature at the middle height
of the step in Figure 5. However, there is a higher intrinsic uncertainty in the dilatometric measurements with
respect to the calorimetric ones which affects the accurateness of TTSSd determinations. This uncertainty increases in the low hydrogen range, where the signal of
hydride dissolution is weak. It becomes evident in Figure 6, from 60 to 130 wppm-H, where the scatter of the
data is large. For these data, TTSSd error varies from
18C to 15C at the upper extreme of the interval. For
higher concentrations (in our case, concentrations higher
than 187 wppm-H) the error decreases to 10C, becoming constant for concentrations higher than 250 wppm-H,
where an error of 8C can be assumed.
Concerning DSC determinations, as it can be inferred
from the criteria commented in 2.4, there are some discrepancies regarding the exact point where TTSSd
should be located in the DSC curve [3-7]. As a brief
summary we can say that: Z. L. Pan, measuring Youngs
modulus as functions of temperature and hold time during quasistatic thermal cycles to Zr-2.5Nb hydrided samples, concluded that the most reliable point to associate
TTSS is msT [5]. D. Khatamian found the best correlation for pT contrasting TTSSd determinations at pT, msT
and cT for unalloyed zirconium and Zr-20wt%Nb hydrided/deuterided samples with neutron diffraction
measurements [6]. Recently, the authors of the present
work determined TTSSd for Zr-2.5Nb with pressure tube
microstructure by DSC using DD as a contrasting technique. In this work, the difficulty of determining the best
point to measure TTSSd on the dissolution curves has
been discussed thoroughly. Yet, since the accurateness of
the DSC data is higher than the DD, it was not possible
to obtain conclusions about the best point for TTSSd
determination on the DSC curve from this comparison
[8].
In any case, it is evident that the selection of one of the
three criteria based on the measurements made with a
contrasting technique does not provide physical meaning
to the choice, turning it into the true dissolution point.
In fact, the certainty of this choice will be strongly dependant on the capability of the contrasting technique to
detect the disappearance of very small hydrides at the
final stage of the dissolution process. In the present cir-
ENG
P. VIZCANO
4.2. Axial Elongation of a Cooling Channel

In order to simulate and determine the total elongation of
the cooling channels due to the hydrogen pick up in service, a similar but faster process must be developed in
the laboratory. Hydrogen should be incorporated into the
bulk, starting from a hydrogen-free material. Instead of
the slow hydrogen incorporation due to the corrosion in
service, the hydrogen in the hydride layer diffuses into
the bulk during the heating ramp and the subsequent isothermal annealing at 550C. At this stage, the hydrogen
in the bulk remains in solid solution and the sample
reaches its maximum length. During cooling, the hydrogen precipitates as hydrides reducing the sample length,
but as it was shown in Figure 4, the final length is larger
than the initial. After cooling, the final dimension of the
sample is measured in situ in the dilatometer, obtaining a
differential value. The length increase due to the hydrogen incorporation into the bulk was measured at room
temperature and reported in a recent paper [1]. A linear
dependence on the hydrogen concentration was found for
fully recrystallized Zircaloy-4. For this type of microstructure, hydrides precipitate on the grain boundaries,
but some tendency to precipitate in the rolling direction
577
500
DSC (cT)
400
Temperature ( C)
cumstances we judged that it would be most advisable to

choose the criteria that better agree with the highly referenced curve of Kearns [9] and the equilibrium solvus
line by Zuzek et al. [10], as done by other authors [3,4].
This comparison is shown in Figure 8. Regarding the
DSC data, the best agreement with Kearns and Zuzek
equilibrium curves corresponds to the completion point
(cT). Beyond the criterion chosen for TTSSd determination, the error at pT, msT or cT is always smaller than
2C and the reproducibility is excellent.
Although the present comparison has shown that DD
is less accurate than DSC for solubility determinations, it
must be alleged in its favor that the technique was capable to detect the hydride dissolution for samples with
concentrations from 60 wppm-H. This implies that the
sensitivity is good enough to detect dimensional changes
at very low concentrations. Since the main objective of
the present work is to detect dimensional changes for
hydrogen contents like 200 or 300 wppm-H, typical of
cooling channels that remained for about 10 years in the
reactor at full power operation [11], the performance of
the technique is suitable for these purposes. This matter
will be developed in the following section.
ET AL.
DSC (pT)
DD (knee)
300
200
Kearns
Zuzek
100
0
50
100
150
200
250
300
[H] (wppm)
350
400
Figure 8. Comparison between the curves obtained in the

present work and classical literature curves.
recalling the cold rolling process still subsists after the

recrystallization treatment. The competition of these two
ways of precipitation generates some scatter in the data.
However, a linear model seemed to be a good choice to
represent the expansion vs. [H]. The linear assumption
was made by simplicity, based on the values of the statistical parameters, considering an error of 0.5 m for the
micrometer, Table 1.
The following relation was obtained:
m
(1)
wppm
Dividing Equation (1) by the initial length of the samples from which Equation (1) was obtained (L0 = 18600
m), the relative expansion is:
L 5.6 m 0.054 [H ]
L
1
) room 3.2 10-4 3.00 10-6 [H ]
L0
wppm
(2)
where (L/Lo)room is the relative increase after the hydrogen diffusion into the bulk at room temperature.
As the cooling channels operate in the two-phase field,
in order to determine the total expansion in service, the
contribution of both, the fraction of H atoms in solid solution and that of the zirconium hydrides at the reactor
operating temperature (300C) should be added to the
growth due to the hydrides already present at room temperature. The measurements made at 300C are listed in
Table 2.
The relation found between the expansions and the
hydrogen concentration is linear too. Both, the data and
Table 1. Interception, slope, standard errors (SE) and lower and upper confidence limit (LCL and UCL). The standard deviation (SD) and R-value (R) are also given (97% of confidence).
Intercept
Slope
Statistics
Value
SE
LCL
UCL
Value
SE
LCL
UCL
SD
5.6
2.0
0.6
10.7
0.054
0.004
0.044
0.064
0.9
ENG
578
P. VIZCANO
Table 2. Relative expansion at 300C.
lo (m)
l/lo
0.85
9781
8.69 E-5
5.0
10332
4.81E-4
398
5.7
10035
5.64E-4
358
5.65
10035
5.63E-4
446
9.5
9784
9.75E-4
650
12.9
9968
1.29E-3
[H] (wppm)
l (m)
128
227
ET AL.
the regression line are shown in Figure 9 and the statistics parameters in Table 3. The linear equation is:
L
1
)300 C -1.54 10-4 2.2110-6 [H ]
(3)
L0
wppm
where (L/Lo)300C is the relative length increase of the
hydrided sample at 300C. Then, combining (2) and (3),
the total length increase is calculated as follows:
L
L
L
)TOTAL
) room
)300 C
(4)
L0
L0
L0
Table 3. Interception, slope, standard errors (SE) and lower and upper confidence limit (LCL and UCL). The standard deviation (SD) and R-value (R) are also given (97% of confidence).
Intercept ( 10-4)
Slope ( 10-6)
Statistics
Value
SE
LCL
UCL
Value
SE
LCL
UCL
SD
1.54
1.3
5.5
2.5
2.21
0.33
1.22
3.2
0.96
1.3 10-4
The relative expansion at room temperature and the

total relative expansion are both plotted in Figure 9 too.
Then, the total relative expansion coefficient along the
axial direction at 300C is 5.21 10-4% per wppm-H.
As it was shown in previous works, the hydrogen isotope concentration of the cooling channels measured at
different positions along its length after 10 years in service varies between 150 and 400 wppm-H [11]. If we
choose a medium concentration of 250 wppm-H for the
whole channel, it is possible to estimate the expansion of
the tube at the operating temperature for this concentration. Following Equation (4), the relative expansion will
be 0.0015 m/m of tube. Then, if we consider the full
length of the tube, 5,300 mm, the total expected expansion in the axial direction will be 8 mm, with an error
estimated in 15%.
0.0035
Rel. expansions at 300 C

Linear regressions
Error bands
0.0030
0.0025
TOTAL
L/Lo
L/Lo
0.0020
room
L/Lo
0.0015
0.0010
0.0005
300 C
L/Lo
0.0000
100
200
300
400
[H] (wppm)
500
600
Figure 9. Relative expansion at 300C, at room temperature [1] and the sum of both effects.
5. Conclusions
The present work was focused on two main objectives:
hydrogen solubility measurements and the determination
of the expansion coefficient of Zircaloy-4 for the axial

direction of a tube. As a brief summary, the following
points must be underlined:
The hydrogen solid solubility curve for Zircaloy-4 was determined by two techniques, differential
scanning calorimetry and differential dilatometry. The
comparison with classical literature curves showed a
good agreement with them. The solubility curves obtained with calorimetry, measuring TTSSd at the peak and
completion temperatures are:
CpT = 1.85 105 exp (4362/pT)
CcT = 1.78 105 exp (4546/cT)
and the one obtained by dilatometry, measuring TTSSd
at the knee temperature is:
CkeT = 2.86 105 exp (4730/keT)
Although the coincidence between them is good, the
calorimetric technique is more accurate for these measurements.
Dilatometry showed good sensitivity to detect the

end of dissolution from concentrations around 60 wppm-H
up to the eutectoid temperature (550C) concentration
(650 wppm-H), which is a good indicator of the aptitude
of the technique to measure dimensional changes in hydrided samples in this concentration interval.
The expansion of Zircaloy-4 homogeneously hydrided
samples was measured at 300C, the typical operating
temperature of a nuclear power reactor, obtaining a relative expansion of 2.21 10-4% per wppm-H. Adding to
this value the relative expansion coefficient at room temperature due to hydriding, the total relative expansion
rate is 5.21 10-4% per wppm-H.
6. References
[1]
J. C. Ovejero, A. D. Banchik and P. Vizcano, Axial/

Tangential Expansion Coefficients of Zircaloy-4 Channels Due to the Hydrogen Pick up, Advanced in TechENG
P. VIZCANO
[2]
[3]
[4]
[5]
[6]
nology of Materials and Materials Processing Journal,

Vol. 10, No. 1, 2008, pp. 1-8.
C. P. Fagundez, P. Vizcano, D. Bianchi and A. D. Banchik, Dilatometra del Sistema Zr-H, Proceedings of
the Congress SAM/CONAMET 2005, Mar del Plata,
18-21 October 2005.
J. P. Giroldi, P. Vizcano, A. V. Flores and A. D. Banchik,
Hydrogen Terminal Solid Solubility Determinations in
Zr-2.5 Nb Pressure Tube Microstructure in an Extended
Concentration Range, Journal of Alloys and Compounds,
Vol. 474, No. 1-2, 2009, pp. 140-146.
D. Khatamian and V. C. Ling, Hydrogen Solubility Limits In - and -Zirconium, Journal of Alloys and Compounds, Vol. 253-254, No. 20, 1997, pp. 162-166.
A. McMinn, E. C. Darby and J. S. Schofield, The Terminal Solid Solubility of Hydrogen in Zirconium Alloys, Proceedings of the 12th International Symposium
of the Zirconium in the Nuclear Industry, Toronto, 2000,
pp. 173-195.
Z. L. Pan and M. P. Puls, Precipitation and Dissolution
Peaks of Hydride in Zr-2.5 Nb during Quasistatic Ther-
ET AL.
579
mal Cycles, Journal of Alloys and Compounds, Vol. 310,

No. 1-2, 2000, pp. 214-218.
[7]
D. Khatamian and J. H. Root, Comparison of TSSD

Results Obtained by Differential Scanning Calorimetry
and Neutron Diffraction, Journal of Nuclear Materials,
Vol. 372, No. 1, 2008, pp. 106-113.
[8]
P. Vizcano, A. D. Banchik and J. P. Abriata, Calorimetric Determination of the -Hydride Dissolution Enthalpy
in Zircaloy-4, Metallurgical and Materials Transactions
A, Vol. 35A, No. 8, 2004, pp. 2343-2349.
[9]
J. Kearns, Terminal Solubility and Partitioning of Hydrogen in the Alpha Phase of Zirconium, Journal of Nuclear Materials, Vol. 22, No. 3, 1967, pp. 292-303.
[10] E. Zuzek, J. P. Abriata and A. San Martn, H-Zr (Hydrogen-Zirconium), Bulletin of Alloy Phase Diagrams,
Vol. 11, No. 4, 1990, pp. 385-395.
[11] P. Vizcano, A. D. Banchik and J. P. Abriata, Solubility
of Hydrogen in Zircaloy-4: Irradiation Induced Increase
and Thermal Recovery, Journal of Nuclear Materials,
Vol. 304, No. 2-3, 2002, pp. 96-106.
ENG
Engineering, 2010, 2, 580-584

Electrochemical Generation of Zn-Chitosan Composite

Coating on Mild Steel and its Corrosion Studies
Kanagalasara Vathsala, Thimmappa Venkatarangaiah Venkatesha, Beekanahalli Mokshanatha
Praveen, Kudlur Onkarappa Nayana
Department of Studies in Chemistry, School of Chemical Sciences, Kuvempu University, Shankaraghatta, India
E-mail: drtvvenkatesha@yahoo.co.uk
Received December 16, 2009; revised February 26, 2010; accepted March 6, 2010
Abstract
A Zinc-Chitosan composite coating was generated on mild steel from zinc sulphate-sodium chloride electrolyte by electrodeposition. The electrolyte constituents were optimized for good composite coating. The corrosion resistance behavior of the composite was examined by weight loss, polarization and impedance
methods using 3.5 wt% NaCl neutral solution as medium. Separate polarization profiles were recorded for
composite coating and compared with zinc coated sample. SEM images of coatings were recorded for the
pure and composite coating.
Keywords: Composite Coating, Chitosan, SEM, Impedance, Electrodeposition
1. Introduction
Zinc electroplating is an industrial process and is widely
used to coat on steel for enhancing its service life. As
zinc is electrochemically more active than steel and
hence it sacrificially protect the steel from corrosion.
However zinc itself undergoes corrosion leading to the
formation of zinc compounds called white rust on its
surface. This tendency of formation of white rust reduces
the life of the coating from the expected period. Therefore to enhance the life span of the zinc coating and to
avoid the white rust formation the alternative methods
like surface modification is adopted. The earlier modification methods are associated with chromate based formulations and the procedure is very simple to generate
passive chromate films on corroding zinc coatings. The
use of chromate passivation is prohibited because of
pollution hazards. An alternate to this chromation is to
generate surface films or surface barriers with specific
organic molecules or with certain addition agents [1-6].
Also the service life of zinc coating is enhanced by including the inert materials in its coating. The inclusion is
done by codeposition of these materials with zinc and
thus generating composite coating. These zinc composite
coatings exhibit better corrosion resistance property.
Nowadays the nanosized materials are codeposited to get
better zinc composite with better corrosion resistance
[7-10].
A survey of literature reveals that the conducting
polymers were used for anticorrosive coatings and as

inhibitor for steel [11-13]. However limited information
connected to zinc - polymer composite coatings on steel
is available in the literature and especially with zinc biopolymer composites.
The chitosan is one such biopolymer used in corrosion
inhibition of mild steel without causing environmental
problems. Chitosan possess good biocompatibility, chemical resistance, mechanical strength, antimicrobial properties and thermal stability and have been utilized successfully in biotechnology, for different applications.
The hydroxyl apatite chitosan nanocomposite was obtained on stainless steel to provide better corrosion protection [14,15]. Chitosan is widely used in industry due
to its film forming and gelation characteristics. In dilute
solutions it is a linear polycation with high charge density. This electrochemical property was utilized in the
present work to get the zinc chitosan composite film on
mild steel from electrolysis and its corrosion resistance
property was tested.
2. Experimental
2.1. Plating Process
Zinc and Zn-chitosan coatings were electrically deposited from sulphate-chloride bath. The constituents of the
bath were 250 g/L ZnSO47H2O, 40 g/L NaCl, 30 g/L
H3BO3 and 0 g/L chitosan (88% deacetylated). In all the
ENG
K. VATHSALA
experiments distilled water and analytical grade reagents

were used. The pH of the bath solution was adjusted to
2.5-3 by adding dil.H2SO4 and NaHCO3. The bath was
stirred for few hours before subjecting it into plating experiments. The cathode was mild steel and anode was
zinc (99.99%). The mild steel surface was polished mechanically, and degreased with trichloroethylene in degreased plant followed by water wash. Before each experiment the zinc surface was activated by dipping in
10% HCl for few seconds and was washed with water.
Equal area of anode and cathode was selected for electrode position process. The bath temperature was at 300
K. The deposition process was carried at 4 A/dm2 and
under mechanical stirring.
2.2. Weight Loss Measurements

The coating thickness prepared for corrosion tests was in
the range of 1015 m. The corrosion rate by weight loss
measurements were performed for mild steel samples
coated with pure zinc and Zn-chitosan composite. The
electrolyte was 3.5 wt% NaCl solution and the test samples were immersed vertically in the solution which was
maintained at room temperature. The difference in weight was measured once in every 24 hours for a period of
15 days. In each weight loss measurement the corroded
samples were rinsed in alcohol, dried with hot air, and
then the weight was noted. The weight loss evaluated
was used for estimating the corrosion rate.
2.3. Salt Spray Test

The salt spray test as per (ASTM B 117) was carried out
in a closed chamber. The deposited plates were freely
hanged inside the chamber and subjected to continuous
spray of neutral 5 wt% NaCl vapors. The specimens
were observed periodically and the duration of the time
for the formation of the white rust was noted.
ET AL.
a frequency response analyzer.

The surface morphology of the coatings was examined
using a JEOL-JEM-1200-EX II scanning electron microscope
3. Results and Discussion

3.1. Corrosion Rate Result
The zinc and composite coatings was generated on separate mild steel plates having the thickness of about 1015 m. The steel panels were immersed completely in
3.5 wt% NaCl solution for different time intervals and
the weight loss values were used to calculate the corrosion rate. Figure 1 represents the corrosion rate (wt loss/
hour) profiles with respect to number of hours. The corrosion rates of both composite and zinc coatings were
very high in the beginning and decrease exponentially in
the middle and it becomes constant after 200 and 150 hrs
for zinc and composite coatings respectively. At any
given time the rate of corrosion for composite was always less than that of zinc coating. This suggests that the
composite coating possess higher corrosion resistance
property. This property was due to the presence of chitosan in the zinc matrix.
3.2. Salt Spray Test Result

The industrial method of testing the corrosion behavior
of zinc-plated objects is salt spray test. The test was conducted by spraying 5 wt% NaCl solution in a chamber.
The NaCl drops accumulated on the surface of the coated
specimens facilitate the corrosion resulting in zinc salts
called white rust. The time taken for the formation of
white rust was the indication of the corrosion rate. The
higher corrosion resistance delays the production of
white rust. In the present case the pure zinc produced the
white rust after 19 hrs and the Zn-chitosan composite
2.4. Electrochemical Measurements
Corrosive velocity (10-5kg/m2.h)
A conventional 3-electrode cell was used for polarization

studies. The zinc coated or Zn-chitosan composite coated
specimen with surface area of 1 cm2 was used as working
electrode. Saturated calomel and platinum foil were employed as reference and counter electrodes respectively.
The electrolyte was 3.5 wt% NaCl solution. The corrosion resistance property of these specimens was evaluated from the anodic polarization curves.
The electrochemical impedance measurements were
performed using AUTOLAB from Eco-chemie made in
Netherlands. The steel specimens and their dimensions
were same as that of polarization experiment. The EIS
was recorded in the frequency range from 100 kHz to 10
MHz with 5 mV AC amplitude sine wave generated by
581
zinc coating
composite coating
8
7
6
5
4
3
2
0
50
100
150
200
250
300
350
400
Time (h)
Figure 1. Variation of the corrosion rate with immersion

time for zinc and composite coated samples in 3.5 wt.%
NaCl solution.
ENG
K. VATHSALA
582
produced the white rust after 28 hrs. This test confirms

the enhancement of corrosion resistance of zinc in the
presence of chitosan in its matrix.
ET AL.
zinc coating
com posite coating
16
14
12
3.3. Electrochemical Result

Figure 2 shows anodic polarization profile of zinc and
Zn-chitosan coated sample in 3.5 wt% NaCl solution.
The linear variation was observed in the beginning up to
1.01 V and afterwards there was gradual increase in
current indicating electrochemical oxidation of zinc.
However in the case of composite coating, the potential
was always more positive for any given current density.
This indicates that the composite requires extra potential
to bring anodic reaction. Thus the composite possess
higher resistance to corrosion process on its surface.
The Nyquist plots for zinc and Zn-chitosan coatings
are shown in Figure 3. The larger loop was produced by
Zn-chitosan coatings whereas smaller semicircle was
obtained for pure zinc. It can be easily observed from the
figure that Rp values are higher for composite coating
than zinc coating. This indicates that composite coating
is more corrosion resistant than zinc coating.
3.4. Surface Morphology

The SEM images at lower and higher magnification were
represented in Figure 4. Also the SEM images of corroded surface of zinc and composite are given in Figure
5 and Figure 6. The SEM images show the practical
evidences on the corrosion protection ability of composite coating.
-Z''(ohm)
10
8
6
4
2
0
-2
5
10
15
20
25
30
35
40
45
Z '(o h m )
Figure 3. Impedance diagrams for pure zinc coated and

Zn-chitosan coated samples in 3.5 wt.% NaCl solution.
Figure 4. SEM images for the two samples. (a) Zinc coating,
(b) composite coating.
4. Discussion.
The experimental results of the present investigations
inferred that the chitosan can be included in the deposit
easily. It acquires a positive charge by protonation in
Current density(cm-2)
500
Figure 5. SEM images for two samples after anodic polarization for (a) zinc coated; (b) composite coated sample.
zinc coating
composite coating
400
300
1m
200
X 4000 1m
X 4000
Figure 6. SEM images for two samples after 15 days weight

loss measurements. (a) zinc coated; (b) composite coated.
100
Acid solution [15].

-1.04
-1.03
-1.02
-1.01
-1.00
-0.99
-0.98
Potential (V)
Figure 2. Anode polarization curves for zinc and composite

coated samples in 3.5 wt.% NaCl solution.
chit NH 2 H 3O chit NH 3 H 2 O
During electrodeposition naturally H2 evolution takes

place and there is increase in OH ions at the close
ENG
K. VATHSALA
vicinity of the cathode. These OH ions combine with

proton of the protonated chitosan at the electrode and get
precipitated. This solid precipitate codeposited with the
zinc. Generally the composite coating of zinc with other
material is generated by dispersing the insoluble particles
in the electrolyte. Here the solid is dispersed in liquid
state. Even the same procedure is followed in polymermetal composite. However in chitosan-metal composite,
the chitosan was codeposited from electrolyte, where in
chitosan and electrolyte were in single phase. Chitosan
exists as polycation in acid solution and reaches the
cathode easily during electrodeposition and it will get
deposited on the cathodic site.
The results of corrosion rates for zinc and composite
coatings from chemical and electrochemical methods are
in agreement with each other. The composite coating
with chitosan provides higher corrosion resistance than
zinc coating. As these molecules possess higher molecular weight and larger molecular size, they cover the corroding surface to larger extent through its cationic point
attached to cathodic site of the surface. Thus there may
be formation of barrier which prevents the direct contact
of corroding metal with the corrosive medium. There are
reports in the literature on the corrosion inhibition of
polymer molecule to metals [16-17]. The weight loss method, impedance, salt spray test results of the present
study revealed higher corrosion resistance property of
composite coating. In all these methods, probably chitosan hinders the anodic reaction and finally the corrosion
rate was decreased. The delayed white rust formation in
salt spray inferred that the inclusion of the chitosan
makes the composite coating to acquire more corrosion
resistant property. Also the higher RP value and more
positive potential of composite coating make the deposit
nobler than zinc coating. The corrosion rate and time
profiles indicate that the corrosion velocity (Figure 1) of
composite was always less than zinc coating.
Figure 4 shows the SEM image of the zinc and Znchitosan composite coatings. Composite coated samples
have a ridge shaped grains on the surface which reveals
the inclusion of chitosan into zinc matrix.
The anodic polarization of zinc and composite (Figure
5(a) and 5(b) showed that zinc coating undergoes more
dissolution than composite. The crystals get dissolved
easily during corrosion (Figure 5(a)). This had not been
observed in composite coating (Figure 5(b)). The SEM
images of samples after 15 days of chemical corrosion
showed larger deep pits arising out of higher corrosion
rate for zinc coating (Figure 6(a)). The composite coating (Figure 6(b)) exhibited small pits which distributed
throughout the surface and resulted uniform corrosion
with lower rate. These experimental results revealed
higher corrosion resistance property of Zn-chitosan composite coating compared to pure zinc coating.
ET AL.
583
5. Conclusions
Zn-chitosan composite was generated by electrodeposition from sulphate bath. The precipitated chitosan was
codeposited along with zinc. The performance of composite coating was established from the results of weight
loss, polarization, impedance and salt spray test. In all
these studies Zn-chitosan composite exhibits better anti
corrosion performance. The SEM images of surface provide an evidence for the presence of chitosan in coating
and crystalline nature. The composite showed uniform
and lower corrosion rates than that of zinc coating.
6. Acknowledgements
The authors are grateful to University Grant Commission,
New Delhi, Govt. of India [Major Research Project F.32220/2006(SR)] for providing financial assistance.
7. References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
K. L. Lin, C. F. Yang and J. T. Lee, Growth Behavior

and Corrosion Resistance of 5% Al-Zn Coating, Corrosion, Vol. 49, No. 9, 1993, pp. 9-12.
G. Barcelo, M. Sarret, C. Mller and J. Pregonas. Corrosion Resistance and Mechanical Properties of Zinc Electrocoatings, Electrochimical Acta, Vol. 43, No. 1-2,
1998, pp. 13-20.
A. Y. Hosny, M. E. El-Rafei, T. A. Ramadan, B. A.
El-Gafari and S. M. Morsy, Corrosion Resistance of
Zinc Coatings Produced from a Sulfate Bath, Metal Finishing, Vol. 93, No. 11, 1995, pp. 55-59.
B. Bozzini, V. Accardi, P. L. Cavallotti and F. Pavan,
Electrodeposition and Plastic Behavior of Low-Manganese Zinc-Manganese Alloy Coatings for Automotive Applications, Metal Finishing, Vol. 97, No. 5, 1999, p. 33.
C. Mller, M. Sarret, E. Garcia and J. A. Ortega, CrFree Passivation on ZnNi Alloys, Journal of the Electrochemical Society, Vol. 151, No. 2, 2004, pp. C149C154.
S. Tamil Selvi, V. Raman and N. Rajendran, Corrosion
Inhibition of Mild Steel by Benzotriazole Derivatives in
Acidic Medium, Journal of Applied Electrochemistry,
Vol. 33, No. 12, 2003, pp. 1175-1182.
B. M. Praveen, T. V. Venkatesha, Y. A. Naik and K.
Prashantha, Corrosion Studies of Carbon Nanotubes-Zn
Composite Coating, Surface and Coating Technology,
Vol. 201, No. 12, 2007, pp. 5836-5842.
B. M. Praveen, T. V. Venkatesha and Y. A. Naik,
Corrosion Behaviour of Zn-TiO2 Composite Coating,
Synthesis and Reactivity in Inorganic, Metal-Organic and
Nano-Metal Chemistry, Vol. 37, No. 6, 2007, pp. 461465.
J. Li, J. Jiang, H. He and Y. Sun, Synthesis, Microstructure, and Mechanical Properties of TiO2/Ni Nanocomposite Coatings, Journal of Materials Science Letters, Vol. 21, No. 12, 2002, pp. 939-941.
ENG
584
K. VATHSALA
ET AL.
[10] M. Musiani, Electrodeposition of Composites: An Expanding Subject in Electrochemical Materials Science,

Electrochimica Acta, Vol. 45, No. 20, 2000, pp. 33973402.
[14] F. B. Waanders, S. W. Vorster and A. J, Geldenhuys,

Biopolymer Corrosion Inhibition of Mild Steel: Electrochemical/Mssbauer Results, Hyperfine Interactions,
Vol. 139-140, No. 1-4, 2002, pp. 133-139.
[11] A. A. Pud, G. S. Shapoval, P. Kamarchik, N. A. Ogurtsov,

V. F. Gromovaya, I. E. Myronyuk and Y. V. Kontsur.
Electrochemical Behavior of Mild Steel Coated by
Polyaniline Doped with Organic Sulfonic Acids, Synthetic Metals, Vol. 107, No. 2, 1999, pp. 111-115.
[15] X. Pang and I. Zhitomirsky, Electrophoretic Deposition

of Composite Hydroxyapatite-Chitosan Coatings, Materials Characterization, Vol. 58, No. 4, 2007, pp. 339-348.
[12] R. C. Patil and S. Radhakrishnan. Conducting Polymer

Based Hybrid Nano-Composites for Enhanced Corrosion
Protective Coatings, Progress in Organic Coatings, Vol.
57, No. 4, 2006, pp. 332-336.
[13] B. Szeptycka and A. Gajewska-MidziaIek, The Influence of the Structure of the Nano-Composite Ni-PTFE
Coatings on the Corrosion Properties, Review of Advanced Material Science, Vol. 14, 2007, pp. 135-140.
[16] H. P. Sachin, G. Achary, Y. Arthoba Naik and T. V.

Venkatesha, Polynitroaniline as Brightener for Zinc
Nickel Alloy Plating from Non-Cyanide Sulphate Bath,
Bulletin of Materials Science, Vol. 30, No. 1, 2007, pp.
57-63.
[17] S. Sathiyanarayanan, S. S. Azim and G. A. Vekatachari,
New Corrosion Protection Coating with PolyanilineTiO2 Composite for Steel, Electrochimica Acta, Vol. 52,
No. 5, 2007, pp. 2068-2074.
ENG
Engineering, 2010, 2, 585-593

Tunable Erbium-Doped Fiber Lasers Using Various

Inline Fiber Filters
1
Shien-Kuei Liaw1,4, Kuei-Chu Hsu2, Nan-Kuang Chen3
Department of Electronics Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan, China
2
Department of Optics and Photonics, National Central University, Jhungli, Taiwan, China
3
Department of Electro-Optical Engineering, National United University, Miaoli, Taiwan, China
4
Department of Mechanical Engineering, National Taiwan University, Taipei, Taiwan, China
E-mail: skliaw@mail.ntust.edu.tw
Received February 2, 2010; revised March 23, 2010; accepted March 27, 2010
Abstract
Several high-performance and tunable erbium-doped fiber lasers are reviewed. They are constructed by using
fiber Bragg gratings (FBGs) or short-wavelength-pass filters (SWPFs) as wavelength tunable components
inside the laser cavity. Broadband wavelength tuning range including C- and/or S-band was achieved, and
tunable laser output with high slope efficiency, high side-mode suppression ratio was obtained. These fiber
lasers can find vast applications in lightwave transmission, optical test instrument, fiber-optic gyros, spectroscopy, material processing, biophotonic imaging, and fiber sensor technologies.
Keywords: Fiber Bragg Grating (FBG), Short-Wavelength-Pass Filter (SWPF), Tunable Fiber Laser, Optical
Communication
1. Introduction
In recent years, fiber lasers have found a variety of
applications in the testing of fiber components, fiber
sensing and wavelength division multipling (WDM)
systems, in which they are used to act as a backup
source with ITU-T grids [1]. Also, fiber lasers are useful for spectroscopy, sensing protection, and fiber-optic
gyro [2]. Partially because of their features, such as
low wavelength sensitivity to temperature, low-intensity noise, and all-fiber construction, their advantages
over non-fiber-based laser sources are potentially lowintensity noise, high output power, and compatibility
with fiber components. Previous works have proposed
design and/or characteristics valuation of fiber lasers, including multiple-ring cavity fiber laser [3], two separate
erbium-doped fiber lasers [4], distributed feedback fiber
lasers [5], and Brillouin erbium-doped fiber laser pumped using fiber Bragg grating (FBG) [6]. These fiber lasers, however, have fixed wavelengths that are not suitable for wavelength routing, reconfigurable switching
and/or network protection. On the contrary, tunable erbium-doped fiber lasers could well fit such requirements.
Nowadays, a variety of tunable fiber lasers have been
demonstrated such as tunable single-frequency fiber lasers [7], coherent combining tunable lasers [8], tunable
fiber-ring laser using bending effect [9], and so on.

In this paper, we overview several works regarding
tunable fiber lasers done by our groups. The first kind is
the FBG-based linear-cavity tunable fiber laser using an
optical circulator (OC) [10], or a broadband fiber mirror
(BFM) [11] as rear cavity end while the front cavity end
is based on tunable FBGs (TFBGs) which could be tuned
by applying strain. FBGs have become an enabling
technology that provides convenient, cost-effective, and
reliable solutions to a multitude of design problems in
fiber module. Using a backward pump scheme at 1480
nm in [11], stable lasing output power of 21.14-mW
measured at 1544.8 nm was obtained with a threshold
pump power of 8.0 mW. A side-mode suppression ratio
as high as 57 dB, wavelength tuning range up to 30 nm
with a step resolution of 0.5 mm/turn, and power variation less than 1.0 dB were achieved. The second kind is
the short-wavelength-pass filter (SWPF)-based tunable
fiber ring laser with lasing wavelength down to S-band
using filters adopting a side-polishing [12] or fusedtapering [13] technique to attain the wavelength-dependent fundamental-mode cutoff concept. The fiber sidepolishing and fused-tapering techniques were both employed to fabricate wavelength tunable fiber filters. The
laser can be tuned close to the short-wavelength edge of
the available erbium gain bandwidth with tuning range of
ENG
586
S.-K. LIAW ET AL.
26 nm, signal-to-amplified-spontaneous-emission (ASE)

ratio of around 40 dB, and the full width half maximum
(FWHM) linewidth of about 0.5 nm. The single-longitude-mode (SLM) operation will be briefly discussed in
Subsection 5.3. These two kinds of tunable in-line filterbased tunable fiber lasers as mentioned may broaden
wavelength tuning range in either C- and/or S-band and
will be addressed in detail. Both of them have graceful
features of simple structure, compactness, ease in connection with fiber components, high-efficiency, and
continuous-tuning, which make them promising for vast
applications.
2. Wavelength Tunable Mechanisms

2.1. Tunable Fiber Bragg Grating
In principle, a wavelength shift in a FBG may be due to
the changes in temperature, strain, pressure and/or other
parameters. The shift in Bragg wavelength with strain
and temperature can be expressed as [14]
tance, therefore up to ten turns can be applied to tune the

FBG reflection wavelength.
The lasing wavelength as a function of turns of screw
is shown in Figure 1(a), with a linear slope of 4.82 nm/
mm translational distance. Thus, B0.00482 nm/m
d, where d is the displacement of the translational stage
in unit of micrometers. Another way is to embed the
FBG in the outer laminar. The composite with the TFBG
embedded within is attached to a 3-point tuning device
by using instant adhesive glue. By tuning the precision
screw of the 3-point bending device either by straining or
compressing, we can apply transverse displacement in
either direction to easily attain a tunable range of 10 nm.
This eliminates any use of complicated or bulky components to perform the tuning function.
Figure 1(b) shows the superimposed tuning spectra of
two homemade tunable FBGs. The demonstrated tuning
range of each FBG is approximately 15 nm with reflectance of 99.9%. Before tuning, FBG1 has a central reflection wavelength of 1540.5 nm while that of FBG2 is
1555
Experimental Date
Filting Curve
Wavelength
1553
1552
1551
1550
1549
-1.0
-0.5
0.0
0.5
1.0
Turns of Screw
(a)
Second grating
First grating
-50
1552.68nm
1540.5nm
-60
Power (dBm)
dn
( )
(1)
n2
d
T
2n {1 ( )[ P12 ( p11 p12 )]} [
]T
2
n
where is the applied strain, is the period of fiber, Pi,j

are the Pockels (piezo) coefficients of the stress-optic
tensor, and is the Poissons ratio. Note that n is the effective refractive index of the fiber core as defined in
Equation (1), is the thermal expansion coefficient of
the silica fiber with a typical value of 0.015 nm/C, and
is the temperature change in degree Celsius. The term
(n2/2) [P12(P11P12)] has a numerical value of 0.22.
The strain can be measured under a constant temperature
according to the following equation:
1 B
0.78 106 1
(2)
B
where B is the Bragg wavelength, and this value gives a
rule-of-thumb measurement of wavelength shift for a
FBG with strain of 1 nm per 1000 at 1.31 m. To design a strain tunable FBG, firstly, it is embedded in a
strip of composite thermal plastic material and then is
attached to L-shaped holders at both ends. The FBG is
then mounted on a precision translational stage with a
high-resolution micrometer. By strained or compressed
tuning of the precise screw of the micrometer, we can
apply both directions in the transverse displacement for
increasing the tuning range up to 8 nm. Two steel rods
are attached to the sides of the FBG composite strip to
confine the applied strain or stress to the longitudinal
direction only. The micrometer has a resolution of 0.5
mm/turn and a full range of 5.0 mm in translational dis-
1554
-70
-80
1540.5nm
-85.3dBm
1535.6nm
-85.23dBm
-90
1530
1547.7nm
-85.21dBm
1545.47nm
-85.73dBm
1540
1560.66nm
-83.93dBm
1557.6nm
1552.68nm -84.06dBm
-86.12dBm
1550
1560
Wavelength (nm)
(b)
Figure 1. (a) Wavelength tunable FBG versus turns of micrometer screw; (b) superposed transmission spectra of two
tunable FBGs (Total tuning range may cover the C band).
ENG
S.-K. LIAW ET AL.
587
1552.7 nm. Fine tune resolution as precise as 0.2 nm

FBG can be realized.
2.2. SWPF-Based Tunable Fiber Laser

The S-band tunable erbium-doped fiber lasers were achieved by connecting the active fiber to the thermooptic tunable SWPFs. The mechanism of the proposed
SWPFs [12,13] is to interact with the guiding optical
fields to cause fundamental mode loss at long-wave
length, and the cutoff wavelength can be tuned when the
heating temperature applying on the filter changes. The
dispersion engineering methods had been employed by
controlling the propagation losses of lights at different
wavelengths. Both the side-polishing and the fusedtapering techniques were adopted in our previous works
[12,13]. When the SWPFs are temperature-tuned to attenuate the wavelengths longer than 1530 nm, the C + L
band ASE is suppressed and the S-band gain is obtained.
The commonly used S-band erbium-doped optical fiber
amplifiers (EDFAs) employ erbium-doped fiber (EDF)
with depressed inner cladding to achieve fundamental-mode cutoff at the longer wavelengths [15,16]. The
cutoff wavelength and the mode field diameter can be
adjusted through bending and local heating, but the fabrication, insertion loss, crosstalk and cost of the filters
using dispersive fibers show great difficulties for practical use. Thus, an alternative way to obtain SWPFs is to
interact with the light through the evanescent field that is
spread out of the waveguide with wavelength-dependent
properties. When the optical fiber is side-polished or
tapered, the mode field is expanded out of the fiber cladding. Using dispersive liquids surrounding the side-polished fiber/taper fiber, the device can be a SWPF if the
dispersion relations are properly designed. The tunable
SWPF was achieved by tuning the temperature of the
dispersive liquids to change the dispersive curves for
obtaining different cutoff wavelengths [17,18].
In Figure 2(a), the blue and red curves are the dispersion curves of the Ge-doped core and the fused silica of
SMF-28, and the black curve is the dispersion for Cargille index-matching liquids. The refractive index dispersion of the core and cladding intersect at a fundamental mode cutoff wavelength which divides the wavelengths into bound and refracting leaky modes, and the
lights with wavelength longer than the cutoff wavelength
can be highly attenuated due to the frustrated total internal reflection. In our measurement, a broadband white
light source was launched into the fabricated SWPF,
where the optical liquid surrounding the SWPF was
heated by a thermoelectric cooler (TEC) to stabilize the
temperature. Figure 2(b) depicts the experimental and
simulated transmission spectra of a SWPF over the tuning temperature of 25C27C, where the simulation results were performed by beam propagation method. The
(a)
(b)
Figure 2. (a) Refractive index dispersion curves for index

matching liquid and materials for original fiber core (GeO2)
and cladding (pure silica); (b) experimental and simulated
spectral responses of tunable short-wavelength-pass fiber
filter at different temperatures [18].
experimental results agree well with the simulation ones.

The results include tuning efficiency of 50 nm/C, cutoff efficiency of 1.2 dB/nm, and rejection efficiency of
55 dB. Based on these experimental and simulated results, the mechanism of the SWPF is proved to be qualified for developing wavelength tunable S-band erbiumdoped fiber lasers.
3. FBG-Based Tunable Fiber Lasers:

Configurations and Experimental/
Simulation Results
3.1. Optical Circulator as Lasers Rear Cavity
End
The proposed configuration of linear cavity for the tunable laser is shown in Figure 3. The linear cavity consists of a 3-port OC, two TFBGs, a segment of highENG
588
S.-K. LIAW ET AL.

EDF=0.8m
EDF=1.9m
EDF=3m
EDF=5m
Signal power (mW)
60
50
40
30
20
10
Figure 3. Configuration of proposed OC-based linear-cavity

tunable fiber laser [10] (ISO: optical isolator; LD: laser
diode; OSA: optical spectrum analyzer).
0
0
20
40
60
80
100
120
140
160
Pump power (mW)
(a)
concentration EDF, a 1480/1550 nm WDM coupler, and

one 1480 nm pumping source. The 3-port OC here acts
as a wavelength router by connecting port 3 with port 1.
In this way, the residual pumping power travels back to
the EDF for twice amplification to increase its pumping
efficiency up to 2 dB difference in laser output power. A
piece of EDF is inserted into the cavity to act as pump
absorber. At the right hand side of this cavity, there is
one 1 2 optical switch (OSW) and two TFBGs (i.e.,
TFBG1, TFBG2) connected to the two switched ports of
OSW. The tuning range could cover the whole C-band
by switching between the two OSW ports connected to
individual TFBG. The original reflected wavelengths of
TFBGs are 1540.5 and 1552.68 nm, respectively. Two
variable optical attenuators (VOAs) are used in each port
for the power equalization.
Figure 4(a) shows signal power versus 1480 nm pump
power for various lengths of EDF. With the selected
lasing signal at 1550 nm, the signal power is linearly
proportional to the pumping power. The pumping efficiency increases from 3.7% to 40% as EDF length increases from 0.8 m to 5 m. Although the longer EDF
length seems to have better conversion efficiency and to
generate higher laser power, it also generates more ASE
noise which results in lower signal-to-noise-ratio (SNR).
As the EDF length increases, the extra gain provided by
the pumping power is smaller than the loss attributed by
the EDF. For this linear-cavity fiber laser, the parameters
for achieving optimum performance are 1.9 m in length
for EDF and 50% reflectance for the TFBGs.
Figure 4(b) shows the superimposed output spectra of
the fiber laser. It is randomly tuned across the C-band
with power equalization function. The power equalization can be realized by using VOAs independently. The
power uniformity within 0.1 dB over the tuning range
has been achieved. No polarization mode competition
effect is observed partially due to the narrow linewidth of
TFBGs. The switching time for OSW is 10 ms in this
case, which depends mainly on the specification of the
mechanical OSW.
(b)
Figure 4. (a) Laser output power versus pump power using

different EDF lengths; (b) superimposed output spectra of
tunable fiber laser as wavelength of TFBGs is tuned across
C-band after power equalization.
3.2. Broadband Fiber Mirror as Rear Cavity

End
The proposed BFM-based liner-cavity tunable fiber laser
in a backward pump scheme is shown schematically in
Figure 5. The laser cavity consists of a BFM, a tunable
FBG, and a piece of EDF. The BFM here acts as a
broadband wavelength reflector integrated with a TFBG
to form a laser cavity. It will lase as long as the reflected
wavelength of TFBG is within the reflective range of the
BFM. Also, a piece of EDF is inserted into the cavity to
act as pump absorber.
Figure 6(a) shows the measured experimental results
for the EDF length versus the output power at 1544.8 nm
when Figure 5 has a constant pump power of 150 mW.
We find that as the EDF length is increased from 0 to 4
m, the output lasing power is increased accordingly. If
the EDF is longer than required, there is a region of the
EDF where the pump power is relatively small, the signal has reached saturation intensity, and the gain is decreased according. The curves for various lengths of
EDF with threshold power of 8 mW are shown in Figure
ENG
S.-K. LIAW ET AL.
589
30
Transfer Efficiency (%)
25
20
15
10
5
Figure 5. Configuration of proposed BFM-based liner-cavity tunable fiber laser in backward pump scheme with use of
residual pump power [11] (PM: power meter).
EDF length = 1m
EDF length = 3m
EDF length = 4m
0
0
25
50
75
100
125
150
Pump Power (mW)
(a)
40
Experimental Curve
@ 1544.3 nm
Output Power (mW)
35
30
25
20
15
10
5
0
0
10
EDF length (m)
(a)
(b)
40
EDF length = 1m
EDF length = 3m
EDF length = 4m
Output Power (mW)
35
Figure 7. (a) Experimental results of transfer efficiency

versus pump power for different lengths of EDF; (b) superposed output spectra of BFM-based tunable fiber laser.
(Wavelength is tuned across C-band using two TFBGs).
30
25
20
15
10
5
0
0
20
40
60
80
100
120
140
Pump Power (mW)
(b)
Figure 6. (a) Experimental results show EDF length against

output power as pump power is set at 150 mW; (b) experimental results of pump power versus lasing output power
for various lengths of EDF in BFM-based fiber laser
scheme.
6(b). The transfer efficiency versus pump power for different lengths of EDF is shown in Figure 7(a). We find
that the transfer efficiency is increased as the EDF length
increases in the beginning. Then it reaches a constant
value of 21.5% as the pump power is larger than 70 mW.
The transfer efficiency here is defined as

P
in Las th
(3)
( PP PP )
where is the laser transfer efficiency, Ppin is the input
pump power, and Ppth is the threshold power. Figure 7(b)
shows the superimposed output spectra of the tunable
fiber laser using two TFBGs with original wavelength of
1539.13 and 1553.0 nm, respectively. High side mode
suppression ratio (SMSR) of around 57 dB was obtained
for the entire C-band.
4. Tunable SWPF-Based Tunable Fiber

Lasers: Configurations and Experimental/
Simulation Results
4.1. SWPF-Based Tunable Fiber Laser
In this section, a continuously tunable erbium-doped
ENG
590
S.-K. LIAW ET AL.
fiber laser is demonstrated by incorporating a tunable

SWPF into ring resonator. The wideband tunable SWPF
is based on dispersive evanescent tunneling from a sidepolished single-mode fiber and a dispersive optical
polymer overlay structure. In fabrication, a portion of the
fiber jacket was stripped off and the section was then
embedded and glued into the curved V-groove on a silicon substrate, as shown in Figure 8. The central cladding
thickness after polishing was around 2.7 m. Finally, the
characteristics of the well-polished fibers were calibrated
by liquid-drop experiments. The effective interaction
length was estimated to be 11 mm at 1550 nm wavelength for side-polished SMF-28. In Figure 8, the tunable
SWPF is incorporated into the resonant cavity to provide
a wideband tunable transmission loss window. The dispersive optical polymer overlaying the side-polished
fiber is OCK-433 (Nye Lubricants) with the thermo-optic
coefficient dnD/dT of 3.6 104/C and is heated by a
dual TEC. The refractive index of the OCK-433 decreases with increasing temperature. The tuning efficiency is 7.65 nm/C and the signal-to-ASE ratio is around
40 dB. The EDF used here has absorption coefficient of
12 and 30 dB/m for 1480 and 1530 nm wavelength, respectively, and is pumped by a 1480 nm pump laser diode (LD) in 250 mW launched power.
To investigate the influences of the sharpness of the
spectral cutoff curve, the Cargille liquids were applied on
SWPF. The spectral responses of the fiber laser are
shown in Figure 9(a). When the refractive indices of
1.456 (nD) and 1.458 (nD) were used, the lasing wavelength moved to shorter wavelengths and the peak power
decreased following the gain profile. Subsequently, the
Cargille liquids were replaced by OCK-433 and the spectral responses are shown in Figure 9(b). When the
temperature cooled down, the lasing wavelengths were
moved toward shorter wavelengths again. As the temperature was tuned to 39.6C, the lasing wavelength was
at 1569.8 nm. Thus, the tuning range of the fiber laser is
26 nm with temperature variation of 3.4C, and typical
signal-ASE-ratio is above 40 dB and the average FWHM
is around 0.5 nm.
Figure 8. Experimental setup of erbium-doped fiber ring

laser using side-polished fiber based tunable SWPF [12].
(a)
4.2. Thermo-Optic Tunable Erbium-Doped

FiberRing Laser
In this subsection, wideband tunable high cutoff-efficiency SWPFs were discretely located in standard silicabased C-band EDF to filter out the C + L band ASE so
that the optical gain for S-band could be acquired to realize fiber laser. To investigate the amplification characteristics in the S-band, a 980-nm pump laser with 135mW output power was launched into EDF in a forward
pumping scheme. The high-cutoff-efficiency short-pass
filters in the 17.5-m-long EDF could discretely suppress
the unwanted C + L band ASE and pass the S-band signal
(b)
Figure 9. (a) Spectral responses of EDF fiber ring laser in

air and using two kinds of Cargille index liquids on SWPF;
(b) spectral responses of wavelength tuning of fiber laser
when OCK-433 polymer was cool down [12].
ENG
S.-K. LIAW ET AL.
and 980-nm pump light. Subsequently, an input power of

25 dBm was launched into the EDF from distributed
feedback laser signals in the S-band. The input signal
spectra and amplified output signal spectra in the S-band
at 28.6C are shown in Figure 10(a). In the S-band the
net signal gain at 1486.9 nm was measured to be 18.92
dB.
The experimental set-up of the tunable EDF ring laser
is shown in Figure 11, where the tunable fused-tapered
SWPF with the use of Cargille index liquid (nD = 1.456)
can provide a sharp filter skirt and a deep stop band rejection efficiency (> 50 dB). However, a single local
SWPF is inefficient for the standard EDF to be operated
at the shorter wavelengths (S-band) of the gain bandwidth. Consequently, we employ four-stage in-line tunable fused-tapered fiber SWPFs discretely located in the
standard silica-based EDF to achieve the tunable S-band
fiber laser. When the SWPF is turned on, the C + L-band
ASE is suppressed to obtain the gain for S-band lasing.
The four filters are discretely located in a 16-m-long stan-
(a)
591
Figure 11. Experimental setup of tunable EDF ring laser

towards short-wavelength limit at 1450 nm (Each 4-m-long
EDF and short-pass filter forms a gain stage and there are
four gain stages totally in the ring cavity. The FP filter is
used for narrowing the laser linewidth down to below 0.2
nm [13]).
dard silica-based C-band EDF to substantially suppress

the ASE at the wavelengths longer than the lasing wavelength which can be tuned by varying the applied temperature on SWPFs. When a 980-nm laser with pump
power of 208 mW launches into the EDF, the laser spectra at different temperatures are shown in Figure 10(b).
When the applied temperature slightly decreases, the
lasing wavelength moves to shorter wavelength. The
average for the tunable laser is measured to be as high
as 57.3 nm/C from 1545.2 to 1451.9 nm ascribing to the
wideband tunable high-cutoff-efficiency SWPFs. The
average FWHM is 0.53 nm and the signal-to-ASE ratio
is above 40 dB.
5. Discussion
5.1. Advantages of FBG-Based Tunable Fiber
Lasers
(b)
Figure 10. (a) Amplification spectra of the signals in S-band

at 28.6C; (b) evolution of output laser spectra by cooling
down optical liquid and bending splicing point using the
first set of tapered fibers [13,17].
A versatile and cost-effective laser source should have

the ability to allow the user to choose which wavelength
is needed or the desired scanning range. The wavelength
tunable FBG-based lasers we presented here can satisfy
such requirement. It is well known that the cavity of a
fiber laser may be designed based on a pair of FBGs that
work as its end mirrors and determine the resonant
wavelength. When one of the resonant wavelengths of
the FBGs is changed slightly by tension or heating, the
reflection power by the FBG pair at a new laser wavelength will decrease due to wavelength misalignment
between them, thus it is difficult to fine-tune the FBG
ENG
592
S.-K. LIAW ET AL.
pair back to the same wavelength. Nevertheless, either

the OC-based or BFM-based laser configuration could
overcome such a problem because one FBG only is used
to tune the lasing wavelength. Other advantages of FBGbased tunable fiber lasers are: 1) Narrow laser linewidth
and near polarization-independent; 2) both the OC- based
and BFM-based tunable fiber lasers improve the pumping efficiency by recycling the residual pump power
back to the gain medium using backward pumping; 3)
the TFBG could be used to tune the desired wavelength
precisely and quickly; 4) the proposed FBG-based tunable fiber lasers may use one OSW pair and a plurality of
tunable FBGs to expand the output wavelength range; 5)
they are simpler and potentially less expensive than other
commercial products; and 6) the sizes are compact and
the weights are light.
5.2. Merits of SWPF-Based Tunable Fiber

Lasers
It is advantageous to explore a widely tuning fiber laser
with lasing wavelength down to S-band at a high tuning
speed. Conventionally, the silica-based EDF at room
temperature can only emit fluorescence at wavelengths
longer than 1490 nm. Thus, achieving high-performance
S-band lasers critically depends on the SWPFs. The sidepolished SWPFs were adopted because they are mechanically strong, and the polishing depth and interaction
length can be precisely determined. From a different
point of view, SWPFs using the fused-tapering technique
are easy, fast, and cost-effective fabrication processes.
An optimized side-polishing/tapered fiber filter structure
can attain high-cutoff efficiency and wide tuning range.
Based on the proposed SWPFs, widely tunable, singlefrequency rare-earth-doped fiber lasers can be achieved.
Besides, the SWPF-based tunable fiber lasers have other
advantages such as: 1) wide tuning range covering the Sand C-bands, 2) high power and low noise, 3) simplicity
and cost-effectiveness, and 4) high index sensitivity up to
1 105 with high Q resonator.
5.3. Single-Frequency Design

To design a single-frequency tunable fiber laser, various
kinds of methods such as multiple ring cavities, FBGs,
microrings, spatial hole burning in unpumped EDF, and
nonlinear loop mirror were proposed. Also, a short cavity
length is usually required to enlarge the mode spacing.
For the linear-cavity fiber lasers as mentioned, a simpler
way to achieve single-longitudinal-mode (SLM) operation is to put a piece of EDF as pump absorber between
the WDM coupler and 1 2 OSW for the OC-based linear-cavity tunable fiber laser as shown in Figure 3; and
between the 1480/1550 nm WDM and TFBG for the
BFM-based linear-cavity tunable fiber laser as shown in
Figure 5, individually. On the other hand, the SWPF in

tunable fiber lasers is naturally a broadband filter that is
obviously difficult for single-longitudinal mode laser
operation. However, a SWPF made of a highly dispersive waveguide structure can introduce high chromatic
dispersion inside the laser cavity to significantly reduce
the cavity modes into one. One suggestion is to concatenate the SWPF and an additional ultra-narrowband filter
inside the cavity to attain SLM operation.
6. Conclusions
Two kinds of tunable fiber-filter-based EDF fiber lasers
have been reviewed. Both of them have broadband wavelength tuning range including C- and/or S-band. Using
FBG in strain mechanism, we have proposed and demonstrated a tunable FBG-based fiber laser that employs
one OC, two homemade TFBGs. The configuration consists of a linear cavity to achieve a wavelength tuning
range of 31.5 nm with 0.05 nm linewidth and over 60 dB
SNR. The power variation over the entire tuning range is
less than 0.1 dB with power equalization by using
low-cost VOAs. Another way is to employ a BFM and
tunable FBG at either cavity end of fiber cavity. The
BFM acts as a broadband rear-end reflector both for
lasing signal and pump source. For wavelength tunable
demonstration, power variation over the whole C-band is
less than 1.0 dB without the usage of power equalization. The time to reach stable laser operation is less than
11 ms after switching between the two FBGs, and the
continuous tuning resolution is less than 0.2 nm in the
whole range. For the SWPF-based tunable fiber laser
using temperature tuning mechanism, two tunable
SPWFs based erbium-doped fiber lasers were reviewed.
The side-polishing and fused-tapering techniques were
used to achieve thermo-optic tunable short-wavelengthpass function based on material dispersion discrepancy
and variations of waveguide structures. The tuning efficiency is 50 nm/C, cut-off efficiency is 1.2 dB/nm, and
rejection efficiency is 55 dB, individually. The widely
tunable SWPFs were applied to achieve broadband and
high-tuning-efficiency S- and/or C-band EDF ring lasers,
which can be tuned close to the short-wavelength edge of
gain bandwidth, and the tuning range is 26 nm with the
signal-to-ASE-ratio of around 40 dB, and the FWHM
linewidth is about 0.5 nm. All of them have graceful
features of simple structure, compactness, ease of connection to fiber components, high-efficiency, and continuous tunability. They are promising for vast applications in lightwave transmission, optical test instrument,
fiber-optic gyros, spectroscopy, material processing, fiber sensing, WDM backup light sources, as well as in biophotonics.
ENG
S.-K. LIAW ET AL.
7. Acknowledgements
The authors were partially supported by the National
Science Council (NSC) (Project Nos. NSC 982221E011-017, NSC 97-2923E-011-001-MY3, NSC 982218
E008-004, NSC 98-2221-E-239-001-MY2). We thank
Jang W. Y., Wang C. J., Hung K. L., Jhong G. S., Chi S.,
Tseng S. M., Huang C. M., Lai Y. for discussion, T.
Wang and Z. G. Shieh for kind help.
8. References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
A. Bellemare, J. F. Lemieux, M. Tetu and S. LaRochelle,

Erbium-Doped Ring Lasers Step-Tunable to Exact
Multiples of 100 Ghz (ITU-GRID) Using Periodic filter,
Proceedings of ECOC98, Madrid, September 1998, pp.
153-154.
C. S. Kim and J. U. Kang, Multiwavelength Switching
of Raman Fiber Ring Laser Incorporating Composite
Polarization-Main Maintaining Fiber Lyot-Sagnac Filter,
Applied Optics, Vol. 43, No. 15, 2004, pp. 3151-3157.
C. C. Lee, Y. K. Chen and S. K. Liaw, Single-Longitudinal-Mode Fiber Laser with Passive Multiple-Ring
Cavity and its Application for Video Transmission,
Optics Letters, Vol. 23, No. 5, 1998, pp. 358-360.
S. Kim, B. Lee and D. H. Kim, Experiments on Chaos
Synchronization in Two Separate Erbium-Doped Fiber
Lasers, IEEE Photonics Technology Letters, Vol. 13, No.
4, 2001, pp. 290-292.
S. Foster, Spatial Mode Structure of the Distributed
Feedback Fiber Laser, IEEE Journal of Quantum
Electronics, Vol. 40, No. 7, 2004, pp. 884-892.
M. K. Abd-Rahman and H. Ahmad, Multiwave-Length
Brillouin Erbium Fiber Laser Pumped from FBG Fiber
Laser Sharing the Same EDF, Proceedings of the 4th
Pacific Rim Conference on Lasers and Electro-Optics,
Chiba, July 2001, pp. 40-41.
H. Chen, F. Babin, M. Leblanc and G. W. Schinn, Widely Tunable Single-Frequency Erbium-Doped Fiber
Lasers, IEEE Photonics Technology Letters, Vol. 15, No.
2, 2003, pp. 185-187.
D. Sabourdy, V. Kermene, A. Desfarges-Berthelemot, L.
Lefort, A. Barthelemy, P. Even and D. Pureur, Efficient
Coherent Combining of Widely Tunable Fiber Lasers,
Optics Express, Vol. 11, No. 2, 2003, pp. 87-97.
[9]
593
Y. C. Zhao, S. Winnall and S. Fleming, Tunable FiberRing Laser Based on Broad-Band Fiber Bragg Grating
and Bending Effects, Microwave and Optical Technology Letters, Vol. 46, No. 6, 2005, pp. 562-563.
[10] S. K. Liaw, W. Y. Jang, C. J. Wang and K. L. Hung,
Pump Efficiency Improvement of a C-Band Tunable

Fiber Laser Using Optical Circulator and Tunable Fiber
Gratings, Applied Optics, Vol. 46, No. 12, 2007, pp.
2280-2285.
[11] S. K. Liaw and G. S. Jhong, Tunable Fiber Laser Using
a Broad-Band Fiber Mirror and a Tunable FBG As

Laser-Cavity Ends, IEEE Journal of Quantum Electronics, Vol. 44, No. 6, 2008, pp. 520-527.
[12] N. K. Chen, S. Chi and S. M. Tseng, An Efficient Local
Fundamental-Mode Cutoff for Thermo-Optic Tunable

Er3+-Doped Fiber Ring Laser, Optics Express, Vol. 13,
No.18, 2005, pp. 7250-7255.
[13] N. K. Chen, C. M. Huang, S. Chi and Y. Lai, Towards
The Short-Wavelength Limit Lasing at 1450 Nm over

I-4(13/2)-> I-4(15/2) Transition in Silica-Based ErbiumDoped Fiber, Optics Express, Vol. 15, No. 25, 2007, pp.
16448-16456.
[14] A. D. Kersey, M. A. Davis, H. J. Patrick, M. LeBlanc, K.
P. Koo, C. G. Askins, M. A .Putnam and E. J. Friebele,

Fiber Grating Sensors, Journal of Lightwave Technology, Vol. 15, No. 8, 1997, pp. 1442-1463.
[15] M. Arbore, Y. Zhou, H. Thiele, J. Bromage and L.
Nelson, S-Band Erbium-Doped Fiber Amplifiers for

WDM Transmission between 1488 and 1508 Nm, Proceedings of Optical Fiber Communication Conference,
Georgia, 23-28 March 2003, pp. 374-376.
[16] M. A. Arbore, Application of Fundamental-Mode Cutoff
for Novel Amplifiers and Lasers, Proceedings of Optical

Fiber Communication Conference, (OFC 2005), Anaheim, Vol. 5, 6-11 March 2005.
[17] N. K. Chen, K. C. Hsu, S. Chi and Y. Lai, Tunable
Er3+-Doped Fiber Amplifiers Covering S and C + L

Bands over 14901610 Nm Based on Discrete Fundamental-Mode Cutoff Filters, Optics Letters, Vol. 31,
No. 19, 2006, pp. 2842-2844.
[18] S. Y. Chou, K. C. Hsu, N. K. Chen, S. K. Liaw, Y. S.
Chih, Y. Lai and S. Chi, Analysis of Thermo-Optic

Tunable Dispersion-Engineered Short-Wavelength-Pass
Tapered-Fiber Filters, Journal of Lightwave Technology,
Vol. 27, No. 13, 2009, pp. 2208-2215.
ENG
Engineering, 2010, 2, 594-601

Behaviour of a Composite Concrete-Trapezoidal

Steel Plate Slab in Fire
Toma Hozjan, Miran Saje, Igor Planinc, Stanislav Srpi, Sebastjan Bratina
University of Ljubljana, Faculty of Civil and Geodetic Engineering, Ljubljana, Slovenia
E-mail: miran.saje@fgg.uni-lj.si
Received February 11, 2010; revised March 28, 2010; accepted April 4, 2010
Abstract
The present paper investigates fire resistance of a simply-supported composite concrete-trapezoidal steel
sheet slab. The objective is to find out if a steel sheet, as a moisture diffusion barrier, may substantially effect
the hydro-thermal situation in the concrete part of the cross-section. The numerical integration of the equations of a coupled hygro-thermal boundary-value problem (Tenchev, R.T., Li, L.Y. and Purkiss, J.A. (2001)
Num. Heat Transfer Part A, 39(7), 685-710), with and without considering the barrier, shows that the barrier
does not really effect the magnitude and the development of temperatures over the cross-section, while there
is a significant effect on the pattern of moisture transport and the magnitude of vapour pressure. Particularly
high magnitudes of vapour pressure (about 4.5 MPa) were shown in cases where the steel sheet was considered in analyses, which indicates a possible micro damage of concrete in the web of the section, although
spalling probably cannot take place due to the steel sheet cover. As the typical composite slab investigated
here is not sufficiently fire resistant without any additional reinforcement bars placed in the web, further investigations are directed to finding an optimal position and area of these bars. Following a simplified procedure given in EC2 (Eurocode 2, Design of Concrete Structures, Part 1.2 (2004) Structural fire design, European Committee for Standardization) and assuming that the present composite slab is subject to the uniform
traction q = 1.52 kN/m, yields that placing one bar with the area 1.153 cm2 4 cm away from the edge suffices
for the 60 min fire resistance of the slab.
Keywords: Fire Analysis, Composite Structures, Heat and Mass Transfer
1. Introduction
Composite concrete-trapezoidal steel plate slabs are widely used structural elements in buildings and bridges.
During the placement of concrete the trapezoidal steel
plate replaces panelling, while upon hardening of concrete the two materials work as a composite slab, the
steel plate representing the reinforcement. For the better
vertical load redistribution, concrete is additionally reinforced with a steel mesh at the upper part of the crosssection (here called the flange). The reinforcement in the
web of the concrete part of the section needs rarely to be
applied for non-accidental actions. By contrast, when the
composite slab is exposed to fire, the steel plate is directly exposed to high temperatures resulting in a substantial decrease of its bearing capacity. The reinforcement in the web and its position within the concrete slab
then become essential. In fact, both the position and the
area of the additional reinforcement turn out to be essenCopyright 2010 SciRes.
tial issues of a safe fire design.

The knowledge of temperature and pore pressure distributions in the slab during fire is the key to fire resistance predictions. There are a number of mathematical
models appropriate for the prediction of temperature
field in the composite concrete-trapezoidal steel plate
slab in fire. Luikov [1] was probably the first to introduce the theoretical basis for the coupled heat and moisture transfer through a permeable porous material such as
concrete. His mathematical model is described by the
system of two non-linear partial differential equations
with permeability coefficients and thermodynamic characteristics of material being functions of the heat and
moisture state. Luikovs model was later on improved by
Baant and Thonguthai [2]. Their model is enhanced by
the capability of considering the dehydration process of
chemically bound water in concrete, while free water
evaporation and vapour condensation in concrete are neglected. The evaporation and condensation were later
ENG
T. HOZJAN ET AL.
on accounted for by Davie, Pearce and Biani [3],

Gawin, Pesavento and Schrefler [4] and Tenchev, Li and
Purkiss [5]. Their models are now considered to be rather
complete for the analysis of concrete structures in fire.
In the present paper, a somewhat modified model of
Tenchev et al. [5] is employed to investigate the hygro
-thermal behaviour of a composite concrete-trapezoidal
steel plate slab in fire. This numerical model enables us
to estimate the distribution of temperature, moisture and
pore pressure over the concrete cross-section at any time
during a fire. These are vital data for predicting the fire
resistance time, and spalling of concrete [4]. A typical,
simply supported composite floor of high-rise office
buildings is being analysed, and the effect of the trapezoidal sheet as a moisture diffusion barrier on the distribution of temperature, pore pressure and free water content over the cross-section of the slab is being presented
and discussed. Once the temperatures within the crosssection have been obtained, the fire resistance time at
failure is estimated on the basis of the temperatures in
the additional reinforcement bars of the concrete webs.
In what follows we make a short overview of the
equations of heat and moisture transport in concrete and
describe relevant variables; subsequently we discuss the
hygro-thermal behaviour of the composite slab in fire.
2. Heat and Moisture Transport in Concrete

A coupled heat and moisture transfer in concrete, when
exposed to fire, can be mathematically described by the
system of mass conservation equations for each phase of
concrete separately and with an energy conservation
equation as follows [5]:
Water conservation:
L
t
J L E L
(1)
Water vapour conservation:

G V
(2)
JV E L
Air conservation:
G A
t
(3)
J A
Energy conservation:
C Tt
k T Cv T E E L D
D
t
(4)
In (1)-(3) i denotes density of phase i, G V and

represent mass concentration of air and water vapour per unit volume concrete, Ji is the mass flux of
phase i, E L is the rate of evaporation of free water (including desorption), t is time. Index i denotes the phase:
L is free water, V is water vapour and A is dry air. In (4)
C is heat capacity of concrete, k is thermal conductiv G A
595
ity of concrete, Cv relates to the energy transferred by

fluid flow, E is the specific heat of evaporation, D
is the specific heat of dehydration, and T is the absolute
temperature in degrees Kelvin.
The mass fluxes of dry air, water vapour and free water can be expressed in terms of pressure and concentration gradients assuming that Darcys and Ficks law are
applicable and that the diffusion of adsorbed water on the
surface of solid cement phase skeleton is negligible:

J A G A v G G G D AV A
G
(5)

J G G V v G G G DVA V
G
(6)
(7)
JL L vL
The fluxes are defined per unit area of concrete. In

Equations (5)-(7), DAV and DVA are the diffusion coefficients of dry air in water vapour and water vapour in dry
air within the porous concrete, and vG and vL are the velocities of the gas and liquid water phases resulting from
a pressure-driven flow as given by Darcys law:
vG
vL
KKG
KK L
PG
(8)
PL
(9)
Here K is the intrinsic permeability of dry concrete, KG

and KL are the relative permeabilities of the gas and liquid phases, G and L are their dynamic viscosities,
and PG and PL are the corresponding pressures. Following the model of Tenchev et al. [5], we at this point assume that the liquid pressure is equal to the gas pressure,
PG = PL. It is also assumed that air and water vapour behave as an ideal gas and that the content of free water
L is determined by the help of the simplified sorption
curves introduced by Baant and Kaplan [6]. To achieve
a better numerical stability and to avoid loss of convergence in the global iteration, we model sorption curves
with polynomials of the third order, while its coefficients
are temperature dependent [7].
After combining (1)-(2), we end up with three coupled
partial differential equations describing the transfer of
dry air and moisture, and energy conservation. The solution is obtained numerically by the finite element method
using our original computer programme, where we consider temperature T, pore pressure PG and water vapour
content V as the primary unknowns of the problem.
Further technical details are given in, e.g. [5].
3. The Transfer of Heat and Moisture over

the Cross-Section of the Composite Slab
The composite slab is exposed to standard fire accordENG
T. HOZJAN
596
ing to ISO 834 [8]. Figure 1 presents the geometric and

loading data along with the 2D finite-element mesh of
the heat and moisture transfer analysis over the crosssection of the composite slab. Due to its symmetry only
one half of one wave of the cross-section is accounted
for. We consider two different boundary conditions. In
the first set of boundary conditions (case A1), we consider the trapezoidal steel sheet on the lower surface
(denoted as edge 1 in Figure 1) to be the diffusion barrier. This way no transfer of moisture through the lower
surface is possible. In the second set of boundary conditions (case A2), we neglect any effect of the steel sheet
on the moisture flux. The moisture flux through the top
surface (denoted as edge 2) is regarded as possible.
The fire is assumed to emerge from below (edge 1),
while the temperature of edge 2 remains constant (20C).
ET AL.
The boundary conditions are displayed in Table 1. The

remaining data needed in the analysis are: density of
3
concrete 2400 kg/m , density of cement 300
3
kg/m , temperature T0 = 20C, initial pore pressure
PG,0 = 0.1 MPa, initial water vapour content
3
V , 0 0.0111 kg/m , water vapour content on boundary
c
cem
3
V , 0.0089 kg/m , initial porosity of concrete por 0.15,
0
16
initial permeability of concrete K 1 10 and initial

free water amount 10 kg/m3 . The heat transfer coefficient and emissivity on edge 1 are assumed to be
2
equal to hq 25 W/m K , while the heat transfer coeffi2
cient on edge 2 is hq 9 W/m K. According to EC2 [9]
emissivity on edge 2 is neglected. The time step employed in the numerical time integration equals 0.5 s.
L,0
Figure 1. The composite plate and 2D finite element mesh over the concrete cross-section.
Table 1. Boundary conditions of composite slab.
case
T
A1
PG
V
A2
edge 1
edge 2
qT qT (TISO 834 )
qT qT (T 20 C)
PG
n
V
symmetry
o
PG
PG 0.1 MPa
n
0
qV qV ( V , )
qT qT (TISO 834 )
qT qT (T 20 C)
PG
PG 0.1 MPa
PG 0.1 MPa
qV qV ( V , )
qV qV ( V , )
n
T
n
PG
ENG
T. HOZJAN ET AL.
The development of the temperature field over the

cross-section of the slab at 15, 30 and 60 min is presented
in Figure 2. Although the slab is exposed to the standard
fire according to ISO 834, in which case the initial heating
rate is rather high, an overall heating of the composite
section is quite slow. Figure 2 shows that at 60 minutes
the isothermal for 500C stays rather close to the lower
edge, i.e., only about 2.5 cm away. According to the simplified method in EC2, Annex B [9], the contour line T =
500C is important in determining the fire resistance of the
composite slab exposed to fire, enabling one to neglect the
part of concrete whose temperature is greater than 500C.
To assess the effect of steel sheet as the diffusion barrier, we present in Figure 3 the increase of temperatures
with time in some characteristic points within the web of
the cross-section, whose positions are convenient for
placing additional reinforcement bars to achieve a sufficient fire resistance. The coordinates of the points are
presented in Table 2. The comparison of the results for
597
cases A1 and A2 shows that the temperatures differ only

a little and that the increase of temperature is slower in
case A1. Similar results have been obtained for the same
class of composite cross-sections of various height dimensions by separate numerical investigations. Thus we
can conclude that the steel sheet essentially does not affect the temperatures.
As observed from Figure 3, the temperature gradient
at characteristic points A, B, C and D at 60 min is about
65C per 1 cm, which results in the temperature decrease
between points A and D to be roughly 200C per 2.5 cm.
It is now clear that we can increase the fire resistance of
the composite slab substantially, if we change the position of the bars for only a few centimeters away from the
lower surface of the composite slab. Note that such a
change of the position results in a somewhat smaller
static height. This should not be critical during fire when
the imposed actions and the safety factor are assumed
smaller than regularly.
(a) temperature T [],case A1

t = 15 min
t = 30 min
t = 60 min
0.14
0.14
0.14
0.12
0.12
0.12
0.1
0.1
0.1
0.08
0.08
0.08
0.06
0.06
0.06
0.04
0.04
0.04
0.02
0.02
0.02
0
0 0.02 0.04 0.06 0.08 0.1 0.12
0 0.02 0.04 0.06 0.08 0.1 0.12
0
0 0.02 0.04 0.06 0.08 0.1 0.12
1000
900
800
700
600
500
400
300
200
Figure 2. Distribution of temperature over the cross-section.
Figure 3. Variation of temperature with time in points A, B, C and D.

ENG
T. HOZJAN
598
Table 2. Coordinates of characteristic points A, B, C, D.

point
yre [cm]
zre [cm]
2.5
2.5
2.5
2.5
2.5
While the steel sheet as a diffusion barrier has only a

small effect on the temperature distributions, its effect on
pore pressures is substantial. This is clearly seen in Figure 4. In case A1, where the transfer of water and water
vapour through edge 1 is blocked, the whole amount of
free water and water vapour is pushed by the temperature
and pore pressure gradients towards edge 2. This is depicted in Figure 5, showing the distribution of free water
content over the cross-section at various instants. Beyond
200C the chemically bounded water starts releasing and
further increases the volume of free water (Figure 5(a)).
At some point a full saturation of pores takes place
which causes the substantial rise of the pore pressure.
This is more pronounced in case A1, where the magnitude of pore pressure at 30 min is about 4.5 MPa and is
(a) pore pressure PG [MPa],case A1
t = 15 min
ET AL.
almost homogeneous over the web of the concrete slab

(Figure 4(a)). By contrast, in case A2 (Figure 4(b)),
vapour can also escape through the lower edge; in fact,
the inspection of the results shows that moisture is
largely released through the lower edge.
As a result, the magnitude of the pore pressure in concrete is in this case significantly lower (less than 3 MPa)
both in the flange and in the web of the cross-section.
Comparing the distributions of free water content over
the cross-section (Figure 5) shows that the patterns are
similar, yet the magnitudes of the water content are different. It is obvious that the effect of the steel sheet on
the water content is high due to impervious edge 1.
The results prove that the steel sheet plays a significant role in the distribution of pore pressures as well as
their magnitudes. While this may result in damage of
concrete and the contact between concrete and steel sheet,
it is still not clear if solely high pressures could be responsible for explosive spalling of concrete. The absence
of explosive spalling may, however, be due to the damaged contact between the steel sheet and concrete. Some
water vapour could therefore escape either through the
lower edge or in the axial direction, causing drop in the
pore pressure. Numerous studies also show that pore
pressures need not be the main reason for explosive
spalling of concrete to occur [4,10].
t = 30 min
t = 60 min
0.14
0.14
0.14
0.12
0.12
0.12
0.1
0.1
0.1
0.08
0.08
0.08
0.06
0.06
0.06
0.04
0.04
0.04
0.02
0.02
0.02
0
0 0.02 0.04 0.06 0.08 0.1 0.12
0
0 0.02 0.04 0.06 0.08 0.1 0.12
(b) pore pressure PG [MPa],case A2

t = 15 min
0
0 0.02 0.04 0.06 0.08 0.1 0.12
t = 30 min
t = 60 min
0.14
0.14
0.14
0.12
0.12
0.12
0.1
0.1
0.1
0.08
0.08
0.08
0.06
0.06
0.06
0.04
0.04
0.04
0.02
0.02
0.02
0
0 0.02 0.04 0.06 0.08 0.1 0.12
0
0 0.02 0.04 0.06 0.08 0.1 0.12
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
0 0.02 0.04 0.06 0.08 0.1 0.12
4.5
4
3.5
3
2.5
2
1.5
1
0.5
Figure 4. Distribution of pore pressures over the cross-section at t = 15, 30, and 60 min. (a) case A1; (b) case A2.
ENG
T. HOZJAN ET AL.
599
(a) free water content FW [kg/m3], case A1

t = 15 min
t = 30 min
t = 60 min
0.14
0.14
0.14
30
0.12
0.12
0.12
25
0.1
0.1
0.1
20
0.08
0.08
0.08
0.06
0.06
0.06
0.04
0.04
0.04
0.02
0.02
0.02
0
0 0.02 0.04 0.06 0.08 0.1 0.12
0
0 0.02 0.04 0.06 0.08 0.1 0.12
15
10
5
0
0 0.02 0.04 0.06 0.08 0.1 0.12
(b) free water content FW [kg/m3],case A2

t = 15 min
t = 30 min
t = 60 min
0.14
0.14
0.14
30
0.12
0.12
0.12
25
0.1
0.1
0.1
20
0.08
0.08
0.08
0.06
0.06
0.06
0.04
0.04
0.04
0.02
0.02
0.02
0
0 0.02 0.04 0.06 0.08 0.1 0.12
0
0 0.02 0.04 0.06 0.08 0.1 0.12
0
0 0.02 0.04 0.06 0.08 0.1 0.12
15
10
5
0
Figure 5. Distribution of free water content over the cross-section at t = 15, 30, and 60 min. (a) case A1; (b) case A2.
Figure 6. (a) Variation of temperature with time for different vertical positions of reinforcement bars; (b) Variation of temperature with vertical position zre at t = 30, 60, and 90 min.
As the steel sheet is directly exposed to high temperatures, it loses its bearing capacity shortly after fire begins.
The presence of additional reinforcement bars, if any, is
essential in such cases. That is why fire resistance of
composite slab in terms of the resistance time can be
rather well estimated on the basis of actual temperatures
in the additional reinforcement bars placed in the web. It
is well known that creep strains of mild steel start increasing at about 400C [11]. At roughly 500C, the
creep strain rates become pronounced and dictate the

failure of the slab. Therefore it is plausible to estimate
the resistance time of the composite slab on the basis of
the critical temperature in the additional reinforcement.
Thus we may assume that loss of resistance is strongly
related to the instant when the temperature of the additional reinforcement bar reaches 500C. Figure 6(a)
shows the time development of temperatures at various
vertical locations, zre, appropriate for placing additional
ENG
600
T. HOZJAN
reinforcement bars in the cross-section. The horizontal

position of the bars is kept at yre = 2.5 cm in all cases. As
expected, the rise of temperature in the reinforcement bar
is smaller for bars placed higher. The critical temperature,
Tcr = 500C, for zre = 4 cm is reached in 58 min and for
zre = 6 cm in 83 min. A big difference in time compared
to a relatively small change in the vertical position is noticeable. This becomes even more clear in Figure 6(b),
where the dependency of temperature on the vertical position zre is presented for t = 30, 60, and 90 min.
The idea of a critical temperature can serve as a means
to develop a method for a simplified design. Once the
fire resistance class (in minutes) and the position of the
additional reinforcement bar have been decided on, the
temperature of the bar is obtained from the temperaturetime distributions over the cross-section found by the numerical analysis. Finally, using the simplified EC2 [9]
Annex procedure gives the minimum area of the steel bar.
The above procedure has been validated by the nonlinear mechanical analysis [7] of the simply supported
beam discussed here (Figure 1), subjected to the uniform
non-accidental external load q 1.52 kN/m [7]. Assuming zre = 4 cm and class R60 (i.e., the 60 min fire resistance time), the above proposed simplified EC2 [9] procedure yields the area of the bar 1.153 cm2. Using the
above reinforcement data and the calculated temperature
distributions in the non-linear mechanical analysis [7]
gives the fire resistance time to be 69 min. This rather
accurate and conservative prediction of the fire resistance
time proves that the graphs given in Figure 6 enable us
to fairly accurately choose the position and the area of
the additional reinforcement bars for a given fire resistance class. The details of the mechanical analysis [7] are
out of the scope of the present paper and are here omitted.
take place due to the steel sheet cover.

Our recent mechanical analyses have shown [7] that
the composite slab investigated here is not sufficiently
fire resistant without placing an additional reinforcement
in the web. The objective of the present study was hence
to find the area and optimal position of the steel reinforcement bar in the web such that its fire resistance is
within a chosen time resistance class. Due to a slow temperature diffusion in concrete, a big difference in the
resistance time follows after a small change in the vertical position of the bar. For the composite slab investigated herein, the area of 1.153 cm2 and the distance of
only 4 cm away from the lower surface suffice for the 60
min fire resistance. The resistance time obtained this way
is well in keeping with the resistance time obtained by
the sophisticated mechanical analysis [7].
5. References
[1]
A. V. Luikov, Systems of Differential Equations of Heat

and Mass Transfer on Capillary-Porous Bodies, International Journal of Heat and Mass Transfer, Vol. 18, No.
1, 1975, pp. 1-14.
[2]
Z. P. Baant and W. Thonguthai, Pore Pressure and

Drying of Concrete at High-Temperature, Journal of
Engineering Mechanics Division-ASCE, Vol. 104, No. 5,
1978, pp. 1059-1079.
[3]
C. T. Davie, C. J. Pearce and N. Biani, Coupled Heat

and Moisture Transport in Concrete at Elevated Temperatures - Effects of Capillary Pressure and Absorbed
Water, Numerical Heat Transfer Part A, Vol. 49, No. 8,
2006, pp. 733-763.
[4]
D. Gawin, F. Pesavento and B. A. Schrefler, Towards

Prediction of the Thermal Spalling Risk through a MultiPhase Porous Media Model of Concrete, Computer
Methods in Applied Mechanics and Engineering, Vol.
195, No. 41-43, 2003, pp. 5707-5729.
[5]
R. T. Tenchev, L. Y. Li and J. A. Purkiss, Finite Element Analysis of Coupled Heat and Moisture Transfer in
Concrete Subjected to Fire, Numerical Heat Transfer
Part A, Vol. 39, No. 7, 2001, pp. 685-710.
[6]
Z. P. Baant and M. F. Kaplan, Concrete at High Temperatures: Material Properties and Mathematical Models, Longman, Harlow, 1996.
[7]
T. Hozjan, Non-Linear Analysis of Composite Planar

Structures Exposed to Fire, Ph.D. Dissertation, University of Ljubljana, Faculty of Civil and Geodetic
Engineering, 2009.
[8]
ISO 834, Fire Resistance Tests-Elements of Building

Constructions, International Standard ISO 834, 1975.
[9]
Eurocode 2, Design of Concrete Structures, Part 1.2,

Structural Fire Design, European Committee for Standardization, 2004.
4. Discussion
The simply-supported composite concrete-trapezoidal
steel sheet slab is a widely used structural element in
engineering. Often its fire resistance applied from below
is of prime importance. The steel sheet acts as a moisture
diffusion barrier. In order to find out if such a barrier
may substantially affect the hydro-thermal situation in
the concrete part of the cross-section, we compared the
numerical results of the coupled hygro-thermal boundary-value problem obtained with and without considering
the barrier. The results showed that the barrier does not
really affect the magnitude and the development of temperatures over the cross-section, while the effect on the
pattern of moisture transport and the magnitude of vapour pressure is significant. Particularly high magnitudes
of vapour pressure (about 4.5 MPa) were experienced in
analyses when considering the steel sheet, which indicates a possible micro damage of concrete in the web of
the section, although explosive spalling probably cannot
ET AL.
ENG
T. HOZJAN ET AL.
[10] G. A. Khoury, C. E. Majorana, F. Pesavento and B. A.
Schrefler, Modelling of Heated Concrete, Magazine of
Concrete Research, Vol. 54, No. 2, 2002, pp. 77-101.
601
[11] G. Williams-Leir, Creep of Structural Steel in Fire: Analytical Expressions, Fire and Materials, Vol. 7, No. 2,
1983, pp. 73-78.
ENG
Engineering, 2010, 2, 602-607

The Effect of Initial Oxidation on Long-Term Oxidation of

NiCoCrAlY Alloy*
Chao Zhu, Xiaoyu Wu, Yuan Wu, Gongying Liang
MOE Key Laboratory for Nonequilibrium Synthesis and Modulation of Condensed Matter,
School of Science, Xian Jiaotong University, Xian, China
E-mail: gyliang@mail.xjtu.edu.cn
Received March 8, 2010; revised June 3, 2010; accepted June 5, 2010
Abstract
The initial oxidation behavior of Ni-6.5Co-17.8Cr-3.7Al-0.5Y alloy is investigated at 800C-1000C. X-ray
diffraction results show that the dominant Cr2O3 phase and secondary -Al2O3 and NiO phases are observed
on the surface of samples at all initial stages (oxidized for 16 hours). YAlO3 and -Al2O3 can only be detected at low temperature (800C) while the spinel NiCr2O4 is only observed at 900C and 1000C. Though
the growth rates of -Al2O3 and Cr2O3 are comparable at 900C, the former becomes much lower than the
latter when the temperature changes to 1000C. Scanning electron microscopy (SEM) images show that the
-Al2O3 grows from some irregular ditches in the chromia scale at 900C. However, cracking and spalling
are more serious at 1000C without -Al2O3-grown-ditches, which is in accordance with the growth rates of
these oxides at different temperatures. The cracking can be explained by the results of Raman determination
which indicate that the stress on the surface of specimen oxidized at 1000C is higher than that at 900C.
Owing to this condition, a preoxidation treatment on the NiCoCrAlY alloy for 16 hours is prepared at 900C,
and then thermal cycling oxidation test is conducted at 1000C for 200 hours. The result indicates that the
initial preoxidation treatment at 900C improves the oxidation resistance of alloy at 1000C.
Keywords: NiCoCrAlY, Oxidation Kinetics, Initial Oxidation, Al2O3, Cr2O3
1. Introduction
NiCrCoAlY alloys are often used as bond coatings of
thermal barrier coatings (TBCs) to protect the substrate
from oxidation at high temperature and to provide the
necessary adhesion of the ceramic to the substrate [1-3].
Some authors [4-7] indicated that the protection offered
by MCrAlY (M=Ni, Co or a combination) alloys against
high temperature oxidation relies on the ability of the
alloy to develop and maintain a continuous, dense and
slow growing -Al2O3 scale. The formation of a continuous of alumina (Al2O3) layer during the oxidation of
the substrate at high temperatures could result in a dramatic slowing down of the oxidation process, because
Al2O3 formation has a slower rate of oxidation compared
to other oxidations. Generally, the protectiveness of the
alloy surface at long-term stage is frequently determined
by the initial stage of oxidation [8-10].
*The project was supported by the State Key Development Program for
Basic Research of China (Grant No. 2007CB707700).
Besides alumina, chromia (Cr2O3) also plays an important role during the high temperature oxidation
[11-13]. However, the effect of the interaction among the
oxides on the oxidation resistance of alloy has not been
discussed in detail. In particular, there has been a lack of
attention to the initial stages of oxidation on NiCoCrAlY
alloy to date.
Raman spectroscopy has been used as a non-destructive technique for determining the stresses in oxide
scales for decades due to the bands in the Raman spectra
of specimens shift with pressure [11-13]. The salient
features of the Raman technique are that it does not require a special environment, and it provides a high resolution.
The aim of this study is to improve the service life of
the alloys. The evolution of the oxide scale on the surface of NiCoCrAlY alloy at initial stage oxidized at
800C-1000C were investigated. After reporting the
experimental results, a method to improve the oxidation
resistance was proposed by utilizing the interaction chaENG
C. ZHU
racteristics of Al2O3 and Cr2O3 growth.
2. Experimental
The original powder was commercially available, and its
component was 6.5%Co, 17.8%Cr, 3.7%Al, 0.5%Y,
balance Ni (wt.%) with an average particle size of 16.34
m.
The powder was heated and compacted into the form
of a cylindrical rod of green density equal to 89 5% of
its theoretical density at inner temperature of 800C for 2
hours with a pressure of 300 MPa. Subsequently, the rod
was annealed in vacuum at 1000C for 1 hour in order to
homogenize and recrystallize the alloy.
Disc shaped specimens (diameter 15 mm and thickness 1 mm) were cut from the rod using spark-machining.
The specimen surface was ground and polished. After
each preparation step the specimens were thoroughly
cleaned ultrasonically with alcohol.
Isothermal oxidation was performed in static air at 1
atm pressure in a resistance furnace which has a maximum operating temperature of 1300C. All the specimens were put into the furnace at the same time after the
test temperature was reached. Then, oxidized specimens
were removed out from the furnace after a chosen time
and air-cooled to room temperature. The initial oxidation
tests were performed at 800C, 900C and 1000C for 2,
4, 8 and 16 hours.
A thermal cycling oxidation test was conducted at
1000C for 200 hours. In order to keep consistent of
oxidation condition in the test, the specimens were preoxidized at 900C and 1000C for 16 hours. Then the
oxidation behavior of the specimens was evaluated by
measuring the weight gains of the samples for 184 hours.
The 12-hour cycle consisted of 11 hours holding at
1000C, followed by cooling in air for 1 hour. The precision of the balance was 0.1 mg.
Raman spectroscopy was used here in order to determine the stresses in chromia scales formed at 900C and
1000C on NiCoCrAlY alloy. Chromia has the same
structure as corundum, and therefore it should have
seven Raman active bands (A1g + 5Eg) [11-13]. The most
intense mode is the 549 cm-1 A1g vibration [11] and this
one was used for monitoring the stress.
The spectroscopy was measured at room temperature
using the Renishaw Ramanscope 1000 (Renishaw,
Gloucestershire, UK) in conjunction with an Olympus
BH-2 microscope. During the measurements, the laser
(HeNe, 632.8 nm) was focused at a position on the surface of the sample and the laser spot size was set about
35 m. The Raman spectroscopy acquired was analyzed
by the commercial Renishaw WiRe software to obtain
the peak shift fitted by GaussianLorentzian function.
The surface morphologies and polished cross sections
of the specimens were observed using a scanning elecCopyright 2010 SciRes.
ET AL.
603
tron microscopy (SEM) (JSM-7000F). The chemical

composition of the oxides was determined qualitatively
by energy-dispersive X-ray analysis (EDX). The phases
in the oxide scales were analyzed using an X-ray diffraction (XRD) (Rigaku D, CuK radiation).

3.1. The Oxides on the Surface after Initial
Oxidation
X-ray diffraction patterns of NiCoCrAlY alloy after oxidation for 16 hours at 800C, 900C and 1000C are
shown in Figure 1. Results from the study show that the
oxides on the surface of alloy which was heated to 800C
for 16 hours are composed of Cr2O3, a few -Al2O3 and
YAlO3, trace -Al2O3 and NiO. It was found that however, oxides -Al2O3 and YAlO3 did not form at 900C
and 1000C. The spinel oxide, NiCr2O4, began to exist
after 16 hours of oxidation at 900C.
According to the intensity of diffraction peaks, the
relative oxidation rates of Cr2O3 and -Al2O3 phases on
the surface of alloy oxidized from 2 hours to 16 hours at
800C, 900C and 1000C are shown in Figure 2.
From Figure 2(a), Figure 2(b), it can be seen that the
relative quantities of Al2O3 and Cr2O3 increased quickly
in the first two hours. After that, the oxidation rate of
Al2O3 rises slowly while the relative quantities of Cr2O3
at 900C and 1000C keep fluctuating. The increase of
quantities of -Al2O3 resulted from dense -Al2O3 oxide
forming and -Al2O3 transforming. The fluctuation of
quantities of Cr2O3 at 900C and 1000C may have been
caused by spinel oxides NiCr2O4 formation which consumed Cr2O3. With the increased temperature or prolonged exposure time, NiO eventually became destabilized and reacted with Cr2O3 to form a thin spinel layer
of NiCr2O4, which was thermodynamically more stable
[9,14-16]. The competition between the consumption of
Figure 1. X-ray diffraction patterns of Ni6.5Co17.8Cr3.7

Al0.5Y alloy (oxidized for 16 h) at (a) 800C; (b) 900C and
(c) 1000C.
ENG
C. ZHU ET AL.
604
growth rate of -Al2O3 is not as much as the growth rate

of Cr2O3.
3.2. The Cracking and Closure of the Oxide

Scale in Initial Oxidation
(a)
A SEM image of the specimen surface oxidized in air for

up to 16 hours at 800C is shown in Figure 3(a). Figure
3(b) is at a higher magnification. These images show that
some protrudes, pores and pits are presented on the surface, but few cracks appeared.
The whisker or needlelike oxide phase is observed.
The EDX analysis (Figure 3(c), Figure 3(d)) of the oxide whiskers produced Al, Cr and O peaks, which is
qualitatively identified as Al2O3 and Cr2O3 phases. As
the -Al2O3 phase usually grows in a needlelike, whiskerlike or bladelike morphology and -Al2O3 grows in a
weblike or dense equiaxed structure [14,17], these bladelike oxides should be -Al2O3.
(b)
(c)
(c)
Figure 2. Relative quantities of Cr2O3 and -Al2O3 phases

on the surface of alloy vs. time: (a) -Al2O3; (b) Cr2O3; (c)
the relative oxidation rates of Cr2O3 and -Al2O3 at the
second hour at 800C, 900C and 1000C.
Cr2O3 and the formation of Cr2O3 determined the outline

of curve.
Figure 2(c) shows that the relative oxidation rates of
-Al2O3 and Cr2O3 increased with temperature in the
second hour from 800C to 1000C. Also, it can be observed that the growth rate of -Al2O3 is similar to that of
Cr2O3 at 900C. However, at 1000C, the increase of
(d)
Figure 3. SEM images of the specimen surface (oxidized for
16 h at 800C) (a) surface image; (b) higher magnification;
(c) and (d) EDX analysis in selected region.
ENG
C. ZHU
The surface images of specimens exposed at 900C

and 1000C in a static atmosphere for 16 hours are
shown in Figure 4(a), Figure 4(b). It was found that
specimens covered with fine oxide particles. It was also
observed irregular ditches on the surface at both temperatures and cracks in the oxide scale. The cracks became more serious as the temperature increased. The
spallation on the surface of the oxidized specimen was
unavoidable.
It may have resulted from the stress during the cooling
and heating process and the mismatch between the expansion coefficients of oxides and alloys.
(e)
(f)
605
EDX analysis (Figure 4(f), 4(g)) shows that the oxides

around the cracks are Cr2O3 (Figure 4(d)). This indicates
that these cracks were caused by the Cr2O3 oxide scale
broke during the oxidation. At the same time, it was observed that some oxides were growing from the cracks in
Figure 4(c). By EDX analysis, it is confirmed that these
oxides are -Al2O3 and Cr2O3. It can be speculated that,
at 900C, -Al2O3 grew form the bottom of the Cr2O3
oxide scale cracks, which filled in the cracks and made
the oxide scale dense. Contrasting Figure 4(a) with Figure 4(b), it is can be seen that the cracks in the oxide
scale at 1000C are more than those at 900C. There are
more irregular ditches observed among the Cr2O3 oxide
scale but not so much -Al2O3 fill in the cracks.
Form Figure 2, we observe that the growth rate of
-Al2O3 is similar to that of Cr2O3 at 900C. Though the
spallation on the Cr2O3 scale was unavoidable, -Al2O3
could preferably nucleate during oxidation on the surface
in some cracks of the oxide scale at 900C where they
could grow and fill in those ditches. With the density of
the oxide scale increased, both the oxygen and cation
diffusion rate decreased. Thus the ability of oxidation
resistance would be improved. However, the growth rate
of Cr2O3 is much larger than that of -Al2O3 at 1000C.
When the oxides on the surface of the alloy grew at a
larger rate, the oxide scale cracked and spalled easily.
Because of the cracks, oxygen diffused through the oxide
scale easily to contact the oxidealloy interface, which
speeded up the oxidation of alloy greatly.
The cross-sectional microstructure and the elemental
maps of the NiCoCrAlY specimens obtained by SEM
and EDX after oxidation at three different temperatures
for 16 hours are shown in Figure 5. The elemental concentration regions of O, Cr and Al are presented in the
Figures 5(a2)-5(c4).
From Figure 5(a2), Figure 5(b2) and Figure 5(c2), it
was found that the thickness of the oxide layer increases
with the temperature rising. A continual oxide layer
formed at 900C, and it became thicker at 1000C. The
Cr2O3 phase dominated in the oxide layers (Figure 5(a3),
Figure 5(b3) and Figure 5(c3)) while the Al2O3 phase
was not abundant in these environments (Figure 5(a4),
(g)
Figure 4. The surface images of specimens (oxidized for 16 h

at 900C and 1000C). (a) specimens oxidized at 900C; (b)
specimens oxidized at 1000C; (c) higher magnification of
Figure 4(a); (d) higher magnification of Figure 4(b); (e), (f)
and (g) EDX analysis in selected regions.
ET AL.
Figure 5. The cross-sectional microstructure and the elemental maps of NiCoCrAlY alloy (oxidized for 16 h) at (a)
800C; (b) 900C and (c) 1000C.
ENG
606
C. ZHU ET AL.
Figure 5(b4) and Figure 5(c4)). This indicated that Cr2O3

formed at the initial oxidation stage. Some brightness
Al-containing regions denote the Al2O3 which filled in
the Cr2O3 ditches (Figure 5(b4)). This is in good agreement with the XRD and surface microstructure results.
Though a continuous layer of Al2O3 was not appeared
at the onset of the oxidation, an inner zone of isolated
Al-containing phase could be observed at all three temperatures in Figure 5(a4), Figure 5(b4) and Figure 5(c4).
This was in good agreement with the literature [10] and
the Al-containing phase should be -Al2O3. The lateral
growth of -Al2O3 precipitates occurred until they coalesce into a continuous -Al2O3 layer.
3.3. Stress Determination

Figure 6 shows the Raman spectrum obtained from the
oxide scale formed on the surface of NiCoCrAlY alloys
oxidized at 900C and 1000C for 16h. The only band
which is well-defined in both spectra is the A1g mode at
549 cm-1 in Mougin et al.s work [11]. In the scales, this
mode shifts to 549.53 cm-1 for 900C specimen and
554.34 cm-1 for 1000C specimen respectively, resulting
in the observed shifts are equal to 0.53 cm-1 and 5.34
cm-1 respectively. Using the law given by Mougin et al.
[11] for the frequency dependence with pressure, i.e.,
0.307 0.005 GPa/cm-1, it gives the stress values of
0.163 0.005 GPa for 900C specimen and 1.639
0.005 GPa for 1000C specimen respectively. The shift
direction corresponds to compressive stress.
The results exhibited here agree with the previous discussion. The higher growth rate of Cr2O3 at 1000C resulted in a higher stress than the stress generated at a lower
temperature. The structure under high stress condition was
easier to crack, spall and fracture and was more difficult
to self-healing by the Al2O3 growth simultaneously.
Figure 6. Raman spectrum for the chromia formed on NiCoCrAlY alloys oxidized for 16 h at 900C and 1000C.
3.4. The Effect of the Preoxidation Treatment on

the Oxidation Resistance
Due to the different growth characteristics of -Al2O3
and Cr2O3 at different temperatures at initial stage of
oxidation, two groups of specimen were conducted. One
was directly oxidized at 1000C for 200 hours. The other
one was subjected to a preoxidation treatment at 900C
for 16 hours first, aiming to repair micro-cracks in the
Cr2O3 scale by subsequent growth of -Al2O3, then oxidized at 1000C for 184 hours. The oxidation behavior of
the specimens was evaluated by a cyclic oxidation test.
Figure 7 represents the weight gain as a function of
time for the cycle oxidation at 1000C. In the figure,
curve (a) indicates the specimen directly oxidized at
1000C for 200 hours and curve (b) indicates the specimen which preoxidixed at 900C. At the onset of the
oxidation, the rate of weight gain of the specimen preoxidized at 900C was slower than that of 1000C. After
the sharp increase of weight gain at the initial oxidation
stage, both of the kinetic curves showed an extensive
period of very slow weight gain. Obviously, the alloy
which preoxidised at 900C showed lower weight gains
than that of 1000C.
With oxidation depth increased, oxygen activity reduced unceasingly, Al2O3 precipitates would nuclear in
subsurface of the alloy but no longer for Cr2O3. This
could explain that why the relative quantity of -Al2O3 at
900C was larger than that at 1000C. At 1000C, a great
amount of Cr2O3 formed by the contact between Cr irons
and the oxygen through the severe cracks.
4. Conclusions
Initial oxidation tests of Ni-6.5Co-17.8Cr-3.7Al-0.5Y
alloy specimens was performed at 800C, 900C and
1000C for 16 hours. Cr2O3 was the predominant phase
at all three temperatures and the dense Cr2O3 scale play-
Figure 7. Weight gain of specimens vs. time in the cyclic

oxidation at 1000C for 200 h. (a) directly oxidized at
1000C; (b) preoxidation treatment at 900C for 16 h.
ENG
C. ZHU
ed an important role in protecting against cracking and

oxidation in the first 16 hours of the isothermal oxidation
at 800C. YAlO3 phase was only observed at 800C.
That the growth rate of -Al2O3 was similar to that of
Cr2O3 at 900C lead to the -Al2O3 could grow and fill in
the ditches on the Cr2O3 scale. However, the growth rate
of Cr2O3 at 1000C was much larger and produced higher
stress than that at 900C, so that the -Al2O3 grown from
the Cr2O3 oxide ditches was not enough to fill in these
cracks. Though the spallation on the surface of oxidized
specimen was unavoidable at the higher temperature
(900C and 1000C), a preoxidation treatment at 900C
for 16 hours can cause -Al2O3-dispersions-in-Cr2O3
scale formed on the surface of the specimen. This could
improve the oxidation resistance of NiCoCrAlY alloy in
the thermal cycling oxidation.
5. Acknowledgements
The project was supported by the State Key Development Program for Basic Research of China (Grant No.
2007CB707700).
6. References
[1]
N. P. Padture, M. Gell and E. H. Jordan, Thermal Barrier Coatings for Gas-Turbine Engine Applications, Science, Vol. 296, No. 5566, 2002, pp. 280-284.
[2]
A. G. Evans, D. R. Mumm, J. W. Hutchinson, G. H.

Meier and F. S. Pettit, Mechanisms Controlling the
Durability of Thermal Barrier Coatings, Progress in
Materials Science, Vol. 46, No. 5, 2001, pp. 505-553.
Y. Fengling and T. D. Bennett, Phase of Thermal
Emission Spectroscopy for Properties Measurements of
Delaminating Thermal Barrier Coatings, Journal of
Applied Physics, Vol. 98, No. 10, 2005, pp. 103501103508.
R. Panat, S. L. Zhang and K. J. Hsia, Bond Coat Surface
Rumpling in Thermal Barrier Coatings, Acta Materialia,
Vol. 51, No. 1, 2003, pp. 239-249.
B. Wang, J. Gong, A. Y. Wang, C. Sun, R. F. Huang and
L. S. Wen, Oxidation Behaviour of Nicraly Coatings on
Ni-Based Superalloy, Surface & Coatings Technology,
Vol. 149, No. 1, 2002, pp. 70-75.
U. Schulz, C. Leyens, K. Fritscher, M. Peters, B.
Saruhan-Brings, O. Lavigne, J. M. Dorvaux, M. Poulain,
R. Mevrel and M. L. Caliez, Some Recent Trends in
Research and Technology of Advanced Thermal Barrier
[3]
[4]
[5]
[6]
ET AL.
607
Coatings, Aerospace Science and Technology, Vol. 7,

No. 1, 2003, pp. 73-80..
[7]
R. A. Mahesh, R. Jayaganthan and S. Prakash, Microstructural Characteristics and Mechanical Properties of

HVOF Sprayed Nicral Coating on Superalloys, Journal of Alloys and Compounds, Vol. 468, No. 1-2, 2009,
pp. 392-405.
[8] T. F. An, H. R. Guan, X. F. Sun and Z. Q. Hu Effect of
the Theta-Alpha-Al2O3 Transformation in Scales on the
Oxidation Behavior of a Nickel-Base Superalloy with an
Aluminide Diffusion Coating, Oxidation of Metals, Vol.
54, No. 3-4, 2000, pp. 301-316.
[9] S. O. Moussa and K. Morsi, High-Temperature Oxidation of Reactively Processed Nickel Aluminide Intermetallics, Journal of Alloys and Compounds, Vol. 426,
No. 1-2, 2006, pp. 136-143.
[10] T. J. Nijdam, N. M. van der Pers and W. G. Sloof, Oxide
Phase Development upon High Temperature Oxidation of
Gamma-Nicral Alloys, Materials and Corrosion-Werkstoffe Und Korrosion, Vol. 57, No. 3, 2006, pp. 269-275.
[11] M. Kemdehoundja, J. F. Dinhut, J. L. Grosseau-Poussard
and M. Jeannin, High Temperature Oxidation of Ni70Cr30
Alloy: Determination of Oxidation Kinetics and Stress
Evolution in Chromia Layers, Materials Science and
Engineering A, Vol. 435-436, No. 5, 2006, pp. 666-671.
[12] G. Calvarin, A. M. Huntz, A. Hugot Le Goff, S. Joiret
and M. C. Bernard, Oxide Scale Stress Determination by
Raman Spectroscopy Application to the Nicr/Cr2O3
System and Influence of Yttrium, Scripta Materialia,
Vol. 38, No. 11, 1998, pp. 1649-1658.
[13] P. Kofstadt, High Temperature Corrosion, Elsevier,
New York, 1988.
[14] D. Lee, M. L. Santella, I. M. Anderson and G. M. Pharr
Thermal Aging Effects on the Microstructure and ShortTerm Oxidation Behavior of a Cast Ni3Al Alloy, Intermetallics, Vol. 13, No. 2, 2005, pp. 187-196.
[15] S. Seal, S. C. Kuiry and L. A. Bracho, Surface Chemistry of Oxide Scale on IN-738LC Superalloy: Effect of
Long-Term Exposure in Air at 1173 K, Oxidation of
Metals, Vol. 57, No. 3-4, 2002, pp. 297-322.
[16] S. C. Choi, H. J. Cho and D. B. Lee, Effect of Cr, Co,
and Ti Additions on the High-Temperature Oxidation
Behavior of Ni3Al, Oxidation of Metals, Vol. 46, No.
1-2, 1996, pp. 109-127.
[17] H. H. Angermann, K. Nishi, Y. Aono, M. Inagaki and H.
Kodama, Evolution of Oxides on Ni-Base ODS Superalloys, Oxidation of Metals, Vol. 48, No. 1-2, 1997, pp.
1-39.
ENG
Engineering, 2010, 2, 608-616

Highly Nonlinear Bending-Insensitive Birefringent

Photonic Crystal Fibres
Huseyin Ademgil, Shyqyri Haxha, Fathi AbdelMalek
Broadband and wireless communication group, School of Engineering and Digital Arts, University of Kent,
Canterbury, UK
E-mail: ademgil@hotmail.com
Received May 20, 2010; revised July 21, 2010; accepted July 23, 2010
Abstract
Highly nonlinear birefringent Photonic Crystal Fibre (PCF) that exhibits low losses and small effective mode
area across a wide wavelength range has been presented. The effects of angular orientation on bending losses
of the proposed PCFs have been thoroughly investigated by employing a full vectorial finite element method
(FEM). It has been demonstrated that it is possible to design a bending-insensitive nonlinear PCF with a birefringence in the order of 10-2 and a nonlinear coefficient of 49 W-1km-1 at the wavelength of 1.55 m. Also,
significant improvements on key propagation characteristics of the proposed PCFs have been demonstrated
by carefully altering the desired air hole diameters and the hole-to-hole spacing. It is demonstrated that two
zero dispersion wavelengths can be achieved by the proposed design.
Keywords: Nonlinear Coefficient, Effective Mode Area, Confinement Loss and Birefringence
1. Introduction
Photonic Crystal Fibers consisting of a central defect
region in a regular lattice of air holes have attracted significant research attention. These fibers provide extra
degrees of freedom in manipulating optical properties
[1-2]. PCFs can be divided into two categories according
to the mechanism used to guide the light: photonic-bandgap (PBG) guidance and effective index guidance. The
PBG fibers use a perfectly periodic structure exhibiting a
PBG effect of the crystal lattice at the operating wavelength to guide light in a low-index core region. In PBG
fibres, the core can be created from the lower refractive
index material, which could be solid glass or a large air
hole (in the case of an air-glass PCF) [3-5]. On the other
hand, the effective index-guiding PCFs rely on total internal reflection (TIR) to confine light in the region of
missing air hole forming a central core. The presence of
air holes decreases the effective index of the cladding,
making light guidance possible by TIR. This guiding
method is more analogous to the operation of a conventional step-index fibre [1,3,6].
PCFs have remarkable properties, strongly depending
on the design details such as low sensitivity to bend losses even for high mode areas, where, low or high mode
areas leading to very strong or weak optical nonlinearities. PCF technology, now allows the fabrication of fiCopyright 2010 SciRes.
bers with very tightly confined modes, and thus very

high optical nonlinearities per unit length. Indeed, indexguided PCFs can have nonlinearity 10-100 times that of a
conventional silica fiber [7,8]. Birefringent PCFs can
simply be realised [9] compared to conventional fibres,
since the refractive index contrast between the core and
the cladding is higher than the refractive index contrast
of conventional fibres. Additionally to nonlinearities,
growing interest is being shown in birefringence study in
PCFs. There are different ways of designing birefringent
fibres, such as use of anisotropic materials. However, for
nominally isotropic silica fibres, the usual method is to
create a spatial asymmetry in the index or shape profile
by applying a stress to the fibre [9-11].
Birefringence is used in many sensing applications and
in applications where light is required to maintain a linear polarization state. In this regard, PCFs are considered
to be good candidates for applications requiring high
temperature insensitivity and high birefringence [12-13].
Indeed, PCFs can have birefringence much larger than
that of the conventional PANDA fibres [10]. To increase
the effective index difference between the two orthogonal polarization modes and achieve birefringent PCFs the
structural asymmetry can be achieved by altering the air
hole sizes near the core area [6,14]. Alternatively, by
distorting the shape of the air holes (elliptical air holes)
[11-12] birefringence can be achieved. Previously pubENG
H. ADEMGIL
lished results by Yue et al. [11] and Sun et al. [12] have
demonstrated that it is possible to design PCFs with relatively large birefringence in the order of 10-3-10-2. To our
knowledge proposed PCFs with elliptical air holes [1112] exhibit the highest birefringence to date. However,
fabrication becomes challenging by the use of several
rings of elliptical air holes in cladding region. Moreover,
controlling the elliptical air holes during the fabrication
process might be difficult [10,15].
The design of PCF structures with small mode areas
that lead to high nonlinear coefficient , is an ongoing
challenge. By varying the size of the air holes in the
cladding region and the hole to hole spacing, desired
effective mode areas can be obtained [7,8]. Small core
diameter that leads to low effective mode area can be
reduced by having a relatively small hole to hole spacing.
Previously published results such as Poli et al. [16] and
Saitoh et al. [17] have demonstrated theoretically, that it
is possible to design PCFs with nonlinear coefficients of
about 30 and 44 W-1km-1, respectively, at 1.55 m telecommunication wavelength. However, these structures
are purely theoretical and the hole to hole spacing, , is
around 0.9 m. From the point of view of fabrication,
small hole to hole spacing might be problematic to
manufacture.
In recent years, highly birefringent PCFs with nonlinear properties have received growing attention in telecommunication and supercontiniuum applications [7,1819]. Previously published results by Lee et al. [7] and
Yamamoto et al. [18], have experimentally demonstrated
that it is possible to design highly nonlinear PCFs with a
relatively large birefringence in the order of 10-3 at 1.55
m telecommunication wavelengths. Lee et al. [7] has
demonstrated a birefringent PCF having nonlinear coefficient , of 31 W-1km-1 for the use of optical
code-division multiple access (OCDMA) applications.
Similarly, Yamamoto et al. [18] has demonstrated highly
birefringent PCF with Ge-doped core having nonlinear
coefficient, , of 19 W-1km-1.
Moreover, recently published papers such as Kudlinski
et al. [20] and Cumberland et al. [21] have shown that
PCFs with two zero dispersion wavelengths (ZDW)
demonstrate stronger power spectral densities than single
ZDW PCFs. Therefore PCFs with two ZDW can be beneficial in supercontinuum applications. Kudlinski et al.
[20] have demonstrated that it is possible to design two
ZDW PCF with a nonlinear coefficient of 31 W-1km-1.
For many applications it is essential to design PCFs
that exhibit simultaneous high birefringence, low losses,
and high nonlinear coefficient across a wide wavelength
window. Additionally, bending losses can be a critical
issue in the sensing and communication applications [22].
Bending is one of the important issues regarding the
practical development of PCFs. When an optical fibre is
bent, the field profile deforms outwards in the direction
ET AL.
609
of the bend and radiation losses occur. Since there are

more holes around the core of the holey fiber, the effecttive refractive index can be designed more flexibly than
that of conventional optical fibers by adjusting the hole
diameter and hole to hole spacing [22-23].
Figure 1 presents our proposed design which looks
similar to a design proposed by Saitoh et al. [24]. However, our design differs from this design in a number of
key areas. In our design we have shifted the first row of
air holes outwards by /2. We have also used different
hole to hole spacing and diameters that in combination
improve the birefringence and reduce the confinement
losses. In ref. [24] the birefringence is around 4 10-3
and the losses are around 0.1 dB/km at 1.55 m wavelength. Compared to ref. [24], at the same wavelength,
lower confinement losses ( 0.001 dB/km) and higher
birefringence ( 8 10-3) can be achieved with our proposed design. Additionally, in our design only 5 air hole
rings are used which makes the design less complex and
potentially easier to fabricate.
The main purpose of the proposed PCF structure is to
simultaneously achieve high birefringence, low confinement loss and a high nonlinear coefficient. In this paper,
we propose a novel type of bending-insensitive highly
birefringent nonlinear PCF. High birefringence in PCFs
can be produced by combining the asymmetric core and
the large core-cladding index contrast. As shown in Figure 1, in order to destroy the symmetry of the fiber core,
the first row of the central air hole group is shifted outwards by /2. Additionally, different air-hole diameters
along the two orthogonal axes are used in the core region.
All the air holes in the cladding region have the same
diameter except for the outermost ring which has larger
air holes in order to reduce the confinement losses. The
current progress in PCF (nanophotonics) technology [15,
25], has demonstrated that fabrication of our proposed
Figure 1. Schematic cross section of PCF, dm/ = 0.941,

d/ = 0.588, d5/ = 0.764.
ENG
H. ADEMGIL ET AL.
610
PCF structure is not an issue. Theoretical and experimenttal investigations by Suzuki et al. [10] have shown
that it is possible to fabricate even complex PCF structures by adapting conventional stack and draw methods.
However, stack and draw methods [15] are limited to
closely-packed geometries such as triangular or honeycomb lattices and cannot easily generate different geometries. Alternatively, drilling methods allow great flexibility for both the hole size and spacing, but the structures are generally limited to a small number of holes.
Another alternative fabrication method is to use the solgel method which allows for independent adjustment of
the hole size and spacing. The sol-gel method [15] provides additional design flexibility that will be necessary for such PCF structures. Additionally, recently published results by Vu et al. [22] demonstrated experimenttally, that the fabrication of bending insensitive PCFs is
possible and that these fibers are robust against high
amounts of bending.
In this work, we have employed full vectorial finite
element method (FEM) to investigate key modal properties of the proposed index guided PCF. The modal solution approach based on FEM is more flexible and reliable
than other techniques. It can represent any arbitrary
cross-section more accurately and has been widely used
to find the modal solutions of a wide range of optical
waveguides [6,26]. The FEM formulation for modal
analysis based on anisotropic perfectly matched layers
(PML) is capable of handling as many modes as required
and analyse leaky modes. By using PMLs boundary condition, propagation characteristics and optical properties
of leaky modes in PCFs, it can be precisely evaluated
[27]. The modal analysis has been applied on the crosssection in the x-y plane of the PCF as the wave is propagating in the z direction.
In this study, birefringence, confinement loss, effective mode area, nonlinear coefficient properties of the
proposed PCFs are reported thoroughly. Also, significant
improvements of propagation characteristics of the PCFs
are demonstrated. Following this introduction, a brief
theoretical analysis is provided in Section 2. The simulation results are reported in Section 3 and, finally, conclusions are drawn in the last section.
wave-number in the vacuum, n is the refractive index of

the domain, [s] is the PML matrix, [s]-1 is an inverse matrix of [s] and is the operating wavelength.
2. Theory
where E is the amplitude of the transverse electric field

propagating inside the fibre.
Study of Aeff is thus an important starting point in the
understanding of the nonlinear phenomena in PCFs. Due
to the high index contrast between silica and air, PCF
technology offers the possibility of much tighter mode
confinement and thereby a lower effective mode area
compared to conventional fibres. An important value for
the calculation of the strength of nonlinear effects, is the
ratio between the nonlinear refractive-index coefficient,
n2 (Kerr constant), and the effective area for a given
wavelength of the optical field. The nonlinear coefficient
The PCF cross section of Figure 1, with a finite number

of air holes is divided into homogeneous subspaces
where Maxwells equations are solved by accounting for
the adjacent subspaces. These subspaces are triangles
that allow a good approximation of the circular structures.
Using the anisotropic PML [6-26] from the Maxwell
equations the following vectorial equation is derived:
( E) k02 n 2 E = 0
(1)
where E is the electric field vector, k0 (= 2/) is the

2.1. Confinement and Bending Loss

Due to a finite number of layers of air holes, it is
inevitable that the optical mode will leak from the core
region into the outer air hole region. Considering the fact
that the jacket of the PCF is far from cladding and core
regions, the light guidance in the core region is exclusively due to a finite number of layers of air holes in the
silica extending to infinity. The amount of leakage constitutes the confinement loss. The confinement loss is
calculated from the imaginary part of the complex
effective index, neff using [5,6,26]:
40
Conf ..Loss
Im neff 103 [dB/km] (2)
ln 10
where Im is the imaginary part of the neff.
The calculations of the bending loss were carried out
using the same formulation. We assume a circular bend
structure where the PML is used along the radiation direction (+x direction) for suppressing spurious reflection.
The curved fiber is replaced by a straight fiber with an
equivalent refractive index distribution defined by [9,2223].
x
(3)
neq = n (x,y)exp
R
where n (x,y) is the refractive index profile of the straight
fiber and R represents the bend radius.
2.2. Effective Mode Area and Nonlinear

Coefficient
Another key factor in designing PCFs is the effective
mode area. The effective mode area, Aeff is related to the
effective area of the core area, which is calculated using
[9,26];
Aeff
dxdy
4
E dxdy
(4)
ENG
H. ADEMGIL
is inversely proportional to the effective mode area and

can be calculated from [7,16-18];
2 n2
( )
(5)
Aeff ( )
where n2 is the nonlinear refractive-index coefficient (n2

= 2.76 10-20 m2/W) [27].
2.3. Chromatic Dispersion

Chromatic dispersion is one of the most important modal
properties of the PCFs. It is the main contributor to optical pulse broadening. Chromatic dispersion is caused by
the combined effects of material and waveguide dispersion. Moreover, the chromatic dispersion consists of material dispersion and waveguide dispersion, which can be
calculated from the real part neff values against the wavelength. The material dispersion given by Sellmeiers
formula is directly included in the calculation [2,6].
2
Re neff
(6)
D
c
2
where c is the velocity of light and Re(neff) is the real part
of the neff. Material dispersion refers to the wavelength
dependence of the refractive index of material caused by
the interaction between the optical mode and ions,
molecules or electrons in the material.
3. Simulation Results
In Figure 1, the proposed PCF design with an indexguiding core surrounded by a triangular array of air holes
is presented. The diameter of air holes and hole-to-hole
spacing is denoted by d and , respectively. The refractive index of pure silica is set equal to 1.45. In order to
reduce the confinement losses, five rings of air holes are
considered.
Previously published results by Ortigosa et al. [14]
have shown that, by varying the hole diameters along the
two orthogonal axes high birefringence can be achieved.
Therefore, in our design in order to achieve ultrahigh
birefringence, the air hole diameter sizes, d1, along the
x-direction are increased. To enhance the birefringence
further, the first column of the air hole group is shifted
outwards by /2. As a result, the PCF core becomes
more asymmetrical which results in a significant increase
in the birefringence.
In order to keep the birefringence at the optimum level
and reduce confinement losses, we next investigate the
size of the air holes in the cladding region. It is known
that confinement losses [2,26] can be reduced by increasing the size of the air holes in the inner cladding
area, d. However, according to our simulations this has a
negative effect on the birefringence. Hence, there is a
tradeoff between ultrahigh birefringence and low confinement losses. Alternatively, by increasing the number
ET AL.
611
of air hole rings [26], ultra low confinement losses can

be realized with negligible reduction in the birefringence.
Next, in order to control the dispersion, a dispersion
management technique [2] is also applied to proposed
PCF design. With this technique different air hole diameters are used in each ring to control the chromatic
dispersion across a wide wavelength range. On the other
hand, it is well known that by altering and d, ZDW can
be controlled [20-21]. Also, control of ZDW is much easier when hole to hole spacing is small [2]. In this regard,
the desired properties of birefringence, nonlinearities,
chromatic dispersion and confinement losses have been
simultaneously achieved in the PCF structure configuration, shown in Figure 1, where, dm/ = 0.941, d/ =
0.588 and d5/ = 0.764.
The polarization dependent confinement losses still
need to be evaluated before one can conclude the fiber
structures to be practical. The confinement losses strongly depend on the number of air hole rings, air hole diameter and hole-to-hole spacing. Due to the number of
air hole rings and their diameters used in our proposed
PCF design, confinement losses are minimized. The confinement feature of the mode to the core region is directly linked to how much the mode is leaking into the
outer air hole region.
Our proposed PCF supports the fundamental mode and
some higher-order modes. In order to clarify this, the
confinement losses of fundamental and first order modes
are investigated and presented in Figure 2. These modes
are approximately linearly polarized and, by analogy to
the elliptical core fiber and other asymmetric waveguides,
may be labeled as LP modes, such as in this case the
fundamental LP01 mode which corresponds to an HE11 .
In this regard, variation of confinement loss as a function
of hole-to-hole spacing , when dm/ = 0.941, d/ =
0.588, d5/ = 0.764 at the operating wavelength = 1.55
m, is shown in Figure 3. The confinement losses for both
Figure 2. Variation of Confinement losses as a function of

the hole to hole spacing, , where = 1.55 m.
ENG
612
H. ADEMGIL ET AL.
Figure 3. Variation effective mode area, Aeff of the fundamental mode as a function of the operating wavelength.
Figure 4. Variation of nonlinear coefficient , as a function

of the operating wavelength.
fundamental and first order modes reduce with increasing hole-to-hole spacing.
As expected, losses of higher order modes are much
higher than the fundamental mode. It is worth noting that,
confinement loss for y-polarized mode of both fundamental and first order mode are higher than x-polarized
mode. In this study we mainly concentrate on the behavior of the fundamental modes. Therefore, further analysis of propagation properties focuses on the fundamental
modes.
Figure 3 shows the variation of effective mode area as
a function of wavelength. It can be noted that Aeff is increasing with the increasing hole to hole spacing. The
effective mode area steadily increases when the wavelength increases. It is worth noting that the effective area
is much smaller than that of the conventional fibres at
1.55 m wavelength. This would contribute to increase
the nonlinearities produced by refractive index power dependence [1].
Having the freedom to control the optical properties of
the PCF by hole to hole spacing and placement whilst
maintaining strong confinement of the mode, allows for
the realization of high nonlinear effects. With ultra-high
nonlinearities, we can generate supercontinuum with relatively low pumping power. This is a very important
advantage. The devices can be made smaller, cheaper
and become more portable [28].
Variation of the nonlinear coefficient as a function of
wavelength is presented in Figure 4. As presented in
Equation (5) the nonlinear coefficient is inversely proportional to the effective area. Small effective mode area
leads to high nonlinear coefficient that would be useful
in the context of supercontiniuum generation and soliton
pulse transmission [7,9]. In this regard, the nonlinear
coefficient steadily increases when the wavelength and
hole to hole spacing, decreases. Our design shows that
the nonlinear coefficient, , for = 1.7 m and = 2 m
at 1.55 m operating wavelength is 26 W-1km-1 and 20
W-1km-1, respectively.
Due to the low effective mode area, our proposed PCF

is expected to be insensitive to bending. Also it is known
that, low effective mode area has a positive effect on
bending loss [9]. Moreover, the impact of angular orientation on bending losses is a critical issue in PCFs.
Figure 5 shows variation of the confinement losses as
a function of bending radius at different angular orientations of the fibre with respect to the bending plane,
where, = 1.9 m at operating wavelength 1.55 m. As
can be seen from figure, three angular orientations = 0,
= 15 and = 30 are investigated and these orientations has a critical effect on the proposed PCF losses
when the fibre is bent. As the bending radius increases,
the effect of increases. According to our simulations,
the effect of angular orientation on confinement losses is
related to core size. As expected, when increases, confinement losses increase marginally. However, the effect
of angular orientation on losses is similar for all values
of bending radii. One can see that, as hole to hole spacing increases (core size increases) the effect of angular
Figure 5. Variation of confinement losses of the HE11x mode

as a function of the bending radius, R, for three different
angular orientation, when = 1.9 m and = 1.55 m.
ENG
H. ADEMGIL
orientation becomes the same for all bending radii. This

phenomenon can be linked to the effect of the core size.
On the other hand, due to leaky nature of the fibre, the
effective mode area plays a key role in confinement. It is
well known that small mode areas are usually the consequence of strong guiding where bend losses and other
effects of external disturbances are weak. Therefore, in
our case it is evident that low effective mode areas minimize the bending effects on the confinement losses. In
other words, our PCF design is insensitive to bending.
Recently, published results by Vu et al. [22] demonstrated experimentally, that fibres can be bent up to 3
mm radius. In our design, fibre size compared to their
design [22] is much smaller and for this reason our proposed fibre can be anticipated to be more flexible and
may be bent further.
Next, we have investigated birefringence properties of
the proposed structure. Our simulated results indicate
that the effective index of the HE11x mode is larger than
that of the HE11y mode. Figure 6 illustrates the variation of the modal birefringence as a function of wavelength for different hole-to-hole spacing. As can be observed from this figure, relatively large birefringence of
the order of 10-3-10-2 is achieved. It can clearly be seen
that the birefringence is sensitive to the varying wavelength . It can be anticipated that as hole-to-hole spacing decreases, the birefringence increases. It can be
noted that birefringence for = 1.7 m and = 2 m at
1.55 m operating wavelength is 9 10-3 and 7.3 10-3,
respectively. Highly birefringent PCFs provide several
advantages for supercontinuum generation. Specifically,
all the spectral components exhibit the same linear polarization and also the power required to generate the
continuum is reduced compared to non-birefringent
PCFs. In addition, the fibre allows for simultaneous generations of two different continua due to the large difference between two polarization modes [19].
Figure 6. Variation of birefringence as a function of the

wavelength.
ET AL.
613
As can be seen from Figure 3, effective mode area can

be minimized by reducing the hole to hole spacing.
Therefore, nonlinear coefficient of proposed PCF can be
improved by reducing the hole to hole spacing. Moreover,
significant increase on birefringence can be observed by
reducing the hole to hole spacing. In this regard, in order
to improve the birefringence and nonlinear coefficient,
the proposed PCF is investigated for different design
parameters (smaller hole to hole spacing). The nonlinear
coefficient of the proposed PCF is illustrated in Figure 7.
It can be seen that, the nonlinear coefficient steadily increases when the wavelength and hole to hole spacing, ,
decreases. Our simulations show that, the highest nonlinear coefficient corresponding to effective area, Aeff =
2.28 m2 at = 1.55 m, is equal to 49 W-1 km-1 for =
1 m.
Next, we have investigated the birefringence properties of the proposed PCF, shown Figure 8. Our simulated
results indicate that the effective index of the HE11x
mode is larger than that of the HE11y mode. As can be
observed from this figure, relatively high birefringence
Figure 7. Variation of nonlinear coefficient , as a function

of the operating wavelength.
Figure 8. Variation of birefringence as a function of the

wavelength.
ENG
614
H. ADEMGIL ET AL.
of the order of 10-2 is achieved. The birefringence is sensitive to the varying wavelength, and it increases as the
wavelength increases. It can be noted that the birefringence for = 1 m and = 1.6 m at 1.55 m operating
wavelength is 2.65 10-2 and 1.01 10-2, respectively.
Finally, the chromatic dispersion profile can be easily
controlled by varying the hole diameter and the hole to
hole spacing [28]. Controllability of chromatic dispersion
in PCFs is a very important problem for practical applications to optical communication systems, dispersion
compensation, and nonlinear optics [12]. At short wavelength, the modal field remains confined to the silica
region, but at longer wavelengths the effective cladding
index decreases. Thus, as we change the size of air hole d
or the separation between them , ZDW can be altered
to any value. This unusual dispersion characteristic of
PCFs allows them to be used in non-linear fiber optics.
One can see that it is possible to shift the zero dispersion
wavelength from visible to near-infrared (IR) regions by
appropriately changing the geometrical parameters (d
and ).
At short wavelength, the modal field remains confined
to the silica region, but at longer wavelengths the effective cladding index decreases. Thus, as we change the
size of air hole d or the separation between them ,
ZDW can be altered to any value. This unusual dispersion characteristic of PCFs allows them to be used in
non-linear fiber optics. One can see that it is possible to
shift the zero dispersion wavelength from visible to
near-infrared (IR) regions by appropriately changing the
geometrical parameters (d and ). As may be seen from
Figure 9 when hole to hole spacing, = 1 m and =
1.2 m the proposed PCF has a single ZDW, 0.8 m and
0.84 m respectively. On the other hand, when = 1 m
and = 1.2 m two ZDW is achieved. The first ZDW
for both cases is around 0.8 m. However, according to
simulation results the second ZDW for = 1 m and
Figure 9. Variation of the chromatic dispersion of HE11x

modes as a function of the wavelength for different hole to
hole spacing.
= 1.2 m is found as 1.36 m and 1.67 m, respectively.

PCFs that have two ZDW have been used previously for
high power supercontinuum generation applications
[20-21]. Also, Cumberland et al. [21] have shown that
two ZDW PCFs can be used to control the long wavelength edge of the continuum when needed for specific
applications.
In nonlinear optics, to maximize the spectral broadening, it is advantageous to have a polarization maintaining
(PM) nonlinear fiber (birefringent nonlinear fiber).
Pumping a PM fiber with the pump source polarization
aligned to one of the principle axes in the fiber yields a
power advantage close to a factor of two compared to a
non-PM fiber. Moreover, the output from the fiber is also
polarized, increasing the usability of the generated light.
Therefore, highly birefringent nonlinear PCFs can be
useful in SC and nonlinear applications.
The birefringence and nonlinear coefficient properties
of the proposed PCF reported in this paper are much larger than that of the conventional fibres. These fibres are
useful to improve the capabilities of optical fibre communication systems and new types of optoelectronic devices. Nonlinear PCFs with highly birefringence and low
confinement losses can be widely used for polarization
control in fibre-optic sensors, precision optical instruments, ultra-short solution pulse transmission and fourwave mixing [8,18-19]. Moreover, reported results can
be useful for optical communication systems [10], optical switching and OCDMA applications [7]. From experimental point of view, the sol-gel fabrication method
offers flexible design freedom with such a lattice structure and is robust against high degrees of bending. This
fabrication method allows the experienced manufacturer
to produce low cost highly advanced PCF structures tailored to the desired propagation properties.
4. Conclusions
In summary, we have presented a highly nonlinear birefringent PCF. Simultaneous, birefringence, and nonlinear
(coefficient) properties of the proposed PCF have been
reported in this paper that to the best of our knowledge,
are much higher than any other results published so far in
literature. Moreover, two ZDW that is beneficial for supercontiniuum applications has been achieved. Also, it is
shown that a low effective area has a positive effect on
the bending losses and the proposed structure is bending
insensitive. The proposed PCF structure configuration is
straightforward when compared to many fabricated PCF
structures in literature. Therefore, fabrication of the proposed PCFs is believed to be possible and is not beyond
the realms of todays existing PCF technology. These
reported results can be widely used for the supercontinuum generation, polarization control in fiber-optic
sensors and telecommunication applications.
ENG
H. ADEMGIL
5. References
ET AL.
615
Photonic Crystal Fibers, Optics Express, Vol. 12, No. 24,

2004, pp. 5931-5939.
[1]
J, C. Knight, T, A. Birks and P. S. J. Russell, All-Silica

Single-Mode Optical Fibre with Photonic Crystal Cladding, Optics Letter, Vol. 21, No. 19, 1996, pp. 15471549.
[14] A. Ortigosa-Blanch, J. C. Knight, W. J. Wadsworth, J.

Arriaga, B. J. Mangan, T. A. Birks, P. St and J. Russell,
Highly Birefringent Photonic Crystal Fibers, Optics
Letter, Vol. 25, No. 18, 2000, pp. 1325-1327.
[2]
G, Renversez, B. Kuhlmey and R. McPhedran, Dispersion Management with Microstructured Optical Fibers:
Ultraflattened Chromatic Dispersion with Low Losses,
Optics Letter, Vol. 28, No. 12, 2003, pp. 989-991.
[3]
J. Broeng, D. Mogilevstev, S. E. Barkou and A. Bjarklev,

Photonic Crystal Fibres: A New Class of Optical Waveguides, Optical Fiber Technology, Vol. 5, No. 3, 1999,
pp. 305-330.
[15] R. T. Bise and D. J. Trevor, Sol-Gel Derived MicroStructured Fiber: Fabrication and Characterization, Optical Society of America, Optical Fiber Communications
Conference (OFC), Washington, DC, Vol. 3, March
2005.
[4]
[5]
[6]
[7]
[8]
[9]
A. Birks, J. C. Knight, B. J. Mangan and P. St. J. Russell,

Photonic Crystal Fibres: An Endless Variety, IEICE
Transactions on Electronics, Vol. E84-C, 2001, pp. 585592.
K. Saitoh and M. Koshiba, Leakage Loss and Group
Velocity Dispersion in Air-Core Photonic Band-Gap Fibres, Optics Express, Vol. 11, No. 23, 2003, pp. 31003109.
H. Ademgil and S. Haxha, Highly Birefringent Photonic
Crystal Fibres with Ultra-Low Chromatic Dispersion and
Low Confinement Losses, IEEE Journal of Lightwave
Technology, Vol. 26, No. 4, 2008, pp. 441-448.
J. H. Lee, P. C. Teh, Z. Yusoff, M. Ibsen, W. Belardi, T.
M. Monro and D. J. Richardson, A Holey Fiber-Based
Nonlinear Thresholding Device for Optical CDMA Receiver Performance Enhancement, IEEE Photonics
Technology Letter, Vol. 14, No. 6, 2002, pp. 876-878.
A. V. Yulin, D. V. Skryabin and P. S. J. Russell, FourWave Mixing of Linear Waves and Solutions in Fibers
with Higher-Order Dispersion, Optics Letter, Vol. 29,
No. 20, 2004, pp. 2411-2413.
T. Nasilowski, P. Lesiak, R. Kotynski, M. Antkowiak, A.
Fernandez, F. Berghmans and H. Thienpont, Birefringent Photonic Crystal Fiber as a Multi Parameter Sensor,
Proceedings Symposium IEEE/LEOS, Benelux Chapter,
Enschede, 2003, pp. 29-32.
[10] K. Suzuki, H. Kubota, S. Kawanishi, M. Tanaka and M.

Fujita, Optical Properties of Low Loss Polarization
Maintaining Photonic Crystal Fibre, Optics Express, Vol.
9, No.13, 2001, pp. 676-680.
[11] Y. Yue, G. Kai, Z. Wang, T. Sun, L. Jin, Y. Lu, C. Zhang,
J. Liu, Y. Li, Y. Liu, S. Yuan and X. Dong, Highly Birefringent Elliptic-Hole Photonic Crystal Fibre with
Squeezed Hexagonal Lattice, Optics Letter, Vol. 32, No.
5, 2007, pp. 469-471.
[16] F. Poli, A. Cucinotta, S. Selleri and A. H. Bouk, Tailoring of Flattened Dispersion in Highly Nonlinear Photonic
Crystal Fibers, IEEE Photonics Technology Letter, Vol.
16, No. 4, 2004, pp. 1065-1067.
[17] K. Saitoh and M. Koshiba, Highly Nonlinear Dispersion-Flattened Photonic Crystal Fibers for Supercontinuum Generation in a Telecommunication Window,
Optics Express, Vol. 12, No. 10, 2004, pp. 2027-2032.
[18] T. Yamamoto, H. Kubota, S. Kawanishi, M. Tanaka and
S. Yamaguchi, Supercontinuum Generation at 1.55 M in
a Dispersion-Flattened Polarization-Maintaining Photonic
Crystal Fiber, Optics Express, Vol. 11, No. 13, 2003, pp.
1537-1540.
[19] M. Lehtonen, G. Genty, M. Kaivola and H. Ludvigsen,
Supercontinuum Generation in a Highly Birefringent
Microstructured Fiber, Applied Physics Letter, Vol. 82,
No. 14, 2003, pp. 2197-2199.
[20] A. Kudlinski, B. A. Cumberland, J. C. Travers, G.
Bouwmans, Y. Quiquempois and A. Mussot, CW Superconttinuum Generation in Photonic Crystal Fibres
with Two Zero-Dispersion Wavelength, AIP Conference
Proceedings, Sao Pedro, August 2008, pp. 15-18.
[21] B. A. Cumberland, J. C. Travers, S. V. Popov and J. R.
Taylor, 29 W High Power CW Supercontinuum
Source, Optics Express, Vol. 16, No. 8, 2008, pp. 59545962.
[22] N. H. Vu, I. K. Hwang and Y. H. Lee, Bending Loss
Analyses of Photonic Crystal Fibers Based on the FiniteDifference Time-Domain Method, Optics Letter, Vol. 33,
No. 2, 2008, pp. 119-121.
[23] T. Martynkien, J. Olszewski, M. Szpulak, G. Golojuch,
W. Urbanczyk, T. Nasilowski, F. Berghmans and H.
Thienpont, Experimental Investigations of Bending Loss
Oscillations in Large Mode Area Photonic Crystal Fibers, Optics Express, Vol. 15, No. 21, 2007, pp. 1354713556.
[24] K. Saitoh and M. Koshiba Single-Polarization SingleMode Photonic Crystal Fibers, IEEE Photonics Technology Letter, Vol. 15, No. 10, 2003, pp. 1384-1386.
[12] Y. S. Sun, Y.-F. Chau, H.-H. Yeh, L.-F. Shen, T.-J. Yang
and D. P. Tsai, High Birefringence Photonic Crystal Fiber with Complex Unit Cell of Asymmetry Elliptical Air
Holes Cladding, Applied Optics, Vol. 46, No. 22, 2007,
pp. 5276-5281.
[25] N. A. Issa, M. A. van Eijkelenborg, M. Fellew, F. Cox, G.

Henry and M. C. J. Large, Fabrication and Study of Microstructured Optical Fibers with Elliptical Holes, Optics Letter, Vol. 29, No. 12, 2004, pp. 1336-1338.
[13] T. Ritari, H. Ludvigsen, M. Wegmuller, M. Legr, N.

Gisin, J. R. Folkenberg and M. D. Nielsen, Experimental
Study of Polarization Properties of Highly Birefringent
[26] S. Haxha and H. Ademgil, Novel Design of Photonic

Crystal Fibres with Low Confinement Losses, Nearly
Zero Ultra-Flatted Chromatic Dispersion, Negative Chro-
ENG
616
H. ADEMGIL ET AL.
matic Dispersion and Improved Effective Mode Area,

Journal of Optics Communications, Vol. 281, No. 2, 2008,
pp. 278-286.
[27] T. Kato, Y. Suetsugu and M. Nishimura Estimation of
Nonlinear Refractive Index in Various Silica-Based Glasses for Optical Fibers, Optics Letter, Vol. 20, No. 22,
1995, pp. 2279-2281.
[28] P.-A. Champert, V. Couderc, P. Leproux, S. Fvrier, V.

Tombelaine, L. Labont, P. Roy, C. Froehly and P. Nrin,
White-Light Supercontinuum Generation in Normally
Dispersive Optical Fiber Using Original Multi-Wavelength Pumping System, Optics Express, Vol. 12, No. 19,
2004, pp. 4366-4371.
ENG
Engineering, 2010, 2, 617-624

Progress in Antimonide Based III-V Compound

Semiconductors and Devices
Chao Liu, Yanbo Li, Yiping Zeng
Key Laboratory of Semiconductor Materials Science, Institute of Semiconductors, Chinese Academy of Sciences,
Beijing, China
E-mail: cliu@semi.ac.cn
Received December 2, 2009; revised February 11, 2010; accepted February 15, 2010
Abstract
In recent years, the narrow bandgap antimonide based compound semiconductors (ABCS) are widely regarded as the first candidate materials for fabrication of the third generation infrared photon detectors and
integrated circuits with ultra-high speed and ultra-low power consumption. Due to their unique bandgap
structure and physical properties, it makes a vast space to develop various novel devices, and becomes a hot
research area in many developed countries such as USA, Japan, Germany and Israel etc. Research progress
in the preparation and application of ABCS materials, existing problems and some latest results are briefly
introduced.
Keywords: Antimonide Based Compound Semiconductors (ABCS), IR Laser, IR Detector, Integrated Circuit,
Functional Device
1. Introduction
Antimonide based compound semiconductors (ABCS)
mainly refer to the antimonide based binary, ternary and
quaternary compound semiconductor materials, consisting of the III-group elements (Ga, In, Al, etc.) and Sb, As
and other V-group elements, such as GaSb, InSb, AlGaSb, InAsSb, AlGaAsSb, InGaAsSb and so on. Their
crystal lattices are around 6.1 and they together with
the InAs-based materials have been routinely called the
6.1 III-V family materials. Antimonide based semiconductors with narrow bandgap as the basic feature, in
the condition of lattice matched or nearly matched with
strain with GaSb, InAs, InP and other commonly used
substrates, their bandgap can be adjusted in a wide range
coveraging from near-infrared wavelength 0.78 m (AlSb)
to far-infrared spectral regions 12 m (InAsSb). The
heterojunctions formed between them can have type-I,
type-II staggered and type-II misaligned band lineups.
The unique band structure and excellent physical properties of ABCS based materials provide great freedom and
flexibility for band engineering and structural design of
materials and create a broad space for development of
high-performance microelectronics, opto-electronic devices and integrated circuits. Applications could include
active-array space-based radar, satellite communications,
ultra-high-speed and ultra-low power integrated circuits,

portable mobile devices, gas environmental monitoring,
chemical detection, bio-medical diagnosis, drug analysis
and other fields [1-8].
2. The Physical Properties and Preparation

Technology of ABCS Based Materials
The in-depth study of antimonide based semiconductor
materials and devices applications was rapidly developed
in recent ten years. Especially after the antimonide based
compound semiconductors program (ABCS program) [9]
was launched by Defense Advanced Research Projects
Agency (DARPA) of USA in 2001, a series of important
developments and breakthroughs have been made in the
study of antimonide based microstructure materials and
device applications worldwide. The narrow bandgap antimonide based compound semiconductors are widely
regarded as the first candidate materials for fabrication of
the third generation infrared photon detectors and integrated circuits with ultra-high speed and ultra-low power
consumption and also as the important materials for
middle and far infrared quantum cascade lasers and
thermophotovoltaic cells suitable for medium and low
temperature heat sources.
The comparison of physical properties of III-V compound semiconductors (at RT) is showed in Table 1. We
ENG
C. LIU
618
ET AL.
Table 1. Comparison of physical properties of III-V compound semiconductors (at RT).

Physical properties
InSb
GaSb
AlSb
InAs
GaAs
InP
GaN
Energy gap/(eV)
0.18
0.70
1.63
0.36
1.42
1.35
3.39
Electron mobility/
(cm2/V.s)
8104
5 000
200
3104
8 500
5 400
900
Electron saturation velocity

( 107cm/s)
4.0
4.0
1.0
1.0
2.7
226
194
80
0.067
0.082 (L)
0.45 (H)
0.077
0.12 (L)
0.55 (H)
0.2
0.6
Electron mean free path

length/nm
0.014
0.018 (L)
0.4 (H)
0.042
0.4
0.12
0.98
0.024
0.025 (L)
0.37 (H)
Thermal conductivity
(W/cm.K)
0.15
0.4
0.7
0.27
0.5
0.7
1.3
Relative dielectric constant
17.9
15.7
12.04
15.1
12.8
12.5
Effective mass
(m0)
Electron
Hole
can see that ABCS have excellent physical properties.

For example the InSb has the smallest bandgap, the
smallest effective mass of carriers, the largest electronic
saturation drift velocity and mobility of any III-V compound semiconductor materials. The relationship between energy gap & spectral wavelength and lattice constant is shown in Figure 1 which also shows the evolution of HEMTs and HBTs transistors for higher frequencies and lower power operation. The relative position
between energy gap and band offset of III-V semiconductors is shown in Figure 2. Thus it can be seen that
there is a considerable band offset and a rich structure of
the energy band alignment in the ABCS heterojunctions.
By regulating the compositions of ABCS multiple compounds, it is convenient to carry out the bandgap engineering of novel devices in the condition of the lattice
match or the strained match.
Antimonide based compound semiconductors can
generally be divided into bulk crystals and film materials.
The most common bulk crystals are GaSbInSb and
InAs. Due to the relatively low melting point of GaSb
and InSb, i.e., 712 and 525 respectively, no dissociation near melting point temperature and small saturation vapor pressure, they can be prepared using the horizontal Bridgman growth of zone melting or vertical
drawn VP method which is similar to the growth of Ge
bulk crystal. While the InAs (melting point 943) bulk
crystal can be grown using liquid covering Czochralski
(LEC) Pulling Method or vertical gradient freeze (VGF)
method which is similar to the growth of GaAs bulk
crystal. Because of their small bandgap, at room temperature, ABCSs intrinsic carrier concentration are too
high to get high resistivity (semi-insulating) substrate
materials which is a serious impediment to the ABCSs
applications in the field of microelectronic devices. At
present the ultra-high pure InSb bulk crystals carrier
concentration can be less than 1013/cm3 and the residual
hole concentration of GaSb bulk crystals is about 2
1016/cm3. Because the growth process is very immature

and there is immiscible gap in the multi-elements antimonide, the ternary, quaternary antimonide bulk crystal
materials are rarely used.
The commonly used methods for preparation of antimonide film materials are liquid phase epitaxy (LPE),
molecular beam epitaxy (MBE) and metal organic
chemical vapor deposition (MOCVD or OMVPE). LPE
method has the advantages of relatively simple process,
less expensive epitaxial equipment, high utilization rate
of the source material, high crystalline quality of the epitaxial films, fast growing, particularly suitable for the
preparation of thick-film materials and so on. LPE
method is a near-thermodynamic equilibrium growth
technology, and therefore can not be used for the growth
of the metastable ternary, quaternary antimonide materials whose components in the immiscible gap. Its growth
rate is generally higher than MOCVD and MBE, and
changes from different substrate crystalline phases with
the typical growth rate of 100 nm/min to a few m/min.
The weakness of LEP is that it can not be used for precision controlled growth of very thin films of nano-scale.
That is to say that it is not applicable to the growth of
superlattices or quantum-well devices and other complex
micro-structure materials. In addition, the morphology of
materials grown by LPE is usually worse than that grown
by MOCVD or MBE. In recent years, a new method
which combines LPE with Zn diffusion technology for
low-cost, high efficient GaSb based InGaAsSb homogeneous pn junction thermophotovoltaic (TPV) cells has
been developed [8]. This method first grows lattice
matched n-In0.15Ga0.85As0.17Sb0.83 (0.55 eV) epitaxial
layer on the Te-doped n-type GaSb substrate associated
with the LPE, then forms the pn homojunction in the
InGaAsSb layer using Zn diffusion method. The external
quantum efficiency of the TPV is as high as 90% at 2 m
radiation wavelength and the cut-off wavelength is 2.3
m, very close to the technical parameters of materiENG
C. LIU
ET AL.
619
Figure 1. Energy gap & spectral wavelength versus lattice constant, showing the evolution of HEMTs and HBTs transistors
for high-frequency and low-power operation [1].
Figure 2. Relative position between energy gap and band

offset of III-V semiconductors.
als grown by MOCVD or MBE. In addition, LPE method

is also used to grow materials of mid-infrared InGaAsSb,
InSb-based infrared detectors, LED and LD. It is a relatively mature, high efficiency, low cost growth technology which is easy to realize the industrialization.
Both MOCVD and MBE are low temperature epitaxial
growth technology of non-thermodynamic equilibrium.
You can grow almost all compositions of the multielements compound thin films including the ternary,
quaternary antimonide which is in the immiscible-gap
and in the metastable state. Both of them can be used for
growth of complex micro-structural materials of ultrathin layers and is very suitable for development of new
optoelectronic devices and circuits. Antimonide based
materials grown by either MOCVD or MBE have their
own characteristics. For a specific device structure, it is
still hard to judge which growth method used for growth
of the device structure is better. In general, MOCVD is
suitable for mass production of epitaxial materials whose
device structure is relatively mature and easy to expand
the size and production capacity. While the MBE is more
suitable for research and development of the novel epitaxial materials with hyperfine and complex structures.
Although production-based MBE equipment has been
developed, it is still not economical using the MBE for
mass production when considering the cost.
The first epitaxial growth of antimonides thin film

materials using MOCVD was done by Manasevit and
Simpson in 1969 who used TMGa and SbH3 (stibine)
source for growing GaSb films [4]. Different from epitaxial materials grown by MBE, The types of metallorganics have a critical influence on the quality of epitaxial
materials grown by MOCVD. The commonly used
III-group metal-organic sources by MOCVD for antimonide based compounds are 3-methyl compound and
3-ethyl compound, such as: TMGa, TMIn, TMAl, TEGa,
TEIn, etc. The commonly used V-group sources are
TMSb, AsH3, PH3, TMBi and RF-N2, etc. Antimonides
are generally low melting point materials and the temperature of epitaxial substrate is generally about 500.
In addition to TMIns lower decomposition temperature
(250-300), the majority of III-group metal-organic
sources can not be completely decomposed below 500.
Therefore, to growing InSb material whose melting point
is only 525, new organic source material with a lower
decomposition temperature must be adopted. At present
the new organic sources which have been successfully
applied for growing antimonides by MOCVD are:
TDMASb (trisdimethylaminoantimony, decomposition
temperature < 300), TBDMSb, TASb (triallyantimony), TMAA (trimethylamine alane), TTBAl (tritertiarybutylaluminum), EDMAA (ethyldimethylaminealane)
and so on. In addition, because of the lack of room temperature chemical stabilized antimony hydride (SbH3),
when growing Al-containing antimonide materials (such
as: AlSb, AlGaSb, AlGaAsSb, etc.), it is easy to appear
carbon and oxygen contamination problem. This phenomenon may be related to the lack of active hydrogen
atoms on the surface of epitaxial materials in which C is
general for p-type doping. Even if the Al content in the
alloy is only 20%, the doping concentrations of C and O
can reach more than 1 1018/cm3 in the epitaxial film.
This causes certain difficulties in growing n-type doping
Al-containing antimonide films. The presence of high
concentration of O impurity in Al-containing antimonide
materials will make these materials have the semiENG
620
C. LIU ET AL.
insulating properties and difficult to measure their electrical properties. The origin of O impurity is very complex, and both the purity of the metal organic sources and
the epitaxial environment and process conditions are
closely related. The development of new organic aluminum source such as TMAA, TTBAl, EDMAA etc. is
precisely in order to inhibit the serious C contamination
problem [4-5]. Thus, growing AlSb and their multielement materials using MOCVD is the most challenging
work in all the III-V epitaxial materials technologies.
The epitaxial growth of antimonide materials using
MBE was following earlier pioneering work of the IBM
group of L.L Chang and L. Esak, first on InAs/GaSb and
InAs/AlSb films [3]. Different from MOCVD process,
MBE uses ultra-high vacuum epitaxial environment with
single-element materials for molecular beam sources and
is easy to implement epitaxy of atomic layer and in situ
real-time monitoring, avoiding the C-pollution problem
which exits in Al-containing materials growing by
MOCVD and greatly reducing the concentration of O
doping. In fact most of the prototype devices having
complex fine structures and low-dimensional structures
(quantum wells, quantum wires and quantum dots) were
first achieved using materials grown by MBE. It is
noteworthy that, no matter MOCVD or MBE method,
the use of substrates whose surface orientation have a
small angle offset (i.e., low-density atomic step on the
surface of the substrate) seem to be more accessible
high-quality epitaxial layers. Experiments confirmed that
the use of GaSb (100) substrates miscut 2 towards (110)
or 6 towards (11) B may get higher crystal quality of
InGaAsSb and AlGaAsSb epitaxial layers [5]. To overcome the difficulty that antimonides have no semi-insulating substrate materials, the use of GaAs, Si and other
heterogeneous substrate materials for epitaxy of ABCS
films have also attracted great attention. H. Toyota, etc.
[10] reported that they grown high-quality GaSb/AlGaSb
multi-quantum well (MQW) structures with a 5nm AlSb
initiation layer and a relatively thick GaSb buffer layer
(0.5-2.0 m) grown on Si (001) substrates by molecular
beam epitaxy. The photoluminescence (PL) emission
around 1.55 m wavelength was observed for GaSb/
AlGaSb MQW structure at room temperature. Low dislocation density, high-quality GaSb epitaxial films on
GaAs (001) substrates stripe-patterned with SiO2 is also
prepared by MOCVD with low temperature epitaxial
lateral overgrowth (ELO) method [11].
Apart from common binary, ternary and quaternary
antimonides being composed of Al, Ga, In, As and Sb, in
order to extend the applications of antimonide-based
materials in the far-infrared band ( 5 m), easy to adjust
the material lattice constant to match the substrates lattice constant of GaSb, InAs et al. and develop new functional materials, recently some ternary, quaternary antimonides containing N ( 2%), P or Bi ( 2%) and fiveelements antimonides such as AlGaInAsSb, GaInNAsSb
etc have also aroused peoples concern and research interest [12-14]. T. Ashlet, etc. [12] found that the addition
of a small percentage of nitrogen ( 2%) to GaSb, InSb,
and GaInSb materials would significantly change their
energy band structures (bandgap become smaller) which
is very conducive to develop multi-band infrared detectors.
3. Application of ABCS Materials

The early focus of antimonide based compound semiconductors comes from its application prospect in midand far-infrared (photon) detectors, but the first to enter
the market and get a large-scale industrial production is
high-sensitivity InSb magnetoresistive Hall sensors. In
2004, Asahi Kasei Electronic (AKE) of Japan which account for 70% of the global market share of Hall sensors
announced that its InSb Hall sensor output had reached
more than 100 million per month. These products are
widely used in small brushless DC motors, automotive
electronics and consumer electronics products and other
fields. InSb-based infrared detector arrays have gained a
market dominant position of ground-infrared applications
and space instrumentation fields. In addition to these
more mature products, antimonide materials have made
great progress in the third-generation infrared focal plane
array detectors, mid and far infrared quantum cascade
lasers, quantum dot lasers, ultra-high-speed, ultra-lowpower and low-noise amplifiers, thermophotovoltaic
devices and so on in recent years. The following describes some latest results and trends of development of
application of ABSC materials.
3.1. Microelectronic Devices and Integrated

Circuits
HEMT and HBT devices and circuits used by millimeter-wave radar and high-frequency digital communications have so far experienced first generation based on
GaAs-based materials, second generation based on InPbased materials and is currently to the development of
third generation of HEMT and HBT devices and circuits
based on antimonide based compound materials with
ultra-high speed, ultra-lower power consumption and
noise factor. After DARPA launched the ABCS projects
in 2001, Rockwell Scientific Company (RSC) starting in
2003, has developed Ka-band (34-36 GHz), W-band
(92-102 GHz) and X-band (8-12 GHz) low noise amplifier microwave monolithic integrated circuit (MMIC)
and the transmit/receive (T/R) integrated modules based
on InAs/AlSb mHEMT through its mature GaAs
pHEMT technology platform. Currently ABCS Integrated Circuit was regarded as a core and key technologies to accelerate the development by DARPA and the
short-term goal is to develop practical ABCS IC products
ENG
C. LIU
with integration of transistors more than 5000 and the

working voltage of about 0.5 V.
The five-stage W-band MMIC LNA chip is shown in
Figure 3 [15]. The compact 1.2 mm2 five-stage W-band
LNA using 0.2-m gate length InAs/AlSb metamorphic
HEMTs demonstrated a 3.9 dB noise-figure at 94 GHz
with an associated gain of 20.5 dB, fT = 142 GHz, fmax =
178 GHz. The measured dc power dissipation of the
ABCS LNA was only 6.0 mW which is less than onetenth the dc power dissipation of a typical equivalent
InGaAs/AlGaAs/GaAs HEMT LNA. The ABCS HEMT
structure [15] is grown using MBE on semi-insulating
GaAs substrates using an AlSb buffer to accommodate
the lattice mismatch and a strained InAlAs cap layer to
provide a chemically stable surface layer and minimize
gate leakage. Hall measurements show 2DEG of InAs
channel concentration and mobility to be 3.7 1012 cm-2
and 19,000 cm2/Vs at 295K.
Growing the Sb-based HEMTs on Si substrate can
combine the high mobility of antimonide based compound materials and excellent features of Si substrate
with broad application prospects. M. K. Kwang et al. [16]
reported their research results of growing AlGaSb/InAs
HEMT structure on Si substrates. By using an AlGaSb
buffer layer containing InSb quantum dots for dislocation
termination, they can effectively terminate the propagation of micro-twin-induced structural defects into overlying layers, resulting in the low defect material grown
on a largely mismatched substrate with a relatively thin
buffer layer. Figure 4 shows the schematic of the AlGaSb/InAs HEMT grown on Si substrate. The high quality AlGaSb/InAs HEMT materials grown on Si (001)
substrate with the electron mobility of higher than 16000
cm2V1s1 at room-temperature and a sheet density of 2.5
1012 cm2 were obtained by using this technique. It
seems to provide a new way of integrating Sb-based devices and circuits on Si substrate.
3.2. Infrared Detectors

There has been more than 60 years in the study of the
infrared photon detectors. The development of the first
Figure 3 A photomicrograph of the five-stage ABCS HEMT

MMIC W-band LNA fabricated by Rockwell Scientific
Company.
ET AL.
621
Si:1 ML InAs
Figure 4. Schematic of the AlGaSb/InAs HEMT grown on

Si substrate [16].
generation of infrared detectors began in the late forties

of the last century, using one-dimensional linear arrays
which were made of lead salt such as PbSe, and PbTe to
detect the mid-infrared (MWIR) (3-5 m). The second
generation infrared detector materials were mainly InSb
and HgCdTe (MCT) for the two atmospheric IR windows of the mid-infrared band and the far-infrared band
(LWIR) respectively [17]. The devices with the focal
plane array structures of one dimension and two dimensions are currently very widely used and more mature
products. In recent years, the third generation infrared
detectors were researched and developed in many countries, their main features are multi-band infrared detection, high-resolution (high pixels and high frame rate),
high operating temperatures, high spatial uniformity,
high stability and low cost [18]. As it is difficult for the
MCT to achieve large area uniformity and stability, the
ABCS superlattice materials is generally considered as
the preferred materials of the third-generation infrared
detectors [6-7]. In principle, the bandgap of the ABCS
superlattice materials can be tailored to cover the entire
spectrum area of infrared detection by adjusting the
thickness and composition of the ABCS materials [19].
In 2007, C.J. Hill et al. of the Jet Propulsion Laboratory [20] reported the GaSb/InAs type-II superlattice
detectors grown on unintentional doped p-type GaSb
(100) substrate designed for 25 m and 812 m bands
infrared absorption. The LWIR detectors have detectivities as high as 8 1010 Jones (cm.H1/2/W) with a differential resistancearea product (RoA) greater than 6 Ohm
cm2 at 80 K with a cutoff wavelength of approximately
12 m. The measured internal quantum efficiency (QEi)
of these front-side illuminated devices is close to 30% in
the 1011 m range. The MWIR detectors have detectivities as high as 8 1013 Jones with a differential resistancearea product greater than 3 107 Ohm cm2 at 80 K
with a cutoff wavelength of approximately 3.7 m. The
measured internal quantum efficiency of these front-side
illuminated MWIR devices is close to 40% in the 23 m
range at low temperature and increases to over 60% near
room temperature. From the RoA and QEi indicators, we
ENG
622
C. LIU ET AL.
can see that the ABCS II-type superlattice mid-infrared

detector will have a great potential for application of
mid-infrared focal plane array devices of non-low-temperature environment. In addition, InAs/InGaSb type-II
superlattice materials have also been widely concerned
and in-depth research and they are considered as candidate materials for the third-generation infrared detectors.
Two-color or dual-band infrared detectors have the
ability of inhibiting the complex background and improving the target detection efficiency and can significantly improve the system performances. Dual-band
LWIR/VLWIR type-II superlattice infrared detectors was
reported by E. H. Aifer et al. [21]. The cut-off wavelengths of the two bands are 11.4 m and 17 m respectively. But the quantum efficiency of the dual-band infrared detectors is too low (only 4-5%) compared to the
single-band type II superlattice infrared detectors and the
device structure needs to be further optimized. High
quality GaSb based two-color 288 384 MWIR InAs/
GaSb type-II SLS FPAs was reported by M. Mnzberg
et al. [22] of the Fraunhofer Institute in Freiburg. First,
the blue channel consisting of 330 periods of p-type of
a 7.5 ML InAs/10 ML GaSb was deposited on the GaSb
substrate for spectral selective detection in the 3.0-4.1
m wavelength range. Next, the red channel consisting
of 150 periods of a 9.5 ML InAs/10 ML GaSb superlattices was deposited for spectral selective detection in the
4.1-5.0 m wavelength range. The thickness of the entire
vertical pixel structure is only 4.5 m, which significantly reduces the technological challenge compared to
dual- band HgCdTe FPAs with a typical total layer
thickness around 15 m. Excellent thermal resolution
with Noise Equivalent Temperature Difference (NETD)
< 17 mK for the red channel and NETD < 30 mK for
the blue channel has been achieved.
3.3. Infrared Lasers

Solid-infrared laser has important applications in gaseous
environmental monitoring, chemical detection, bio-medical diagnosis, satellite remote sensing technology and so
on. Antimonide based compound semiconductor with
bandgap corresponding to just 2-5 m mid-infrared atomspheric window is an important material of mid-infrared lasers. Research and development of new highperformance antimonide-based infrared laser are very
active research subjects in recent years and researchers
have made a series of important research results such as
AlGaAsSb/InGaAsSb multi-quantum well lasers [23],
AlSb/InAs/InGaSb type-II quantum cascade lasers [24],
W-shaped mid-infrared laser [25], InGaSb quantum
dot lasers [26].
Antimonide-based interband cascade laser combining
the advantages of quantum cascade (QC) laser and typeII quantum well interband laser has potential to achieve
continuous output of high-power infrared laser at room

temperature and is an international hot subject of research and development. Mid-infrared interband cascade
laser made from InAs/Ga(In)Sb/AlSb muti-quantum
wells was reported by C. J. Hill et al. of Jet Propulsion
Laboratory [27]. This laser structure was grown on pGaSb001 substrate by MBE as follows sequence:
0.3 m GaSb buffer layer, 2-3 m InAs/AlSb superlattice
bottom claddings, multi-quantum well InAs/Ga(In)Sb
/AlSb active layers (be repeated 12-35 times), InAs/
AlSb superlattice top claddings and finally an n-type
InAs cap layer. The total thickness of epitaxial layers
was more than 8 m. A 15 m 1.5 mm laser made
from sample J377 lased in cw mode up to 212 K with an
emission wavelength near 3.3 m. Significant output
power (over 30 mW/facet at 140 K) has been obtained
from the laser with relatively low injection currents and
the laser was able to operate in pulsed mode up to 325 K.
A 15 m 1 mm laser made from sample J435 lased in
cw mode at temperatures up to 165 K with a lasing
wavelength of 5.43 m at a current of 70.5 mA. The threshold current density increased from 43 A/cm2 at 80 K to
470 A/cm2 at 165 K. The laser was able to operate in
pulsed mode up to 325 K with an emission wavelength
of 5.7 m. However, at temperatures higher than 230 K,
the spectral linewidth is relatively broad with operation
voltages higher than 10 V.
GaInSb quantum dot surface-emitting laser (QDVCSEL) operating in optical communication wavelength
band of 1.3-1.55 m with continuous emission at room
temperature by either optical pumping or current injection was reported by researchers of Japans National Institute of Information and Communication Technology
[26]. This laser mainly consists of an antimonide-based
quantum dot active layer and two AlAs/GaAs superlattice distributed Bragg reflectors (DBRs). With the development of antimonide-based quantum dots, they have
overcome the technical difficulty of preparing a material
that emits light in the entire optical communication
wavelength bands of 1.3 to 1.55 m on a GaAs substrate
through conventional technologies. In particular, the obtained wavelength of 1.55 m represents the worlds
longest emission wavelength of existent surface-emitting
laser structures based on GaAs substrate. It has great
significance for mass production of low-cost surfaceemitting lasers used in next-generation ultra-high-speed
optical communication technology.
High-power optically pumped semiconductor vertical
external cavity surface emitting laser (VECSEL) operating at 2-m wavelength was reported by A. Hrknen et
al. [28]. The device material was grown on GaSb substrate by MBE and consisted of 15 Ga0.78In0.22Sb quantum-wells placed within a three-lambda GaSb cavity and
grown on the top of an 18-pairs AlAsSb/GaSb Bragg
reflector. When cooled down to 5 and using 790-nm
ENG
C. LIU
diode laser for optical pumping, this laser emitted up to 1

W of optical power in a nearly diffraction-limited Gaussian beam demonstrating the high potential of antimonide material for VECSEL fabrication. LED devices based
on InGaAsSb/AlGaAsSb multi-quantum well active region sandwiched between two AlAsSb/GaSb n- and pdoped Bragg mirrors structure has realized operation in
continuous wave mode under electrical injection at room
temperature and exhibited a bright emitting peak near 2.3
m with an external quantum efficiency of 0.16% at 34
A/cm2 [29]. It shows that antimonides have enormous
potential in the development of new high-power, electrical injection and continuous-wave emission mid-infrared
optoelectronic devices.
3.4. Thermophotovoltaic Cells

Thermophotovoltaic cells are similar to the solar cells
that utilize the thermal infrared radiation of a heated
source to directly generate electric power. The current
trend of development of TPV is to develop high efficiency, low cost, narrow-bandgap (0.6 eV or less) thermophotovoltaic materials and components applicable to
the mid- and low-temperature radiation source (< 1500).
It appears that antimonide based compounds have been
one of the leading material systems for thermophotovoltaic device applications and the most studied TPV is
GaSb-based InGaAsSb p-n cells fabricated by LPE,
MOCVD, MBE and other methods. TPV cells based on
InAsSbP, grown on InAs substrate, can have spectral
responses in the 2.5-3.4 m wavelength range and it is a
hopeful research direction of having great potentials. For
further details, please refer to M.G. Mauks review paper
[8].
4. Conclusions
As the first candidate materials for fabrication of the
third generation large-scale focal plane arrays infrared
(photon) detectors, integrated circuits with ultra-high
speed and ultra-low power consumption and new high
efficiency thermophotovoltaic devices, the research and
development of antimonide based compound semiconductor materials and device applications are in the ascendant, attracting increasingly widespread concern and
research interests of researchers and institutions in the
world. Compared to currently more mature GaAs-based
and InP-based materials growth and device manufacturing process, the growth technology of antimonide based
micro-structure materials such as heterojunctions, superlattice quantum wells and self-aligned quantum dots continues to face considerable great difficulties and technical
challenges and the manufacturing process of various antimonides devices are far from mature. Therefore, there
ET AL.
623
are tremendous opportunities for R&D and innovations

in this area. With the gradual suppression or elimination
of the adverse factors affecting device performance in
narrow bandgap antimonide based compounds (such as
composition segregation, Auger recombination, surface
recombination, carrier absorption, etc.) by continuous
optimization of material growth techniques, improving
the device structure design and manufacturing processes
and other technologies, we believe that in the near future,
new types of high-performance antimonide devices and
integrated circuits will get a wide range of important
applications in the infrared imaging technologies, atmospheric environmental monitoring, biomedical diagnostics,
multi-function digital radar systems, mobile communications, thermophotovoltaic power generation systems, and
many other high-tech fields.
5. Acknowledgements
This work was supported by the National Natural Science Foundation of China (Grant No. 60876004).
6. References
[1]
B. R. Bennett, R. Magno, J. B. Boos, et al., Antimonide-Based Compound Semiconductors for Electronic Devices: A Review, Solid-State Electronics, Vol. 49, No. 12,
2005, pp. 1875-1895.
[2]
M. Razeghi, Overview of Antimonide Based III-V

Semiconductor Epitaxial Layers and Their Applications
at the Center for Quantum Devices, The European
Physical Journal Applied Physics, Vol. 23, No. 3, 2003,
pp. 149-205.
[3]
A. H. Kroemer, The 6.1 family (InAs, GaSb, AlSb)

and its Heterostructures: A Selective Review, Physica E,
Vol. 20, No. 3-4, 2004, pp. 196-203.
[4]
R. M. Biefeld, The Metal-Organic Chemical Vapor

Deposition and Properties of III-V Antimony-Based
Semiconductor Materials, Materials Science and Engineering R, Vol. 36, No. 4, 2002, pp. 105-142.
[5]
C. A. Wang, Progress and Continuing Challenges in

Gasb-Based III-V Alloys and Heterostructures Grown by
Organometallic Vapor-Phase Epitaxy, Journal of Crystal Growth, Vol. 272, No. 1-4, 2004, pp. 664-681.
[6]
A. Rogalshi, Material Considerations for Third Generation Infrared Photon Detectors, Infrared Physics &
Technology, Vol. 50, No. 2-3, 2007, pp. 240-252.
[7]
O. Nesher and P. C. Klipstein, High-Performance IR

Detectors at SCD Present and Future, Opto-Electronics
Review, Vol. 14, No. 1, 2006, pp. 61-70.
[8]
M. G. Mauk and V. M. Andreev, Gasb-Related Materials for TPV Cells, Semiconductor Science and Technology, Vol. 18, No. 5, 2003, pp. S191-S201.
[9]
M. Rosker and J. Shah, DARPAs Program on Antimonide Based Compound Semiconductors (ABCS),
IEEE GaAs Digest, 2003, p. 293.
ENG
624
C. LIU ET AL.
[10] H. Toyota, T. Sasaki and Y. Jinbo, Growth and Characterization of Gasb/Algasb Multi-Quantum Well Structures on Si (0 0 1) Substrates, Journal of Crystal Growth,
Vol. 310, No. 1, 2008, pp. 78-82.
[11] K. Zaima, R. Hasimoto, M. Ezaki, et al., Dislocation
Reduction of Gasb on Gaas by Metalorganic Chemical
Vapor Deposition with Epitaxial Lateral Overgrowth,
Journal of Crystal Growth, Vol. 310, No. 23, 2008, pp.
4843-4845.
[12] T. Ashley, L. BuckleG. W. Smith, et al., Dilute Antimonide Nitrides for Very Long Wavelength Infrared
Applications, Proceedings of SPIE, Orlando, 17 May
2006, pp. 62060L.
[13] P. H. Jefferson, L. Buckle , B. R. Bennett, et al., Growth
of Dilute Nitride Alloys of Gainsb Lattice-Matched to
Gasb, Journal of Crystal Growth, Vol. 304, No. 2, 2007,
pp. 338-341.
[14] D. Jackrel, A. Ptak, B. Seth, et al., Gainnassb Solar
Cells Grown by Molecular Beam Epitaxy, IEEE 4th
World Conference on Photovoltaic Energy Conversion.
Waikoloa, HI, 2006, pp. 783-786.
[15] J. B. Hacker, J. Bergman, G. Nagy, et al., An UltraLow Power Inas/Alsb HEMT W-Band Low-Noise Amplifier, IEEE MTT-S International Microwave Symposium, June 2005, pp. 1029-1032.
[16] M. K. Kwang, H. S. Jung, E. K. Dong, et al., The
Growth of a Low Defect Inas HEMT Structure on Si by
Using an Algasb Buffer Layer Containing Insb Quantum
Dots for Dislocation Termination, Nanotechnology, Vol.
20, No. 22, 2009, p. 225201.
[17] A. Rogalski. New Material Systems for Third Generation Infrared Photo Detectors, Opto-Electron Review,
Vol. 16, No. 4, 2008, pp. 458-482.
[18] M. Walther, R. Rehm, J. Schmitz, et al., AntimonyBased Superlattices for High-Performance Infrared Imagers, Proceedings of SPIE, Orlandoof, 17 March 2008, pp.
69400A-69400A-8.
[19] E. Corbin, M. J. Shaw, M. R. Kitchin, et al., Systematic
Study of Type II Ga1Xinxsb/Inas Superlattices for Infra-Red Detection in the 10-12 M Wavelength Range,
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
Semiconductor Science and Technology, Vol. 16, No. 4,

2001, pp. 263-272.
C. J. Hill, J. V. Li, J. M. Mumolo, et al., MBE Grown
Type-II MWIR and LWIR Superlattice Photodiodes, Infrared Physics & Technology, Vol. 50, No. 2-3, 2007, pp.
187-190.
E. H. Aifer, J. G. Tischler, J. H. Warner, et al., Dual
Band LWIR/VLWIR Type-II Superlattice Photodiodes,
Proceedings of SPIE, Orlando, 28 March 2005, Vol. 5783,
pp. 112-122.
M. Mnzberg, R. Breiter, W. Cabanski, et al., Inas/Gasb
Type-II Short-Period Superlattices for Advanced Single
And Dual-Color Focal Plane Arrays, Proceedings of
SPIE, Orlando, 9 April 2007, Vol. 6542, p. 654206.
Y. G. Zhang, Y. L. Zheng, C. Lin, et al., Continuous
Wave Performance and Tunability of MBE Grown 2.1
M Ingaassb/Algaassb MQW Lasers, Chinese Physics
Letters, Vol. 23, No. 8, 2006, pp. 2262-2265.
A. Bauer, F. Langer, M. Dallner, et al., Emission Wavelength Tuning of Interband Cascade Lasers in the 3-4 M
Spectral Range, Applied Physics Letters, Vol. 95, No. 25,
2009, p. 251103.
W. W. Bewley, J. R. Lindle, C. S. Kim, et al., Lifetimes
and Auger Coefficients in Type-II W Interband Cascade
Lasers, Applied Physics Letters, Vol. 93, No. 4, 2008, p.
041118.
N. Yamamoto, Next-Generation Optical Communications through Nanotechnology, NICT News, Vol. 353,
2005, pp. 3-4.
C. J. Hill and R. Q. Yang, MBE Growth Optimization of
Sb-Based Interband Cascade Lasers, Journal of Crystal
Growth, Vol. 278, No. 1-4, 2005, pp. 167-172.
A. Hrknen, M. Guina, O. Okhotnikov, et al., 1-W
Antimonide-Based Vertical External Cavity Surface Emitting Laser Operating At 2-M, Optics Express, Vol. 14,
No. 14, 2006, pp. 6479-6484.
A. Ducanchez, L. Cerutti, P. Grech, et al., Room Temperature Continuous Wave Operation of Electrically-Injected Sb-Based RC-LED Emitting Near 2.3 M, Superlattices and Microstructures, Vol. 44, No. 1, 2008, pp.
62-69.
ENG
Engineering, 2010, 2, 625-634

Lie Group Analysis for the Effects of Variable Fluid

Viscosity and Thermal Radiation on Free Convective Heat
and Mass Transfer with Variable Stream Condition
1
P. Loganathan1, P. Puvi Arasu2
Department of Mathematics, Anna University, Chennai, India

2
Erode Sengunthar Engineering College, Thudupathi, India
E-mail: puviarasup@gmail.com
Received December 23, 2009; revised February 26, 2010; accepted March 6, 2010
Abstract
Natural convective boundary layer flow and heat and mass transfer of a fluid with variable viscosity and
thermal radiation over a vertical stretching surface in the presence of suction/injection is investigated by Lie
group analysis. Fluid viscosity is assumed to vary as a linear function of temperature. The symmetry groups
admitted by the corresponding boundary value problem are obtained by using a special form of Lie group
transformations viz. scaling group of transformations. An exact solution is obtained for translation symmetry
and numerical solutions for scaling symmetry. The effects of fluid viscosity and thermal radiation on the dimensionless velocity, temperature and concentration profiles are shown graphically. Comparisons with previously published works are performed and excellent agreement between the results is obtained. The conclusion is drawn that the flow field and temperature profiles are significantly influenced by these parameters.
Keywords: Scaling Group of Transformations, Free Convective Flow, Temperature-Dependent Fluid
Viscosity, Suction/Blowing, Thermal Radiation
1. Introduction
The study of natural convection flow for an incompressible viscous fluid past a heated surface has attracted the
interest of many researchers in view of its important applications to many engineering problems such as cooling
of nuclear reactors, the boundary layer control in aerodynamics, crystal growth, food processing and cooling
towers. In this paper, symmetry methods are applied to a
natural convection boundary layer problem. The main
advantage of such methods is that they can successfully
be applied to non-linear differential equations. The symmetries of differential equations are those continuous
groups of transformations under which the differential
equations remain invariant, that is, a symmetry group
maps any solution to another solution. The symmetry
solutions are quite popular because they result in the reduction of the number of independent variables of the
problem.
A class of flow problems with obvious relevance to
polymer extrusion is the flow induced by the stretching
motion of a flat elastic sheet. In a melt-spinning process,
the extrudate from the die is generally drawn and simulCopyright 2010 SciRes.
taneously stretched into a filament or sheet, which is

thereafter solidified through rapid quenching or gradual
cooling by direct contact with water or chilled metal rolls.
In fact, stretching imports a unidirectional orientation to
the extrudate, thereby improving its mechanical properties and the quality of the final product greatly depends
on the rate of cooling. Crane [1] was the first who studied the motion set up in the ambient fluid due to a linearly stretching surface. Several authors see e.g., the references cited in [2], have subsequently explored various
aspects of the accompanying heat transfer occurring in
the infinite fluid medium surrounding the stretching
sheet. The hydrodynamics of a finite fluid medium, i.e., a
thin liquid film, on a stretching sheet was first considered
by Wang [3] who by means of a similarity transformation reduced the unsteady NavierStokes equations to a
non-linear ordinary differential equation. The accompanying heat transfer problem was solved more recently by
Andersson et al. [2]. In these studies the film surface was
planar and free of any stresses.
The production of sheeting material arises in a number
of industrial manufacturing processes and includes both
metal and polymer sheets. It is well known that the flow
ENG
626
P. LOGANATHAN
in a boundary layer separates in the regions of adverse

pressure gradient and the occurrence of separation has
several undesirable effects in so far as it leads to increase
in the drag on the body immersed in the flow and adversely affects the heat transfer from the surface of the
body. Several methods have been developed for the purpose of artificial control of flow separation. Separation
can be prevented by suction as the low-energy fluid in
the boundary layer is removed [4,5]. On the contrary, the
wall shear stress and hence the friction drag is reduced
by blowing. Free convective phenomenon has been the
object of extensive research. The importance of this
phenomenon is increasing day by day due to the enhanced concern in science and technology about buoyancy induced motions in the atmosphere, the bodies in
water and quasisolid bodies such as earth. Natural convection flows driven by temperature differences are very
much interesting in case of Industrial applications. Buoyancy plays an important role where the temperature differences between land and air give rise to a complicated
flow and in enclosures such as ventilated and heated
rooms (Elbashbeshy and Bazid [6]).
So such type of problem, which we are dealing with, is
very much useful to polymer technology and metallurgy.
Cheng and Minkowycz [7] and Cheng [8] studied the
free convective flow in a saturated porous medium.
Wilks [9] had studied the combined forced and free convection flow along a semi-infinite plate extending vertically upwards with its leading edge horizontal. Boutros
et al. [10] solved the steady free convective boundary
layer flow on a nonisothermal vertical plate. Recently,
any studies were made on the steady free convective
boundary layer flow on moving vertical plates considering the effect of buoyancy forces on the boundary layer
Chen and Strobel [11], Ramachandran et al. [12], Lee
and Tsai [13]. The radiative effects have important applications in physics and engineering particularly in
space technology and high temperature processes. But
very little is known about the effects of radiation on the
boundary layer. Thermal radiation effects may play an
important role in controlling heat transfer in polymer
processing industry where the quality of the final product
depends on the heat controlling factors to some extent.
High temperature plasmas, cooling of nuclear reactors,
liquid metal fluids, power generation systems are some
important applications of radiative heat transfer from a
vertical wall to conductive gray fluids. The effect of radiation on heat transfer problems have studied by Hossain and Takhar [14], Takhar et al. [15], Hossain et al.
[16]. In all of the above mentioned studies, fluid viscosity was assumed to be constant. However, it is known
that the physical properties of fluid may change significantly with temperature. For lubricating fluids, heat generated by the internal friction and the corresponding rise
in temperature affects the viscosity of the fluid and so the
fluid viscosity can no longer be assumed constant. The
ET
AL.
increase of temperature leads to a local increase in the

transport phenomena by reducing the viscosity across the
momentum boundary layer and so the heat transfer rate
at the wall is also affected. Therefore, to predict the flow
behaviour accurately it is necessary to take into account
the viscosity variation for incompressible fluids. Gary et al.
[17] and Mehta and Sood [18] showed that, when this
effect is included the flow characteristics may changed
substantially compared to the constant viscosity assumption. Mukhopadhyay et al. [19] investigated the MHD
boundary layer flow with variable fluid viscosity over a
heated stretching sheet. Recently, Mukhopadhyay and
Layek [20] studied the effects of thermal radiation and
variable fluid viscosity on free convective flow and heat
transfer past a porous stretching surface.
Many authors have constructed an exponential type of
exact solution using the translation symmetry and a series type of approximate solution using the scaling symmetry and also discussed some boundary value problems.
So far no attempt has been made to study the heat and
mass transfer in a vertical stretching surface using Lie
groups and hence we study the problem of natural convection heat and mass transfer flow past a stretching
sheet for various parameters using Lie group analysis.
2. Mathematical Analysis
We consider a free convective, laminar boundary layer
flow and heat and mass transfer of viscous incompressible fluid over a vertical stretching sheet emerging out of
a slit at origin (x = 0, y = 0) and moving with non-uniform velocity U(x) in the presence of thermal radiation
(Figure 1).
The governing equations of such type of flow are, in
the usual notations,
u v
0
(1)
x y
Figure 1. Physical model of boundary layer flow over a vertical stretching surface.
ENG
P. LOGANATHAN
u
u
v
x
y
1 T u 2 u
[ g (T T ) g * (C C )]
T y y x 2
T
T
T
1 qr
v
x
y c p y 2 c p y
(3)
C
C
2C
v
D 2
x
y
y
(4)
(2)
u U ( x), v V ( x), C Cw , T Tw at y 0
u 0, C C , T T as y
(5)
when the viscous dissipation term in the energy equation
is neglected (as the fluid velocity is very low). Here u
and v are the components of velocity respectively in
the x and y directions, is the coefficient of fluid
viscosity, is the fluid density (assumed constant), T
is the temperature, is the thermal conductivity of the
fluid, D is diffusional coefficient, is the volumetric
coefficient of thermal expansion, * is the volumetric

coefficient of concentration expansion, g is the gravity
field, T is the temperature at infinity, where U ( x) is
the stream wise velocity and V ( x) is the velocity of
suction of the fluid, Tw is the wall temperature.
Using Rosseland approximation for radiation (Brew4 T 4
where 1 is the
ster [21]) we can write qr *1
3k y
StefanBoltzman constant, k* is the absorption coefficient.
Assuming that the temperature difference within the
flow is such that T 4 may be expanded in a Taylor series and expanding T 4 about T and neglecting
higher orders we get T 4 4T 3T 3T 4
Therefore, the Equation (3) becomes
2T 16 1T3 2T
T
T
u
v
x
y c p y 2 3 c p k * y 2
(6)
We now introduce the following relations for

u, v, and as
C C
T T
,v
,
and
(7)
x
y
C w C
Tw T
where u is the stream function. The stream wise
velocity and the suction/injection velocity are taken
as
U ( x) c x , V ( x) V0 x
m
m1
2
(8)
Here c 0 is constant, Tw is the wall temperature,

the power-law exponent m is also constant. In this
study we take c 1 .
ET AL.
627
The temperature-dependent fluid viscosity is given by

(Batchelor [22]),
* [a b(Tw T )]
where * is the constant value of the coefficient of
viscosity far away from the sheet and a, b are constants
and b( 0) . For a viscous fluid, Ling and Dybbs [23]
suggest a viscosity dependence on temperature T of the
form
[1 (T T )]
where c is a thermal property
of the fluid and is the viscosity away from the hot

sheet. This relation does not differ at all with our formulation. The range of temperature, i.e., (Tw T ) studied
here is (0 230 )C .
Using the relations (5) in the boundary layer Equation
(2) and in the energy and diffusion Equations (3) and (4)
we get the following equations
2 2
y xy x y 2
2
3
v
v* [a (1 )] 3 g ( * )
2
b
y y
y
(9)
16 1T3 2

)
c p 3 c p k * y 2
y x x y
(10)
D 2
y x x y
y
(11)
where b(Tw T ), v*
The boundary conditions Equation (5) become

m 1
2
xm ,
V0 x ,
y
x
1 at y 0;
0, 0, 0 as y
y
(12)
We now introduce the simplified form of Lie-group
transformations namely, the scaling group of transformations (Mukhopadhyay et al. [19]),
: x* x e 1 , y* y e 2 , * e 3 ,
u* u e 4 , v* v e 5 , * e 6 , * e 7
(13)
Equation (13) may be considered as a point-transformation which transforms co-ordinates ( x, y, , u, v, , )

to the coordinates ( x* , y* , * , u * , v* , * , * ) .
Substituting (13) in (9), (10) and (11) we get,
ENG
P. LOGANATHAN
628

v*e (3 2 3 6 ) *
*2
y y
(14)
3 *
3 *
v* [a ]e (3 2 3 ) *3 v*e (3 2 3 6 ) * *3
y
y
g
( e
* 7
e
*
)
*
1
1
These relations give 6 7 0, 2 1 3 .
4
3
1
The boundary conditions yield 4 m1 1 ,
2
m 1
1
1
5
1 1 (as m ) .
2
4
2
In view of these, the boundary conditions become
x ,
V0 x
*
x
y
1
( )
* 4
, y* y e
1
4
, * e
3
1
4
u u u
*
1
2
, v v v
*
1
4
1
4
, *
*
3F 0
4V0
, 1 at
3
0 and F 0, 0, 0 as
F 1, F
31
,
4
(19)
(20)
(21)
(22)
Again, we introduce the following transformations for

, F , and in Equations (19), (20) and (21):
g
g
( 1 )1 v*b1 * , F ( 1 )1 v*b1 F * ,
b
b
g 1 1 *b1 *
g 1 1 *b1 *
*
) v , (
) v where 1
(
2
b
b
(23)
F * f , and the Equations
Taking
(19), (20) and (21) finally take the following form
4(a ) F 4 F 4 F
4
4
1
3F 0
Pr 3N
4
3F 0
Sc
, 0
*
(18)
The boundary conditions take the following form
(17)
1
4
x* x x 1 , y* y y
( * )
2 F 2 3FF 4 ( )
u * u e , v* v e
, * ,, *
Expanding by Taylors method in powers of and
keeping terms up to the order we get
2
4 D 3F 0
*
0, * 0, * 0 as y*
0 and
y*
The set of transformations reduces to
4(a )v* F 4 v* F 4 g
, 1 at y
*

16 T
4
c p 3 c k
1 2 2 2 3 3 2 3 6 3 2 3
6 7 ; 1 2 3 6
2 2 6 and 1 2 3 7 2 2 7
3
4
, x F ( ), * ( ), * ( )
*
3
1
*
p
(16)
2 *
De
y*2
The system will remain invariant under the group of
transformations , we would have the following relations
among the parameters, namely
x* x e
y x
1
4
* * * *
*
*
*
x y*
y x
(1 2 3 7 )
1
*2
* *
(15)
(2 2 7 )
Solving the above equations we get,
2 F 2 3FF 4 v* F

16 1T3 (2 2 6 ) 2 *
e
c p 3 c k *
y*2
p
dx
dy
d
du
dv
d d
3
x 1
0
0
y 1 1 u 1 v 1
4
4
2
4
With the help of these relations, the (14), (15) and (16)
become

e (1 2 3 6 ) *
*
*
x y *
y x
*
AL.
In terms of differentials these yield
* 2 * * 2 *
*
e (1 2 2 23 ) *
*
*
x y*2
y x y
*
ET
(25)
(26)
*c p
is the Prandtl number,
k*
v*
N
is the Radiation parameter, Sc is the
3
D
4 1T
Schmidt number. The boundary conditions take the following forms.
where
Pr
v* c p
(24)
f 1, f S , 1,
at * 0 and f 0, 0, 0 as *
(27)
ENG
P. LOGANATHAN
1
g 4 2
4
where S V0 ( 1 ) v , S 0 corresponds to suction
3
b
and S 0 corresponds to injection.
3. Numerical Solution
The set of non-linear ordinary differential Equations (24)
to (26) with boundary conditions (27) have been solved
by using the R. K. Gill method, (Gill [24]) along with
Shooting Technique with , Pr, Sc, a and N as prescribed parameters. The numerical solution was done
using Matlab computational software. A step size of
= 0.001 was selected to be satisfactory for a convergence criterion of 10 7 in nearly all cases. The value of
was found to each iteration loop by assignment
statement = + . The maximum value of ,
to each group of parameters , Pr, Sc, a and N, determined when the values of unknown boundary conditions
at = 0 not change to successful loop with error less
ET AL.
629
Prandtl number Pr , Schmidt number Sc and radiationparameter N . For illustrations of the results, numericalvalues are plotted in the Figures 2-9. In all cases we
take a 1.0 .
In the absence of diffusion equations, in order to ascertain the accuracy of our numerical results, the present
study is compared with the available exact solution in the
literature. The temperature profiles for Prandtl number
Pr are compared with the available exact solution of
1.2
1
0.8
0.6
( )
Pr = 0.3
Pr = 0.5
0.4
0.2
than 10 7 . Effects of heat and mass transfer are studied

for different values of temperature-dependent viscosity at
the wall of the surface and the strength of thermal radiation. In the following section, the results are discussed in
detail.
Pr = 1.0
0
0.5
1.5
Figure 2. Influence of Prandtl number over the temperature profiles. Sc 0.0, N 0.1, a 1.0, S 0.5, 0.5 .
To analyze the results, numerical computation has been

carried out using the method described in the previous
section for various values of the temperature-dependent
viscosity parameter , suction/injection parameter S ,
Comparison of the temperature profiles ( present result )

with Mukhopadhyay and Layek [20]
Symbol : Re sult for Mukhopadhyay and Layek [20];
Solid line : Current result .
1.2
1.0
0.8
0.6
( )
0.4
N = 0.5
0.2
0.0
N = 1.0
N = 3.0
0.0
-0.2
0.5
1.0
1.5
Figure 3. Effects of thermal radiation over the temperature profiles. Pr 0.3, a 1.0, S 0.5, 0.5, Sc 0.62 .
ENG
P. LOGANATHAN
630
ET
AL.
1.6
1.4
= 1.0
1.2
= 0.5
1
f '() 0.8
= 0.1
0.6
0.4
0.2
0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
Figure 4. Effects of fluid viscosity over the velocity profiles. Sc 0.62, N 0.1, a 1.0, S 0.5, Pr 0.71 .
1.2
0.8
0.6
( )
0.4
= 0.1
= 0.5
= 1.0
0.2
0.5
-0.2
1.5
Figure 5. Effects of fluid viscosity over the temperature profiles. Sc 0.62, N 0.1, a 1.0, S 0.5, Pr 0.71 .
Mukhopadhyay and Layek [20], is shown in Figure 2. It

is observed that the agreements with the theoretical solution of temperature profiles are excellent. For a given N ,
it is clear that there is a fall in temperature with increaseing the Prandtl number. This is due to the fact that
there would be a decrease of thermal boundary layer
thickness with the increase of Prandtl number as one can
see from Figure 3 by comparing the curves with
Pr 0.3 and Pr 1.0 . This behavior implies that fluids
having a smaller Prandtl number are much more responsive to thermal radiation than fluids having a larger
Prandtl number.
Figure 3 illustrates the typical temperature profiles for

various values of the thermal radiation parameter N . At
a particular value of N , the temperature decreases with
accompanying decreases in the thermal boundary layer
thickness by increasing the values of Pr . Further, it is obvious that for a given Pr , the temperature is decreased
with an increase in N . This result can be explained by
*
the fact that a decrease in the values of ( N k )
4 1T3
for given k * and T means a decrease in the Rosseland
radiation absorptivity. According to Equations (2) and (3)
ENG
P. LOGANATHAN
ET AL.
631
1.6
1.4
1.2
S = 3.0
S = 5.0
f '()
0.8
S = 8.0
0.6
0.4
0.2
0
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
Figure 6. Effect of suction over the velocity profiles. S 1.0, N 0.1, a 1.0, 0.5, Pr 0.71 .
1.2
0.8
0.6
( )
S = 3.0
0.4
S = 5.0
0.2
S = 8.0
0
0
0.5
1.5
-0.2
Figure 7. Influence of suction over the temperature profiles. Sc 0.62, N 0.1, a 1.0, 0.5, Pr 0.71 .
ENG
P. LOGANATHAN
632
ET
AL.
1.2
0.8
0.6
( )
S = 3.0
0.4
S = 5.0
0.2
S = 8.0
0
0
0.5
-0.2
1.5
Figure 8. Effects of suction over the concentration profiles. Sc 0.62, N 0.1, a 1.0, 0.5, Pr 0.71 .
1.2
0.8
Sc = 0.22
0.6
( )
0.4
Sc = 0.62
0.2
Sc = 0.78
0
0
-0.2
0.5
1.5
Figure 9. Effects of Schmidt number over the concentration profiles. S 1.0, N 0.1, a 1.0, 0.5, Pr 0.71 .
ENG
P. LOGANATHAN
qr
increases
y
as decreases which in turn increases the rate of radiative heat transferred to the fluid and hence the fluid temperature increases. In view of this explanation, the effect
of radiation becomes more significant as N 0 ( N
0) and can be neglected when N . Also, it is seen
from Figure 3 that the larger the N , the thinner the
thermal boundary layer thickness for both values of Pr .
In addition, radiation demonstrates a more pronounced
influence on the temperature distribution of ( Pr 0.3 )
than that of ( Pr 1.0 ). It is noticed from the figure that
the temperature decreases with the increasing value of
the radiation parameter N . The effect of radiation parameter N is to reduce the temperature significantly in
the flow region. The increase in radiation parameter
means the release of heat energy from the flow region
and so the fluid temperature decreases as the thermal
boundary layer thickness becomes thinner.
Figure 4 exhibits the velocity profiles for several values of with Pr = 0.71 in presence of suction (S = 0.5)
when N = 0.1. In the case of uniform suction, the velocity of the fluid is found to increase with the increase of
the temperature-dependent fluid viscosity parameter
at a particular value of except very near the wall as
well as far away of the wall (at = 5). This means that
the velocity decreases (with the increasing value of )
at a slower rate with the increase of the parameter at
very near the wall as well as far away of the wall. This
can be explained physically as the parameter increases, the fluid viscosity decreases the increment of the
boundary layer thickness.
In Figure 5, variations of temperature field ( ) with
for several values of (with Pr = 0.71 and N = 0.1)
in presence of suction (S = 0.5) are shown. It is very
clear from the figure that the temperature decreases with
the increasing of whereas the concentration of the
fluid is not significant with the increasing of . The
increase of temperature-dependent fluid viscosity parameter ( ) makes decrease of thermal boundary layer
thickness, which results in decrease of temperature profile ( ) . Decrease in ( ) means a decrease in the
velocity of the fluid particles. So in this case the fluid
particles undergo two opposite forces: one increases the
fluid velocity due to decrease in the fluid viscosity (with
increasing ) and other decreases the fluid velocity due
the divergence of the radiative heat flux
to decrease in temperature ( ) (since ( ) decreases with increasing ). Near the surface, as the
temperature ( ) is high so the first force dominates and
far away from the surface ( ) is low and so the second
ET AL.
633
force dominates here. Now we concentrate in the velocity and temperature distribution for the variation of suction parameter S in the absence and presence of temperature-dependent fluid viscosity parameter .
Figure 6 presents the effects of suction on fluid velocity when the fluid viscosity is uniform, i.e., = 0.5.
With the increasing value of the suction (S > 0) ( =
0.5, Pr = 0.71 and N = 0.1), the velocity is found to decrease (Figure 6), i.e., suction causes to decrease the
velocity of the fluid in the boundary layer region. The
physical explanation for such a behavior is as follows. In
case of suction, the heated fluid is pushed towards the
wall where the buoyancy forces can act to retard the fluid
due to high influence of the viscosity. This effect acts to
decrease the wall shear stress. Figures 7 and 8 exhibit
that the temperature ( ) and concentration ( ) in
boundary layer also decrease with the increasing suction
parameter S (S > 0) ( = 0.5, Pr = 0.71 and N = 0.1).
The thermal and solutal boundary layer thickness decrease with the suction parameter S which causes an increase in the rate of heat and mass transfer. The explanation for such behavior is that the fluid is brought closer to
the surface and reduces the thermal boundary layer
thickness in case of suction. As such then the presence of
wall suction decreases velocity boundary layer thicknesses but decreases the thermal and solutal boundary
layers thickness, i.e., thins out the thermal and solutal
boundary layers.
Figure 9 illustrates the influence of Schmidt number
Sc on the concentration. As Schmidt number Sc increases,
the mass transfer rates increases. Hence, the concentration decreases with increasing Sc. It is evident from this
figure that the concentration ( ) takes its limiting
value C, for higher values of the dimensionless distance
. From this figure, we observe that when the concentration difference C is maintained constant, the dimensionless concentration profile decreases, in the since that
the values of the Schmidt number increases. The variation in the thermal boundary layer is very small corresponding to a moderate change in Schmidt number.
There are very small changes in velocity and temperature
distributions when moderate changes in Schmidt number.
5. Conclusions
By using Lie group analysis, first find the symmetries of
the partial differential equations and then reduce the
equations to ordinary differential equations by using
scaling and translational symmetries. Exact solutions for
translation symmetry and numerical solution for scaling
symmetry are obtained. From the numerical results, it is
predict that the effect of increasing temperaturedependent fluid viscosity parameter on a viscous incompressible fluid is to increase the flow velocity which in
ENG
P. LOGANATHAN
634
turn, causes the temperature to decrease. It is interesting

to note that the temperature of the fluid decreases at a
very fast rate in the case of water in comparison with air.
So, the thermal radiation effects in the presence of fluid
viscosity have a substantial effect on the flow field and,
thus, on the heat and mass transfer rate from the sheet to
the fluid. Decrease of the concentration field due to increase in Schmidt number shows that it increases gradually as we replace Hydrogen (Sc = 0.22) by water vapour
(Sc = 0.67) and Ammonia (Sc = 0.78) in the said sequence.
6. References
[1]
L. J. Crane, Flow Past a Stretching Plate, The Journal

of Applied Mathematics and Physics, Vol. 21, No. 4, pp.
645-647.
[2]
1. H. I. Andersson, J. B. Aarseth and B. S. Dandapat,

Heat Transfer in a Liquid Film on an Unsteady Stretching Surface, International Journal of Heat and Mass
Transfer, Vol. 43, No. 1, 2000, pp. 69-74.
[3]
C. Y. Wang, Liquid Film on an Unsteady Stretching

Surface, Quarterly of Applied Mathematics, Vol. 48, No.
4, 1990, pp. 601-610.
[4]
P. Saikrishnan and S. Roy, Non-Uniform Slot Injection

(Suction) into Steady Laminar Water Boundary Layers
over (i) a Cylinder and (ii) a Sphere, International
Journal of Engineering Science, Vol. 41, No. 12, 2003,
pp. 1351-1365.
[5]
S. Roy and P. Saikrishnan, Non-Uniform Slot Injection

(Suction) into Steady Laminar Water Boundary Layer
Flow over a Rotating Sphere, International Journal of
Heat and Mass Transfer, Vol. 46, No. 18, 2003, pp.
3389-3396.
[6]
E. M. A. Elbashbeshy and M. A. A. Bazid, Heat Transfer in a Porous Medium over a Stretching Surface with
Internal Heat Generation and Suction or Injection, Applied Mathematics and Computation, Vol. 158, No. 3,
2004, pp. 799-807.
[7]
P. Cheng and W. J. Minkowycz, Free Convection about

a Vertical Flat Plate Embedded in a Porous Medium with
Application to Heat Transfer from a Disk, Journal of
Geophysical Research, Vol. 82, No. 14, 1963, pp. 20402044.
[8]
P. Cheng, The Influence of Lateral Mass Flux on a Free

Convection Boundary Layers in Saturated Porous Medium, International Journal of Heat and Mass Transfer,
Vol. 20, No. 3, 1977, pp. 201-206.
[9]
G. Wilks, Combined Forced and Free Convection Flow

on Vertical Surfaces, International Journal of Heat and
Mass Transfer, Vol. 16, No. 10, 1973, pp. 1958-1964.
[10] Y. Z. Boutros, M. B. Abd-el-Malek and N. A. Badran,

Group Theoretic Approach for Solving Time-Independent Free Convective Boundary Layer Flow on a NonIsothermal Vertical Flat Plate, Archives of Mechanics,
Vol. 42, No. 3, 1990, pp. 377-395.
ET
AL.
[11] T. S. Chen and F. A. Strobel, Buoyancy Effects in

Boundary Layer Adjacent to a Continuous Moving Horizontal Flat Plate, Journal of Heat Transfer, Vol. 102,
1980, p. 170.
[12] N. Ramachandran, B. F. Armaly and T. S. Chen, Correlation for Laminar Mixed Convection on Boundary Layers Adjacent to Inclined Continuous Moving Sheets, International Journal of Heat and Mass Transfer, Vol. 30,
No. 10, 1987, pp. 2196-2199.
[13] S. L. Lee and J. S. Tsai, Cooling of a Continuous Moving Sheet of Finite Thickness in the Presence of Natural
Convection, International Journal of Heat and Mass
Transfer, Vol. 33, No. 3, 1990, pp. 457-464.
[14] M. A. Hossain and H. S. Takhar, Radiation Effect on
Mixed Convection along a Vertical Plate with Uniform
Surface Temperature, International Journal of Heat and
Mass Transfer, Vol. 31, No. 2, 1996, pp. 243-248.
[15] H. S. Takhar, R. S. R. Gorla and V. M. Soundalgekar,
Radiation Effects on MHD Free Convection Flow of a
Gas Past a Semi-Infinite Vertical Plate, The International Journal of Numerical Methods for Heat & Fluid
Flow, Vol. 6, No. 2, 1996, pp. 77-83.
[16] M. A. Hossain, M. A. Alim and D. A. Rees, The Effect
of Radiation on Free Convection from a Porous Vertical
Plate, International Journal of Heat and Mass Transfer,
Vol. 42, No. 7, 1999, pp. 181-191.
[17] J. Gary, D. R. Kassoy, H. Tadjeran and A. Zebib, The
Effects of Significant Viscosity Variation on Convective
Heat Transport in Water-Saturated Porous Media, Journal of Fluid Mechanics, Vol. 117, 1982, pp. 233-249.
[18] K. N. Mehta and S. Sood, Transient Free Convection
Flow with Temperature-Dependent Viscosity in a Fluid
Saturated Porous Medium, International Journal of Engineering Science, Vol. 30, No. 8, 1992, pp. 1083-1087.
[19] S. Mukhopadhyay, G. C. Layek and S. A. Samad, Study
of MHD Boundary Layer Flow over a Heated Stretching
Sheet with Variable Viscosity, The International Journal of Numerical Methods for Heat & Fluid Flow, Vol.
48, No. 21-22, 2005, pp. 4460-4466.
[20] S. Mukhopadhyay and G. C. Layek, Effects of Thermal

Radiation and Variable Fluid Viscosity on Free Convective Flow and Heat Transfer Past a Porous Stretching
Surface, International Journal of Heat and Mass Transfer, Vol. 51, No. 2, 2008, pp. 2167-2178.
[21] M. Q. Brewster, Thermal Radiative Transfer Properties, John Wiley and Sons, Chichester, 1972.
[22] G. K. Batchelor, An Introduction to Fluid Dynamics,
Cambridge University Press, London, 1987.
[23] J. X. Ling and A. Dybbs, Forced Convection over a Flat
Plate Submersed in a Porous Medium: Variable Viscosity
Case Paper 87-WA/HT-23, American Society of Mechanical Engineers, New York, 1987.
[24] S. Gill, A Process for the Step-by-Step Integration of
Differential Equations in an Automatic Digital Computing Machine, Proceedings of the Cambridge Philosophical Society, Vol. 47, No. 1, 1951, pp. 96-108.
ENG
Engineering, 2010, 2, 635-640

Statistical Modeling of Pin Gauge Dimensions of Root of

Gas Turbine Blade in Creep Feed Grinding Process
Ahmad Reza Fazeli
Manufacture and Production Engineer, Mapna Group (Mavadkaran Engineering Co.), Tehran, Iran
E-mail: fazeli@ mavadkaran.com
Received January 28, 2010; revised March 9, 2010; accepted March 13, 2010
Abstract
Creep feed grinding is a recently invented process of material handling. It combines high quality of the piece
surface, productivity, and the possibility of automatic control. The main objectives of this research is to study
the influences of major process parameters and their interactions of creep feed grinding process such as
wheel speed, workpiece speed, grinding depth, and dresser speed on the pin gauge dimensions of root of gas
turbine blade by design of experiments (DOE). Experimental results are analyzed by analysis of variance
(ANOVA) and empirical models of pin gauge dimensions of root are developed. The study found that higher
wheel speed along with slower workpiece speed, lower grinding depth and higher dresser speed, cause to
obtain best conditions for pin gauge dimensions of root.
Keywords: Creep Feed Grinding, Pin gauge dimension, Analysis of Variance, Regression,
Interactive Effect
1. Introduction
Grinding has traditionally been associated with small
rates of material removal and fine finishing operations.
Using an approach known as creep-feed grinding (Figure 1), a large-scale metal removal similar to milling can
be achieved. Using this approach, higher material removal rates can be performed by selection of a higher
depth of cut and lower workpiece speed. The correct
selection of the cutting conditions and the wheel specifications can provide a greater material removal rate and a
finer surface quality. One of the most important applications of creep-feed grinding is the production of the aerospace parts used in jet engines such as turbine vanes,
and blades where parts should have high strength to the
fatigue loads and creep strains. These parts are made
from nickel-based super-alloys such as Inconel, Udimet,
Rene, Waspaloy, and Hastelloy. They provide a higher
strength to weight ratio, and maintain high resistance to
corrosion, mechanical thermal fatigue, and mechanical
and thermal shocks [1].
Vafaeesefat modeled and predicted the grinding forces
of the creep feed grinding of supper-alloy materials using
neural network. This model was then used to select the
working conditions (such as depth of cut, the wheel
speed, and workpiece speeds) to prevent the surface
burning and to maximize the material removal rate. The
results showed that the combination of neural network

and an optimization system is capable of generating optimal process parameters [1].
Wange et al. [2] provided a thermal model that focused on the heat transfer to the fluid, workpiece and
grain for creep feed grinding. In their model, the conduction effect in the moving direction of the workpiece was
considered which was found to be very significant, especially for creep feed grinding. Moreover, the thermal
partition ratios to the workpiece, fluid and grain were
well defined and discussed. The results revealed that the
cooling effect of the fluid is more crucial especially at
larger grinding depth.
Hann [3], Malkin and Anderson [4], Malkin [5], Rowe
and Morgan [6] derived thermal partition and workpiece
Figure 1. Creep-feed grinding of gas turbine blade.

ENG
A. R. FAZELI
636
temperature in dry grinding that failed to take into account the cooling effect of the fluid.
Lavine et al. [7] presented a conical grain model, with
grain slope set to one. Lavine and Jen [8] derived a separate thermal model among the fluid, wheel and workpiece to predict the occurrence of boiling.
Wange et al. [9] depicted that the grinding energy
when the fluid begins to cause boiling is defined as the
critical grinding energy for the workpiece burning. The
results showed that the workpiece burning can be predicted or evaluated to avoid the working conditions of
burning occurrence.
Shafto et al. [10] proposed that workpiece burning
could be explained by the phenomenon of fluid film
boiling.
Ohishi and Furukawa [11] derived the relationship
between the grinding heat flux and grinding zone temperature at burning using the fraction of the grinding
energy entering into the workpiece at 10%.
Wange et al. [12] modeled the grinding force of the
creep feed grinding using the improved back propagation
neural (BPN) network in view of avoiding the workpiece
burning. The results showed that the grinding energy can
be accurately predicted by the application of the grinding
force model and that a larger size of wheel is available to
have a better working efficiency.
Pin gauge dimension is one of the important geometrical dimensions in root of gas turbine blade that plays an
important role in correct assembly of blade on disk of
turbine. If this important dimension is not properly controlled and goes out of its tolerance range (within 0.062),
the blade will not fit on the disk of turbine.
Figure 2 shows the gas turbine blade. Figure 3 illustrates the pin gauge dimension that is one of the important geometrical dimensions in root of gas turbine blade.
In this research, the influences of major process parameters and their interactions of creep feed grinding
process such as wheel speed, workpiece speed, grinding
depth and dresser speed on the root geometrical dimensions of gas turbine blade is studied using design of experiments (DOE).
It is desirable to know the effects of the major parameters and interactive influences among the process
Figure 3. Gas turbine blade.
parameters on root geometrical dimensions and relationship between root geometrical dimensions and process
parameters to obtain the best conditions of parameters
for optimum production.
For modeling and determining the influences of main
parameters and interaction effects among parameters of
the process on root geometrical dimensions, design of
experiments method (DOE) has been employed. DOE is
a statistical method which is used to find the significance
of interactive effects among variables and relations
among process parameters using variance analysis. Finally, using this model and the suitable pin gauge dimension, input parameters has been achieved for optimum
production.
2. Description of Material
We have chosen Inconel 738 LC supper-alloy as the experimental sample. This supper-alloy provides higher
strength to weight ratio, and maintains high resistance to
corrosion, mechanical thermal fatigue, and mechanical
and thermal shocks. The chemical composition of this
supper-alloy is presented in Table 1.
3. Experimental Modeling
3.1. The Output Parameters
Output parameter, pin gauge dimension measured in
terms of mm with inside micrometer 0-25, 0.01 mm precision.
3.2. The Input Parameters
Figure 2. Pin gauge dimension.

Input parameters were selected from the various parameters of creep feed grinding process such as the properties of the work piece material, tools, dresser rotational
speed, rigidity of machine tools and type of coolant. The
selected parameters are:
ENG
A. R. FAZELI
Table 1. The chemical compositions of Inconel 738 LC supper-alloy.
Min
Max
Element
Min
Element
Percentage
C
Max
0.09
0.13
Percentage
Nb
0.6
1.1
Cr
15.7
16.3
Ta
1.5
Co
2.4
2.8
Al
3.2
3.7
Fe
0.3
Ti
3.2
3.7
Si
0.05
(AI + Ti)
6.5
7.2
Mn
0.05
0.007
0.009
0.003
Zr
0.03
0.06
Mo
1.5
Ni
Bal.
- The wheel speed.

- The workpiece speed.
- The grinding depth.
- The dresser speed.
637
Table 2. The parameter levels.
Parameters
Low
Level
High
Level
Wheel speed ( m/s ) V
17
25
workpiece speed (mm/min ) f
100
180
dresser speed (m/rev ) E
0.05
0.15
Grinding depth (mm) P1
0.6
0.9
grinding depth (mm) P2
0.3
0.6
grinding depth (mm) P3
0.04
0.08
formed in three levels. Table 3 indicates the center

points. Figure 4 shows the flow chart of the analysis.
This subject has to attend that we use absolute value of
difference between the measured dimensions and nominal dimension of pin gauge in statistical analysis. Therefore using design of experiments and ANOVA analysis,
according to the input parameters, this absolute value is
minimized.
4. Analysis of the Experimental Results

The analysis of variance (ANOVA) is a statistical method to investigate the importance and effect of the parameters. After statistical calculations and implementa-
3.3. The Experiment Conditions

Grinding wheel type was Strato/Tyrolit (F13A70FF1)
and coolant type was Cutzol zt 130 (oil Canada).
3.4. The Experimental Design

It is difficult and expensive to perform all experiments.
The DOE method can be employed as an efficient technique to accomplish the suitable and necessary experiments with high accuracy. To investigate multiple interactions between parameters [13] in this study, a fractional-factorial design was employed with two levels for
each parameter (+,-), quadrant fraction with resolution
(IV).
Since we have several steps of grinding to accomplish
the grinding of root, steps of grinding are divided to three
sections (P1, P2, and P3) and in each section, we use a
constant grinding depth.
Table 2 shows the input parameters of the process.
The procedure includes 16 experiments. Since the considered levels for each of the input parameters are two
levels, the number of experiments is conducted to determine whether three levels is necessary for each parameter or not, which is called the center points. If these
points were recognized as the effective points by the
analysis of variance, then the experiments should be perCopyright 2010 SciRes.
Figure 4. Flow chart of the analysis.

ENG
A. R. FAZELI
638
Table 3. The applied center points in this study.

Run
P1 (mm)
P2 (mm)
P3 (mm)
V (m/s)
F (mm/min)
E (m/rev)
0.75
0.45
0.06
140
0.1
21
0.75
0.45
0.06
140
0.1
21
0.75
0.45
0.06
140
0.1
21
tion of the F-test on the experimental data by ANOVA,

probability values of each parameter are extracted from
the table of variance analysis. The risk level is considered as 0.05 for the ANOVA.
Once the experimental results are obtained, the coefficients and analysis of variance (ANOVA) are calculated
with MINI TAB software to determine the significance
of the parameters, and P-Values are used to determine
which parameter is most significant. The F-ratio test is
conducted to check the adequacy for the proposed model.
Through experiments, pin gauge dimensions are collected and then fed into a DOE/STAT program to construct statistical regression equations in order to achieve
the initializing of input parameters for optimum production.
After the initial variance analysis and elimination of
the unimportant parameters (with low effect coefficient)
and use of projection (due to lack of repeat), and with
regards to the calculated values of F and P for each one
of the effective parameters which is extracted from the
table of variance analysis, it can be concluded that the
center points have no effect (P = 0.175). Therefore, the
two levels design is appropriate and we do not need to
consider the effective parameters in three or more levels.
The risk level of less than 0.05 for parameters in Table 4 shows that the related parameter is significant. The
R squared and the adjusted R squared are shown in bottom of the Table 4. Also, the lack of fitness is insignificant which shows the adequacy of the developed model.
Figure 5 indicates the residuals analysis graph of the
regression model. As it is observed, the residuals have a
normal distribution.
Figure 6 shows the graphs of each input parameter
effect on the pin gauge dimension. Figure 7 indicates
interactions effects of the parameters on the pin gauge
dimension.
Figure 7 shows that for the pin gauge dimension there
are significant interactive influences among grinding
depth (first section) and workpiece speed.
According to the graphs of mean parameter effect and
the graphs of the parametric interactions effect, higher
wheel speed, dresser speed, grinding depth (first section)
and lower grinding depth (second and third section) and
slower workpiece speed, lead to smaller absolute value
of pin gauge dimension.
Finally, a hierarchical model was developed for pin
gauge dimension by multiple linear regression technique.
The insignificant terms were removed from the model
Figure 5. Residuals analysis graph of the regression model.
Figure 6. The graphs of mean parameter effect on the pin

gauge dimension.
Figure 7. The graphs of the parametric interactions effect

on the pin gauge dimension.
ENG
A. R. FAZELI
Parameters
Coefficient
P-Value
Constant
0.3
0.001
P1
0.683
0.174
P2
0.067
0. 758
P3
0.460
0.001
0.04
0.74
0.017
0.012
0.012
P1 P2
0.222
0.19
P1 P3
1.166
0.322
P1 f
0.001
0.032
P1 E
1.366
0.04
P1 V
0.028
0.01
R-Sq = %97.89
sion.
Figure 10 summarizes the dresser speed on the pin
gauge dimension at grinding depth (first section). The
results show that an increase of dresser speed combined
with the increase of grinding depth (first section), produces small absolute value of difference between the
measured dimensions and nominal dimension of pin
gauge dimension.
Reasonably, with higher wheel speed, grinding depth
(first section) and slower workpiece speed, machining
forces apply equally on the other side of root. Therefore
Pin gauge dimension R1
Table 4: The variance analysis (ANOVA) for the pin gauge

dimension of root of blade
R-Sq(adj) = %89.43
639
Wheel speed (m/s)

Grinding depth (mm)
R1 0.3 0.683(P1) 0.067(P2) P3 0.001(f )

0.74(E) 0.012(V) 0.222(P1 P2) 1.166(P1 P3)
0.001(P1 f ) 1.366(P1 E) - 0.028(P1 V)
(1)
R 2 0.33 0.738(P1) 0.56(P2) 0.075P3 0.0002(f )
0.21(E) 0.014(V) 1.105(P1 P2) 0.031(P1 V)
0.019(P2 V)
Figure 8. Effect of the wheel speed on the pin gauge dimension at grinding depth (first section).
Pin gauge dimension R1
and the final models were developed with significant

terms which were determined by ANOVA Equation (1)
for pin gauge dimension.
(2)
R3 0.13 0.453(P1) 0.033(P2) 0.0005(f )
0.255(E) 0.006(V) 0.683(P1 E) 0.272(P1 P2)
0.0008(P1 f ) 0.11(P2 V) - 0.018(P1 V)
(3)
workpiece speed (mm/min)

Grinding depth (mm)
Figure 9. Effects of the workpiece speed on the pin gauge

dimension at grinding depth (first section).
5. Discussion
Figure 8 summarizes the wheel speed on the pin gauge
dimension at grinding depth (first section). The results
show that an increase of wheel speed combined with the
increase of grinding depth (first section), produces small
absolute value of pin gauge dimension.
Figure 9 shows the effect of workpiece speed on the
pin gauge dimension at grinding depth (first section).
The results show that a decrease of workpiece speed
combined with the increase of grinding depth (first section) produces small absolute value of pin gauge dimenCopyright 2010 SciRes.
Grinding depth (mm)
Figure 10. Effects of the dresser speed on the pin gauge

dimension at grinding depth (first section).
ENG
A. R. FAZELI
640
absolute value of pin gauge dimension decreases.
6. Conclusions
In this study, the creep feed grinding process has been
optimized by selection of significant input parameters
including the wheel speed, dresser speed, grinding depth
and slower workpiece speed. Finally, by means of
ANOVA, the main effects of the input parameters and
their interactions on the pin gauge dimension were determined. Based on the statistical analysis of the experimental data the following conclusions can be obtained.
1) According to the variance analysis and the effect of
interactions between the input parameters, it can be concluded that with higher wheel speed, dresser speed,
grinding depth (first section) and lower grinding depth
(second and third section) and slower workpiece speed,
the pin gauge dimension decreases and as a result the pin
gauge dimension reaches a suitable level.
2) In the creep feed grinding process center points
have insignificant effects on the pin gauge dimension. It
means that the process can be modeled with two levels
for each input parameters.
3) Finally, with the large number of effective parameters in the creep feed grinding process, consideration of
the creep feed grinding process through the design of
experiments is shown to be the efficient method for
achieving the acceptable results.
7. References
[1]
A. Vafaeesefat, Optimum Creep Feed Grinding Process

Conditions for Rene 80 Supper Alloy Using Neural Network, International Journal of Precision Engineering
and Manufacturing, Vol. 10, No. 3, 2009, pp. 5-11.
[2]
S.-B. Wang and H.-S. Kou, Selections of Working Conditions for Creep Feed Grinding. Part (I)Thermal Partition ratios, International Journal of Advanced Manu-
[3]
[4]
facturing Technology, Vol. 23, No. 6, 2004, pp. 700-706.

R. S. Hahn, On the Nature of the Grinding Process, In:
Proceedings of 3rd MTDR Conference, London, 1963, pp.
129-154.
S. Malkin and R. B. Andersson, Thermal Analysis of the
Grinding. Part I: Energy Partition, Journal of Industrial
Engineering, Vol. 96, 1974, pp. 1177-1183.
[5]
S. Malkin, Thermal Analysis of the Grinding. Part II

Surface Temperature and Workpiece Burn, Journal of
Engineering for Industry, Vol. 96, 1974, pp. 1184-1191.
[6]
W. B. Rowe and M. N. Morgan, A Simplified Approach

to Control of Thermal Damage in Grinding, Annals
CIRP, Vol. 45, No. 1, 1996, pp. 299-302.
[7]
S. Lavine, S. Malkin and T. C. Jen, Thermal Aspects of

Grinding with CBN Wheels, Annals CIRP, Vol. 38, No.
1, 1989, pp. 557-560.
[8]
S. Lavine and T. C. Jen, Coupled Heat Transfer to

Workpiece, Wheel, and Fluid in Grinding, and the Occurrence of Workpiece Burn, International Journal of
Heat Mass Transfer, Vol. 34, No. 45, 1991, pp. 983-992.
[9]
S.-B. Wang and H.-S. Kou, Selections of Working Conditions for Creep Feed Grinding. Part (II): Workpiece
Temperature and Critical Grinding Energy for Burning,
International Journal of Advanced Manufacturing Technology, Vol. 28, No. 1-2, 2006, pp. 38-44.
[10] G. R. Shafto, T. D. Howes and C. Andrew, Thermal

Aspects of Creep Feed Grinding, 16th Machine Tool
Design Research Conference, Manchester, England, 1975,
pp. 31-37.
[11] S. Ohishi and Y. Furukawa, Analysis of Workpiece
Temperature and Grinding Burn in Creep Feed Grinding, Bull JSME, Vol. 28, No. 242, 1985, pp. 1775-1781.
[12] S.-B. Wang and H.-S. Kou, Selections of Working Conditions for Creep Feed Grinding. Part (III): Avoidance of
the Workpiece Burning by Using Improved BP Neural
Network, International Journal of Advanced Manufacturing Technology, Vol. 28, No. 1, 2006, pp. 31-37.
[13] C. Montgomery, Design of Experimental & Statistical
Modeling, McGraw Hill, Inc., New York, 2005.
ENG
Engineering, 2010, 2, 641-6647

doi:10.4236/engg.2010.28082 Puublished Onlinee August 2010 (hhttp://www.SciR
RP.org/journal/eeng).
Wind Turbine Tower

W
T
O
Optimizzation under
u
V
Various
Requiremen
nts by U
Using Geenetic Algorith
A
hm
Serdar
S
Yldrm, brahim
m zkol
Faculty of A
Aeronautics and
d Astronauticss, Istanbul Tecchnical Universsity, Istanbul, Turkey
E-m
mail: serdar.yild
dirim-2@postggrad.manchester.ac.uk, ozko
ol@itu.edu.tr
Receivved February 24, 2010; revissed April 7, 20010; accepted April
A
11, 20100
Abstract
The purpose of this studyy is to optim
mize the mass of 1.5 MW wind turbinee steel tower performing Genetic
Algorithm method
m
(GA). In accordancce with ASCE
E 7-98, AISC
C-89 and IEC
C61400-1 , the impact of lloads on
tower is calculated withinn the highest safety condittions against buckling
b
strength of each sections of toower by
means of GA
A codes. Thee stifness alon
ng tower is eensured entirrely while thee mass of tow
wer is mitigaated and
optimized.
Keywords: Mass
M Optimizzation, Genetiic Algorithm, Wind Turbiine Tower
1. Introducction
2. Alloowable Streess Design (ASD)

(
The cost of wiind turbine tow

wers can amou
unt nearly 20-225%
of the total investment cost for wind enerrgy plant. Minnim of wind tuurbine tower haas become morre
mization of mass
crucial job for
fo last two ddecades. Mostt modern winnd
turbines are installed
i
with tubular coniccal steel towerrs
from the aspeect of aestheticcs. They are id
deally manufacctured in 20-330 meters lonng welded secctions and theen
bolted each otther on site [1]]. Steel tubularr conical towerrs
are manufactuured as the tapered steel tower namely thesse
towers have a conical shape with a wider base
b than the toop
in general. Suuch designs inccrease their streength and savees
material. The tower of winnd turbine gaathers net loadds
wer head and transmits thesse loads to thhe
from the tow
foundation. Thhe main load is the axial load on the rotoor.
Dynamic loadding is generaated by wind turbulence annd
constantly byy blade tower interaction. The
T stiffness oof
tower is baseed on the tower top weightt and the toweer
height. Additiional design reequirements have
h
to be satisfied with adeqquate strength since admissiible stresses arre
not exceeded and for conicaal towers, shell buckling muust
be prevented. The existing pre-sized tow
wer is tackled tto
p analysis aand design co
onditions. Steel
evaluate as per
tower is assuumed to be loocated in Balkesir-Bandrm
ma
region in Turkkey with 52 m tower height and the 54.7 m
tower hub heiight. The top diameter of to
ower is 2.56 m
and the base diameter
d
towerr is 4.3 m.
The steeel towers are primarily

p
desiggned and sizedd to meet
the AIS
SC strength design
d
criteria.. Allowable sttress design method
m
(ASD) is used in lieeu of AISC-899 for the
steel tuubular tower deesign. The load combinationn method
for the service load (characteristic
(
load) conditioon is carE-7-98. The allowable
ried ouut with refereence to ASCE
bendingg stress Fb fo
or noncompactt section is 0.6 Fy, in
which the
t yielding sttress Fy of the steel tubular sstructure
is typiccally 345 MPaa (50 ksi). The allowable sheear stress
Fv is 0.4
0 Fy. The allowable
a
com
mpression stresss Fa is
represeented by the following formuula [2]:
L 2
(K r )
1
..Fy
2.Cc 2
Fa
(1)
L
L
3( K ) ( K )3
5
r
r
3
8Cc
8Cc 3
Copyright 2010
2
SciRes.
L
r
of steel ttower, K
) is thee slenderness ratio
r
= 2 forr the cantileveer type of structure, and L aand r are
the lenggth of the toweer and radius of
o section, resppectively.
The maaterial coefficient Cc is calculated by:
where ( K
Cc
2. 2 .E
Fy
(2)
ENG
S. YILDIRIM ET AL.
642
where E is the steel modulus. When KL/r is greater than

Cc , the allowable compression stress Fa shall be recalculated by:
12. 2 .E
Fa
(3)
L
23.( K )2
r
Typically, the ratio of the applied axial compression
stress fa to the allowable compression stress Fa of the
steel tower is less than 0.15. The combined stress for the
applied bending stress fb acting on the steel tower shall
be satisfied with interaction equation.
fa fb
1
Fa Fb
(4)
The applied shear stress fv from the torsion and the

shear force on the tower shall also be less than the allowable shear stress Fv [2].
3. Fatigue Load
The DEL (Damage Equivalent Load) method facilitates
to determine the steel tower preliminary dimensions in
any circumstances which fatigue load histogram data
does not exist. The SN curve for the DEL method can be
expressed in the following [2]:
loglog s n loglog 80 MPa
2 106 n
m
(5)
The number of cycles corresponding to the withstand

limit along the tower height z can be calculated by using
DEL method.
N z
Mf z
rmax S z .m N0
(6)
where:
Mf ( z ) is the moment produced by the fatigue DEL
thrust along steel tower. S z is the section modulus
that varies along the height of tower. rmax is the
maximum allowable stress range at N 0 cycles (typi-
cally 104 ) . m is the slope of the curve [2].
3.1. Damage Equivalent Load for Steel

Tower
There are many fatigue calculation methods. One of
which is DEL method in most cases where the full histogram of fatigue cycles is available but only a DEL specified. The DEL is added by a value of SNslope (m = 4
used in this circumstance and a number of cycles (Ne)).
Total moment range along the tower is calculated as
follows:
max max Mx, yT max max Mx, yB z

Mx, y z
h
max max Mx, yB
(7)
max(Mx,yT) = Maximum moment range at tower top
x or y direction .
max(Mx,yB) = Maximum moment range at tower
base x or y direction.
Safety Factor of DEL is 1.0.
Consequence failure factor and material factor: sd.m
= 1.15 1.1 = 1.265
Number of cycles: 5.29 108 for 1.5 MW turbine
This represents a 20 year lifetime.
4. Local Buckling Stress

The strength of the tubular steel tower in axial compression is the lesser of the yield strength and the elastic critical buckling stress cr is calculated:
t
(8)
cr 0.605 E.
r
where r is the cylinder radius and t is the wall thickness.
However, the presence of imperfections, particularly
those introduced by welding, will significantly reduce
the tower wall resistance to buckling. As per steel tower
design, the reduction coefficient
for axial load is
found by:
0.83
1 0.01. r if r 212
t
t
0
(9)
0.70
r
212
if
t
r
0.1
0.01
.
The reduction coefficient B for bending load is calculated as follows:

B 0.1887 0.8113o
(10)
The buckling stress u can be computed in terms of the
yielding stress Fy:
0.6

Fy if B . cr Fy
Fy 1 0.4123

2 (11)
au
B . cr
Fy
if B . cr
0.175. B . cr
2
The maximum applied stress a combined with normal
stress and shear stress is calculated by a :
fa fb
3 fv 2
(12)
The unity ratio for the combined stresses is found as

follows. Now that the steel tower is liable to combined
stress with axial compression and bending moment, the
steel tower is designed to satisfy the combined stress
check. This check named unity check interaction equaENG
S. YILDIRIM ET AL.
tion is carried out in accordance with the AISC manual

(ASD 9th Edition).
fa fb
1 fb 0.15 Fb
(13)
Fa Fb
where:
fa
is the applied compression stress
Fa is the allowable stress
fb
is the applied bending stress
Fb is the allowable bending stress.
5. Earthquake Load
This section is based on ASCE 7-98 Earthquake Load
Specification. Even though earthquake load seems to be
not much significant effect on design of steel tower because of the fact that wind turbine towers are placed in
low seismic areas, earthquake load should be taken into
consideration so as to be more precise in designing of
steel tubular tower (see Figure 1).
The maximum considered earthquake spectral response acceleration for short periods ( SMS ) and at 1
adjusted for site class effects, should be
second
determined by [3]:
S MS Fa.Ss; S M 1 Fv.S1
(14)
where :
Mapped maximum considered earthquake spectral response acceleration at a period of 1s as determined
in accordance with Section 9.4.1 (ASCE 7-89).
Ss Mapped maximum considered earthquake spectral response acceleration at short periods as determined
in accordance with Section 9.4.1 (ASCE 7-89).
Fa and Fv are defined in Tables 9.4.1.2.4a and b
respectively in accordance with Section 9.4.1 (ASCE
7-89). According to the ASCE 7-89 9.4.1.2.5 design
spectral response acceleration at short periods, S DS and
at 1 s period S D1 shall be determined from Equations
9.4.1.2.5-1 and 9.4.1.2.5-2 respectively:
S DS
2
2
S MS ; S D1 S M 1
3
3
Figure 1. Earthquake overturning moment.
(15)
643
From the earthquake geographic map, the maximum

considered earthquake (MCE) ground motion for soil site
Category B with 5% damping is 1.5 g (Ss) for structures
with a short period of 0.2 s and 0.6 g (S1) for structures
with a period of 1 s. The wind turbine towers are typically located in open areas away from population centers
with very low occupancy. Because, the occupancy importance factor (I) is equal to 1.0. Site Classi- fication D
is assumed for Balkesir-Bandrma/Marmara Region.
Site Classification D is typified by stiff soils with shear
velocity (Vs in soil) typically 6001,200 fps (183366
m/s).
2
2
S DS Fs S S 1.0 g ; S DI Fv S I 0.6 g
(16)
3
3
Fa and Fv can be defined according to the ASCE
7-98 Tables 9.4.1.2.4a-4b respectively. Fa is the site
coefficient as a function of site class and short period
MCE. Fv is the site coefficient as a function of site
class and a 1 second period MCEg is the acceleration
caused by gravity.
S DI
, if T Ts
T
(17)
S a T S DS 0.4 0.6
T T0
T
0 , if
S
Otherwise
DS
TS S DI / S DS ; T0 0.2.Ts
(18)
T is the structural period.
Spectral acceleration response can be found as in Figure 2.
5.1. Design Earthquake Load

The earthquake lateral load affects the whole tower
height h as per its weight distribution [2].
h
W w z dz WHead Mass
(19)
w(z) is weight distribution as a function of height. W is

the total weight of steel tower with Turbine Head. Base
Figure 2. Spectra acceleration (earthquake).
ENG
S. YILDIRIM ET AL.
644
shear coefficient is,
where z g the nominal height of the atmospheric boun-
I
(20)
Cs T Sa t .
R
I is the importance factor. R is the reduction factor
equal to 1.
Base Shear is V C s T W (Figure 3)
(21)
dary layer is 213 m and 1 is 11.5 for exposure D category in accordance with ASCE 7-98.
The towers are assumed to be located in flat unobstructed area for direct wind exposure Category D where
wind flows over the open water and flat terrain.
Importance factor is 1.0 for low occupancy concerning
the wind turbine erection and installation.
The direct wind load on the tower is not only based on

the direct wind pressure on the tower but also on the gust
factor G f and the force coefficient C f .
G f is calculated by the following equation [2] :
1 1.7 I g 2 Q 2 g 2 R 2
z
Q
R
G f 0.925
1 1.7 g v I z
6. Wind Velocity Pressure

The velocity pressure: (Figure 4)
N
qz 0.613K z K zt K d V 2 2
m
6.1. Direct Wind Load on Tower
(22)
where:
The topographic factor K zt is 1.0 for the flat area.
K d is 0.95 for a round cylinder tower in accordance
with Table 6-6 in ASCE 7-98.
The terrain exposure coefficient is determined as per
Table 6-5 of ASCE 7-98 or by the following formula [3]:
2
15 ft 1
) ifz 15 ft
2.01(
zg
Kz z
(23)
2
z 1
otherwise
2.01( z )
g
(24)
where
The intensity factor of turbulence :
I z 0.15(33 ft / z )1/ 6
(25)
The background response Q and the resonant response

are given in accordance with Eq.6.4 of ASCE 7-98.
1
(26)
B h 0.63
1 0.63(
)
Lz
where B is the horizontal dimension of tower measured
normal to wind direction, Lz is the integral length scale
of the turbulence at the equivalent height given by
Lz l ( z / 33 ft ) . l and are constants listed in Table
6.4 of ASCE 7-98.
g R and gQ shall be taken as the constant value of
Q
3.4 and g v is given by:

g v 2 ln(3600n1 )
0.577
2 ln(3600n1 )
(27)
R , the resonant response factor is given by:

1
Rn Rh RB 0.53 0.47 RL
7.47 N1
Rn
(1 10.3N1 )5/3
Figure 3. Earthquake shear force.
(28)
(29)
The force coefficient C f is determined as per Tables

6-10 of ASCE 7-98. Lateral wind load along the tower is
calculated by the direct pressure on the projected area
which differs with respect to diameter distribution d(z) of
tower. Fz ( z ) is determined in the following equation.
Fz z q z G f C f d z
(Figure 5)
(30)
Vz z Fz x dx
(Figure 6)
(31)
Figure 4. Wind velocity pressure on tower.

Mz z Fz z .( x z ) dx
(Figure 7)
(32)
ENG
S. YILDIRIM ET AL.
645
the height of tower. E and I are elastic modulus and moment of inertia of the tower. Wt and Wtow are the
weight of head mass and tower mass respectively.The
natural frequency in Hz is calculated by: ft wt / 2 .
8. Genetic Algorithm Approach

Figure 5. Force distributions because of direct wind effect on
tower.
Figure 6. Wind shear force along tower.
Genetic Algorithms is one of the methods used in optimization problems. Particularly, it is based on natural
selection. Genetic Algorithms is dependent on that the
best generation has to live in nature. Although many genetic algorithms have been said with different structures,
all of them comprises of three basic operations. Genetic
Algorithm uses reproduction, crossover and mutation
operators to define fitness and to create new solutions.
Reproduction is simply a process to make decision which
strings should remain and how many copies of them
should be produced in the pool. The decision is made by
comparing the fitness of each string. The fitness indicates
survival potential and reproduction efficiency of the
string in the next generations. For an Optimization problem, the fitness function is the objective function of optimization problem as shown in Figure 8.
9. Optimization Problem
Figure 7. Wind effect moment along tower.
7. Dynamic Behavior of Steel Tower

Main key consideration in wind turbine design is the
avoidance of resonant tower oscillations excited by rotor
thrust fluctuations at rotational or blade passing frequency. Natural frequency of tower greater than the
blade-passing frequency is said to be stiff on the other
hand towers natural frequency between rotational frequency and blade passing frequency are regarded as soft.
If the natural frequency, the tower is said to be soft-soft
[4].
EIg
(33)
wt 1.75
H 3 Wt 0.25Wtow
where
wt is the estimated natural frequency of the tower is
Optimization problems are generally expressed as given

in the following:
p1(x): Margins of safety combined stress (bending
stress and shear stress for torsion) because of the wind
load effect.
p2(x): Combined stress (bending stress and shear
stress) because of the earthquake load effect.
p3(x): Natural frequency for the 1st mode bending.
p4(x): Fatigue stress.
Mtower: Mass of turbine tower.
Minimize
(34)
f(x) = Mtower(x).(1+p1(x) +p2(x) + p3(x) + p4(x))
Constraints 12 t xi t xi 1 .. t xi 51 26 the thicknesses of sections
(35)
i = 1 ,2 ,.........................,51
Figure 8. Evaluation of fitness function.
ENG
S. YILDIRIM ET AL.
646
U_C_WL( x) DCRsw( x) (Figure 9).

Unity check of critical combined buckling stress ratio
due to wind effect load against combined stress ratio:
U_C_EQ ( x) DCRsq ( x) (Figure 10).
Unity check of critical combined buckling stress ratio

U_C_EQ( x) due to earthquake effect load against
combined stress ratio DCRsq ( x) . wt wr , wt Natural
Figure 9. Buckling unity check for wind turbine effect and

direct wind load.
Figure 10. Buckling unity check for earthquake load (EQ).
WIND TURBINE DATA

1) Thrust Force and Moment
2) Initial Tower
INPUT DATA
1) General Specifications
2) Material Characteristics
3) Structural Parameters
4) Load conditions
5) GA Parameters
Load Calculation
6) ASCE 7-98/IEC/Eurocode
Loads
7) Safety Factors
OPTIMIZATION PROGRAM
1) Natural frequency
2) Extreme Loads
3) Fatigue damage
4) Fitness function
OUTPUT PARAMETERS
1) Wall thicknesses of each segment
Figure 11. Design flow.
ENG
S. YILDIRIM ET AL.
frequency of tower1st mode wr Operation frequency of
of turbine .
Design flow is shown as follows. This GA structure
minimizes tower mass subject to general dimensions,
design loads and some design restrictions. Load calculation depends on wind turbine design requirements of the
standard IEC61400-1 and ASCE 7-98. All extreme loads
of tower sections are calculated by the load combination
[5] as shown in Figure 11.
10. Optimization Results

Figure 12. Thickness distribution.
Score in figure above represents mass of steel tower in kg.

After 40th generation of GA, value of mass levels out. In
other words, optimization result has been reached at this
point.
further as longas stiffness is obtained.
11. Conclusions
[1]
R. Y. Redlinger, P. D. Andersen and P. E. Morthorst,

Wind Energy in the 21st Century, Economics, Policy,
Technology and the Cahnging Electricity Industry, New
York, 2002.
[2]
LWST Phase 1. Project Conceptual Design Study:

Evaluation of Design and Construction Approaches for
Economical Hybrid Steel/Concrete Wind Turbine Towers, 28 June 2002-31 July 2004, NREL Subcontractor
Report, January 2005 NREL/SR-500-36777.
[3]
ASCE 7-98, Minimum Design Loads for Buildings and

Other Structures, 1998.
[4]
R. Harrison, E. Hau and H. Snel, Large Wind Turbines,

Design and Economics, 2000.
[5]
IEC 61400-1, Wind Turbine Generator System Part 1

Safety Requirements, 1999.
Each 1 m section along tower represented a chromosome

in GA. Each section has been evaluated step by step on
the basis of buckling strength in GA. An objective
function has been produced by using a genetic algorithm.
This optimizes the thickness of steel tower ranging from
top 12 mm to base 26 mm in the distribution of pattern.
All in all, the thicknesses of tower have been evaluated
separately in each 1 m section along the height of the
tower. Within the best solution of current conditions, the
weight of tower has been obtained 63000 kg with a type
of S355J0 material quality. Andit gives results for the
best solution as indicated above in Figure 12. Moreover,
the upcoming studies can be carried out and developed-
12. References
ENG
Engineering, 2010, 2, 648-657

A Device that can Produce Net Impulse Using Rotating

Masses
Christopher G. Provatidis
Laboratory of Dynamics & Structures, Mechanical Design & Control Systems Section, School of Mechanical
Engineering, National Technical University of Athens, Athens, Greece
E-mail: cprovat@central.ntua.gr
Received March 20, 2010; revised July 6, 2010; accepted July 9, 2010
Abstract
This paper describes a device capable of producing net impulse, through two synchronized masses, which
move along a figure-eight-shaped orbit. In addition to the detailed description of the mechanical components
of this device, particular attention is paid to the theoretical treatment of the innovative principle on which the
device is based. In more details, the mechanical system consists of two independent but simultaneous rotations, the former being related to the formation of the figure-eight-shaped path and the latter to an additional
spinning. Based on the parametric equations of motion of the lumped masses, and considering semi-static
tensile deformation of the connecting rods carrying them, it was found that the resultant impulse towards the
direction of the spin vector includes a non-vanishing term that is linearly proportional to the time. In addition,
reduced but encouraging experimental results are reported. These findings sustain the capability of the proposed mechanism to achieve propulsion.
Keywords: Inertial Propulsion, Centrifugal Force, Net Impulse, Rotating Figure-Eight, Mechanism
1. Introduction
The matter of the inertial propulsion utilizing eccentric
masses aiming at producing limited motion of the object
to which they are attached, is an old topic [1-4]. The
general impression is that these masses lead to periodic
oscillations in which the synchronization of participating
masses plays a significant role [5], while chaotic phenomena may also appear [6,7].
In the particular case of possible space propulsion,
fifty years ago Norman Dean (a civil service employee
residing in Washington DC) proposed the use of two
contra-rotating eccentric masses in order to convert rotary motion to unidirectional motion [4]. He claimed that
in this way one could achieve thrust thus producing motion of the object to which this system was attached.
Since then, Deans mechanism was internationally named as Dean drive or Dean space drive (the interested
reader may consult, for example, http://en.wikipedia.
org/wiki/Dean_drive). However, despite the extremely
high number of internet references as well as the many
articles sited in popular mechanics or science fiction magazines, a very small number of scientific papers exist in
the open literature. A careful search reveals an old paper
in the Russian language [8], two remarks in a textbook

[9], as well as a general NASA report dealing with many
possible mechanical antigravity concepts (including
Deans drive) [10] and a relevant review paper [11].
Moreover, quite recently Provatidis [12] has shown that
Deans drive practically works like a catapult while a
variable angular velocity can only control the smoothness of the object velocity to which the drive is attached.
In brief, as it will be discussed below, the main disadvantage of Deans drive is due to the circular paths on
which the eccentric masses move.
Motivated by the abovementioned findings on Deans
drive [12], this paper revisits the subject and shows in a
theoretical way that, the shape of the path on which the
lumped masses move, is of major importance. For example, if one considers a lumped mass at the end of an elastic bar of radius r , which rotates at a constant angular
velocity on a vertical plane, the mass moves along
an ideally circular path (C) because the induced centrifugal force has a constant value thus leading to a permanent tensile elongation r of the rod (final radius:
r r r ). In this case, a rigid-body analysis is allowable, and also, due to the geometrical symmetry of the
path as well as due to the fact that every 360 degrees the
ENG
C. G. PROVATIDIS
mass takes exactly the same position possessing exactly

the same velocity, the net impulse caused by the centrifugal force vanishes. In more details, when the mass
draws the upper half of the circle the corresponding impulse is positive, while when it draws the lower half it
becomes negative (both of equal absolute value).
Within this context, this paper investigates the possibility of strengthening the impulse on the upper part of
the circular path (appearing in Dean drive) with respect
to the lower one. A possible solution to this problem,
which has been previously presented [13,14] but it is
fully explained here for the first time, is to transform the
circle to a different shape. It is proposed to achieve it
by deforming the circle in two successive ways. First
the circle is folded by rotating its lower part around the
vertical axis of symmetry thus producing a crossed figure-eight-shape, which entirely lies on the vertical plane.
Second the latter planar path is further bent in such a way
that it perfectly lies over the surface of a hemisphere, the
latter having a center O and a radius r . These two
successive deformation steps lead to a new, fully threedimensional, curvilinear path that lies entirely above or
entirely below the center of the hemisphere; henceforth it
is called figure-eight-shaped path. It is clarified that in
this final configuration of the mechanism, the immobile
end of every connecting bar is pinned to the centre of the
hemisphere while the second end carries the corresponding mass. Consequently, one could say that-in this waythe proposed procedure achieves to create a new path on
which only the upper, or only the lower half of the initially considered circular path (C), operate. Despite this
fact, it has been theoretically verified that the maximum
upward force is equal and opposite to the maximum
downward force thus net propulsion is still impossible
[15]. A mechanical device capable of producing the aforementioned figure-eight-shaped path is presented in Section 2 and it is shown in Figure 1.
649
In order to increase the maximum upward force with

respect to the maximum downward one, the abovementioned figure-eight-shaped curve is further modified as
follows. Another rotation z is imposed around the
vertical axis of symmetry, which obviously passes through the center O of the hemisphere. Due to the additional
centrifugal forces ought to the rotation z , when the
elastic rod is found at the horizontal position is suffers
much larger tensile force than that it suffers at the vertical position; therefore, the instantaneous elongated radius
at the horizontal position is higher than that of the vertical position. As a result, the induced vertical inertial
forces - which are always proportional to the instantaneous radius r (cf. Equation (23)) - at the horizontal position are higher than those corresponding to the vertical
position (a full explanation will be provided in Sections
5-7). In this way, the proposed mechanism that consists
of a rotating figure-eight is capable of producing net
impulse, a finding that constitutes the novel feature of
this paper. The full description of the demonstration device, an experimental result, and particularly the theoretical treatment of the involved mechanics, are presented here for the first time.
The structure of this paper is as follows. Section 2 presents details of the proposed device concerning its construction features, operation and design parameters. Then,
a theoretical treatment follows (Section 3 presents analytical closed-form expressions of the figure-eight-shaped
path, Section 4 presents the velocities and the accelerations of the lumped masses, Section 5 presents the expression for the resultant inertial forces, Section 6 deals
with the elastic deformation of the connecting bars and
Section 7 presents closed-form expressions for the resultant impulse). Experimental results are presented in Section 8, while a discussion follows in Section 9.
2. Presentation of the Innovative Device

2.1. Construction Features
Figure 1. The prototype mechanism. In the left part the

arrows show the directions of the two simultaneous rotations, while in the right part the figure-eight-shaped path is
clearly illustrated for two different views.
An overview of the proposed device, which has been

developed so as to produce the abovementioned figure-eight-shaped path and its further spinning, is shown
in Figure 2. The mechanism consists of a frame (1) on
which two electric motors (M1, M2) are attached. The
mechanism consists of a horizontal shaft (2) supported
by the frame (1) guided by the electric motor (M1); in
this implementation through the shaft (7) and belt (8)
towards the end (6) of the shaft. The shaft (2) includes in
its interior other shafts that drive an attached planetary
system (3) that is positioned preferably in its middle.
Two rods (4,5) are attached in the ends of the spin gears
axis of revolution (9) where masses (a, b) are attached
and regularly move on the path. During the rotation of
the electric motor (M1), the masses (a, b) move along
ENG
650
C. G. PROVATIDIS
a figure-eight-shaped path (shown in the right part of

Figure 1), which entirely belongs to the surface of a
sphere the center of which is the intersection of the horizontal axis (2) with the axis of the spin gears (9); in other
words, the center of the planetic system (3). Finally, the
end of the shaft (2) includes a wheel (10), while the other
end is driven by the motor (M1); in this specific case
through the belt (8). Also, a second electric motor (M2)
is illustrated at the bottom of the frame (1), which causes
the rotation of the entire frame and consequently of the
shaft (2) as well as of the attached masses (a, b).
The central part of the abovementioned device is the
planetary system. To make it more clearly to the reader,
Figure 3 is a cross-section of the horizontal shaft (2) and
the attached planetary system (3). The latter consists of a
casing (11) that expands almost until the ends of the
shaft (2). Internally, the planetary system (3) consists of
some planet gears and spin gears. Again, the case of two
planet gears (P1,P2) and two spin gears (S1,S2) is quite
indicative. The planet gears and the spin gears are supported on the casing through the rolling bearings
(211,212) and (213,214), respectively (212 and 214 are
not shown in Figure 3). In this case, the planet gear (P1)
is firmly fixed to the right half of the internal shaft (22)
that ends to the point (6), while the planet gear (P2) is
firmly fixed to the other half of the internal shaft (21)
that ends to the wheel (10). Finally, the rods (4,5) are
again shown together with the corresponding attached
masses (a,b), which have been already mentioned in
Figure 2.
locity shaft motor where 1 is the speed reducetion of the transmission between the motor M1 and the
right half of the shaft (2). Thus power transmission is
performed through P1-S1 towards the mass a. Similarly,
the rest half of the power produced by the motor M1 is
transmitted through P1-S2 towards the other mass b. A
characteristic of this mechanism is that the second planet
gear, P2, is fixed thus causing rolling of the spin gears S1
and S2 on P2. Obviously, the rotation of the planet gear
P1 enforces the spin gear S1 to rotate about its local axis
(initially coinciding with the global z-axis) and also enforces the casing (11) to rotate around x-axis.
2.2. Operation
In brief, the motor M1 rotates at an angular velocity
motor and drives the planet gear (P1) at an angular ve-
Figure 2. Sketch of the proposed mechanism, in which the

lumped masses a and b are driven by the electric motors
M1 and M2 through a motion transmission system.
Figure 3. Abstractive sketch of the planetary system. The mechanical power flows through the planet gear (P1) to the spin
gears (S1) and (S2). The lumped masses a and b are attached to the free ends of the rods (4) and (5), respectively, which
are firmly connected with the aforementioned spin gears (S1) and (S2).
ENG
C. G. PROVATIDIS
cos t
R 01 sin t
When assuming the same diameters Rm of the four

gears (P1, P2, S1 and S2), due to the aforementioned
rolling at the interface between P2 and S1, the following
conditions occur:
the spin gear S1 rotates at an angular velocity
, which is half of that of the shaft ( shaft 2
motor 2 ),
the spin gear S2 rotates at the same angular velocity
but of opposite sign, ,
the casing rotates at the same angular velocity, .
2.3. Design Parameters

Referring again to Figure 3, the characteristic dimensions of the mechanism are:
the radius r of the level (rod length) where the masses are attached and
the radius R of the casing; more accurately it should
be the distance between the centroids of the masses
a and b.
The position of the concentrated masses, a and b,
are determined through the angular position t of
the a-rod (No. 4) with respect to the negative x-axis
(convention of the positively oriented angle is (Ox,
Oy). The configuration in Figure 3 corresponds to the
initial time, t = 0. In this case, the coordinates of the
masses a and b are xa , ya , za t 0 r , 0, R and
xb , yb , zb t 0 r , 0, R , respectively, corresponds to the
initial time, t = 0. In more details, during the time interval t, not only the two rods rotate around the axes of the
spin gears but also the casing in such a way that
casing . In the general case dealt in this paper, the
entire system rotates about the z-axis at an angular velocity z , using a second motor M2 shown in Figure 1 as
well as in Figure 2.
cos t
0
1
R12 0 cos t
0 sin t
cos z t
R 23 sin z t
(2)
sin t
cos t
0
sin z t
cos z t
0
(3)
(4)
Therefore, the coordinates of the mass a are analytically given by:
x t
y t R
z t
x
y
t 0
r cos t cos z t r sin t cos t R sin t sin z t
(5)
r cos t sin z t r sin t cos t R sin t cos z t
r sin t R cos t
2
Similarly, the coordinates of the mass b are analytically given by:
x t
y t R
z t
x
y
t 0
r cos t cos t r sin t cos t R sin t sin t
r cos t sin t r sin t cos t R sin t cos t
r sin t R cos t
(6)
Details about the figure-eight-shaped curve are provided in Appendix A.
Differentiating (5) in time, the velocity components of

the mass particle a become:
At any time instance, in order to reach the final position,

starting from the abovementioned position xa , ya , z a t 0
r , 0, R , the mass a undertakes three simultaneous
motions. The first motion is the rotation of the spin gear
S1 at an angle 1 t about z-axis, the second one is a
rotation of the casing around xaxis at an angle 2 1 ,
while the third one is a rotation around the vertical
z -axis at an angle 3 z t . As a result, when performing all three aforementioned rotations, at any time instance t the resultant rotation matrix becomes:
where
sin t
4. Velocities and Accelerations
3. Parametric Equations of the

Figure-Eight-Shaped Path
R a R 23 R 12 R 01
651
(1)
x a t r sin t cos z t z r cos t sin z t

r cos 2 t R cos t sin z t
(7)
z r cos t R sin t cos z t
y a t r sin t sin z t - z r cos t cos z t

r cos 2 t R cos t cos z t
(8)
z r cos t R sin t sin z t

za t rsin2 t Rsin t ,
(9)
Further differentiating in time, the acceleration components of the mass a become:
ENG
C. G. PROVATIDIS
652
z
2
R 4
rcost 2 r
z
(10)
sin t sin z t +2z rcos2t +Rcost cosz t
Substituting (10)-(12) into (19) and (16)-(18) into (20),

the resultant force components are given by:
4 sin t sin 2t 4 cos t cos 2t

2
2rsin2 t Rsint cos t

+2 rcos2t +Rcost sin t
(21)
+z
Fxa Fxb mr
2
2
ya t +z rcost sinz t +2z rsint cosz t
(20)
Fxb mxb , Fyb myb , Fzb mzb
2
2
xa t z r cos t cos z t
(11)
Fya Fyb mr
rcost +R sin tcos t
z cos z t sin 2t 4 z sin z t cos 2 t

2
(22)
za t
2r cos 2t Rcost
(12)
Differentiating (6) in time, the velocity components of

the mass b are given by:
xb t r sin t cos z t z r cos t sin z t
r cos 2 t R cos t sin z t
(13)
z r cos t R sin t cos z t

y b t r sin t sin z t z r cos t cos z t
r cos 2 t R cos t cos z t
(14)
z r cos t R sin t sin z t
zb t rsin2t Rsint
(15)
By further differentiation in time, the acceleration components of the mass b become:
xb t 2 r cos t cos z t 2 z r sin t sin z t
4r sin t cos t R sin t sin z t

2 z r cos 2t R cos t cos z t
z 2 r sin t cos t R sin t sin z t
2
Fza Fzb 4m r cos 2 t
(16)
ideally cancel those of the first couple (a,b).
6. Elastic Deformation of Rods
In order to determine the axial component of the inertial

force vector F F F F , k a, b , at a certain
T
2r sin 2 t R sin t cos z t
(17)
(25)
nrzb zb zb / r
where
Based on the abovementioned kinematics, the components of the inertial force exerted on the mass a can be
calculated by:
while those on the mass b:
nrxb xb xb / r
nryb yb yb / r
(18)
5. Inertial Forces
Fxa mxa , F ya mya , Fza mza
(24)
nrza za za / r
and
r cos t R sin t cos t
zb t 2 2r cos 2t +Rcost
zk
nrya ya ya / r
2 z r cos 2 t R cos t sin z t

2
yk
nrxa xa xa / r
2 z r sin t cos z t
xk
mass, a or b, it is necessary to consider the direction

cosines along the rod axes, which are given by:
(23)
Equation (23) depicts that:

1) the z-component of the resultant inertial force is
proportional to the radius r, and
2) the maximum upward force Fz (appearing at
0 ) is equal and opposite to the maximum
downward (appearing at 1t 2 ) value.
It can be also noticed that the geometrical parameter R
is not included in (21)-(23).
Remark: When considering a second couple of equal
masses a' and b' at a phase-difference 2 , it is
trivial to verify that the corresponding forces Fx and
Fy , which are given by (21) and (22) respectively,
2
2
yb t z r cos t sin z t
6.1. Axial Forces
z 2 r cos t cos z t
(19)
xa R sin t sin z t
, xb R sin t sin z t
ya R sin t cos z t , yb R sin t cos z t

za R cos t
(26)
, zb R cos t
denote the coordinates of those ends of the rods that do

not carry the concentrated masses.
ENG
C. G. PROVATIDIS
6.2. Rod Deformation (Quasi-Static Analysis)
Fz Fza Fzb 2m
Considering an instantaneous deformation of the elastic

rod a according to Hookes law (semi-static assumption), at every position the elongation of the rod that corresponds to the mass particle a is given by:
Fra r
ra
(27)
EA
(28)
Similarly, for the rod that corresponds to the mass particle mass b it holds:
Frb r
rb
(29)
EA
r
2
z
Frb mr z cos t
2
1 sin t 2
2
cos t (31)
3
cos t
3
(32)
Also, the semi-static deformations become:

ra
EA
z cos t
2
m
EA
z r
2
1 sin t
(33)
t 2 cos 2 t 2 z r 2 cos 3 t
(34)
Similarly, for the rod that corresponds to the mass particle mass b it holds:
rb
EA
(35)
rb r
EA
z r
2
cos
r
4
cos
t 2
Fza mza t 2m ra cos 2 t
(37)
Fzb mzb t 2m rb cos 2 t
(38)
4m
2
(40)
(41)
(42)
EA
cos 2 t
r 2 cos t r cos t 2 cos t

2
2
z
(43)
representing the contribution of the tension at the two

elastic rods, which carry the mass particles a and b.
7. Impulse
The impulse caused by the vertical resultant force
is given by:
Fz t
(44)
Integrating (40) over time, after manipulations the

impulse (i.e., (44)) can be analytically expressed in closed form, using one term for the rigid-body part and another for the tensile part, as follows:
I z I z , rigid I z , tension
(45)
where
I z , rigid 2 m r sin 2
(46)
and
(47)
with
m
2
Consequently, due to the change in the radii, z and

rb , the updated vertical forces become:
2
t 2 cos t
2
representing the vertical resultant force due to the rigidbody motion of the rods, and
c0
(36)
t 2 cos t 2 z r cos t
2
2 cos t
and therefore the updated variable radius is given by:

m
Fz, rigid 4m 2 r cos 2 t
2 z cos
I z , tension c0 c2 sin 2 c4 sin 4 c6 sin 6
cos t 1 sin t
2
with
2 r 2 cos 2 t 2
cos
cos

ra r
EA
I z t Fz d
2 z cos t
(39)
Equation (40) can be split into two parts as follows:
(30)
1 sin t 2
Fz 4m cos 2t r
Fz, tension
In order to simplify the subsequent analysis, without

loss of generality we assume that R 0 . In this case, the
radial forces become:
Fra mr z cos t
ra rb cos 2t
Substituting (34) and (36) into (39), one obtains:

rb r rb
Fz Fz, rigid Fz,tension

ra r ra
653
EA
2
z
m 2 9
2
2 2
z 3 r
EA 8
2
m 2
c4
z 2 r 2
c2
(48)
4 EA
m
2
c6
24 EA
z2 r 2
Obviously, the harmonics ( sin 2 , sin 4 , sin 6 ) in (47)
ENG
C. G. PROVATIDIS
654
lead to zero values every 180, 90 and 60 degrees, respectively, thus not practically contributing to the propulsion.
Concerning particularly the second harmonic ( sin 2 ),
not only the elastic part of the impulse vanishes every
180 degrees, but also that caused by the rigid-body motion (cf. Equation (46)). However, in addition to the three
harmonic terms, (47) includes also the term c t ,
which increases linearly with the time t. The analytical
expression of the term c in (48) depicts that for a given
angular frequency and given elasticity properties of
the two rods, the value of the net impulse is fully controlled by the angular frequency .
0
8. A Preliminary Experimental Result

The prototype device weights approximately 22 kg including all its structural members and the electric motors.
For reasons of functionality, the elastic bars (member No.
4 and member No. 5 in Figure 2) were manufactured
adequately thin, of 5.5 mm diameter and of 200 mm length, made of steel. Each lumped mass was taken approximately equal to 20 grams. The prototype device was
put in the center of the horizontal platform of an electronic scale, which was equipped by four identical straingages at its four corners. The mean average of these four
sensors was shown on a digital display, with an accuracy
of 10 gr.
The experimental validation was a very difficult task,
mainly due to the high angular velocities required to overcome the gyroscopic phenomena appearing in the prototype and the high bending occurring in the particular
choice of thin cylindrical bars. Nevertheless, even for the
abovementioned small masses, and even for the very
slow angular velocities (100 300 rpm) that were allowed so as to avoid collision (of the members No. 4 and
No. 5 on the horizontal shaft No. 2, shown in Figure 2),
the resultant impulse obtains a nonzero value. Clearly,
for a time-interval of 124 seconds in which 618 measurements were automatically recorded as shown in Figure 4, the relative difference of the sum of the negative
values with respect to the sum of the positive values was
found about 11.6%, a fact that sustains the findings of
the abovementioned theoretical analysis.
9. Discussion
It is remarkable that the findings of this paper are in consistency with previous experimental results (8% weightreduction) related to ideally rigid gyroscopes [16]. Since
the proposed device is essentially a flexible (elastic) gyroscope of which the masses operate in a hemisphere, it
is believed that similar accurate experiments with those of [16] should be performed for the current device.
Figure 4. A digital record of the ground reaction (in dN or

kg-force), for a time interval of approximately 2 minutes,
using two lumped masses each of 20 gr and low angular
velocities: approximately = 100 rpm and z = 320
rpm. The amplitude of the ground force is close to 0.2
kg-force (i.e., 0.2 dN, equivalently, 2 N).
It is also believed that the proposed idea is a mechanical

alternative to older relativistic thoughts of 1960s, which
have recently revived [17].
In addition to the experimental shortcomings mentioned in Section 8, the weaknesses of the approximate
model presented in this study are as follows:
Only the action of the concentrated masses has been
considered, while the moment of inertia of the rods
has been omitted.
The axial deformation of the rods has been assumed
to coincide with that of static conditions immediately imposed at every time instance, while a more
accurate semi-analytical approach had to consider
dynamic response caused by longitudinal and transverse traveling elastic waves along the rods.
The influence of the bending deformation of the elastic bars has not been considered.
The angular velocities and have been considered to be constant while an accurate simulation
model would require the dynamic modeling of the
motors themselves or/and the use of power-to-angular velocity curves [18,19].
For the particular setup of this paper, at
2 , 3 2 the masses a and b are found at the
point of intersection I R 0, 0, r . This collision
could be theoretically avoided considering one mass
being of male and the other of female type. Obviously, in the general case where R 0 this shortcoming is easily overcome but this selection leads
to more complicated analytical expressions that do
not offer further insight or any other essential advantages.
Concerning the mathematical analysis of this work, of
course a more accurate analysis has to be performed on
z
ENG
C. G. PROVATIDIS
the basis of a finite element elastodynamic model of the

entire structure in which not only tension but also bending and torsion will be automatically considered. A special feature of such an analysis is the dependence of the
inertial forces on the radius, which is not a-priori known.
Furthermore, concerning the experimental part of this
work, it is worth-mentioning that not in all cases the experiments were of the same quality as that presented in
Figure 4. In general, in order to achieve a remarkable net
impulse, a proper synchronization between the two angular velocities, and , is required. In order to increase the magnitude of the net impulse, one could increase either the angular velocities or the magnitude of
the lumped masses. In both cases the high bending involved requires thick rods (high diameters) as well as
motors of high electric power. Therefore, due to the limited power capacity and, since the overall mechanical
strength of the experimental setup could not be improved
we were forced to stay with preliminary measurements
only. Of course, a future study should consider an entirely different experimental setup.
Summarizing, the aforementioned advanced elastodynamic model has also to be compared with high precision
experiments on a well-designed experimental device, a
fact that presupposes high technical skills and high-tech
premises, probably within an advanced industrial environment.
z
10. Conclusions
This work contributes to the field of inertial propulsion,
proposing a new concept for the production of net impulse through rotating masses. In the beginning, it was
shown that when a lumped mass moves along a circumference, it repeats its position every 360 degrees, thus its
initial linear momentum is repeated and no net impulse is
finally produced for a whole revolution; this case corresponds to the notorious Dean drive. Then, it was theoretically shown that when two lumped masses move along
a specific figure-eight-shaped path at a phase difference
of 180 degrees, in such a way that the latter path lays on
the surface of a hemisphere that additionally spins about
its axis of symmetry, the involved inertial forces lead to a
non-vanishing net impulse. Intuitively, this claim is true
because when the ratio of the spinning angular velocity
over the first one (formation of the figure-eight-shaped
path) is not an integer number, every mass does not repeat its initial position; in other words, when the mass
completes the figure-eight-shaped path (every 360 degrees), the linear momentum has a different value that
what it had at the initial position. In terms of combined
structural mechanics and kinematics, every mass is connected to the center of the hemisphere through an elastic
rod that is imposed to highly variable tensile deformation.
This finding implies a variable radius of the hemisphere
655
and is decisive to produce net impulse towards the axis

of symmetry of the hemisphere. In addition to the theoretical findings, for purposes of demonstration and validation, a prototype mechanical device, capable of producing the two aforementioned rotations, was manufactured and fully described in this paper. Although the
need for more realistic experimental tests has been discussed, preliminary measurements are in consistency
with the proposed theory and sustain the production of
net impulse using rotating masses.
11. References
[1]
H. Yoshikawa, T. Kagiwada, H. Harada and M. Mimura,

Improvement of Propulsion Mechanism Based on the
Inertial Force, In: F. Kimura and K. Horio, Eds., Towards Synthesis of Micro-/Nano-Systems, Springer, London, 2007, pp. 333-334.
[2]
J. M. Gilbert, Gyrobot: Control of Multiple Degree of

Freedom Underactuated Mechanisms Using a Gyrating
Link and Cyclic Braking, IEEE Transactions on Robotics, Vol. 23, No. 4, 2007, pp. 822-827.
[3]
P. R. Ouyang, Q. Li and W. J. Zhang, Integrated Design

of Robotic Mechanisms for Force Balancing and Trajectory Tracking, Mechatronics, Vol. 13, No. 8-9, October
2003, pp. 887-905.
[4]
N. L. Dean, System for Converting Rotary Motion into

Unidirectional Motion, US Patent 2886976, 19 May
1959.
[5]
I. I. Blekhman, Synchronization in Science and Technology, ASME Press, New York, 1988.
[6]
I. I. Blekhman, P. S. Landa and M. G. Rozenblum,

Synchronization and Chaotization in Interacting Dynamical Systems, Applied Mechanics Reviews, Vol. 48,
No. 11, November 1995, pp. 733-752.
[7]
I. I. Blekhman, A. L. Fradkov, H. Nijmeijer and A. Yu.

Pogromsky, On Self-Synchronization and Controlled
Synchronization, Systems and Control Letters, Vol. 31,
No. 5, October 1997, pp. 299-305.
[8]
G. Y. Stepanov, Why is it Impossible to Have Deans

Apparatus? Journal Priroda (in Russian), Vol. 7, 1963,
pp. 85-91.
[9]
I. I. Blekhman, Vibrational Mechanics: Nonlinear Dynamic Effects, General Approach, Applications, World
Scientific, Singapore, 2000.
[10] M. G. Millis and N. E. Thomas, Responding to Mechanical Antigravity, NASA/TM-2006-214390, AIAA2006-4913, December 2006. http://gltrs.grc.nasa.gov/reports/2006/TM-2006-214390.pdf
[11] M. G. Millis, Assessing Potential Propulsion Breakthroughs, In: E. Belbruno, Ed., Annals of the New York
Academy of Sciences, Vol. 1065, New York, December
2005, pp. 441-461.
[12] C. G. Provatidis, Some Issues on Inertia Propulsion
Mechanisms Using Two Contra-Rotating Masses, Theory of Mechanisms and Machines, Vol. 8, No. 1, April
ENG
C. G. PROVATIDIS
656
2010, pp. 34-41.
[13] C. G. Provatidis and V. T. Tsiriggakis, A New Kinematics Theory in Physics and Presentation of a Device for
Gravity Studies, Proceedings 9th International Scientific-Practical Conference on Research, Development and
Applications of High Technologies in Industry, A. P.
Kudinov, Ed., Vol. 1, St. Petersburg, April 2010, pp. 386393.
[14] C. G. Provatidis and V. T. Tsiriggakis, A New Concept
and Design Aspects of an Antigravity Propulsion
Mechanism Based on Inertial Forces, Proceedings 46th
AIAA/ASME/SAE/ASEE Joint Propulsion Conference &
Exhibit, Nashville, July 2010.
[15] C. G. Provatidis, A Novel Mechanism to Produce Fig-
ure-Eight-Shaped Closed Curves in the Three-Dimensional Space, In: D. Tsahalis, Ed., Proceedings of 3rd
International Conference on Experiments/Process/System
Modeling/Simulation & Optimization, Athens, July 2009.
[16] R. Wayte, The Phenomenon of Weight-Reduction of a
Spinning Wheel, Meccanica, Vol. 42, No. 4, August
2007, pp. 359-364.
[17] M. Tajmar, Homopolar Artificial Gravity Generator Based on Frame-Dragging, Acta Astronautica, Vol. 66, No.
9-10, May-June 2010, pp. 1297-1301.
[18] V. O. Kononenko Vibrating Systems with a Limited
Power Supply, Iliffe Books Ltd, London, 1969.
[19] A. H. Nayfeh and D. T. Mook Nonlinear Oscillations,
John Wiley & Sons, Inc., New York, 1979.
ENG
C. G. PROVATIDIS
10) All points of the abovementioned path belong to a

sphere of radius, i.e.:
APPENDIX A
Remarks concerning the figure-eight-shaped curve
For the purpose of completeness, details are provided
here for the figure-eight-shaped curve.
First, in the absence of the spinning motion, i.e., when
0 , a careful inspection of (5) and (6) reveals that:
1) At the initial time instance ( t 0 ), in fact the rods
obtain their horizontal position parallel to x-axis,
i.e., a on the left and b on the right side.
2) At the time instance given by t 2 , both
masses obtain their vertical position.
3) At the time instance given by t , the masses
mutually interchange their (horizontal) position.
4) At the time instance given by t 3 2 , the
masses are again found at their vertical position.
5) At the time instance given by t 2 , the masses
obtain their initial (horizontal) position, and so on.
6) Therefore, the distance between the two masses
varies from the minimum 2 R (vertical position) to
z
the maximum value 2 r 2 R 2 (horizontal position).

7) Both mass particles, a and b, share a common
path. This happens because when putting
t t t in Equation (5), then it becomes
identical with Equation (6). In other words, the
masses a and b move on the same path and appear a constant phase difference of 180 degrees,
thus they are mutually interchanged.
8) During the first 90 degrees ( 0 ), the mass
a draws the blue line while b draws the red line
of the common path (Figure 5(a)). In the next 90
degrees the situation is reversed, so as the mass a
follows the read and the mass b follows the blue
line.
9) The common path intersects itself at the unique
point I R, 0, r , which corresponds to 2
or 3 2 , and so on.
a
657
xa ya z a xb yb zb rsphere
, with rsphere r 2 R 2
Second, in the case of a non-vanishing spinning angular velocity ( 0 ), the lumped masses do not generally
follow the same path, as clearly shown in Figures
5(b)-5(f), particularly when the ratio z of the angular velocities is not an integer number.
z
1a
1a
1a
Figure 5. Perspective view of the paths drawn by the

lumped masses (r = 80 mm, R = 0), which are produced at
different synchronizations (ratio of angular velocities,
z ): (a) ratio = 0; (b) ratio = 0.5; (c) ratio = 1.0; (d) ratio
= 2.0; (e) ratio = 3.0 and (f) ratio = 3.5. In all cases, the blue
and red lines correspond to the masses a and b (shown in
Figure 2), respectively. The cases (a) and (d) are shown for
the first 180 degrees ( 0 t ) so as to avoid overlapping, while the rest cases for the first 720 degrees
( 0 t 4 ).
ENG
Engineering, 2010, 2, 658-664

Computer-Aided Solution to the Vibrational Effect of

Instabilities in Gas Turbine Compressors
1
Ezenwa Alfred Ogbonnaya1, Hyginus Ubabuike Ugwu2, Charles Agbeju Nimibofa Johnson3
Department of Marine Engineering, Rivers State University of Science and Technology, Port Harcourt, Nigeria
Department of Mechanical Engineering, Michael Okpara University of Agriculture (MOUA), Umuahia, Nigeria
3
Department of Marine Engineering, Niger Delta University, Bayelsa, Nigeria
E-mail: {ezenwaogbonnaya, canjohnson2000}@yahoo.com, ubadyke2001@yahoo.com.au
Received March 25, 2010; revised June 24, 2010; accepted June 28, 2010
Abstract
Surge and stall are the two main types of instabilities that often occur on the compressor system of gas turbines. The effect of this instability often leads to excessive vibration due to the back pressure imposed to the
system by this phenomenon. In this work, fouling was observed as the major cause of the compressor instability. A step to analyze how this phenomenon can be controlled with the continuous examination of the vibration amplitude using a computer approach led to the execution of this work. The forces resulting to vibration in the system is usually external to it. This external force is aerodynamic and the effect was modeled
using force damped vibration analysis. A gas turbine plant on industrial duty for electricity generation was
used to actualize this research. The data for amplitude of vibration varied between 15 and 15 mm/s while
the given mass flow rate and pressure ratio were determined as falling between 6.1 to 6.8 kg/s and 9.3 to 9.6
respectively. A computer program named VICOMS written in C++ programming language was developed.
The results show that the machine should not be run beyond 14.0 mm vibration amplitude in order to avoid
surge, stall and other flow-induced catastrophic breakdown.
Keywords: Computerized Solution, Instabilities, Vibration, Gas Turbine Compressors, Operational Limits
1. Introduction
The economics of power generation with gas turbines
(GT) is now quite attractive due to its low capital cost, its
high reliability and flexibility in operation. Another outstanding feature is its capability of quick starting and
using wide variety of fuel from natural gas to residual oil
or powdered coal. Due to better material being made
available and with the use of adequate blade cooling, the
inlet gas temperature of the turbine blades can now exceed 1200C as a result of which the overall efficiency of
GT plant can be 35%. This is almost the same as that of a
conventional steam power plant. Based on these developments, occurrence of instabilities in the compressor
system would no doubt result to performance deterioration of the overall efficiency of the GT. There are two
basic types of instabilities that could be encountered in
the GT compressor system namely the rotating stall and
surge. Both types of instabilities have damaging consequences to the compressor.
According to Iwakiri, et al. [1], rotating stall causes
the compressor to operate with extremely low frequencies resulting in excessive high internal temperature that
has an adverse effect on blade life. Surge causes severe
problems such as excessive pressure built-up at the inlet
and cyclic loading on the compressor. These instabilities
might lead to the inability of the compressor blade to
produce the required loading and the engine might sustain catastrophic damage as a result of the excessive vibration [2].
1.1. Other Approaches/Techniques

1) Lu et al. [3] and Okada et al. [4] presented a draft on
Stall Inception in Axial Flow Compressor. A comprehensive measurement and theoretical analyses was used
to determine which of the two types of instabilities (rotating stall or surge) would occur in a particular situation.
2) Ogbonnaya and Johnson [5] dealt specifically on
surge and rotating stall. In their work, a theoretical compressor system was modeled followed by experimental
results and comparison with theory was presented to
ENG
E. A. OGBONNAYA ET AL.
analyze the characteristics of these instabilities. The

model was used to predict whether a surge or stall would
occur at stall limit. Similarly, Iwakiri et al. [1] and
Huabing et al. [6] carried out similar work on rotating
stall on centrifugal and axial compressors respectively.
1.2. Approach Used in This Present Work

This paper provides a computerized approach to monitor
the vibration effects of instabilities generally in a GT
compressor system. A program named VICOMS written in the C++ programming language was used to bring
it to fruition. VICOMS stands for Vibration Instabilities
of Condition Monitoring System. It further provides a
detailed description of these phenomenon/threats and
their consequences. Also, this present work looks at general instabilities including galloping, flutter in GT compressors. Accoding to Rao [7], Flutter is a form of self
excited stall which can occur when the section of the
blade is just beginning to stall.
1.3. Causes of Instabilities

The degradation of fouling is one of the causes of GT
performance deterioration. It results to instability in the
compressor system. Fouling is known as the source of
about 70-85% of performance deterioration of GT engine
[8]. Morini, et al. [9] developed a stage by stage model
to investigate the effect of compressor and turbine stage
deterioration. It was observed that compressor fouling is
the most common source of loss in a GT system performance.
Fouling is defined as the deposition process of air
borne particle on the blade surfaces. In GT compressor,
foulant tends to deposit on the compressor blades as the
air flowing into the compressor get contaminated which
may cause malfunction of the blade profile and as well
affect the compressor flow coefficient [10]. The rate at
which this fouling takes place is difficult to quantify because it depends not only on the types and quantities of
materials ingested, but also on the peculiar properties of
the substances that cause them to stick [11].
Under design operating condition, most stage would
operate at design flow coefficient and at a high isentropic efficiency. When the flow coefficient is to the
right of the characteristics curve as shown in Figure 1,
the stage is lightly loaded and extreme right point is
known as choke point. To the left of the characteristic
curve is a region where aerodynamic stall occurs (surge
region).
As fouling drops, the mass flow rate (flow coefficient)
in the first stage affects the performance of the later
stages as the operating point on the first stage characteristic curve move toward the left, thus increasing the
659
Surge line
Fouled
Design point
Pa
P1
Heavy stage
loading
Light stage
.
m To
Po
Figure 1. Compressor stage characteristics during fouling

[12].
pressure ratio as shown in Figure 1. This causes a high

density at the inlet to the second stage. Thus, there will
be further reduction in second stage flow coefficient
(mass flow rate). If this effect progresses throughout the
successive stages, a later stage will stall aerodynamically
and trigger surge.
2. Materials and Method

Data were collected on hourly basis for a period of ten
months from an operational GT used for electricity generation. The data were sampled and the mean taken for
monthly basis. The GT is a 75MW plant called AFAM
III, GT17, TYPE 13D located near Port Harcourt in Rivers State of Nigeria. The characteristics of the GT is
shown in Appendix A.
Any machine handling fluid will vibrate due to reaction of the blade and vanes of the fan or impeller striking
the media of operation. This vibration is rarely troublesome except that they exert some part of the machine or
dotting to resonance and it is vibration due to aerodynamic force. When a system is subjected to a force harmonic excitation, its vibration response takes place at the
same frequency as that of the excitation. Common sources of harmonic excitation are imbalance in rotating machines, forces produced by reciprocating machines, or
the motion of the machine itself. Figures 2 and 3 show a
compressor model and the free body diagram of a GT
engine, respectively [13,14].
According to Ogbonnaya [14], Rao [15], Dukkipati
and Srinivas [16], the equation of motion that leads to the
circumstances on instabilities of s single degree of freedom system is considered as follows:
mx cx kx 0
(1)
For the model shown in Figures 2 and 3 the equaENG
E. A. OGBONNAYA
660
ET
AL.
c
k
2
2 ,
n
m
m
K
Substituting back into Equation (6), we have
C-damper
.
..
F cos t
x 2 x n 2 x o
m
The complementary function, i.e., solution of
..
x 2 x n2 x 0 is;
Direction of
shaftrotation
Figure 2. Compressor Model.
X 1 C1e
2 - t
C2 e
2 - 2 t
X 2 C1 C 2t e t
.
Cx
Kx
(7)
X3 e
t
A cos
2 t B sin
2 t
The general solution of Equation (7) can be obtained

thus:
D
F0cost
F0
cos t
M
2 D 2 x
F0
cost
M
and x 2
D 2 D 2
since f (D2) cost = f ( 2) cost,
..
mx
Figure 3. Free body diagram of the Compressor Model.
tion of motion may be expressed as follows:

..
F = ma = m x
..
F0cost + k (dst + x) = m x
..
2 D - 2 2
X 2 2 n 2
2
4 D n
F0
cos t
M
F0 2 Dsin t - n2 2 cos
2
M
4 2 2 n2 2
2 2
2
2 2
F0 4 n cos t -

2
M
4 2 2 n2 2
F0cost + kdst + kx = m x
(2)
But = kdst substituting into Equation (2)

..
F0cost + kdst kdst + kx = m x
(3)
where tan
2 2
n
(4)
The total solution is the sum of the transient solution

(complementary function) and general solution (steady
state solution) but the transient solution decreases exponentially with time (refer to X1, X2, X3). Thus, when
harmonic solution is considered, we have:
..
F0cost + kx = m x
..
and
m x + kx = F0cost
.
For damping force analysis C x
Putting C x into Equation (4)

Ferdinand and Johnston [17] showed that
.
..
m x + C x + kx = Focost
Dividing through by m, we have
.
cos t
c
k ..
x x x Fo
m
m
m
where:
(5)
(6)
2 2
2
2
F0 4 n
M 4 2 2 n2 2 2
which,
Frequency, F =
cost
Hz and vibration displacement
ENG
amplitude.
661
Start
F0
M 4 2 2 n2 2
(8)
From the referenced GT plant, it was possible to read

out the vibration amplitude directly from the machine
other than the aerodynamic force resulting to the vibration. It is therefore necessary to determine the corresponding aerodynamic force resulting to vibration. Hence,
from Equation (8) making aerodynamic force, F0, the
subject of formula we have:
F0 XM
4
2

2
n
(9)
where;
F0 = aerodynamic force (N)
M = mass of the shaft (kg)
X = vibration displacement amplitude (mm/s)
= force frequency (Hz)
n = natural frequency (rad/s)
= product of coefficient of damping (Nsm-1).
This is the equation used to model the flowchart and
consequently design the program to simulate a solution
to the vibration effects of instabilities in GT compressors.
Figure 4 shows the flowchart for VICOMS written in
C++ language for obtaining the aerodynamic force resulting to vibration in the GT system. It has one loop as
shown and can go round several iterations until the first
speed is equal to or less than the last speed to make the
program stop. This is when the surge would have been
uncontrollable as to cause damage to the plant. Hence,
VICOMS would predict when the GT should undergo
maintenance check.
The flowchart in Figure 4 led to the evaluation of the
computer program code written in C++ programming
language. The program helped in the calculation of aerodynamic force as stated in Equation (9).

The readings of vibration amplitude of the two end bearings of the compressor unit in a GT plant on industrial
duty for electricity generation was taken with the corresponding mass flow rate, pressure ratio, shaft speed and
active load. These readings are shown in Table 1. It was
observed that the GT was run above its critical speed
value of 3000 rpm.
Figure 5 shows the vibration amplitude as a function
of time. From this result, it is observed that the disturbing force has an oscillatory nature. The force varies as a
Declare / Define
variables used
1 1, n 314
Input first, speed last speed
and step value
n = 2N / 60
Input active load
Input X1 [ i ] and X2 [ i ]
F01 i X1 i m 4 2 2 n2 2
2
F i X i m 4 2 2 n2 2
2

02
Print speed, frequency, active load X1,

X2, and F01 and F02
1=1+1
First speed = first speed + step Val
NO
Is
Firstspeed < = lastspeed
Yes
Stop
Figure 4. Program flow chart for a VICOMS.
sinusoidal function of time and the wave form.

The wave form shows that it is a steady state vibration. The graph further depicts that the maximum amplitude which the engine can withstand is 15 mm despite running above its critical speed. Figure 6 depicts
the general operation of a compressor under surging
condition. It also shows that the region from point A to
B implies a stable operation without surge or stall.
There is a reverse flow from B, which would lead to
surge. The flow again recovers from C to D yielding a
normal flow; which within the compressor depicts a
normal operating range.
The graph obtained is also in conformity with that given in Ogbonnaya and Johnson [5]. The minor difference
in profile between B and C could be attributed to the size
of the engine and environmental conditions [14] where
the engines are being used.
ENG
E. A. OGBONNAYA
662
Table 1. Reading showing the data taken from AFAM III,

GT 17, TYPE 13D.
Shaft
Speed
rp
(kg/s)
(RPM)
Active
Load
Fre-
Vibration
quency
Amplitude
(Hz)
Brg.2
15.0
6.49
9.30
3005
50
51.2
4.8
6.7
9.40
3053
50
51.3
4.8
15.0
6.2
9.30
3051
50
51.2
4.6
15.0
6.8
9.50
3080
50
51.3
4.9
10.0
6.4
9.60
3063
50
51.0
4.9
15.0
6.3
9.50
3063
50
50.4
5.1
15.0
6.1
9.40
3074
45
51.1
5.1
15.0
6.3
9.50
3076
40
51.3
5.3
7.0
6.5
9.40
3077
40
56.7
5.0
15.0
6.66
9.40
3081
36
51.5
5.1
6.40
would correspond to speeds between 3051 and 3053 rpm.
A work has been carried out on the computerized solution of instabilities in GT compressor. The test engine is
AFAM III, GT 17, TYPE 13D.
The compressor system suffered surge and stall, which
resulted to instabilities in the test engine due to fouling.
It was shown that fouling leads to the stuffing of the
compressor stages. This also results in the reduction of
the compressor surge margin and dramatic instability of
the operation of the whole GT compressor system, culminating to vibration.
The mass flow rate, pressure ratio, shaft speed and vibration amplitude in the system were collected from the
two end bearings of the compressor system in the GT
plant. A model was consequently developed to analyze
the data collected in order to determine the corresponding aerodynamic force causing vibration in the system.
The mathematical model was used to run a program code
named VICOMS written in C++ programming language.
The results and the graph showed that the GT should not
be run beyond 14 mm.
20
15
Amplitude of Vibration (mm)
AL.
4. Conclusions
(mm/sec)
Brg1
ET
10
5
0
-5
5. Acknowledgements
-10
The authors wishes to acknowledge the efforts of Messrs.

Woji, John and Ebunuoha, Chigozie for their immense
contributions to the success of this research project. They
visited the thermal stations for the experimentations and
data collection associated with this work. They are
equally appreciative of the efforts of Mr. Jeffrey Mukoro
and Chike in typeseting, proof-reading and editing the
manuscript.
-20
Time in Hours
Figure 5. Amplitude of vibration as a function of time.

3090
3080
Compressor Speed (rpm)
3070
3060
6. References
3050
[1]
K. Iwakiri, M. Furukawa, S. Ibaraki and I. Tomita,

Unsteady and 3D Flow Phenomena in Transonic
Centrifugal Compressor Impeller at Rotating Stall, Proceedings of ASME Turbo Expo, Orlando, Florida, 8-12
June 2009. www.teurbexpo.org
[2]
R. Kurz, Surge Control Design System Design, Proceedings of ASME Turbo Expo, Orlando, Florida, 8-12
June 2009. www.teurbexpo.org
J.-L. Lu, W. Chu and K. Peng, Numerical and Experimental Research of Stall Inception on Subsonic Axial
Flow Compressor Rotor, Journal of Aerospace Engineering, Vol. 48, No. 2, 2010, pp. 3-4.
T. Okada, A. Kawajiri, O. Yataka and O. Eisuke, Stall
Inception Process and Prospects for Active Hub-Flap
Control in Three Stage Axial Flow Compressor, Journal
of Thermal Science, Co-Published with Springer-Varlag
GmbH, 2008, pp. 4-8.
3040
3030
3020
3010
[3]
3000
6
6.2
6.4
6.6
6.8
Mass flow Rate (kg/sec)
Figure 6. Speed against flow rate.
This analysis shows that running the compressor within

the regions of B and C should be avoided. This region
[4]
ENG
[5]
[6]
[7]
[8]
[9]
E. A. Ogbonaya, Modeling Vibration-Based Faults in

Rotor Shaft of Gas Turbine, Ph.D. Dessertation, Department of Marine Engineering, Nigeria, 2004, pp. 20-22,
92-95, 180-181.
J. Huabing, Y. Wei, L. Yagun and L. Quishi, Experimental Investigation of the Influence of Inlet Distortion
on the Stall Inception in a Low Speed Axial Compressor, Proceedings of ASME Turbo Expo, Orlando, Florida, 8-12 June 2009. www.Teurbex-po.org
S. S. Rao, Mechanical Vibration, 4th Edition, Pvt Ltd.,
Dorling Kingsley, Licenses of Pearson Education in
South Asia.
C. B. Meher-Homji, M. Chaker and A. E. Brouley, The
Fouling of Axial Flow Compressor-Causes, Effects Susceptibility and Sensitivity, Proceedings of ASME Turbo
Expo, Orlando, Florida, 8-12 June 2009. www.teurbexpo.
org
M. Morini, M. Pinelli, P. R. Spina and M. Venturini,
Influence of Blade Deterioration on Compressor and
Turbine Performance, ASME Paper GT2008-50043,
2008.
[10] T. W. Song, J. L. Sohn, T. S. Kim and T. R. O. Sung,

An Improved Analytic Model to Predict Fouling Phenomena in the Axial Flow Compressor of Gas Turbine
Engines, Proceedings of the International Gas Turbine
Congress, Tokyo, 2-7 November 2003.
663
[11] A. D. Mezherisky and A. V. Sudarev, The Mechanism

of Fouling and the Cleaning Technique in Application to
Flow Parts of Power Generation Plant Compressors,
ASME Paper 90-GT-103, 1990.
[12] C. B. Meher-Homji, Gas Turbine Axial Compressor
Fouling: A Unified Treatment of its Effect, Detection and
Control, International Journal of Turbo and Jet Engines,
Vol. 9, No. 4, 1992, pp. 99-111.
[13] E. D. Bently, C. T. Hatch and B. Grissom, Fundamental
of Rotating Machinery Diagnostics, Bentley Pressurized
Bearing Press, Minden, 2002.
[14] E. A. Ogbonnaya and K. T. Johnson, Modeling Vibrational Effects of Surge and Stall on GT Compressor,
Journal of International Research and Development Institute, 2010, pp. 142-146.
[15] N. S. V. K. Rao, Mechanical Vibration of Elastic Systems, 1st Edition, Asian Books Private Limited, Mahavirlane, Vardhan House, Darya Lzanj, New Delhi,
2006.
[16] R. V. Dukkipati and J. Srivinas, Textbook of Mechanical Vibrations, 3rd Edition, Prentice-Hall of India Private Limited, Connaught Circus, New Delhi, 2007.
[17] P. B. Ferdinand and R. E. Jr. Johnston, Vector Mechanics for Engineers: Static and Dynamic, 6th Edition,
McGraw-Hill Companies Inc., Boston, 2004.
ENG
E. A. OGBONNAYA
664
ET
AL.
Appendix A: Characteristics of afam iii, gt 17, type 13d

NAME OF EQUIPMENT:
MANUFACTURER:
TYPE:
CAPACITY:
CRITICAL SPEED:
TYPE OF COMPRESSOR:
DESISGN OF COMPRESSOR:
NO. OF STAGES:
AIR PUMPING CAPABILITY:
Nomenclature:
M = Lumped mass of shaft (kg)

X = Vibration displacement amplitude (mm)
= 22
7
N = Cycle per minute

F0 = Aerodynamic force (N)
X = Vibration velocity amplitudes (mm/s)
S = Boundary condition (varying between 1 and 2)
n = Natural damped frequency (rad/s)
BROWN BOVIERI- SULZER TURBO MESCHINER;

ABB now ALSTOM
13D
75MW
3000RPM
VA 14017
AXIAL
17
295 m3/s
W = Weight of the compressor rotor shaft (KN/kg)

K = Shaft stiffness (KN/m)
C = Damping coefficient (KN s/m)
st = Elongation (m)
= Frequency (Hz)
= Inlet flow angle o (degree)
Eo = Cascade collection efficiency (%)
rp = Pressure ratio
m = Mass flow rate (kg/s)
ENG
Engineering, 2010, 2, 665-667

Flipped Voltage Follower Design Technique for Maximised

Linear Operation
Ching-Mei Chen, Khaled Hayatleh, Bryan Hart, John Lidgey
School of Technology, Oxford Brookes University, Oxford, UK
E-mail: khayatleh@brookes.ac.uk
Received July 27, 2010; revised August 3, 2010; accepted August 6, 2010
Abstract
The results of comparative DC simulation tests confirm that a proposed modification to the feedback circuit
of a Flipped Voltage Follower (FVF), to produce a type of Folded Flipped Voltage Follower (FFVF), is
capable of maximising the linear DC operating range for given values of supply rail voltage and operating
current.
Keywords: Analog, Voltage Follower
1. Introduction
The name Flipped Voltage Follower (FVF) was coined
by Carvajal et al. [1] to describe a class of pre-existing,
and new, low power/low voltage analogue circuits.
A prototype FVF is a two transistor source-follower in
which the input mosfet is forced to operate at a sensibly
constant DC drain current, set by ancillary circuitry, despite variation in input voltage or load current. This is
achieved by the action of shunt negative feedback.
The overall result is a source-follower with decreased
output impedance and increased linearity in its voltage
transfer characteristic. The so-called Super Source-Follower [2] can be regarded as a member of the FVF
family: in fact, it has been called a Folded Flipped Voltage
Follower (FFVF) [3].
In Figure 1, M1 and M2 are inter-connected to form an
N-channel FVF the operating current for which in supplied by MX, the output mosfet of a simple 1:1 current
mirror formed from MW and MX. The mirror input current,
IX, is set by choice of RB.
A capacitor, CS [1], may be required to produce a
specified phase margin in the loop-gain frequency response.
M1 passes an effectively constant current so the incremental voltage gain of the FVF is close to unity providing it operates in its linear region. Unfortunately, as
has been noted in [1], the valid linear range decreases with
threshold voltage. This is most easily seen by applying
Equations (1) and (2), which follow to the case in which
the characteristics of M1 and M2 are identical.
From [4],
ID
N
2
VGS VTN
2
(1)
2I D
N
(2)
and, VDS(min)
+VDD
1:1
MX
MW
VD
IX
RB
M1
VG
VS
M2
CS
-VSS
Figure 1. A prototype Flipped Voltage Follower (FVF).
ENG
C.-M. CHEN ET AL.
666
In these equations the symbols have their usual mosfet

meanings:
ID = Drain Current; VGS(VG-VS) = Gate-source voltage;
VDS(VD-VS) = Drain-source voltage; VTN = Threshold
voltage; N = NCOX(W/L).
Linear operation requires both M1 and M2 to operate in
the saturation region. Using Equations (1) and (2) the
conditions for this for the circuit of Figure 1 are,
2I X
VG VSS 2
VTN
(3)
N
and, VG VSS
2I X
2VTN
N
(4)
If VG denotes the linear range then, from (3), (4),

2I X
VG VTN
(5)
N
The problem, now, is that VG may be unacceptably
small for the devices of modern CMOS technology, at
even low values of IX.
This problem does not arise in the new Folded Flipped
Voltage Follower design technique described here because Equations (4) and (5) no longer apply.
For space reasons, the DC operating mode, only, is
outlined here: small-signal performance is the subject of a
future publication.
2. Proposed Circuit
Figure 2 shows the proposed FFVF circuit. It differs from
that of Figure 1 (and that of [3]), by the way in which the
feedback connection is made from the drain of M1 to the
gate of M2. Instead of the direct link of Figure 1 an additional mosfet, M3, is included and forced to operate at a
sensibly constant current, IZ, provided by the high output
resistance Widlar-type current mirror formed from MY, MZ
and RZ. MX performs the same function as Figure 1 but, in
this case its output current is IY so in normal operation the
current in M1 is (IY-IZ).
M3 performs two functions. The first is to provide a
feedback current, which is converted to a feedback voltage at the gate of M2. The second function of M3 is to
keep the drain source voltage of M1 constant, preferably
at the minimum level for saturation, with variation in VG.
Using Equations (1) and (2), that can be achieved if,
2I Z
VTP
P
2 IY I Z
N
(6)
in which subscript P refers to P-channel mosfet M3.

A design requirement for the maximum linear range for
VD, and hence VS, is for the equality sign in Equation (6) to
hold under worst case operating conditions, i.e., IY and P
high but IZ, VTP and N low.
+VDD
1:1
MW
MX
VD
VG
M1
IY
RB
VS
M3
M2
IZ
MY
MZ
CS
RZ
Figure 2. Proposed circuit: A Folded Flipped Voltage Follower (FFVF).
ENG
C.-M. CHEN ET AL.
The upper limit to the linear range is governed by the

onset of triode region operation in MX.
Thus,
2 IY I Z
2 IY
VS(max) VDD
(7)
P
N
From a DC standpoint the choice of the ratio IZ/IY is not
critical provided Equation (6) is satisfied. However, from
a small signal viewpoint the choice of IZ affects the
loop-gain characteristics via its effect on the dynamic
parameters of M3.
667
1.5V
1.0V
0.5V
0V
VVDD,
VZZ
V
VGG
-0.5V
3. Results
VVZZ
The circuits of Figure 1 and Figure 2 were simulated for

operation at 27C. All the mosfets, except M3, had L =
0.13 u, W = 10 u. It was assumed that for low voltagelow
power operation VDD and VSS would not exceed 1.5 V and
IX would not exceed 1 mA, so these values were used in
tests. For a fair comparison, M1 was made to operate at
the same current in both circuits. For M3, the choices IZ =
50 uA (so IY = 1.05 mA), L = 0.13 u, W = 50 u satisfied
Equation (6).
Simulated test results, displayed for comparison, in
Figure 3, Figure 4 refer, respectively, to the circuits of
Figure 1 and Figure 2.
When VG is such that M1 is passing only a small
leakage current the curves for VD in Figures 3 and 4 are
similar, as are those for VS.
However, once M1 commences conduction differences
appear. In Figure 3 there is no region for which the
voltage trace for VS is parallel to that for VG as would be
the case for M1, M2 both operating in their saturated
regions.
In Figure 4 there is an extended region, above VG
-1.0V
VSS
V
-1.5V
-1.5V
VVDD
VGG
V
0V
V
VSS
-1.0V
-0.5V
-0.0V
0.5V
1.0V
-0.0V
0.5V
1.0V
1.5V
4. Conclusions
The superior DC performance of the proposed FFVF,
compared with that of the FVF is clearly evident.
5. References
[1]
R. G. Carvajal, J. Ramrez-Angulo, A. J. Lpez-Martin, A.

Torralba, J. A. Gmez Galn, A. Carlosena and F. M.
Chavero, The Flipped Voltage Follower: A Useful Cell
for Low-Voltage Low-Power Circuit Design, IEEE
Transactions on Circuits and Systems-I: Regular Papers,
Vol. 52, No. 7, July 2005, pp. 1276-1291.
[2]
P. R. Gray, P. J. Hurst, S. H. Lewis and R. G. Meyer,

Analysis and Design of Analog Integrated Circuits, 4th
Edition, John Wiley and Sons Inc., New York, 2001.
[3]
I. Padilla, J. Ramrez-Angulo, R. G. Carvajal and A. J.

Lpez-Martin, Highly Linear V/I Converter with Programmable Current Mirrors, Circuits and Systems, ISCAS
2007, IEEE International Symposium, 27-30 May 2007.
[4]
P. E. Allen and D. R. Holberg, CMOS Analog Circuit

Design, 2nd Edition, Oxford University Press, Oxford,
2002.
-0.5V
-1.0V
-0.5V
0.5 V where the voltage traces for VD (< VG) and VS are
parallel to that of VG, in accordance with the theory presented. (Above VG = 1 V the onset of triode behavior in
MX causes non linearity)
1.0V
0.5V
-1.0V
VG
VG
Figure 4. Voltage traces for Figure 2 (See text for circuit
details).
1.5V
-1.5V
-1.5V
VVDD
1.5V
VG
VG
Figure 3. Voltage traces for Figure 1 (See text for circuit details).
ENG
Engineering, 2010, 2, 668-672


Yeong Lee Foo, Ah Heng You, Chee Wen Chin
Faculty of Engineering and Technology, Multimedia University, Jln Ayer Keroh Lama, Melaka, Malaysia
E-mail: foo.yeong.lee08@mmu.edu.my
Received February 5, 2010; revised March 22, 2010; accepted March 25, 2010
Abstract
Manual process of creating bonding diagram is known to be time consuming and error prone. In comparison,
offline programming (OLP) provides a much more viable option to reduce the wire bonding creation time
and error. OLP is available in two versions, i.e., vendor specific OLP and direct integration offline programming (Di-OLP). Both versions utilize the bonding diagram and computer aided design data to speed up
bonding program creation. However, the newly proposed Di-OLP is more flexible as it can be used to create
bonding program for multiple machine platforms in microelectronics industry. Some special features of
Di-OLP method are presented. The application of generic OLP however, is applicable to machines that recognize ASCII text file. The user needs to know the data format accepted by machine and convert the data
accordingly to suit its application for different machine platforms. Di-OLP is also a practical method to replace the time consuming manual method in production line.
Keywords: Wire Bonding, Offline Programming, Computer Aided Design, Direct Integration Offline
Programming, Bondlist
1. Introduction
Semiconductor industry is moving in the trend of increased integration and miniaturization. This has resulted in increasing number of bond pads on a chip. These
pads will later be wire bonded to a leadframe via a process called wire bonding [1,2]. Wire bonding process is
basically a process where interconnection between chip
and leadframe is established via thin gold wires. Wire
bond machines utilize precise control of bonding force,
ultrasonic vibration, bonding temperature and bonding
time to establish the connection between gold wire to
bond pad or leadframe. The trend of increased integration has resulted in new challenges for wire bond process;
mainly because more wires are bonded on a chip and pad
pitch has become smaller [3]. A single semiconductor
product can contain as much as 600 wires and pitch distance can be as low as 50 micron or smaller [4]. One of
the main challenges from this trend is the traditional
method to manually prepare the bonding program has
become very time consuming and error prone.
In order to carry out automatic wire bonding, a wire
bond machine requires a set of pre-program instruction.
These instructions will be saved as a wire bond program
(WBP). The WBP is also called wire bond recipe. The
WBP mainly consists of three sets of bonding instruction.
They are material handling, bonding parameter and bonding path instruction. The material handling information
such as magazine dimension, leadframe-indexing pitch
can be keyed into machine relatively fast as in most production floor these dimensions are standardized. Bonding
parameters on the other hand do require slightly more
effort if optimization is required. However when proper
characterization is carried out, such as grouping of the
bonding parameter by different types of bond pad material, bonding capillary, etc, it allows user to re-use the
bond parameter when coming across the same bonding
condition. This will allow wire bond parameter to be
keyed into WBP with relative ease.
The standardization and characterization option, however do not apply to the bonding path component of
bonding program. Bonding path component is required
to guide bond head to the correct position during the
bonding process. Bonding path component can be represented by a set of bonding coordinates with each consists
of two points that are connected to form a representing
connectivity between bond pad and leadfinger as shown
in Figure 1. According to the conventional manual
method, every new product would require the user to
manually input all bonding path coordinates into the
bonding program one by one.
The conventional manual method of inputting bonding
coordinates into bonding program is suitable for product
ENG
Y. L. FOO ET AL.
Wire,
W1 = xb1, yb1, xw1, yw1
W2 = xb2, yb2, xw2, yw2
W3 = xb3, yb3, xw3, yw3
= .
= .
= .
= .
= .
= .
Wn = xbn, ybn, xwn, ywn
Figure 1. A bonding diagram and a set of bonding coordinates representing bonding path component. Cross symbol shows the coordinate system origin at the center of the
chip and leadframe.
with relatively low pin count (less than 48 wires). However as wire count increases and bond pitch becomes
smaller the manual bonding program preparation method
becomes very tedious and time consuming. Moreover
errors are likely to occur with this method. Errors are
likely to occur due to the fact that operator only has access to locally enlarge image of bond pad and leadframe
during the manual bonding program preparation process.
This narrow vision makes it very difficult for operator to
recognize the exact bonding bond pad position where
bonding wire is to be placed. The risk of misplacing the
wire on the adjacent bond pad increases, as bond pad
pitch becomes smaller.
These limitations of manual bonding program creation
process have driven user and equipment vendor to explore other option, to speed up bonding program creation process and reduce human errors. The solution is
known as offline programming (OLP) method. This method extracts the bonding coordinates from the computer
aided design (CAD) drawing and utilizes it to simultaneously program all wires [5]. It greatly reduces the time
required to create the bonding program [6-8].
2. Vendor Specific Offline Programming

Currently most machine vendors have their own version
669
of OLP programs. OLP is a more practical approach of

inputting coordinate into bonding program. OLP is a
concept where CAD data is directly extracted from
bonding diagram drawing. The CAD data consists of x
and y coordinates of each bonding point which is used by
OLP software to create workable wire bonding program.
The OLP method is known to speed up bonding program
creation process and improve accuracy of bond program
created [6,9]. Today, most wire bond machine vendors
have their own version of OLP programs. These vendor
specific OLP programs are usually created as add-on
features to CAD application in the market such as AutoCAD. The application requires the engineer to convert
the bonding diagram drawing from *.gds (graphic data
system) format into the *.dwg (drawing) format, dwg
format is standard file format used by AutoCAD for
saving vector graphics file. AutoCAD is used to open the
bonding diagram in *.dwg format. The engineer then
selects the bonding reference points on the chip and
leadframe. The reference points will enable the machine
to create a coordinate system origin where all bonding
coordinates will be referred. Reference points also enable
machine to precisely compensate for any die placement
variations or orientation that occurs during the die-attached process.
Once coordinate reference points are defined, user will
trigger the wire bond OLP procedure to extract two end
points from each line and uses it to create WBP. The
WBP created by OLP procedure can be directly loaded to
wire bond machine and allows all wires to be created in
bonding program simultaneously as shown in Figure 2.
This helps to address the problem of long programming
time associated with manual bonding program creation
process.
However, the vendor specific OLP program does have
it disadvantages. The vendor specific OLP is usually
very rigid as they only create WBP usable for specific
type of wire bond machine, as data format acceptable by
wire bond machine is different from one vendor to the
other. If a production line consists of wire bond machine
from different vendors, one must procure multiple ver-
Create reference system for

chip and leadframe.
Loading of computer aided

design wire coordinate to
create wire bond program.
Figure 2. OLP allows all wires to be created in bonding program simultaneously.

ENG
670
Y. L. FOO ET AL.
sions of OLP softwares from different wire bond machine vendor. This will increase the implementation cost
of vendor specific OLP. The other disadvantage of vendor specific OLP is, it only works on specific version of
CAD application. OLP from wire bond vendor A might
requires AutoCAD 2005 to work with, while OLP from
wire bond vendor B might requires AutoCAD 2006 to
create WBP. This means if a production line consists of
wire bond machine from different vendors, one might
need to license two or more versions of CAD software.
This would translate into additional licensing fees. The
cost of implementation and the inflexibility of vendor
specific OLP are key factors that hinder wide spread of
vendor specific OLP.
Another alternative to the vendor specific OLP is to
use the bondlist created by bonding diagram creation
software. Bondlist information can be converted into
machine recognizable format and carries out OLP. The
method on how bondlist information can be utilized for
OLP is presented in the following section. The term direct integration offline programming (Di-OLP) is used to
differentiate this method from vendor specific OLP.
filled. Bonding diagram must be created in scale of 1:1

and coordinates system origin of all bonding points must
be referred to the center of chip or leadframe.
First step of Di-OLP involves extraction of bondlist
from bonding diagram as shown in Figure 3. Bondlist is
a text file containing coordinates representing end points
of all lines drawn to represent wires connecting chip
bond pad to leadframe leadfinger. For bonding locations
on leadframe, coordinate system origin is at the center of
the leadframe/package drawing, thus all bonding location
coordinates are referred to this origin and no transformation effort is required. Figure 4 shows the bonding loca-
3. Direct Integration Offline Programming

Di-OLP involves creating bonding program from bondlist data. Bondlist is a text file containing all the bonding coordinates in a bonding diagram that defines the
wire connectivity. All these coordinates are referring to
the bonding diagram coordinate system origin. Although
bondlist contains all the coordinates of bonding diagram,
it cannot be directly uploaded to the machine to create
bonding program. The first reason is the coordinate system origin for bondlist file does not match wire bond
machine coordinate system origin. Second, the data format of bondlist is different from machine recognizable
structure. Understanding of coordinate system origin of
bonding diagram and data format acceptable by machine
is important for successful implementation of Di-OLP.
Successful implementation requires bonding diagram
created to adhere to the relevant rules. Bonding diagram
must be drawn in scale 1:1, and the coordinate system
origin of all bonding co-ordinates must be referred to the
center of the chip and leadframe in Figure 2 [2,6]. Setting the coordinate system origin to the center of package
enables user to easily match the bonding diagram coordinate system origin to machine coordinate system origin.
The vendor specific OLP only needs the bonding diagram to fulfill the first requirement. For the second requirement, vendor specific OLP allows user to set coordinate system origin on any location of chip and leadframe. Coordinates of all wires in vendor specific OLP
will be referred to the user defined coordinate system
origin.
In Di-OLP, however both conditions need to be fulCopyright 2010 SciRes.
Figure 3. Example of Bondlist extraction from bonding

diagram.
Center of leadframe
PIN1 , , 5073.28, -3230.81,

PIN2 , , 5090.99, -2964.26,
PIN3 , , 5100.44, -2715.11,
PIN4 , , 5125.39, -2472.47,
PIN5 , , 5142.19, 2232.36,
In bondlist all the bonding location on the leadfinger is represented by a set of x and y coordinate.
Figure 4. Bonding location on leadframe is referred to coordinate system origin at the center of the leadframe drawing.
ENG
Y. L. FOO ET AL.
tion on leadframe is referred to coordinate system origin

at the center of the leadframe drawing. However, for
bonding coordinates on the chip, it is found that the coordinate system origin is not referred to the center but
instead to the lower left corner of the chip. As a result the
data needed to be transformed to the center of package
before it can be used for Di-OLP.
Just like vendor specific OLP, Di-OLP helps to reduce
time needed to create WBP. It minimizes error such as
missing wire or misplaced wire. It also improves machine utilization as bonding program can be created offline
instead of using productive machine operation time.
Besides the advantages described earlier Di-OLP is more
flexible compared to vendor specific OLP. It utilizes
bondlist created by bonding diagram creation software.
This allows the OLP application without the need to
license for specific CAD software or updated software
version. Di-OLP will enable user who is not familiar
with CAD software to carry out OLP as bondlist text
format can be converted to machine usable and recognizable data format using worksheet application. Di-OLP
is suitable for implementation in production line with
multiple machine platforms as long as machines accept
the coordinates in ASCII format. This eliminates the
needs for multiple vendor specific OLP softwares thus
reduces the cost of OLP implementation across multiple
machine platforms.
4. Transformation of Bonding System

Coordinates
4.1. Transformation from Lower Left Side of
Chip to Center of Chip
Any two-dimensional coordinate point in a Cartesian
coordinate system can be represented by x and y coordinates by referring to a system origin, (0, 0). A vector can
be used to represent a point in a coordinate system, i.e.,
xn
Pn .
yn
When a new origin point is to be used (x, y) coordinate
point is then translated to (x, y) and the coordinates of
x and y refer to this new origin can be obtained using
the transformation vector. Figure 5 shows the transformation of coordinate origin from lower left corner of
chip to center of chip using vectors.
Poo is a vector representing the transformation of new
origin from the initial origin point. It is given by
xoo'
Poo'
yoo'
(1)
Thus, the new coordinate Pno with refer to the center

of package can be obtained by solving the following
vector equation
671
o
Poo
Pno
Pno
Figure 5. Using vector to transform coordinate origin from

lower left corner of chip to center of chip.
Pno = Pno Poo.
(2)
In the matrix form, it is given as

xno' xno xoo'
(3)
.
yno' yno yoo'
With this method, a new set of coordinates can be obtained by transforming the origin of all the bonding coordinates from lower left corner to the center of the chip.
4.2. Transformation Due to Chip Orientation

When chip layout drawing is merged with leadframe
drawing, the chip might require to be turned to 90 to
ensure that the chip can be fitted into the island pad of
leadframe. This chip might also be turned 90, 180 or
270 if the pad one is required to be located at a predefined location with reference to the leadframe. The
bondlist text file contains original coordinates from GDS
(Graphic Data System) II file that have not been transformed. If the untransformed data is loaded to the machine the wire bond machine cannot interpret the coordinate correctly.
The orientation transformation can be achieved by the
transformation equation.
For rotation by an angle counterclockwise about the
origin, the functional forms are x' = xcos ysin and y'
= xsin + ycos as shown in Figure 6. The equations can
be written in matrix form as given below:
x ' cos sin x

(4)
y ' sin cos y

By solving the matrix, it will provide the coordinates
for orientation in z-axis at the center of the package.
The two sets of transformation equations described
above will enable coordinates origin being transfered to
center of package and enable all coordinates being
oriented to appropriate angle. The matrix transformations
allow user to create a set of data that can be interpreted
by wire bond machine. Once the transformations are
ENG
Y. L. FOO ET AL.
672
machines that recognize ASCII text file. The user needs

to understand the data format accepted by machine and
converts the data in order to suit its application to different machine platforms. Di-OLP is a more suitable
method to replace the time consuming manual method.
y
P (x, y)
6. References
P (x, y)
[1]
Figure 6. A set of coordinates P being rotated in degree.
completed the data will be converted to ASCII format

recognizable by the machine. Different type of machine
is known to accept different text format. Some machines
accept data in comma delimited format while others
accept data in space delimited format. User of Di-OLP
need to understand the exact format accepted by a
particular machine type and carry out conversion of data
accordingly. These data can then be loaded to the wire
bond machine to simultaneously create all the connections required.
5. Conclusions
Manual process of creating bonding diagram is found to
be time consuming and error prone. OLP provides a
much more viable option to reduce the wire bonding
creation time and error. OLP is available in two versions,
vendor specific OLP and Di-OLP as described. Both
versions utilize the bonding diagram CAD data to speed
up bonding program creation. However, the proposed
Di-OLP is more flexible as it can be used to create
bonding program for multiple machine platforms. The
application of generic OLP however, is applicable to
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
S. Kalpakjian, Manufacturing Engineering and Technology, 3rd Edition, Surface Technology, Kansas, 1995.
S. DiBartolomeo, Advance Packaging, Penn Well,
Nashua, 2000.
R. R. Tummala, V. Sudaram, F. Liu, G. White, S. Bhattacharya, R. M. Pulugurtha, M. Swaminathan, J. Laskar,
N. M. Jokerst and S. Y. Chow, High Density Packaging
in 2010 and Beyond, IEEE International Symposium on
Electronic Materials and Packaging, Taiwan, 2002, pp.
30-36.
L Nguyen, I. Singh, C. Murray, J. Jackson, J. DeRosa and
D. Ho, 70 m Fine Pitch Wire Bonding, IEEE International Electronics Manufacturing Technology Symposium, Adelaide, 1998, pp. 394-400.
T. C. Chang, R. A. Wysk and H. P. Wang, ComputerAided Manufacturing, 3rd Edition, Prentice Hall, New
York, 1998.
S. K. Prasad, Advanced Wire Bond Interconnection,
Springer, Berlin, 2004.
C. J. Oh, Y. J. Lee, Y. J. Han and C. S. Ahn, A New
System for Reducing the Bonding Process Cycle Time
and Increasing the Accuracy of Bonding Diagram, IEEE
International Conference on System, Man and Cybernetics, Vol. 5, 2004, pp. 4301-4305.
Y. L. Foo, A. H. You and C. W. Chin, Direct Integration
Offline Programming Method in Wire Bonding Process,
11th International Conference on Electronic Materials
and Packaging, Taiwan, 2009, pp. 1-5.
G. G. Harman, Wire Bonding in Microelectronics Materials, Processes, Reliability and Yield, 2nd Edition,
McGraw-Hill, New York, 1997.
ENG
Call for Papers
ENGINEERING
A Journal Published by Scientific Research Publishing, USA
www.scirp.org/journal/eng
9 771947 393005
Editor-in-Chief
08
Editorial Board
Dr. Rehan Ahmed
Prof. MohD A. M. Al-Nimr
Prof. Jan Awrejcewicz
Prof. Shahnor Basri
Prof. Alain. Bernard
Dr. P. Chandramohan
Dr. Hongyang Chen
Prof. Ming Chen
Dr. Rahat Iqbal
Prof. Shouetsu Itou
Prof. Jae Moung Kim
Prof. Chui-Chi Lee
Prof. Yong-Gang Lv
Prof. John Marsh
Dr. Mehrdad Massoudi
Dr. Koduri Venkata Surya Ramam
Prof. Sergey M. Smolskiy
Prof. Hongbin Sun
Prof. Baolin Wang
Prof. Chengshan Wang
Prof. Gaofeng Wang
Prof. Zhao Xu
Dr. Wei Yan
Dr. V.P. Yashnikov
Prof. Tsutomu Yoshihara
Prof. Xiangjun Zeng
Dr. Hongyu Zhang
Prof. Luowei Zhou
H e r i o t-Wa t t U n i v e r s i t y, U K
J o r d a n U n i v e r s i t y o f S c i e n c e a n d Te c h n o l o g y, J o r d a n
Te c h n i c a l U n i v e r s i t y o f L o d z , Po l a n d
P u t ra U n i v e r s i t y, M a l a y s i a
E c o l e C e n t ra l e d e N a n t e s , F ra n c e
A n n a U n i v e r s i t y, I n d i a
T h e U n i v e r s i t y o f To k y o , J a p a n
S o u t h e a s t U n i v e r s i t y, C h i n a
C o v e n t r y U n i v e r s i t y, U K
Ka n a g a w a U n i v e r s i t y, J a p a n
I N H A U n i v e r s i t y, Ko r e a ( S o u t h )
S H U -T E U n i v e r s i t y, Ta i w a n ( C h i n a )
I E E E P h o t o n i c s S o c i e t y, U K
U . S . D e p a r t m e n t o f E n e r g y / N E T L- P i t t s b u r g h , U S A
Universidad of Concepcion, Chile
M o s c o w Po w e r E n g i n e e r i n g U n i v e r s i t y, R u s s i a
Ts i n g h u a U n i v e r s i t y, C h i n a
H a r b i n I n s t i t u t e o f Te c h n o l o g y, C h i n a
T i a n j i n U n i v e r s i t y, C h i n a
Wu h a n U n i v e r s i t y, C h i n a
Te c h n i c a l U n i v e r s i t y o f D e n m a r k , D e n m a r k
Tr e n d M i c r o , U S A
Russian Academy of Sciences, Russia
Wa s e d a U n i v e r s i t y, J a p a n
C h a n g s h a U n i v e r s i t y o f S c i e n c e a n d Te c h n o l o g y, C h i n a
C e r e s I n c ., T h o u s a n d O a k s , U S A
ENGINEERING is an international journal dedicated to the latest advancement of engineering. The goal of
this journal is to provide a platform for engineers and academicians all over the world to promote, share,
and discuss various new issues and developments in different areas of engineering. All manuscripts must
be prepared in English, and are subject to a rigorous and fair peer-review process. Accepted papers will
immediately appear online followed by printed hard copy. The journal publishes original papers including
but not limited to the following fields:
Aerospace Engineering
Agricultural Engineering
Chemical Engineering
Civil Engineering
Electrical Engineering
Environmental Engineering
Industrial Engineering
Materials Engineering
Mechanical Engineering
Mining Engineering
Nanotechnology
Nuclear Engineering
Power Engineering
Test Engineering
Transportation Engineering
We are also interested in: 1) Short Reports2-5 page papers where an author can either present an idea
with theoretical background but has not yet completed the research needed for a complete paper or
preliminary data; 2) Book ReviewsComments and critiques.
Notes for Intending Authors

Submitted papers should not be previously published nor be currently under consideration for publication
elsewhere. Paper submission will be handled electronically through the website. For more details, please
access the website.
Website and E-Mail

Hh ttp://www. scirp.org/journal/eng
eng@scirp.org
Engineering, 2010, 2, 559-672
Published Online August 2010 in SciRes (http://www.scirp.org/journal/eng/)
TABLE OF CONTENTS
August 2010
Volume 2 Number 8
Experimental Comparative and Numerical Predictive Studies on Strength Evaluation of

Cement Types: Effect of Specimen Shape and Type of Sand
H. Hodhod, M. A. M. Abdeen........................................................................................................559
Hydrogen Pick up in Zircaloy-4: Effects in the Dimensional Stability of Structural

Components under Nuclear Reactor Operating Conditions
P. Vizcano, C. P. Fagundez, A. D. Banchik....................................................................................573
Electrochemical Generation of Zn-Chitosan Composite Coating on Mild Steel and its

Corrosion Studies
K. Vathsala, T. V. Venkatesha, B. M. Praveen, K. O. Nayana............................................................580
Tunable Erbium-Doped Fiber Lasers Using Various Inline Fiber Filters

S.-K. Liaw, K.- C. Hsu, N.- K. Chen............................................................................................... 585
Behaviour of a Composite Concrete-Trapezoidal Steel Plate Slab in Fire

T. Hozjan, M. Saje, I. Planinc, S. Srpcic, S. Bratina.......................................................................594
The Effect of Initial Oxidation on Long-Term Oxidation of NiCoCrAlY Alloy

C. Zhu, X. Y. Wu, Y. Wu, G. Y. Liang.............................................................................................602
Highly Nonlinear Bending-Insensitive Birefringent Photonic Crystal Fibres

H. Ademgil, S. Haxha, F. AbdelMalek...........................................................................................608
Progress in Antimonide Based III-V Compound Semiconductors and Devices

C. Liu, Y. B. Li, Y. P. Zeng............................................................................................................617
Lie Group Analysis for the Effects of Variable Fluid Viscosity and Thermal Radiation on
Free Convective Heat and Mass Transfer with Variable Stream Condition
P. Loganathan, P. P. Arasu............................................................................................................ 625
Statistical Modeling of Pin Gauge Dimensions of Root of Gas Turbine Blade in Creep
Feed Grinding Process
A. R. Fazeli.................................................................................................................................635
Wind Turbine Tower Optimization under Various Requirements by Using Genetic

Algorithm
S. Yildirim, I. zkol.....................................................................................................................641
A Device that can Produce Net Impulse Using Rotating Masses

C. G. Provatidis...........................................................................................................................648
Computer-Aided Solution to the Vibrational Effect of Instabilities in Gas Turbine

Compressors
E. A. Ogbonnaya, H. U. Ugwu, C. A. N. Johnson............................................................................658
Flipped Voltage Follower Design Technique for Maximised Linear Operation

C. - M. Chen, K. Hayatleh, B. L. Hart, F. J. Lidgey...........................................................................665

Y. L. Foo, A. H. You, C. W. Chin....................................................................................................668
Copyright
2010 SciRes.
Engineering, 2010, 2, 559-672

Engineering 02 08 2010092609263635 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Engineering 02 08 2010092609263635 PDF

Uploaded by

Copyright:

Available Formats

ISSN: 1947-3931 Volume 2, Number 8, August 2010

Wake Forest University, USA

Journal Editorial Board

Wake Forest University, USA

Prof. MohD A. M. Al-Nimr

Prof. Jan Awrejcewicz

Prof. Shahnor Basri

Prof. Alain. Bernard

Dr. Hongyang Chen

Prof. Ming Chen

Dr. Rahat Iqbal

Prof. Shouetsu Itou

Prof. Jae Moung Kim

Prof. Chui-Chi Lee

Prof. John Marsh

Dr. Mehrdad Massoudi

Dr. Koduri Venkata Surya Ramam

Universidad of Concepcion, Chile

Prof. Sergey M. Smolskiy

Prof. Hongbin Sun

Prof. Baolin Wang

Prof. Chengshan Wang

Prof. Gaofeng Wang

Dr. Wei Yan

Dr. V.P. Yashnikov

Russian Academy of Sciences, Russia

Prof. Tsutomu Yoshihara

Prof. Xiangjun Zeng

Dr. Hongyu Zhang

Prof. Luowei Zhou

Scientific Research Publishing, USA

Engineering, 2010, 2, 559-672

Experimental Comparative and Numerical Predictive Studies on Strength Evaluation of Cement

Hydrogen Pick up in Zircaloy-4: Effects in the Dimensional Stability of Structural Components

Tunable Erbium-Doped Fiber Lasers Using Various Inline Fiber Filters

Behaviour of a Composite Concrete-Trapezoidal Steel Plate Slab in Fire

The Effect of Initial Oxidation on Long-Term Oxidation of NiCoCrAlY Alloy

Highly Nonlinear Bending-Insensitive Birefringent Photonic Crystal Fibres

Progress in Antimonide Based III-V Compound Semiconductors and Devices

A Device that can Produce Net Impulse Using Rotating Masses

Computer-Aided Solution to the Vibrational Effect of Instabilities in Gas Turbine Compressors

Flipped Voltage Follower Design Technique for Maximised Linear Operation

Wire Bonding Using Offline Programming Method

Copyright 2010 SciRes.

Engineering, 2010, 2, 559-572

Experimental Comparative and Numerical Predictive

Hossam Hodhod1, Mostafa A. M. Abdeen2

Department of Structural Engineering, Faculty of Engineering, Cairo University, Giza, Egypt

ening cement had the higher early strength than others.

about the actual composition of cement and its properties.

4. Materials and Specimens

ing compressive strength of mortar with mix proportions

Figure 1. Bags of CEN sand.

Sieve Opening Size (mm)

4.5. Specimens (Cubes and Prisms)

Figure 2. Grading of local sand used in concrete mixes.

Table 1. Investigated types of cement.

Strength Evaluation (on Mortar)

Copyright 2010 SciRes.

CEN Upper Limit

Sieve Opening Size - mm

Figure 3. Grading limits of CEN sand and grading of locally

5.1. Cement Setting Time