ISSN: 19473931 Volume 2, Number 8, August 2010
Engineering Prof. David L. Carroll
Wake Forest University, USA
ISSN: 19473931
www.scirp.org/journal/eng/ 9 771947 393005
08
Journal Editorial Board ISSN: 19473931 (Print), 1947394X (Online) H http://www.scirp.org/journal/ eng
EditorinChief Prof. David L. Carroll
Wake Forest University, USA
Editorial Board Dr. Rehan Ahmed
H e r i o tWa t t U n i v e r s i t y, U K
Prof. Moh’D A. M. AlNimr
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. ChuiChi 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. YongGang 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
C h o n g q i n g U n i v e r s i t y, C h i n a
Editorial Assistant Yanna Li
Scientific Research Publishing, USA
[email protected]
Engineering, 2010, 2, 559672 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. Abdeen……………………………………………………………………………………………………559
Hydrogen Pick up in Zircaloy4: Effects in the Dimensional Stability of Structural Components under Nuclear Reactor Operating Conditions P. Vizcaíno, C. P. Fagundez, A. D. Banchik………………………………………………………………………………………573
Electrochemical Generation of ZnChitosan Composite Coating on Mild Steel and its Corrosion Studies K. Vathsala, T. V. Venkatesha, B. M. Praveen, K. O. Nayana…………………………………………………………………580
Tunable ErbiumDoped Fiber Lasers Using Various Inline Fiber Filters S.K. Liaw, K.C. Hsu, N.K. Chen…………………………………………………………………………………………585
Behaviour of a Composite ConcreteTrapezoidal Steel Plate Slab in Fire T. Hozjan, M. Saje, I. Planinc, S. Srpčič, S. Bratina……………………………………………………………………………594
The Effect of Initial Oxidation on LongTerm Oxidation of NiCoCrAlY Alloy C. Zhu, X. Y. Wu, Y. Wu, G. Y. Liang……………………………………………………………………………………………602
Highly Nonlinear BendingInsensitive Birefringent Photonic Crystal Fibres H. Ademgil, S. Haxha, F. AbdelMalek…………………………………………………………………………………………608
Progress in Antimonide Based IIIV 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. Yıldırım, İ. Özkol………………………………………………………………………………………………………………641
A Device that can Produce Net Impulse Using Rotating Masses C. G. Provatidis…………………………………………………………………………………………………………………648
ComputerAided 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
Wire Bonding Using Offline Programming Method Y. L. Foo, A. H. You, C. W. Chin………………………………………………………………………………………………668
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, Email:
[email protected]
SERVICES Advertisements Advertisement Sales Department, Email:
[email protected] Reprints (minimum quantity 100 copies) Reprints Coordinator, Scientific Research Publishing, Inc., USA. Email:
[email protected]
COPYRIGHT Copyright©2010 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: Email:
[email protected]
Engineering, 2010, 2, 559572 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 Email: {hossamhodhod, mostafa_a_m_abdeen}@hotmail.com Received May 6, 2010; revised July 23, 2010; accepted July 25, 2010
2
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 EN196 and EN197 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 CEMI 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 CEMI 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 [13]. 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 skijump bucket using Neural Networks (NN). Abdeen [13] presented a study for the development of ANN Copyright © 2010 SciRes.
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 loadresisting system for multistory 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 EN197. 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 EN196. 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
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 I52.5N
SINAI
ElMASRIYA
HELWAN
3
CEM I42.5N (SRC)(**) CEM I42.5N (SRC)
4
CEM I42.5N
ElMASRIYA
5
CEM I42.5R
HELWAN
2
CEM I42.5N HELWAN (White) 7 CEM I32.5R ELKAUMIYA (NCC) 8 CEM I32.5R (SRC) ASSIUT (CEMEX) 9 CEM IIBS 32.5N HELWAN 10 CEM IIBL 32.5N ELKAUMIYA (NCC) (*) Sand having a grading similar to CEN sand. (**) This cement will be denoted in figures as SRC1. 6
Copyright © 2010 SciRes.
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
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 BS32.5 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 EN197.
5.2. Mortar Compressive Strength Compressive strength measured for all specimens and Copyright © 2010 SciRes.
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 noteworthy 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 68 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.
Copyright © 2010 SciRes.
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 Copyright © 2010 SciRes.
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: 040 mm, 4080 mm, and 80120 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.
Copyright © 2010 SciRes.
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 42.5NSRC
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 IIBS32.5N CEM II BL32.5N
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 human’s 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.
Copyright © 2010 SciRes.
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 42.5NSRC
15
CEM I 42.5N CEM I 42.5R
10
CEM I 42.5N white
5 0 0
10
20
ET AL. Concrete Compressive Strength (MPa)
Concrete Compressive Strength (MPa)
H. HODHOD
567
30 25 20
CEM I 32.5R CEM I 32.5RSRC
15
CEM IIBS32.5N CEM II BL32.5N
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 32.5RSRC
10 5 0 0
10
20
30
Concrete Compressive Strength (MPa)
Concrete Compressive Strength (MPa)
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 42.5NSRC CEM I 42.5N
15
CEM I 42.5N white CEM IIBS32.5N
10
CEM II BL32.5N
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 Copyright © 2010 SciRes.
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
1
CEM I52.5N CEM I42.5N (SRC) (**) CEM I42.5N (SRC) CEM I42.5N CEM I42.5R CEM I42.5N (White) CEM I32.5R CEM I32.5R (SRC) CEM IIBS32.5N CEM IIBL32.5N
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
1
52.5
19
100
0
1
42.5
19
100
0
60 67
0
1
42.5
19
100
0
1 1
42.5 42.5
19 23
100 100
0 0
1
42.5
19
100
0
50
0
1
32.5
23
100
0
60
0
1
67
2
60 60
0 0
1
32.5
23
100
0
67
0
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)
Copyright © 2010 SciRes.
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 neuron’s 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 (RMSError). 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 agedays 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)
Copyright © 2010 SciRes.
Input Layer
No. of Neurons in each Layer Second First Hidden Third Hidden Hidden
Output Layer
4
7
5
3

1
4
7
5
3

1
5
8
6
4
2
1
5
4
4
3
2
1
5
8
6
4
2
1
5
4
4
3
2
1
5
8
6
4
2
1
ENG
H. HODHOD ET AL.
570
Table 5. Parameters used in the developed neural network models.
Training Epochs MPRE RMSError
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 Copyright © 2010 SciRes.
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 M I 52.5 N
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 M I 52.5 N
25 20 15 10
Experiment ANN Training
5
C E M I 52.5 N
30.0 Compressive Strength Mpa
30 Compressive Strength Mpa
No. 1
Exp.
ANN Prediction
0
25.0 20.0 15.0 10.0
Experiment ANN Training
5.0
ANN Prediction
0.0 0
10
20
30
40
50
0
10
Age  days
20
30
40
50
Age  days
Figure 18. Prism compressive strength (left: cen sand, right: regenerated sand).
C E M I 52.5 N
12 8 Experiment ANN Training
4
C E M I 52.5 N
16 Flexural Strength  MP
Flexural Strength  MP
16
ANN Prediction
12 8 Experiment ANN Training
4
ANN Prediction
0
0 0
10
20
30
40
Age  days
50
0
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 Copyright © 2010 SciRes.
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
201206.
range of the experiment or in the future).
10. Acknowledgements The Authors would like to express their gratitude towards Prof. Dr. Farouk ElHakim of 15th May institute for Civil and Arch. Engineering, and undergraduate students (4th year–civil) 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. 2333.
[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.
Copyright © 2010 SciRes.
[9]
K. A. Kheireldin, “Neural Network Application for Modeling Hydraulic Characteristics of Severe Contraction,” Proceedings of the 3rd Internetional Conference, Hydroinformatics, Copenhagen, 2426 August 1998, pp. 4148.
[10] M. A. M. Abdeen, “Neural Network Model for Predicting Flow Characteristics in Irregular Open Channel,” Scientific Journal, Faculty of EngineeringAlexandria University, Vol. 40, No. 4, 2001, pp. 539546. [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. 898908. [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 LoadResisting 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. 18301842. [16] Y. Shin, “NeuralystTM User’s Guide—Neural Network Technology for Microsoft Excel,” Cheshire Engineering Corporation Publisher, Pasadena, 1994.
ENG
Engineering, 2010, 2, 573579 doi:10.4236/eng.2010.28073 Published Online August 2010 (http://www.SciRP.org/journal/eng).
Hydrogen Pick up in Zircaloy4: Effects in the Dimensional Stability of Structural Components under Nuclear Reactor Operating Conditions Pablo Vizcaíno, Cintia Paola Fagundez, Abraham David Banchik Centro Atómico Ezeiza, Comisión Nacional de Energía Atómica, Presbítero J. González y Aragón Nº 15, Buenos Aires, Argentina Email:
[email protected] 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 Zircaloy4 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 Zircloy4 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 wppmH up to 650 wppmH, 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 Zircaloy4 homogeneously hydrided samples was measured at 300oC, the typical operating temperature of a nuclear power reactor, obtaining a relative expansion of 2.21 × 104% per wppmH. Considering the relative expansion observed for Zircaloy4 at room temperature due to hydriding, starting from a hydrogen free sample, the total relative expansion rate is calculated to be 5.21 × 104% per wppmH. Keywords: Thermal Analysis, Dimensional Change, Hydrides, Zircaloy4
1. Introduction Most of the core structural components of the nuclear power reactors are made of Zicaloy4, 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 Copyright © 2010 SciRes.
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 Zircaloy4 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 quasiradial 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. VIZCAÍNO
574
aim of the present work is to determine the expansion coefficient of Zircaloy4 for the axial direction of the channel at the reactor operating temperature (300°C) [2].
2. Experimental Procedure 2.1. Material and Sample Preparation The Zircaloy4 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 coldshaped and welded (by the tungsten inert gas method) from fully recrystallized Zircaloy4 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 = 0.220.26, Fradial = 0.670.73. 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 1520 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 ± 2°C. 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 RH404 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 wppmH.
2.3. Differential Dilatometry A Shimadzu TMA60H 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 Zircaloy4 sample was used, containing about 20 wppmH, 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 5°C/min. After keeping the samples 30 minutes at the maximum temperature they were cooled down at 5°C/min. To avoid the effect of the transformation, the maximum temperature was a few degrees below 550°C.
2.4. Differential Scanning Calorimetry
20 m
Figure 2. Fully recrystallized microstructure of a Zircaloy4 cooling channel. The typical size of the equiaxed grains is 20 ± 6 m.
Copyright © 2010 SciRes.
The calorimetric experiments were made using a thermal flux differential scanning calorimeter Shimadzu, model DSC60. 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 5°C/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. VIZCAÍNO
0.20
Extrapolation lines
0.05
2
DSC curve Derived curve
4.0
2
0.10 3.0
2
0.15 2.0
d q/dt (mJ/sec )
msT
1.0
dq/dt (mJ/sec)
0.25
0.00
pT
5.0
o
25 C
100
200
300
400 o
500
0.05
Temperature ( C)
Figure 3. Calorimetric curve of a sample containing 480 wppmH, 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 hydrogenfree reference sample (in fact it contains 20 wppmH). 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 (550°C), 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 wppmH, at 403°C. 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 40°C to 50°C. 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
8
300
6 200
4 Lengh increase due to the hydrogen incorporation
100 0
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.
Copyright © 2010 SciRes.
o
0.0020
403 C keT
0.0015
2.0
L(m)
10
400
Expansion curve Derived curve
2.5
12
L (m)
Temperature(o(篊 C))
500
575
dL/dt (m/sec)
cT
0.0
ET AL.
1.5
0.0010
o
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 wppmH. ENG
P. VIZCAÍNO
576
for other zirconium alloys [3], but no difference was observed between them for Zircaloy4. Thus, TTSSd was measured at the knee point (keT). In any case, it was observed that within an uncertainty interval of 24°C, 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)
5
(3)
CcT = 1.78 × 10 exp (4546/cT) 500
Temperature(o(C 癈) )
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癈 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.
Copyright © 2010 SciRes.
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 50°C, an interval larger than the 25°C 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 wppmH, where the scatter of the data is large. For these data, TTSSd error varies from 18°C to 15°C at the upper extreme of the interval. For higher concentrations (in our case, concentrations higher than 187 wppmH) the error decreases to 10°C, becoming constant for concentrations higher than 250 wppmH, where an error of 8°C 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 [37]. As a brief summary we can say that: Z. L. Pan, measuring Young’s modulus as functions of temperature and hold time during quasistatic thermal cycles to Zr2.5Nb 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 Zr20wt%Nb hydrided/deuterided samples with neutron diffraction measurements [6]. Recently, the authors of the present work determined TTSSd for Zr2.5Nb 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. VIZCAÍNO
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 hydrogenfree 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 550°C. 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 Zircaloy4. 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
o
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 2°C 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 wppmH. 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 wppmH, 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: μm (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: ΔL 5.6 μm 0.054 [H ]
(
ΔL 1 ) room 3.2 104 3.00 106 [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 twophase 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 (300°C) should be added to the growth due to the hydrides already present at room temperature. The measurements made at 300°C 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 Rvalue (R) are also given (97% of confidence). Intercept
Slope
Statistics
Value
SE
LCL
UCL
Value
SE
LCL
UCL
R
SD
5.6
2.0
0.6
10.7
0.054
0.004
0.044
0.064
0.9
5
Copyright © 2010 SciRes.
ENG
578
P. VIZCAÍNO Table 2. Relative expansion at 300°C. lo (m)
l/lo
0.85
9781
8.69 E5
5.0
10332
4.81E4
398
5.7
10035
5.64E4
358
5.65
10035
5.63E4
446
9.5
9784
9.75E4
650
12.9
9968
1.29E3
[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: ΔL 1 )300 º C 1.54 104 2.21106 [H ] (3) L0 wppm where (L/Lo)300°C is the relative length increase of the hydrided sample at 300°C. Then, combining (2) and (3), the total length increase is calculated as follows: ΔL ΔL ΔL )TOTAL ) room )300 º 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 Rvalue (R) are also given (97% of confidence). Intercept (× 104)
Slope (× 106)
Statistics
Value
SE
LCL
UCL
Value
SE
LCL
UCL
R
SD
−1.54
1.3
−5.5
2.5
2.21
0.33
1.22
3.2
0.96
1.3 × 104
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 300°C is 5.21 × 104% per wppmH. 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 wppmH [11]. If we choose a medium concentration of 250 wppmH 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
o
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
o
300 C
L/Lo
0.0000 100
200
300
400
[H] (wppm)
500
600
Figure 9. Relative expansion at 300°C, 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 Copyright © 2010 SciRes.
of the expansion coefficient of Zircaloy4 for the axial direction of a tube. As a brief summary, the following points must be underlined: The hydrogen solid solubility curve for Zircaloy4 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 wppmH up to the eutectoid temperature (550°C) concentration (650 wppmH), 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 Zircaloy4 homogeneously hydrided samples was measured at 300°C, the typical operating temperature of a nuclear power reactor, obtaining a relative expansion of 2.21 × 104% per wppmH. Adding to this value the relative expansion coefficient at room temperature due to hydriding, the total relative expansion rate is 5.21 × 104% per wppmH.
6. References [1]
J. C. Ovejero, A. D. Banchik and P. Vizcaíno, “Axial/ Tangential Expansion Coefficients of Zircaloy4 Channels Due to the Hydrogen Pick up,” Advanced in TechENG
P. VIZCAÍNO
[2]
[3]
[4]
[5]
[6]
nology of Materials and Materials Processing Journal, Vol. 10, No. 1, 2008, pp. 18. C. P. Fagundez, P. Vizcaíno, D. Bianchi and A. D. Banchik, “Dilatometría del Sistema ZrH,” Proceedings of the Congress SAM/CONAMET 2005, Mar del Plata, 1821 October 2005. J. P. Giroldi, P. Vizcaíno, A. V. Flores and A. D. Banchik, “Hydrogen Terminal Solid Solubility Determinations in Zr2.5 Nb Pressure Tube Microstructure in an Extended Concentration Range,” Journal of Alloys and Compounds, Vol. 474, No. 12, 2009, pp. 140146. D. Khatamian and V. C. Ling, “Hydrogen Solubility Limits In  and Zirconium,” Journal of Alloys and Compounds, Vol. 253254, No. 20, 1997, pp. 162166. 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. 173195. Z. L. Pan and M. P. Puls, “Precipitation and Dissolution Peaks of Hydride in Zr2.5 Nb during Quasistatic Ther
Copyright © 2010 SciRes.
ET AL.
579
mal Cycles,” Journal of Alloys and Compounds, Vol. 310, No. 12, 2000, pp. 214218. [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. 106113.
[8]
P. Vizcaíno, A. D. Banchik and J. P. Abriata, “Calorimetric Determination of the Hydride Dissolution Enthalpy in Zircaloy4,” Metallurgical and Materials Transactions A, Vol. 35A, No. 8, 2004, pp. 23432349.
[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. 292303.
[10] E. Zuzek, J. P. Abriata and A. San Martín, “HZr (HydrogenZirconium),” Bulletin of Alloy Phase Diagrams, Vol. 11, No. 4, 1990, pp. 385395. [11] P. Vizcaíno, A. D. Banchik and J. P. Abriata, “Solubility of Hydrogen in Zircaloy4: Irradiation Induced Increase and Thermal Recovery,” Journal of Nuclear Materials, Vol. 304, No. 23, 2002, pp. 96106.
ENG
Engineering, 2010, 2, 580584 doi:10.4236/eng.2010.28074 Published Online August 2010 (http://www.SciRP.org/journal/eng).
Electrochemical Generation of ZnChitosan 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 Email:
[email protected] Received December 16, 2009; revised February 26, 2010; accepted March 6, 2010
Abstract A ZincChitosan composite coating was generated on mild steel from zinc sulphatesodium 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 [16]. 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 [710]. A survey of literature reveals that the conducting Copyright © 2010 SciRes.
polymers were used for anticorrosive coatings and as inhibitor for steel [1113]. 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 Znchitosan coatings were electrically deposited from sulphatechloride bath. The constituents of the bath were 250 g/L ZnSO4·7H2O, 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 2.53 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 10–15 µm. The corrosion rate by weight loss measurements were performed for mild steel samples coated with pure zinc and Znchitosan 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 JEOLJEM1200EX 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 µm. 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 zincplated 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 Znchitosan composite
2.4. Electrochemical Measurements
Copyright © 2010 SciRes.
9
Corrosive velocity (105kg/m2.h)
A conventional 3electrode cell was used for polarization studies. The zinc coated or Znchitosan 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 Ecochemie 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 Znchitosan 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 Znchitosan coatings are shown in Figure 3. The larger loop was produced by Znchitosan 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 Znchitosan 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(cm2)
500
Figure 5. SEM images for two samples after anodic polarization for (a) zinc coated; (b) composite coated sample.
a
b
zinc coating composite coating
400
300
1m
200
X 4000 1m
X 4000
Figure 6. SEM images for two samples after 15 day’s weight loss measurements. (a) zinc coated; (b) composite coated.
100
0
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.
Copyright © 2010 SciRes.
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 chitosanmetal 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 [1617]. 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 Znchitosan composite coating compared to pure zinc coating. Copyright © 2010 SciRes.
ET AL.
583
5. Conclusions Znchitosan 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 Znchitosan 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% AlZn Coating,” Corrosion, Vol. 49, No. 9, 1993, pp. 912. G. Barcelo, M. Sarret, C. Müller and J. Pregonas. “Corrosion Resistance and Mechanical Properties of Zinc Electrocoatings,” Electrochimical Acta, Vol. 43, No. 12, 1998, pp. 1320. A. Y. Hosny, M. E. ElRafei, T. A. Ramadan, B. A. ElGafari and S. M. Morsy, “Corrosion Resistance of Zinc Coatings Produced from a Sulfate Bath,” Metal Finishing, Vol. 93, No. 11, 1995, pp. 5559. B. Bozzini, V. Accardi, P. L. Cavallotti and F. Pavan, “Electrodeposition and Plastic Behavior of LowManganese ZincManganese Alloy Coatings for Automotive Applications,” Metal Finishing, Vol. 97, No. 5, 1999, p. 33. C. Müller, 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. 11751182. B. M. Praveen, T. V. Venkatesha, Y. A. Naik and K. Prashantha, “Corrosion Studies of Carbon NanotubesZn Composite Coating,” Surface and Coating Technology, Vol. 201, No. 12, 2007, pp. 58365842. B. M. Praveen, T. V. Venkatesha and Y. A. Naik, “Corrosion Behaviour of ZnTiO2 Composite Coating,” Synthesis and Reactivity in Inorganic, MetalOrganic and NanoMetal 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. 939941.
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/Mössbauer Results,” Hyperfine Interactions, Vol. 139140, No. 14, 2002, pp. 133139.
[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. 111115.
[15] X. Pang and I. Zhitomirsky, “Electrophoretic Deposition of Composite HydroxyapatiteChitosan Coatings,” Materials Characterization, Vol. 58, No. 4, 2007, pp. 339348.
[12] R. C. Patil and S. Radhakrishnan. “Conducting Polymer Based Hybrid NanoComposites for Enhanced Corrosion Protective Coatings,” Progress in Organic Coatings, Vol. 57, No. 4, 2006, pp. 332336. [13] B. Szeptycka and A. GajewskaMidziaIek, “The Influence of the Structure of the NanoComposite NiPTFE Coatings on the Corrosion Properties,” Review of Advanced Material Science, Vol. 14, 2007, pp. 135140.
Copyright © 2010 SciRes.
[16] H. P. Sachin, G. Achary, Y. Arthoba Naik and T. V. Venkatesha, “Polynitroaniline as Brightener for Zinc– Nickel Alloy Plating from NonCyanide Sulphate Bath,” Bulletin of Materials Science, Vol. 30, No. 1, 2007, pp. 5763. [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. 20682074.
ENG
Engineering, 2010, 2, 585593 doi:10.4236/eng.2010.28075 Published Online August 2010 (http://www.SciRP.org/journal/eng).
Tunable ErbiumDoped Fiber Lasers Using Various Inline Fiber Filters 1
ShienKuei Liaw1,4, KueiChu Hsu2, NanKuang 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 ElectroOptical Engineering, National United University, Miaoli, Taiwan, China 4 Department of Mechanical Engineering, National Taiwan University, Taipei, Taiwan, China Email:
[email protected] Received February 2, 2010; revised March 23, 2010; accepted March 27, 2010
Abstract Several highperformance and tunable erbiumdoped fiber lasers are reviewed. They are constructed by using fiber Bragg gratings (FBGs) or shortwavelengthpass filters (SWPFs) as wavelength tunable components inside the laser cavity. Broadband wavelength tuning range including C and/or Sband was achieved, and tunable laser output with high slope efficiency, high sidemode suppression ratio was obtained. These fiber lasers can find vast applications in lightwave transmission, optical test instrument, fiberoptic gyros, spectroscopy, material processing, biophotonic imaging, and fiber sensor technologies. Keywords: Fiber Bragg Grating (FBG), ShortWavelengthPass 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 ITUT grids [1]. Also, fiber lasers are useful for spectroscopy, sensing protection, and fiberoptic gyro [2]. Partially because of their features, such as low wavelength sensitivity to temperature, lowintensity noise, and allfiber construction, their advantages over nonfiberbased 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 multiplering cavity fiber laser [3], two separate erbiumdoped fiber lasers [4], distributed feedback fiber lasers [5], and Brillouin erbiumdoped 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 erbiumdoped fiber lasers could well fit such requirements. Nowadays, a variety of tunable fiber lasers have been demonstrated such as tunable singlefrequency fiber lasers [7], coherent combining tunable lasers [8], tunable Copyright © 2010 SciRes.
fiberring 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 FBGbased linearcavity 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, costeffective, 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 21.14mW measured at 1544.8 nm was obtained with a threshold pump power of 8.0 mW. A sidemode 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 shortwavelengthpass filter (SWPF)based tunable fiber ring laser with lasing wavelength down to Sband using filters adopting a sidepolishing [12] or fusedtapering [13] technique to attain the wavelengthdependent fundamentalmode cutoff concept. The fiber sidepolishing and fusedtapering techniques were both employed to fabricate wavelength tunable fiber filters. The laser can be tuned close to the shortwavelength edge of the available erbium gain bandwidth with tuning range of
ENG
586
S.K. LIAW ET AL.
26 nm, signaltoamplifiedspontaneousemission (ASE) ratio of around 40 dB, and the full width half maximum (FWHM) linewidth of about 0.5 nm. The singlelongitudemode (SLM) operation will be briefly discussed in Subsection 5.3. These two kinds of tunable inline filterbased tunable fiber lasers as mentioned may broaden wavelength tuning range in either C and/or Sband and will be addressed in detail. Both of them have graceful features of simple structure, compactness, ease in connection with fiber components, highefficiency, and continuoustuning, 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, ΔλB＝0.00482 nm/μm × 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 3point tuning device by using instant adhesive glue. By tuning the precision screw of the 3point 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
B
Copyright © 2010 SciRes.
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 ε is the applied strain, Λ is the period of fiber, Pi,j are the Pockel’s (piezo) coefficients of the stressoptic tensor, and ν is the Poisson’s ratio. Note that n is the effective refractive index of the fiber core as defined in Equation (1), α is the thermal expansion coefficient of the silica fiber with a typical value of 0.015 nm/ºC, and ΔΤ is the temperature change in degree Celsius. The term (n2/2) [P12－ν(P11－P12)] 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 λB is the Bragg wavelength, and this value gives a “ruleofthumb” measurement of wavelength shift for a FBG with strain of 1 nm per 1000 με at 1.31 μm. To design a strain tunable FBG, firstly, it is embedded in a strip of composite thermal plastic material and then is attached to Lshaped holders at both ends. The FBG is then mounted on a precision translational stage with a highresolution 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. SWPFBased Tunable Fiber Laser The Sband tunable erbiumdoped 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 longwave 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 sidepolishing and the fusedtapering techniques were adopted in our previous works [12,13]. When the SWPFs are temperaturetuned to attenuate the wavelengths longer than 1530 nm, the C + L band ASE is suppressed and the Sband gain is obtained. The commonly used Sband erbiumdoped optical fiber amplifiers (EDFAs) employ erbiumdoped fiber (EDF) with depressed inner cladding to achieve fundamentalmode 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 wavelengthdependent properties. When the optical fiber is sidepolished or tapered, the mode field is expanded out of the fiber cladding. Using dispersive liquids surrounding the sidepolished 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 Gedoped core and the fused silica of SMF28, and the black curve is the dispersion for Cargille indexmatching 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 25ºC–27ºC, where the simulation results were performed by beam propagation method. The Copyright © 2010 SciRes.
(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 shortwavelengthpass 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 Sband erbiumdoped fiber lasers.
3. FBGBased Tunable Fiber Lasers: Configurations and Experimental/ Simulation Results 3.1. Optical Circulator as Laser’s Rear Cavity End The proposed configuration of linear cavity for the tunable laser is shown in Figure 3. The linear cavity consists of a 3port 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 OCbased linearcavity 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 3port 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 Cband 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 signaltonoiseratio (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 linearcavity 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 Cband 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. Copyright © 2010 SciRes.
(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 Cband after power equalization.
3.2. Broadband Fiber Mirror as Rear Cavity End The proposed BFMbased linercavity 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 BFMbased linercavity 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
1
2
3
4
5
6
7
8
9
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 BFMbased tunable fiber laser. (Wavelength is tuned across Cband 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 BFMbased 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. Copyright © 2010 SciRes.
The transfer efficiency here is defined as P in Las th (3) ( PP PP ) where η 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 Cband.
4. Tunable SWPFBased Tunable Fiber Lasers: Configurations and Experimental/ Simulation Results 4.1. SWPFBased Tunable Fiber Laser In this section, a continuously tunable erbiumdoped 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 singlemode 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 Vgroove on a silicon substrate, as shown in Figure 8. The central cladding thickness after polishing was around 2.7 μm. Finally, the characteristics of the wellpolished fibers were calibrated by liquiddrop experiments. The effective interaction length was estimated to be 11 mm at 1550 nm wavelength for sidepolished SMF28. 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 sidepolished fiber is OCK433 (Nye Lubricants) with the thermooptic coefficient dnD/dT of –3.6 × 10–4/°C and is heated by a dual TEC. The refractive index of the OCK433 decreases with increasing temperature. The tuning efficiency is 7.65 nm/°C and the signaltoASE 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 OCK433 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 signalASEratio is above 40 dB and the average FWHM is around 0.5 nm.
Figure 8. Experimental setup of erbiumdoped fiber ring laser using sidepolished fiber based tunable SWPF [12].
(a)
4.2. ThermoOptic Tunable ErbiumDoped FiberRing Laser In this subsection, wideband tunable high cutoffefficiency SWPFs were discretely located in standard silicabased Cband EDF to filter out the C + L band ASE so that the optical gain for Sband could be acquired to realize fiber laser. To investigate the amplification characteristics in the Sband, a 980nm pump laser with 135mW output power was launched into EDF in a forward pumping scheme. The highcutoffefficiency shortpass filters in the 17.5mlong EDF could discretely suppress the unwanted C + L band ASE and pass the Sband signal Copyright © 2010 SciRes.
(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 OCK433 polymer was cool down [12].
ENG
S.K. LIAW ET AL.
and 980nm pump light. Subsequently, an input power of −25 dBm was launched into the EDF from distributed feedback laser signals in the Sband. The input signal spectra and amplified output signal spectra in the Sband at 28.6°C are shown in Figure 10(a). In the Sband the net signal gain at 1486.9 nm was measured to be 18.92 dB. The experimental setup of the tunable EDF ring laser is shown in Figure 11, where the tunable fusedtapered 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 (Sband) of the gain bandwidth. Consequently, we employ fourstage inline tunable fusedtapered fiber SWPFs discretely located in the standard silicabased EDF to achieve the tunable Sband fiber laser. When the SWPF is turned on, the C + Lband ASE is suppressed to obtain the gain for Sband lasing. The four filters are discretely located in a 16mlong stan
(a)
591
Figure 11. Experimental setup of tunable EDF ring laser towards shortwavelength limit at 1450 nm (Each 4mlong EDF and shortpass 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 silicabased Cband 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 980nm 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 η 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 highcutoffefficiency SWPFs. The average FWHM is 0.53 nm and the signaltoASE ratio is above 40 dB.
5. Discussion 5.1. Advantages of FBGBased Tunable Fiber Lasers
(b)
Figure 10. (a) Amplification spectra of the signals in Sband at 28.6°C; (b) evolution of output laser spectra by cooling down optical liquid and bending splicing point using the first set of tapered fibers [13,17]. Copyright © 2010 SciRes.
A versatile and costeffective laser source should have the ability to allow the user to choose which wavelength is needed or the desired scanning range. The wavelength tunable FBGbased 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 finetune the FBG ENG
592
S.K. LIAW ET AL.
pair back to the same wavelength. Nevertheless, either the OCbased or BFMbased 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 polarizationindependent; 2) both the OC based and BFMbased 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 FBGbased 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 SWPFBased Tunable Fiber Lasers It is advantageous to explore a widely tuning fiber laser with lasing wavelength down to Sband at a high tuning speed. Conventionally, the silicabased EDF at room temperature can only emit fluorescence at wavelengths longer than 1490 nm. Thus, achieving highperformance Sband 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 fusedtapering technique are easy, fast, and costeffective fabrication processes. An optimized sidepolishing/tapered fiber filter structure can attain highcutoff efficiency and wide tuning range. Based on the proposed SWPFs, widely tunable, singlefrequency rareearthdoped fiber lasers can be achieved. Besides, the SWPFbased tunable fiber lasers have other advantages such as: 1) wide tuning range covering the Sand Cbands, 2) high power and low noise, 3) simplicity and costeffectiveness, and 4) high index sensitivity up to 1 10–5 with high Q resonator.
5.3. SingleFrequency Design To design a singlefrequency 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 linearcavity fiber lasers as mentioned, a simpler way to achieve singlelongitudinalmode (SLM) operation is to put a piece of EDF as pump absorber between the WDM coupler and 1 × 2 OSW for the OCbased linearcavity tunable fiber laser as shown in Figure 3; and between the 1480/1550 nm WDM and TFBG for the BFMbased linearcavity tunable fiber laser as shown in Copyright © 2010 SciRes.
Figure 5, individually. On the other hand, the SWPF in tunable fiber lasers is naturally a broadband filter that is obviously difficult for singlelongitudinal 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 ultranarrowband filter inside the cavity to attain SLM operation.
6. Conclusions Two kinds of tunable fiberfilterbased EDF fiber lasers have been reviewed. Both of them have broadband wavelength tuning range including C and/or Sband. Using FBG in strain mechanism, we have proposed and demonstrated a tunable FBGbased 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 lowcost VOAs. Another way is to employ a BFM and tunable FBG at either cavity end of fiber cavity. The BFM acts as a broadband rearend reflector both for lasing signal and pump source. For wavelength tunable demonstration, power variation over the whole Cband 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 SWPFbased tunable fiber laser using temperature tuning mechanism, two tunable SPWFs based erbiumdoped fiber lasers were reviewed. The sidepolishing and fusedtapering techniques were used to achieve thermooptic tunable shortwavelengthpass function based on material dispersion discrepancy and variations of waveguide structures. The tuning efficiency is 50 nm/C, cutoff efficiency is –1.2 dB/nm, and rejection efficiency is 55 dB, individually. The widely tunable SWPFs were applied to achieve broadband and hightuningefficiency S and/or Cband EDF ring lasers, which can be tuned close to the shortwavelength edge of gain bandwidth, and the tuning range is 26 nm with the signaltoASEratio 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, highefficiency, and continuous tunability. They are promising for vast applications in lightwave transmission, optical test instrument, fiberoptic 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 98–2221–E011017, NSC 972923–E011001MY3, NSC 98–2218 –E–008004, NSC 982221E239001MY2). 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, “ErbiumDoped Ring Lasers StepTunable to Exact Multiples of 100 Ghz (ITUGRID) Using Periodic filter,” Proceedings of ECOC’98, Madrid, September 1998, pp. 153154. C. S. Kim and J. U. Kang, “Multiwavelength Switching of Raman Fiber Ring Laser Incorporating Composite PolarizationMain Maintaining Fiber LyotSagnac Filter,” Applied Optics, Vol. 43, No. 15, 2004, pp. 31513157. C. C. Lee, Y. K. Chen and S. K. Liaw, “SingleLongitudinalMode Fiber Laser with Passive MultipleRing Cavity and its Application for Video Transmission,” Optics Letters, Vol. 23, No. 5, 1998, pp. 358360. S. Kim, B. Lee and D. H. Kim, “Experiments on Chaos Synchronization in Two Separate ErbiumDoped Fiber Lasers,” IEEE Photonics Technology Letters, Vol. 13, No. 4, 2001, pp. 290292. S. Foster, “Spatial Mode Structure of the Distributed Feedback Fiber Laser,” IEEE Journal of Quantum Electronics, Vol. 40, No. 7, 2004, pp. 884892. M. K. AbdRahman and H. Ahmad, “MultiwaveLength Brillouin Erbium Fiber Laser Pumped from FBG Fiber Laser Sharing the Same EDF,” Proceedings of the 4th Pacific Rim Conference on Lasers and ElectroOptics, Chiba, July 2001, pp. 4041. H. Chen, F. Babin, M. Leblanc and G. W. Schinn, “Widely Tunable SingleFrequency ErbiumDoped Fiber Lasers,” IEEE Photonics Technology Letters, Vol. 15, No. 2, 2003, pp. 185187. D. Sabourdy, V. Kermene, A. DesfargesBerthelemot, 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. 8797.
Copyright © 2010 SciRes.
[9]
593
Y. C. Zhao, S. Winnall and S. Fleming, “Tunable FiberRing Laser Based on BroadBand Fiber Bragg Grating and Bending Effects,” Microwave and Optical Technology Letters, Vol. 46, No. 6, 2005, pp. 562563.
[10] S. K. Liaw, W. Y. Jang, C. J. Wang and K. L. Hung,
“Pump Efficiency Improvement of a CBand Tunable Fiber Laser Using Optical Circulator and Tunable Fiber Gratings,” Applied Optics, Vol. 46, No. 12, 2007, pp. 22802285. [11] S. K. Liaw and G. S. Jhong, “Tunable Fiber Laser Using
a BroadBand Fiber Mirror and a Tunable FBG As LaserCavity Ends,” IEEE Journal of Quantum Electronics, Vol. 44, No. 6, 2008, pp. 520527. [12] N. K. Chen, S. Chi and S. M. Tseng, “An Efficient Local
FundamentalMode Cutoff for ThermoOptic Tunable Er3+Doped Fiber Ring Laser,” Optics Express, Vol. 13, No.18, 2005, pp. 72507255. [13] N. K. Chen, C. M. Huang, S. Chi and Y. Lai, “Towards
The ShortWavelength Limit Lasing at 1450 Nm over I4(13/2)> I4(15/2) Transition in SilicaBased ErbiumDoped Fiber,” Optics Express, Vol. 15, No. 25, 2007, pp. 1644816456. [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. 14421463. [15] M. Arbore, Y. Zhou, H. Thiele, J. Bromage and L.
Nelson, “SBand ErbiumDoped Fiber Amplifiers for WDM Transmission between 1488 and 1508 Nm,” Proceedings of Optical Fiber Communication Conference, Georgia, 2328 March 2003, pp. 374376. [16] M. A. Arbore, “Application of FundamentalMode Cutoff
for Novel Amplifiers and Lasers,” Proceedings of Optical Fiber Communication Conference, (OFC 2005), Anaheim, Vol. 5, 611 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 1490−1610 Nm Based on Discrete FundamentalMode Cutoff Filters,” Optics Letters, Vol. 31, No. 19, 2006, pp. 28422844. [18] S. Y. Chou, K. C. Hsu, N. K. Chen, S. K. Liaw, Y. S.
Chih, Y. Lai and S. Chi, “Analysis of ThermoOptic Tunable DispersionEngineered ShortWavelengthPass TaperedFiber Filters,” Journal of Lightwave Technology, Vol. 27, No. 13, 2009, pp. 22082215.
ENG
Engineering, 2010, 2, 594601 doi:10.4236/eng.2010.28076 Published Online August 2010 (http://www.SciRP.org/journal/eng).
Behaviour of a Composite ConcreteTrapezoidal Steel Plate Slab in Fire Tomaž Hozjan, Miran Saje, Igor Planinc, Stanislav Srpčič, Sebastjan Bratina University of Ljubljana, Faculty of Civil and Geodetic Engineering, Ljubljana, Slovenia Email:
[email protected] Received February 11, 2010; revised March 28, 2010; accepted April 4, 2010
Abstract The present paper investigates fire resistance of a simplysupported composite concretetrapezoidal steel sheet slab. The objective is to find out if a steel sheet, as a moisture diffusion barrier, may substantially effect the hydrothermal situation in the concrete part of the crosssection. The numerical integration of the equations of a coupled hygrothermal boundaryvalue problem (Tenchev, R.T., Li, L.Y. and Purkiss, J.A. (2001) Num. Heat Transfer Part A, 39(7), 685710), with and without considering the barrier, shows that the barrier does not really effect the magnitude and the development of temperatures over the crosssection, 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 concretetrapezoidal 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 nonaccidental 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 concretetrapezoidal 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 nonlinear partial differential equations with permeability coefficients and thermodynamic characteristics of material being functions of the heat and moisture state. Luikov’s model was later on improved by Bažant 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 Bičanić [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 concretetrapezoidal steel plate slab in fire. This numerical model enables us to estimate the distribution of temperature, moisture and pore pressure over the concrete crosssection 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 highrise 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 crosssection 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 hygrothermal 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
D
(1)
t
Water vapour conservation: G V
t
(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
Copyright © 2010 SciRes.
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 Darcy’s and Fick’s 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 pressuredriven flow as given by Darcy’s law: vG vL
KKG
G
KK L
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 Bažant 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 CrossSection 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 finiteelement 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 crosssection 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 (20°C).
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 = 20ºC, 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 crosssection. Table 1. Boundary conditions of composite slab. case T A1
PG V
A2
Copyright © 2010 SciRes.
edge 1
edge 2
qT qT (TISO 834 )
qT qT (T 20 C)
PG
0
n V
0
n
symmetry o
T
PG
PG 0.1 MPa
0
n 0
n
qV qV ( V , )
T
qT qT (TISO 834 )
qT qT (T 20 C)
PG
PG 0.1 MPa
PG 0.1 MPa
V
qV qV ( V , )
qV qV ( V , )
o
V
0
n T
0
n PG
0
n
V
0
n
ENG
T. HOZJAN ET AL.
The development of the temperature field over the crosssection 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 500ºC 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 = 500ºC 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 500ºC. 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 crosssection, 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 crosssections 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 65°C per 1 cm, which results in the temperature decrease between points A and D to be roughly 200ºC 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 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 crosssection.
Figure 3. Variation of temperature with time in points A, B, C and D. Copyright © 2010 SciRes.
ENG
T. HOZJAN
598
Table 2. Coordinates of characteristic points A, B, C, D. point
yre [cm]
zre [cm]
A
2.5
2.5
B
2.5
3
C
2.5
4
D
2.5
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 crosssection at various instants. Beyond 200ºC 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 crosssection. Comparing the distributions of free water content over the crosssection (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 crosssection at t = 15, 30, and 60 min. (a) case A1; (b) case A2.
Copyright © 2010 SciRes.
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
0
(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 crosssection 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 400ºC [11]. At roughly 500ºC, the Copyright © 2010 SciRes.
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 500ºC. 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 crosssection. 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 = 500ºC, 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 crosssection 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 nonaccidental 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 nonlinear 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 CapillaryPorous Bodies,” International Journal of Heat and Mass Transfer, Vol. 18, No. 1, 1975, pp. 114.
[2]
Z. P. Bažant and W. Thonguthai, “Pore Pressure and Drying of Concrete at HighTemperature,” Journal of Engineering Mechanics DivisionASCE, Vol. 104, No. 5, 1978, pp. 10591079.
[3]
C. T. Davie, C. J. Pearce and N. Bičanić, “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. 733763.
[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. 4143, 2003, pp. 57075729.
[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. 685710.
[6]
Z. P. Bažant and M. F. Kaplan, “Concrete at High Temperatures: Material Properties and Mathematical Models,” Longman, Harlow, 1996.
[7]
T. Hozjan, “NonLinear 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 TestsElements 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 simplysupported composite concretetrapezoidal 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 hydrothermal situation in the concrete part of the crosssection, we compared the numerical results of the coupled hygrothermal boundaryvalue 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 crosssection, 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 Copyright © 2010 SciRes.
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. 77101.
Copyright © 2010 SciRes.
601
[11] G. WilliamsLeir, “Creep of Structural Steel in Fire: Analytical Expressions,” Fire and Materials, Vol. 7, No. 2, 1983, pp. 7378.
ENG
Engineering, 2010, 2, 602607 doi:10.4236/eng.2010.28077 Published Online August 2010 (http://www.SciRP.org/journal/eng).
The Effect of Initial Oxidation on LongTerm 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, Xi’an Jiaotong University, Xi’an, China Email:
[email protected] Received March 8, 2010; revised June 3, 2010; accepted June 5, 2010
Abstract The initial oxidation behavior of Ni6.5Co17.8Cr3.7Al0.5Y alloy is investigated at 800°C1000°C. Xray diffraction results show that the dominant Cr2O3 phase and secondary αAl2O3 and NiO phases are observed on the surface of samples at all initial stages (oxidized for 16 hours). YAlO3 and θAl2O3 can only be detected at low temperature (800°C) while the spinel NiCr2O4 is only observed at 900°C and 1000°C. Though the growth rates of αAl2O3 and Cr2O3 are comparable at 900°C, the former becomes much lower than the latter when the temperature changes to 1000°C. Scanning electron microscopy (SEM) images show that the αAl2O3 grows from some irregular ditches in the chromia scale at 900°C. However, cracking and spalling are more serious at 1000°C without αAl2O3grownditches, 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 1000°C is higher than that at 900°C. Owing to this condition, a preoxidation treatment on the NiCoCrAlY alloy for 16 hours is prepared at 900°C, and then thermal cycling oxidation test is conducted at 1000°C for 200 hours. The result indicates that the initial preoxidation treatment at 900°C improves the oxidation resistance of alloy at 1000°C. 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 [13]. Some authors [47] 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 αAl2O3 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 longterm stage is frequently determined by the initial stage of oxidation [810]. *The project was supported by the State Key Development Program for Basic Research of China (Grant No. 2007CB707700).
Copyright © 2010 SciRes.
Besides alumina, chromia (Cr2O3) also plays an important role during the high temperature oxidation [1113]. 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 nondestructive technique for determining the stresses in oxide scales for decades due to the bands in the Raman spectra of specimens shift with pressure [1113]. 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 800°C1000°C 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 μm. 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 800°C for 2 hours with a pressure of 300 MPa. Subsequently, the rod was annealed in vacuum at 1000°C 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 sparkmachining. 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 1300°C. 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 aircooled to room temperature. The initial oxidation tests were performed at 800°C, 900°C and 1000°C for 2, 4, 8 and 16 hours. A thermal cycling oxidation test was conducted at 1000°C for 200 hours. In order to keep consistent of oxidation condition in the test, the specimens were preoxidized at 900°C and 1000°C for 16 hours. Then the oxidation behavior of the specimens was evaluated by measuring the weight gains of the samples for 184 hours. The 12hour cycle consisted of 11 hours holding at 1000°C, 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 900°C and 1000°C on NiCoCrAlY alloy. Chromia has the same structure as corundum, and therefore it should have seven Raman active bands (A1g + 5Eg) [1113]. The most intense mode is the 549 cm1 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 BH2 microscope. During the measurements, the laser (He–Ne, 632.8 nm) was focused at a position on the surface of the sample and the laser spot size was set about 3–5 μm. The Raman spectroscopy acquired was analyzed by the commercial Renishaw WiRe software to obtain the peak shift fitted by Gaussian–Lorentzian 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) (JSM7000F). The chemical composition of the oxides was determined qualitatively by energydispersive Xray analysis (EDX). The phases in the oxide scales were analyzed using an Xray diffraction (XRD) (Rigaku D, CuKα radiation).
3. Results and Discussion 3.1. The Oxides on the Surface after Initial Oxidation Xray diffraction patterns of NiCoCrAlY alloy after oxidation for 16 hours at 800°C, 900°C and 1000°C are shown in Figure 1. Results from the study show that the oxides on the surface of alloy which was heated to 800°C for 16 hours are composed of Cr2O3, a few θAl2O3 and YAlO3, trace αAl2O3 and NiO. It was found that however, oxides θAl2O3 and YAlO3 did not form at 900°C and 1000°C. The spinel oxide, NiCr2O4, began to exist after 16 hours of oxidation at 900°C. According to the intensity of diffraction peaks, the relative oxidation rates of Cr2O3 and αAl2O3 phases on the surface of alloy oxidized from 2 hours to 16 hours at 800°C, 900°C and 1000°C 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 900°C and 1000°C keep fluctuating. The increase of quantities of αAl2O3 resulted from dense αAl2O3 oxide forming and θAl2O3 transforming. The fluctuation of quantities of Cr2O3 at 900°C and 1000°C 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 [9,1416]. The competition between the consumption of
Figure 1. Xray diffraction patterns of Ni6.5Co17.8Cr3.7 Al0.5Y alloy (oxidized for 16 h) at (a) 800°C; (b) 900°C and (c) 1000°C. ENG
C. ZHU ET AL.
604
growth rate of αAl2O3 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 800°C 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 θAl2O3 phase usually grows in a needlelike, whiskerlike or bladelike morphology and αAl2O3 grows in a weblike or dense equiaxed structure [14,17], these bladelike oxides should be θAl2O3.
(b)
(c) (c)
Figure 2. Relative quantities of Cr2O3 and αAl2O3 phases on the surface of alloy vs. time: (a) αAl2O3; (b) Cr2O3; (c) the relative oxidation rates of Cr2O3 and αAl2O3 at the second hour at 800°C, 900°C and 1000°C.
Cr2O3 and the formation of Cr2O3 determined the outline of curve. Figure 2(c) shows that the relative oxidation rates of αAl2O3 and Cr2O3 increased with temperature in the second hour from 800°C to 1000°C. Also, it can be observed that the growth rate of αAl2O3 is similar to that of Cr2O3 at 900°C. However, at 1000°C, the increase of Copyright © 2010 SciRes.
(d) Figure 3. SEM images of the specimen surface (oxidized for 16 h at 800°C) (a) surface image; (b) higher magnification; (c) and (d) EDX analysis in selected region.
ENG
C. ZHU
The surface images of specimens exposed at 900°C and 1000°C 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 αAl2O3 and Cr2O3. It can be speculated that, at 900°C, αAl2O3 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 1000°C are more than those at 900°C. There are more irregular ditches observed among the Cr2O3 oxide scale but not so much αAl2O3 fill in the cracks. Form Figure 2, we observe that the growth rate of αAl2O3 is similar to that of Cr2O3 at 900°C. Though the spallation on the Cr2O3 scale was unavoidable, αAl2O3 could preferably nucleate during oxidation on the surface in some cracks of the oxide scale at 900°C 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 αAl2O3 at 1000°C. 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 oxide–alloy interface, which speeded up the oxidation of alloy greatly. The crosssectional 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 900°C, and it became thicker at 1000°C. 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 900°C and 1000°C). (a) specimens oxidized at 900°C; (b) specimens oxidized at 1000°C; (c) higher magnification of Figure 4(a); (d) higher magnification of Figure 4(b); (e), (f) and (g) EDX analysis in selected regions.
Copyright © 2010 SciRes.
ET AL.
Figure 5. The crosssectional microstructure and the elemental maps of NiCoCrAlY alloy (oxidized for 16 h) at (a) 800°C; (b) 900°C and (c) 1000°C. 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 Alcontaining 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 Alcontaining 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 Alcontaining phase should be αAl2O3. The lateral growth of αAl2O3 precipitates occurred until they coalesce into a continuous αAl2O3 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 900°C and 1000°C for 16h. The only band which is welldefined in both spectra is the A1g mode at 549 cm1 in Mougin et al.’s work [11]. In the scales, this mode shifts to 549.53 cm1 for 900°C specimen and 554.34 cm1 for 1000°C specimen respectively, resulting in the observed shifts are equal to 0.53 cm1 and 5.34 cm1 respectively. Using the law given by Mougin et al. [11] for the frequency dependence with pressure, i.e., 0.307 ± 0.005 GPa/cm1, it gives the stress values of 0.163 ± 0.005 GPa for 900°C specimen and 1.639 ± 0.005 GPa for 1000°C 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 1000°C 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 selfhealing by the Al2O3 growth simultaneously.
Figure 6. Raman spectrum for the chromia formed on NiCoCrAlY alloys oxidized for 16 h at 900°C and 1000°C.
Copyright © 2010 SciRes.
3.4. The Effect of the Preoxidation Treatment on the Oxidation Resistance Due to the different growth characteristics of αAl2O3 and Cr2O3 at different temperatures at initial stage of oxidation, two groups of specimen were conducted. One was directly oxidized at 1000°C for 200 hours. The other one was subjected to a preoxidation treatment at 900°C for 16 hours first, aiming to repair microcracks in the Cr2O3 scale by subsequent growth of αAl2O3, then oxidized at 1000°C 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 1000°C. In the figure, curve (a) indicates the specimen directly oxidized at 1000°C for 200 hours and curve (b) indicates the specimen which preoxidixed at 900°C. At the onset of the oxidation, the rate of weight gain of the specimen preoxidized at 900°C was slower than that of 1000°C. 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 900°C showed lower weight gains than that of 1000°C. 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 αAl2O3 at 900°C was larger than that at 1000°C. At 1000°C, 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 Ni6.5Co17.8Cr3.7Al0.5Y alloy specimens was performed at 800°C, 900°C and 1000°C 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 1000°C for 200 h. (a) directly oxidized at 1000°C; (b) preoxidation treatment at 900°C 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 800°C. YAlO3 phase was only observed at 800°C. That the growth rate of αAl2O3 was similar to that of Cr2O3 at 900°C lead to the αAl2O3 could grow and fill in the ditches on the Cr2O3 scale. However, the growth rate of Cr2O3 at 1000°C was much larger and produced higher stress than that at 900°C, so that the αAl2O3 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 (900°C and 1000°C), a preoxidation treatment at 900°C for 16 hours can cause αAl2O3dispersionsinCr2O3 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 GasTurbine Engine Applications,” Science, Vol. 296, No. 5566, 2002, pp. 280284.
[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. 505553. 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. 239249. B. Wang, J. Gong, A. Y. Wang, C. Sun, R. F. Huang and L. S. Wen, “Oxidation Behaviour of Nicraly Coatings on NiBased Superalloy,” Surface & Coatings Technology, Vol. 149, No. 1, 2002, pp. 7075. U. Schulz, C. Leyens, K. Fritscher, M. Peters, B. SaruhanBrings, 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]
Copyright © 2010 SciRes.
ET AL.
607
Coatings,” Aerospace Science and Technology, Vol. 7, No. 1, 2003, pp. 7380.. [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. 12, 2009, pp. 392405. [8] T. F. An, H. R. Guan, X. F. Sun and Z. Q. Hu “Effect of the ThetaAlphaAl2O3 Transformation in Scales on the Oxidation Behavior of a NickelBase Superalloy with an Aluminide Diffusion Coating,” Oxidation of Metals, Vol. 54, No. 34, 2000, pp. 301316. [9] S. O. Moussa and K. Morsi, “HighTemperature Oxidation of Reactively Processed Nickel Aluminide Intermetallics,” Journal of Alloys and Compounds, Vol. 426, No. 12, 2006, pp. 136143. [10] T. J. Nijdam, N. M. van der Pers and W. G. Sloof, “Oxide Phase Development upon High Temperature Oxidation of GammaNicral Alloys,” Materials and CorrosionWerkstoffe Und Korrosion, Vol. 57, No. 3, 2006, pp. 269275. [11] M. Kemdehoundja, J. F. Dinhut, J. L. GrosseauPoussard 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. 435436, No. 5, 2006, pp. 666671. [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. 16491658. [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. 187196. [15] S. Seal, S. C. Kuiry and L. A. Bracho, “Surface Chemistry of Oxide Scale on IN738LC Superalloy: Effect of LongTerm Exposure in Air at 1173 K,” Oxidation of Metals, Vol. 57, No. 34, 2002, pp. 297322. [16] S. C. Choi, H. J. Cho and D. B. Lee, “Effect of Cr, Co, and Ti Additions on the HighTemperature Oxidation Behavior of Ni3Al,” Oxidation of Metals, Vol. 46, No. 12, 1996, pp. 109127. [17] H. H. Angermann, K. Nishi, Y. Aono, M. Inagaki and H. Kodama, “Evolution of Oxides on NiBase ODS Superalloys,” Oxidation of Metals, Vol. 48, No. 12, 1997, pp. 139.
ENG
Engineering, 2010, 2, 608616 doi:10.4236/eng.2010.28078 Published Online August 2010 (http://www.SciRP.org/journal/eng).
Highly Nonlinear BendingInsensitive 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 Email:
[email protected] 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 bendinginsensitive nonlinear PCF with a birefringence in the order of 102 and a nonlinear coefficient of 49 W1km1 at the wavelength of 1.55 μm. Also, significant improvements on key propagation characteristics of the proposed PCFs have been demonstrated by carefully altering the desired air hole diameters and the holetohole 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 [12]. PCFs can be divided into two categories according to the mechanism used to guide the light: photonicbandgap (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 lowindex 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 airglass PCF) [35]. On the other hand, the effective indexguiding 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 stepindex 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 10100 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 [911]. 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 [1213]. 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) [1112] 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 103102. 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 γ, 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 W1km1, respectively, at 1.55 μm telecommunication wavelength. However, these structures are purely theoretical and the hole to hole spacing, Λ, is around 0.9 μm. 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 103 at 1.55 μm telecommunication wavelengths. Lee et al. [7] has demonstrated a birefringent PCF having nonlinear coefficient γ, of 31 W1km1 for the use of optical codedivision multiple access (OCDMA) applications. Similarly, Yamamoto et al. [18] has demonstrated highly birefringent PCF with Gedoped core having nonlinear coefficient, γ, of 19 W1km1. 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 W1km1. 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 Copyright © 2010 SciRes.
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 [2223]. 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 Λ/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 × 103 and the losses are around 0.1 dB/km at 1.55 μm wavelength. Compared to ref. [24], at the same wavelength, lower confinement losses (≈ 0.001 dB/km) and higher birefringence (≈ 8 × 103) 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 bendinginsensitive highly birefringent nonlinear PCF. High birefringence in PCFs can be produced by combining the asymmetric core and the large corecladding 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 Λ/2. Additionally, different airhole 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/Λ = 0.941, d/Λ = 0.588, d5/Λ = 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 closelypacked 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 solgel 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 crosssection 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 xy 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.
wavenumber 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 λ 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 refractiveindex 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 Maxwell’s 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 [626] 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 (= 2π/λ) is the Copyright © 2010 SciRes.
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
E
2
dxdy 4
E dxdy
2
(4)
ENG
H. ADEMGIL
is inversely proportional to the effective mode area and can be calculated from [7,1618]; 2 n2 ( ) (5) Aeff ( )
where n2 is the nonlinear refractiveindex coefficient (n2 = 2.76 × 1020 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 Sellmeier’s formula is directly included in the calculation [2,6]. 2 λ Re neff (6) D c λ 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 holetohole spacing is denoted by d and Λ, 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 xdirection are increased. To enhance the birefringence further, the first column of the air hole group is shifted outwards by Λ/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 Copyright © 2010 SciRes.
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 Λ and d, ZDW can be controlled [2021]. 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/Λ = 0.941, d/Λ = 0.588 and d5/Λ = 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 holetohole 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 higherorder 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 holetohole spacing Λ, when dm/Λ = 0.941, d/Λ = 0.588, d5/Λ = 0.764 at the operating wavelength λ = 1.55 μm, 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, Λ, where λ = 1.55 μm.
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 γ, as a function of the operating wavelength.
fundamental and first order modes reduce with increasing holetohole spacing. As expected, losses of higher order modes are much higher than the fundamental mode. It is worth noting that, confinement loss for ypolarized mode of both fundamental and first order mode are higher than xpolarized 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 μm 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 ultrahigh 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, Λ decreases. Our design shows that the nonlinear coefficient, γ, for Λ = 1.7 μm and Λ = 2 μm at 1.55 μm operating wavelength is 26 W1km1 and 20 W1km1, 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, Λ = 1.9 μm at operating wavelength 1.55 μm. As can be seen from figure, three angular orientations φ = 0°, φ = 15° and φ = 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 φ increases. According to our simulations, the effect of angular orientation on confinement losses is related to core size. As expected, when φ 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
Copyright © 2010 SciRes.
Figure 5. Variation of confinement losses of the HE11x mode as a function of the bending radius, R, for three different angular orientation, when Λ = 1.9 μm and λ = 1.55 μm.
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 holetohole spacing. As can be observed from this figure, relatively large birefringence of the order of 103102 is achieved. It can clearly be seen that the birefringence is sensitive to the varying wavelength λ. It can be anticipated that as holetohole spacing Λ decreases, the birefringence increases. It can be noted that birefringence for Λ = 1.7 μm and Λ = 2 μm at 1.55 μm operating wavelength is 9 × 103 and 7.3 × 103, 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 nonbirefringent 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.
Copyright © 2010 SciRes.
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, Λ, decreases. Our simulations show that, the highest nonlinear coefficient corresponding to effective area, Aeff = 2.28 μm2 at λ = 1.55 μm, is equal to 49 W1 km1 for Λ = 1 μm. 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 γ, 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 102 is achieved. The birefringence is sensitive to the varying wavelength, λ and it increases as the wavelength increases. It can be noted that the birefringence for Λ = 1 μm and Λ = 1.6 μm at 1.55 μm operating wavelength is 2.65 × 102 and 1.01 × 102, 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 Λ, ZDW can be altered to any value. This unusual dispersion characteristic of PCFs allows them to be used in nonlinear fiber optics. One can see that it is possible to shift the zero dispersion wavelength from visible to nearinfrared (IR) regions by appropriately changing the geometrical parameters (d and Λ). 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 Λ, ZDW can be altered to any value. This unusual dispersion characteristic of PCFs allows them to be used in nonlinear fiber optics. One can see that it is possible to shift the zero dispersion wavelength from visible to nearinfrared (IR) regions by appropriately changing the geometrical parameters (d and Λ). As may be seen from Figure 9 when hole to hole spacing, Λ = 1 μm and Λ = 1.2 μm the proposed PCF has a single ZDW, 0.8 μm and 0.84 μm respectively. On the other hand, when Λ = 1 μm and Λ = 1.2 μm two ZDW is achieved. The first ZDW for both cases is around 0.8 μm. However, according to simulation results the second ZDW for Λ = 1 μm and Λ
Figure 9. Variation of the chromatic dispersion of HE11x modes as a function of the wavelength for different hole to hole spacing.
Copyright © 2010 SciRes.
= 1.2 μm is found as 1.36 μm and 1.67 μm, respectively. PCFs that have two ZDW have been used previously for high power supercontinuum generation applications [2021]. 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 nonPM 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 fibreoptic sensors, precision optical instruments, ultrashort solution pulse transmission and fourwave mixing [8,1819]. Moreover, reported results can be useful for optical communication systems [10], optical switching and OCDMA applications [7]. From experimental point of view, the solgel 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 today’s existing PCF technology. These reported results can be widely used for the supercontinuum generation, polarization control in fiberoptic sensors and telecommunication applications.
ENG
H. ADEMGIL
5. References
ET AL.
615
Photonic Crystal Fibers,” Optics Express, Vol. 12, No. 24, 2004, pp. 59315939.
[1]
J, C. Knight, T, A. Birks and P. S. J. Russell, “AllSilica SingleMode Optical Fibre with Photonic Crystal Cladding,” Optics Letter, Vol. 21, No. 19, 1996, pp. 15471549.
[14] A. OrtigosaBlanch, 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. 13251327.
[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. 989991.
[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. 305330.
[15] R. T. Bise and D. J. Trevor, “SolGel 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. E84C, 2001, pp. 585592. K. Saitoh and M. Koshiba, “Leakage Loss and Group Velocity Dispersion in AirCore Photonic BandGap Fibres,” Optics Express, Vol. 11, No. 23, 2003, pp. 31003109. H. Ademgil and S. Haxha, “Highly Birefringent Photonic Crystal Fibres with UltraLow Chromatic Dispersion and Low Confinement Losses,” IEEE Journal of Lightwave Technology, Vol. 26, No. 4, 2008, pp. 441448. J. H. Lee, P. C. Teh, Z. Yusoff, M. Ibsen, W. Belardi, T. M. Monro and D. J. Richardson, “A Holey FiberBased Nonlinear Thresholding Device for Optical CDMA Receiver Performance Enhancement,” IEEE Photonics Technology Letter, Vol. 14, No. 6, 2002, pp. 876878. A. V. Yulin, D. V. Skryabin and P. S. J. Russell, “FourWave Mixing of Linear Waves and Solutions in Fibers with HigherOrder Dispersion,” Optics Letter, Vol. 29, No. 20, 2004, pp. 24112413. 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. 2932.
[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. 676680. [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 EllipticHole Photonic Crystal Fibre with Squeezed Hexagonal Lattice,” Optics Letter, Vol. 32, No. 5, 2007, pp. 469471.
[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. 10651067. [17] K. Saitoh and M. Koshiba, “Highly Nonlinear DispersionFlattened Photonic Crystal Fibers for Supercontinuum Generation in a Telecommunication Window,” Optics Express, Vol. 12, No. 10, 2004, pp. 20272032. [18] T. Yamamoto, H. Kubota, S. Kawanishi, M. Tanaka and S. Yamaguchi, “Supercontinuum Generation at 1.55 M in a DispersionFlattened PolarizationMaintaining Photonic Crystal Fiber,” Optics Express, Vol. 11, No. 13, 2003, pp. 15371540. [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. 21972199. [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 ZeroDispersion Wavelength,” AIP Conference Proceedings, Sao Pedro, August 2008, pp. 1518. [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 TimeDomain Method,” Optics Letter, Vol. 33, No. 2, 2008, pp. 119121. [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 “SinglePolarization SingleMode Photonic Crystal Fibers,” IEEE Photonics Technology Letter, Vol. 15, No. 10, 2003, pp. 13841386.
[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. 52765281.
[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. 13361338.
[13] T. Ritari, H. Ludvigsen, M. Wegmuller, M. Legré, 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 UltraFlatted Chromatic Dispersion, Negative Chro
Copyright © 2010 SciRes.
ENG
616
H. ADEMGIL ET AL.
matic Dispersion and Improved Effective Mode Area,” Journal of Optics Communications, Vol. 281, No. 2, 2008, pp. 278286. [27] T. Kato, Y. Suetsugu and M. Nishimura “Estimation of Nonlinear Refractive Index in Various SilicaBased Glasses for Optical Fibers,” Optics Letter, Vol. 20, No. 22, 1995, pp. 22792281.
Copyright © 2010 SciRes.
[28] P.A. Champert, V. Couderc, P. Leproux, S. Février, V. Tombelaine, L. Labonté, P. Roy, C. Froehly and P. Nérin, “WhiteLight Supercontinuum Generation in Normally Dispersive Optical Fiber Using Original MultiWavelength Pumping System,” Optics Express, Vol. 12, No. 19, 2004, pp. 43664371.
ENG
Engineering, 2010, 2, 617624 doi:10.4236/eng.2010.28079 Published Online August 2010 (http://www.SciRP.org/journal/eng).
Progress in Antimonide Based IIIV 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 Email:
[email protected] 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 ultrahigh speed and ultralow 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 IIIgroup elements (Ga, In, Al, etc.) and Sb, As and other Vgroup elements, such as GaSb, InSb, AlGaSb, InAsSb, AlGaAsSb, InGaAsSb and so on. Their crystal lattices are around 6.1Å and they together with the InAsbased materials have been routinely called the “6.1Å IIIV 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 nearinfrared wavelength 0.78 m (AlSb) to farinfrared spectral regions 12 m (InAsSb). The heterojunctions formed between them can have typeI, typeII staggered and typeII 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 highperformance microelectronics, optoelectronic devices and integrated circuits. Applications could include activearray spacebased radar, satellite communications,
Copyright © 2010 SciRes.
ultrahighspeed and ultralow power integrated circuits, portable mobile devices, gas environmental monitoring, chemical detection, biomedical diagnosis, drug analysis and other fields [18].
2. The Physical Properties and Preparation Technology of ABCS Based Materials The indepth 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 ultrahigh speed and ultralow 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 IIIV compound semiconductors (at RT) is showed in Table 1. We
ENG
C. LIU
618
ET AL.
Table 1. Comparison of physical properties of IIIV 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)
8×104
5 000
200
3×104
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
9
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 IIIV 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 IIIV 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 GaSb、InSb 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, ABCS’s intrinsic carrier concentration are too high to get high resistivity (semiinsulating) substrate materials which is a serious impediment to the ABCS’s applications in the field of microelectronic devices. At present the ultrahigh pure InSb bulk crystal’s carrier concentration can be less than 1013/cm3 and the residual hole concentration of GaSb bulk crystals is about 2 × Copyright © 2010 SciRes.
1016/cm3. Because the growth process is very immature and there is immiscible gap in the multielements 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 thickfilm materials and so on. LPE method is a nearthermodynamic 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 μm/min. The weakness of LEP is that it can not be used for precision controlled growth of very thin films of nanoscale. That is to say that it is not applicable to the growth of superlattices or quantumwell devices and other complex microstructure 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 lowcost, high efficient GaSb based InGaAsSb homogeneous pn junction thermophotovoltaic (TPV) cells has been developed [8]. This method first grows lattice matched nIn0.15Ga0.85As0.17Sb0.83 (0.55 eV) epitaxial layer on the Tedoped ntype 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 cutoff 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 highfrequency and lowpower operation [1].
Figure 2. Relative position between energy gap and band offset of IIIV semiconductors.
als grown by MOCVD or MBE. In addition, LPE method is also used to grow materials of midinfrared InGaAsSb, InSbbased 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 nonthermodynamic equilibrium. You can grow almost all compositions of the multielements compound thin films including the ternary, quaternary antimonide which is in the immisciblegap and in the metastable state. Both of them can be used for growth of complex microstructural 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 productionbased MBE equipment has been developed, it is still not economical using the MBE for mass production when considering the cost. Copyright © 2010 SciRes.
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 IIIgroup metalorganic sources by MOCVD for antimonide based compounds are 3methyl compound and 3ethyl compound, such as: TMGa, TMIn, TMAl, TEGa, TEIn, etc. The commonly used Vgroup sources are TMSb, AsH3, PH3, TMBi and RFN2, etc. Antimonides are generally low melting point materials and the temperature of epitaxial substrate is generally about 500℃. In addition to TMIn’s lower decomposition temperature (250300℃), the majority of IIIgroup metalorganic 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 Alcontaining 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 ptype 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 ntype doping Alcontaining antimonide films. The presence of high concentration of O impurity in Alcontaining 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 [45]. Thus, growing AlSb and their multielement materials using MOCVD is the most challenging work in all the IIIV 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 ultrahigh vacuum epitaxial environment with singleelement materials for molecular beam sources and is easy to implement epitaxy of atomic layer and in situ realtime monitoring, avoiding the Cpollution problem which exits in Alcontaining materials growing by MOCVD and greatly reducing the concentration of O doping. In fact most of the prototype devices having complex fine structures and lowdimensional 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., lowdensity atomic step on the surface of the substrate) seem to be more accessible highquality epitaxial layers. Experiments confirmed that the use of GaSb (100) substrates miscut 2° towards (110) or 6° towards (1ī1) B may get higher crystal quality of InGaAsSb and AlGaAsSb epitaxial layers [5]. To overcome the difficulty that antimonides have no semiinsulating 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 highquality GaSb/AlGaSb multiquantum well (MQW) structures with a 5nm AlSb initiation layer and a relatively thick GaSb buffer layer (0.52.0 µm) grown on Si (001) substrates by molecular beam epitaxy. The photoluminescence (PL) emission around 1.55 µm wavelength was observed for GaSb/ AlGaSb MQW structure at room temperature. Low dislocation density, highquality GaSb epitaxial films on GaAs (001) substrates stripepatterned 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 antimonidebased materials in the farinfrared 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 Copyright © 2010 SciRes.
etc have also aroused people’s concern and research interest [1214]. 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 multiband infrared detectors.
3. Application of ABCS Materials The early focus of antimonide based compound semiconductors comes from its application prospect in midand farinfrared (photon) detectors, but the first to enter the market and get a largescale industrial production is highsensitivity 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. InSbbased infrared detector arrays have gained a market dominant position of groundinfrared applications and space instrumentation fields. In addition to these more mature products, antimonide materials have made great progress in the thirdgeneration infrared focal plane array detectors, mid and far infrared quantum cascade lasers, quantum dot lasers, ultrahighspeed, ultralowpower and lownoise 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 millimeterwave radar and highfrequency digital communications have so far experienced first generation based on GaAsbased 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 ultrahigh speed, ultralower power consumption and noise factor. After DARPA launched the ABCS projects in 2001, Rockwell Scientific Company (RSC) starting in 2003, has developed Kaband (3436 GHz), Wband (92102 GHz) and Xband (812 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 shortterm 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 fivestage Wband MMIC LNA chip is shown in Figure 3 [15]. The compact 1.2 mm2 fivestage Wband LNA using 0.2μm gate length InAs/AlSb metamorphic HEMTs demonstrated a 3.9 dB noisefigure 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 semiinsulating 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 cm2 and 19,000 cm2/Vs at 295K. Growing the Sbbased 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 microtwininduced 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 cm2V−1s−1 at roomtemperature and a sheet density of 2.5 × 1012 cm−2 were obtained by using this technique. It seems to provide a new way of integrating Sbbased 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 fivestage ABCS HEMT MMIC Wband LNA fabricated by Rockwell Scientific Company.
Copyright © 2010 SciRes.
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 onedimensional linear arrays which were made of lead salt such as PbSe, and PbTe to detect the midinfrared (MWIR) (35 m). The second generation infrared detector materials were mainly InSb and HgCdTe (MCT) for the two atmospheric IR windows of the midinfrared band and the farinfrared 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 multiband infrared detection, highresolution (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 thirdgeneration infrared detectors [67]. 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 typeII superlattice detectors grown on unintentional doped ptype GaSb (100) substrate designed for 2–5 μm and 8–12 μm bands infrared absorption. The LWIR detectors have detectivities as high as 8 × 1010 Jones (cm.H1/2/W) with a differential resistance–area product (RoA) greater than 6 Ohm cm2 at 80 K with a cutoff wavelength of approximately 12 μm. The measured internal quantum efficiency (QEi) of these frontside illuminated devices is close to 30% in the 10–11 μm range. The MWIR detectors have detectivities as high as 8 × 1013 Jones with a differential resistance–area product greater than 3 × 107 Ohm cm2 at 80 K with a cutoff wavelength of approximately 3.7 μm. The measured internal quantum efficiency of these frontside illuminated MWIR devices is close to 40% in the 2–3 μm 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 IItype superlattice midinfrared detector will have a great potential for application of midinfrared focal plane array devices of nonlowtemperature environment. In addition, InAs/InGaSb typeII superlattice materials have also been widely concerned and indepth research and they are considered as candidate materials for the thirdgeneration infrared detectors. Twocolor or dualband infrared detectors have the ability of inhibiting the complex background and improving the target detection efficiency and can significantly improve the system performances. Dualband LWIR/VLWIR typeII superlattice infrared detectors was reported by E. H. Aifer et al. [21]. The cutoff wavelengths of the two bands are 11.4 μm and 17 μm respectively. But the quantum efficiency of the dualband infrared detectors is too low (only 45%) compared to the singleband type II superlattice infrared detectors and the device structure needs to be further optimized. High quality GaSb based twocolor 288 × 384 MWIR InAs/ GaSb typeII SLS FPAs was reported by M. Münzberg et al. [22] of the Fraunhofer Institute in Freiburg. First, the “blue channel” consisting of 330 periods of ptype of a 7.5 ML InAs/10 ML GaSb was deposited on the GaSb substrate for spectral selective detection in the 3.04.1 μm 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 4.15.0 μm wavelength range. The thickness of the entire vertical pixel structure is only 4.5 μm, which significantly reduces the technological challenge compared to dual band HgCdTe FPAs with a typical total layer thickness around 15 μm. 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 Solidinfrared laser has important applications in gaseous environmental monitoring, chemical detection, biomedical diagnosis, satellite remote sensing technology and so on. Antimonide based compound semiconductor with bandgap corresponding to just 25 m midinfrared atomspheric window is an important material of midinfrared lasers. Research and development of new highperformance antimonidebased infrared laser are very active research subjects in recent years and researchers have made a series of important research results such as AlGaAsSb/InGaAsSb multiquantum well lasers [23], AlSb/InAs/InGaSb typeII quantum cascade lasers [24], “W”shaped midinfrared laser [25], InGaSb quantum dot lasers [26]. Antimonidebased interband cascade laser combining the advantages of quantum cascade (QC) laser and typeII quantum well interband laser has potential to achieve Copyright © 2010 SciRes.
continuous output of highpower infrared laser at room temperature and is an international hot subject of research and development. Midinfrared interband cascade laser made from InAs/Ga(In)Sb/AlSb mutiquantum wells was reported by C. J. Hill et al. of Jet Propulsion Laboratory [27]. This laser structure was grown on pGaSb（001） substrate by MBE as follows sequence: 0.3 μm GaSb buffer layer, 23 μm InAs/AlSb superlattice bottom claddings, multiquantum well InAs/Ga(In)Sb /AlSb active layers (be repeated 1235 times), InAs/ AlSb superlattice top claddings and finally an ntype InAs cap layer. The total thickness of epitaxial layers was more than 8 μm. A 15 μm × 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 μm × 1 mm laser made from sample J435 lased in cw mode at temperatures up to 165 K with a lasing wavelength of 5.43 μm 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 μm. However, at temperatures higher than 230 K, the spectral linewidth is relatively broad with operation voltages higher than 10 V. GaInSb quantum dot surfaceemitting laser (QDVCSEL) operating in optical communication wavelength band of 1.31.55 μm with continuous emission at room temperature by either optical pumping or current injection was reported by researchers of Japan’s National Institute of Information and Communication Technology [26]. This laser mainly consists of an antimonidebased quantum dot active layer and two AlAs/GaAs superlattice distributed Bragg reflectors (DBRs). With the development of antimonidebased 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 μm on a GaAs substrate through conventional technologies. In particular, the obtained wavelength of 1.55 μm represents the world’s longest emission wavelength of existent surfaceemitting laser structures based on GaAs substrate. It has great significance for mass production of lowcost surfaceemitting lasers used in nextgeneration ultrahighspeed optical communication technology. Highpower optically pumped semiconductor vertical external cavity surface emitting laser (VECSEL) operating at 2μm wavelength was reported by A. Härkönen et al. [28]. The device material was grown on GaSb substrate by MBE and consisted of 15 Ga0.78In0.22Sb quantumwells placed within a threelambda GaSb cavity and grown on the top of an 18pairs AlAsSb/GaSb Bragg reflector. When cooled down to 5℃ and using 790nm ENG
C. LIU
diode laser for optical pumping, this laser emitted up to 1 W of optical power in a nearly diffractionlimited Gaussian beam demonstrating the high potential of antimonide material for VECSEL fabrication. LED devices based on InGaAsSb/AlGaAsSb multiquantum 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 highpower, electrical injection and continuouswave emission midinfrared 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, narrowbandgap (0.6 eV or less) thermophotovoltaic materials and components applicable to the mid and lowtemperature 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 GaSbbased InGaAsSb pn cells fabricated by LPE, MOCVD, MBE and other methods. TPV cells based on InAsSbP, grown on InAs substrate, can have spectral responses in the 2.53.4 m wavelength range and it is a hopeful research direction of having great potentials. For further details, please refer to M.G. Mauk’s review paper [8].
4. Conclusions As the first candidate materials for fabrication of the third generation largescale focal plane arrays infrared (photon) detectors, integrated circuits with ultrahigh speed and ultralow 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 GaAsbased and InPbased materials growth and device manufacturing process, the growth technology of antimonide based microstructure materials such as heterojunctions, superlattice quantum wells and selfaligned 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 Copyright © 2010 SciRes.
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 highperformance antimonide devices and integrated circuits will get a wide range of important applications in the infrared imaging technologies, atmospheric environmental monitoring, biomedical diagnostics, multifunction digital radar systems, mobile communications, thermophotovoltaic power generation systems, and many other hightech 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., “AntimonideBased Compound Semiconductors for Electronic Devices: A Review,” SolidState Electronics, Vol. 49, No. 12, 2005, pp. 18751895.
[2]
M. Razeghi, “Overview of Antimonide Based IIIV Semiconductor Epitaxial Layers and Their Applications at the Center for Quantum Devices,” The European Physical Journal Applied Physics, Vol. 23, No. 3, 2003, pp. 149205.
[3]
A. H. Kroemer, “The 6.1 Å family (InAs, GaSb, AlSb) and its Heterostructures: A Selective Review,” Physica E, Vol. 20, No. 34, 2004, pp. 196203.
[4]
R. M. Biefeld, “The MetalOrganic Chemical Vapor Deposition and Properties of IIIV AntimonyBased Semiconductor Materials,” Materials Science and Engineering R, Vol. 36, No. 4, 2002, pp. 105142.
[5]
C. A. Wang, “Progress and Continuing Challenges in GasbBased IIIV Alloys and Heterostructures Grown by Organometallic VaporPhase Epitaxy,” Journal of Crystal Growth, Vol. 272, No. 14, 2004, pp. 664681.
[6]
A. Rogalshi, “Material Considerations for Third Generation Infrared Photon Detectors,” Infrared Physics & Technology, Vol. 50, No. 23, 2007, pp. 240252.
[7]
O. Nesher and P. C. Klipstein, “HighPerformance IR Detectors at SCD Present and Future,” OptoElectronics Review, Vol. 14, No. 1, 2006, pp. 6170.
[8]
M. G. Mauk and V. M. Andreev, “GasbRelated Materials for TPV Cells,” Semiconductor Science and Technology, Vol. 18, No. 5, 2003, pp. S191S201.
[9]
M. Rosker and J. Shah, “DARPA’s 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 MultiQuantum Well Structures on Si (0 0 1) Substrates,” Journal of Crystal Growth, Vol. 310, No. 1, 2008, pp. 7882. [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. 48434845. [12] T. Ashley, L. Buckle，G. 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 LatticeMatched to Gasb,” Journal of Crystal Growth, Vol. 304, No. 2, 2007, pp. 338341. [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. 783786. [15] J. B. Hacker, J. Bergman, G. Nagy, et al., “An UltraLow Power Inas/Alsb HEMT WBand LowNoise Amplifier,” IEEE MTTS International Microwave Symposium, June 2005, pp. 10291032. [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,” OptoElectron Review, Vol. 16, No. 4, 2008, pp. 458482. [18] M. Walther, R. Rehm, J. Schmitz, et al., “AntimonyBased Superlattices for HighPerformance Infrared Imagers,” Proceedings of SPIE, Orlandoof, 17 March 2008, pp. 69400A69400A8. [19] E. Corbin, M. J. Shaw, M. R. Kitchin, et al., “Systematic Study of Type II Ga1−Xinxsb/Inas Superlattices for InfraRed Detection in the 1012 ΜM Wavelength Range,”
Copyright © 2010 SciRes.
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
Semiconductor Science and Technology, Vol. 16, No. 4, 2001, pp. 263272. C. J. Hill, J. V. Li, J. M. Mumolo, et al., “MBE Grown TypeII MWIR and LWIR Superlattice Photodiodes,” Infrared Physics & Technology, Vol. 50, No. 23, 2007, pp. 187190. E. H. Aifer, J. G. Tischler, J. H. Warner, et al., “Dual Band LWIR/VLWIR TypeII Superlattice Photodiodes,” Proceedings of SPIE, Orlando, 28 March 2005, Vol. 5783, pp. 112122. M. Münzberg, R. Breiter, W. Cabanski, et al., “Inas/Gasb TypeII ShortPeriod Superlattices for Advanced Single And DualColor 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 µM Ingaassb/Algaassb MQW Lasers,” Chinese Physics Letters, Vol. 23, No. 8, 2006, pp. 22622265. A. Bauer, F. Langer, M. Dallner, et al., “Emission Wavelength Tuning of Interband Cascade Lasers in the 34 µM 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 TypeII W Interband Cascade Lasers,” Applied Physics Letters, Vol. 93, No. 4, 2008, p. 041118. N. Yamamoto, “NextGeneration Optical Communications through Nanotechnology,” NICT News, Vol. 353, 2005, pp. 34. C. J. Hill and R. Q. Yang, “MBE Growth Optimization of SbBased Interband Cascade Lasers,” Journal of Crystal Growth, Vol. 278, No. 14, 2005, pp. 167172. A. Härkönen, M. Guina, O. Okhotnikov, et al., “1W AntimonideBased Vertical External Cavity Surface Emitting Laser Operating At 2ΜM,” Optics Express, Vol. 14, No. 14, 2006, pp. 64796484. A. Ducanchez, L. Cerutti, P. Grech, et al., “Room Temperature Continuous Wave Operation of ElectricallyInjected SbBased RCLED Emitting Near 2.3 ΜM,” Superlattices and Microstructures, Vol. 44, No. 1, 2008, pp. 6269.
ENG
Engineering, 2010, 2, 625634 doi:10.4236/eng.2010.28080 Published Online August 2010 (http://www.SciRP.org/journal/eng).
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 Email:
[email protected] 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, TemperatureDependent 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 nonlinear 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 meltspinning 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 Navier–Stokes equations to a nonlinear 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 lowenergy 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 semiinfinite 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 Copyright © 2010 SciRes.
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 nonuniform 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
u
Figure 1. Physical model of boundary layer flow over a vertical stretching surface. ENG
P. LOGANATHAN
u
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
(2)
u
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 Stefan–Boltzman 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
U ( x) c x , V ( x) V0 x m
m1 2
(8)
Here c 0 is constant, Tw is the wall temperature, the powerlaw exponent m is also constant. In this study we take c 1 . Copyright © 2010 SciRes.
ET AL.
627
The temperaturedependent 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)
2 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 Liegroup 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 pointtransformation which transforms coordinates ( 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
b
( e
6
* 7
e *
2
*
) *
*
*
*
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
*
1
, y* y e
1
1 4
, * e
3 1 4
u u u *
1 2
, v v v *
Copyright © 2010 SciRes.
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
b
( * )
2 F 2 3FF 4 ( )
u * u e , v* v e , * ,, * Expanding by Taylor’s 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 e
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
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 nonlinear 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 29. 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 temperaturedependent 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
0.5
1
1.5
4. Results and Discussion
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 temperaturedependent 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 .
Copyright © 2010 SciRes.
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
1
0.8
0.6 ( ) 0.4 = 0.1 = 0.5 = 1.0
0.2
0
0
0.5
0.2
1
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. Copyright © 2010 SciRes.
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
1
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
1
0.8
0.6
( ) S = 3.0
0.4
S = 5.0
0.2 S = 8.0
0 0
0.5
1
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 .
Copyright © 2010 SciRes.
ENG
P. LOGANATHAN
632
ET
AL.
1.2
1
0.8
0.6 ( )
S = 3.0
0.4
S = 5.0
0.2
S = 8.0
0 0
0.5
1
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
1
0.8
Sc = 0.22
0.6 ( ) 0.4
Sc = 0.62
0.2 Sc = 0.78
0 0 0.2
0.5
1
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 .
Copyright © 2010 SciRes.
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 temperaturedependent 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 temperaturedependent 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 Copyright © 2010 SciRes.
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 temperaturedependent 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 ΔC 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. 645647.
[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. 6974.
[3]
C. Y. Wang, “Liquid Film on an Unsteady Stretching Surface,” Quarterly of Applied Mathematics, Vol. 48, No. 4, 1990, pp. 601610.
[4]
P. Saikrishnan and S. Roy, “NonUniform 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. 13511365.
[5]
S. Roy and P. Saikrishnan, “NonUniform 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. 33893396.
[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. 799807.
[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. 201206.
[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. 19581964.
[10] Y. Z. Boutros, M. B. AbdelMalek and N. A. Badran, “Group Theoretic Approach for Solving TimeIndependent Free Convective Boundary Layer Flow on a NonIsothermal Vertical Flat Plate,” Archives of Mechanics, Vol. 42, No. 3, 1990, pp. 377395.
Copyright © 2010 SciRes.
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. 21962199. [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. 457464. [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. 243248. [15] H. S. Takhar, R. S. R. Gorla and V. M. Soundalgekar, “Radiation Effects on MHD Free Convection Flow of a Gas Past a SemiInfinite Vertical Plate,” The International Journal of Numerical Methods for Heat & Fluid Flow, Vol. 6, No. 2, 1996, pp. 7783. [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. 181191. [17] J. Gary, D. R. Kassoy, H. Tadjeran and A. Zebib, “The Effects of Significant Viscosity Variation on Convective Heat Transport in WaterSaturated Porous Media,” Journal of Fluid Mechanics, Vol. 117, 1982, pp. 233249. [18] K. N. Mehta and S. Sood, “Transient Free Convection Flow with TemperatureDependent Viscosity in a Fluid Saturated Porous Medium,” International Journal of Engineering Science, Vol. 30, No. 8, 1992, pp. 10831087. [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. 2122, 2005, pp. 44604466.
[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. 21672178. [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 87WA/HT23,” American Society of Mechanical Engineers, New York, 1987. [24] S. Gill, “A Process for the StepbyStep Integration of Differential Equations in an Automatic Digital Computing Machine,” Proceedings of the Cambridge Philosophical Society, Vol. 47, No. 1, 1951, pp. 96108.
ENG
Engineering, 2010, 2, 635640 doi:10.4236/eng.2010.28081 Published Online August 2010 (http://www.SciRP.org/journal/eng).
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 Email:
[email protected] 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 creepfeed grinding (Figure 1), a largescale 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 creepfeed 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 nickelbased superalloys 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 supperalloy 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 Copyright © 2010 SciRes.
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. Creepfeed 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 supperalloy as the experimental sample. This supperalloy 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 supperalloy 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 025, 0.01 mm precision.
3.2. The Input Parameters
Figure 2. Pin gauge dimension. Copyright © 2010 SciRes.
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 supperalloy. 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
2
Co
8
9
W
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
B
0.007
0.009
S

0.003
Zr
0.03
0.06
Mo
1.5
2
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 (µm/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 fractionalfactorial 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 (µm/rev)
1
0.75
0.45
0.06
140
0.1
21
2
0.75
0.45
0.06
140
0.1
21
3
0.75
0.45
0.06
140
0.1
21
tion of the Ftest 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 PValues are used to determine which parameter is most significant. The Fratio 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 Copyright © 2010 SciRes.
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
PValue
Constant
−0.3
0.001
P1
0.683
0.174
P2
−0.067
0. 758
P3
1
0.460
f
−0.001
0.04
E
0.74
0.017
V
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
RSq = %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
RSq(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. 511.
[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
Copyright © 2010 SciRes.
[3]
[4]
facturing Technology, Vol. 23, No. 6, 2004, pp. 700706. R. S. Hahn, “On the Nature of the Grinding Process,” In: Proceedings of 3rd MTDR Conference, London, 1963, pp. 129154. S. Malkin and R. B. Andersson, “Thermal Analysis of the Grinding. Part I: Energy Partition,” Journal of Industrial Engineering, Vol. 96, 1974, pp. 11771183.
[5]
S. Malkin, “Thermal Analysis of the Grinding. Part II Surface Temperature and Workpiece Burn,” Journal of Engineering for Industry, Vol. 96, 1974, pp. 11841191.
[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. 299302.
[7]
S. Lavine, S. Malkin and T. C. Jen, “Thermal Aspects of Grinding with CBN Wheels,” Annals CIRP, Vol. 38, No. 1, 1989, pp. 557560.
[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. 983992.
[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. 12, 2006, pp. 3844.
[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. 3137. [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. 17751781. [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. 3137. [13] C. Montgomery, “Design of Experimental & Statistical Modeling,” McGraw Hill, Inc., New York, 2005.
ENG
Engineering, 2010, 2, 6416647 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 Yıldıırım, İbrahim m Özkol Faculty of A Aeronautics and d Astronauticss, Istanbul Tecchnical Universsity, Istanbul, Turkey Em mail: serdar.yild
[email protected], ozko
[email protected] 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 798, AISC C89 and IEC C614001 , 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 20225% 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 20330 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 presized tow wer is tackled tto p analysis aand design co onditions. Steel evaluate as per tower is assuumed to be loocated in BalııkesirBandırm 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 AISC899 for the steel tuubular tower deesign. The load combinationn method for the service load (characteristic ( load) conditioon is carE798. 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.π 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.π 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 σs 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 σ 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. σ 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: Copyright © 2010 SciRes.
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: γsd.γm = 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 σcr is calculated: t (8) σ 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 α0 (9) 0.70 r 212 if t r 0.1 0.01 . t The reduction coefficient αB for bending load is calculated as follows: αB 0.1887 0.8113αo (10) The buckling stress σu can be computed in terms of the yielding stress Fy: 0.6 Fy if α B .σ cr Fy Fy 1 0.4123 2 (11) au α B .σ cr Fy if αB .σ cr 0.175.α B .σ cr 2 The maximum applied stress σa combined with normal stress and shear stress is calculated by σ a :
σa
fa fb
2
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 798 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 789). Ss Mapped maximum considered earthquake spectral response acceleration at short periods as determined in accordance with Section 9.4.1 (ASCE 789). Fa and Fv are defined in Tables 9.4.1.2.4a and b respectively in accordance with Section 9.4.1 (ASCE 789). According to the ASCE 789 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 9.4.1.2.51 and 9.4.1.2.52 respectively: S DS
2 2 S MS ; S D1 S M 1 3 3
Figure 1. Earthquake overturning moment.
Copyright © 2010 SciRes.
(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 occu pancy importance factor (I) is equal to 1.0. Site Classi fication D is assumed for BalıkesirBandırma/Marmara Region. Site Classification D is typified by stiff soils with shear velocity (Vs in soil) typically 600–1,200 fps (183–366 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 798 Tables 9.4.1.2.4a4b 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 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)
z
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 α1 is 11.5 for exposure D category in accordance with ASCE 798.
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 66 in ASCE 798. The terrain exposure coefficient is determined as per Table 65 of ASCE 798 or by the following formula [3]: 2 15 ft α1 ) ifz 15 ft 2.01( zg Kz z (23) 2 z α1 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 798. 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 798. 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 β 7.47 N1 Rn (1 10.3N1 )5/3
R
Figure 3. Earthquake shear force.
(28) (29)
The force coefficient C f is determined as per Tables 610 of ASCE 798. 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)
h
Vz z Fz x dx
(Figure 6)
(31)
z
h
Figure 4. Wind velocity pressure on tower. Copyright © 2010 SciRes.
Mz z Fz z .( x z ) dx
(Figure 7)
(32)
z
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 / 2π .
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 bladepassing 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 softsoft [4]. EIg (33) wt 1.75 H 3 Wt 0.25Wtow where wt is the estimated natural frequency of the tower is Copyright © 2010 SciRes.
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 798/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.
Copyright © 2010 SciRes.
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 IEC614001 and ASCE 798. 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 200231 July 2004, NREL Subcontractor Report, January 2005 NREL/SR50036777.
[3]
ASCE 798, “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 614001, “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
Copyright © 2010 SciRes.
12. References
ENG
Engineering, 2010, 2, 648657 doi:10.4236/eng.2010.28083 Published Online August 2010 (http://www.SciRP.org/journal/eng).
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 Email:
[email protected] 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 figureeightshaped 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 figureeightshaped path and the latter to an additional spinning. Based on the parametric equations of motion of the lumped masses, and considering semistatic tensile deformation of the connecting rods carrying them, it was found that the resultant impulse towards the direction of the spin vector includes a nonvanishing 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 FigureEight, 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 [14]. 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 contrarotating 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, Dean’s 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
Copyright © 2010 SciRes.
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 Dean’s drive) [10] and a relevant review paper [11]. Moreover, quite recently Provatidis [12] has shown that Dean’s 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 Dean’s drive is due to the circular paths on which the eccentric masses move. Motivated by the abovementioned findings on Dean’s 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 rigidbody 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 figureeightshape, 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 ‘figureeightshaped’ 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 thatin 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 figureeightshaped 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 figureeightshaped 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 57). In this way, the proposed mechanism that consists of a rotating ‘figureeight’ 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 closedform expressions of the figureeightshaped 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 closedform 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 figureeightshaped path is clearly illustrated for two different views.
Copyright © 2010 SciRes.
An overview of the proposed device, which has been developed so as to produce the abovementioned figureeightshaped 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 figureeightshaped 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 crosssection 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 P1S1 towards the mass ‘a’. Similarly, the rest half of the power produced by the motor M1 is transmitted through P1S2 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 zaxis) and also enforces the casing (11) to rotate around xaxis.
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).
Copyright © 2010 SciRes.
ENG
C. G. PROVATIDIS
cos t R 01 sin t 0
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 xaxis (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 zaxis at an angular velocity z , using a second motor M2 shown in Figure 1 as well as in Figure 2.
Copyright © 2010 SciRes.
cos t
0
0
1
0 1 R12 0 cos t 0 sin t
cos z t
R 23 sin z t
0
0
(2)
sin t cos t 0
sin z t cos z t 0
(3)
0
0
(4)
1
Therefore, the coordinates of the mass ‘a’ are analytically given by:
x t y t R z t a
a
a
a
x y z
a
a
a
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 b
b
b
x y z
b
b
b
b
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 z
(6)
z
z
z
2
Details about the figureeightshaped 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 zaxis, the second one is a rotation of the casing around x–axis 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 FigureEightShaped 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
2
R 4
2
z
2
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:
F
4 sin t sin 2t 4 cos t cos 2t 2
2
z
2rsin2 t Rsint cos t +2 rcos2t +Rcost sin t
z
z
(21)
z
2
z
z
+z
Fxa Fxb mr
x
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
2
F
(11)
Fya Fyb mr
y
4
z
rcost +R sin tcos t
2
z cos z t sin 2t 4 z sin z t cos 2 t 2
(22)
z
za t
2
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
z
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)
z
Copyright © 2010 SciRes.
(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
z
xk
mass, ‘a’ or ‘b’, it is necessary to consider the direction cosines along the rod axes, which are given by:
2
(23)
Equation (23) depicts that: 1) the zcomponent 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 phasedifference 2 , it is trivial to verify that the corresponding forces Fx and Fy , which are given by (21) and (22) respectively,
k
2 2 yb t z r cos t sin z t
2
6.1. Axial Forces
z 2 r cos t cos z t
F
(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 (QuasiStatic Analysis)
Fz Fza Fzb 2m
Considering an instantaneous deformation of the elastic rod ‘a’ according to Hooke’s law (semistatic 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
2
2
2
2
r 2 z
Frb mr z cos t 2
2
2
2
1 sin t 2 2
cos t (31) 3
z
cos t 3
z
(32)
Also, the semistatic deformations become: ra
m
2
r
EA
2
z cos t 2
2
3
m EA
z r 2
1 sin t
4
(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
m
r
EA
2
2
2
(35)
3
rb r
EA
z r 2
2
2
cos
r 4
2
cos
2
t 2
2
3
Fza mza t 2m ra cos 2 t
(37)
Fzb mzb t 2m rb cos 2 t
(38)
2
Copyright © 2010 SciRes.
4m 2
(40)
(41) (42)
2
EA
cos 2 t
r 2 cos t r cos t 2 cos t 2
2
2
2 z
2
4
(43)
2
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)
0
Integrating (40) over time, after manipulations the impulse (i.e., (44)) can be analytically expressed in closed form, using one term for the rigidbody 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
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
t
and therefore the updated variable radius is given by: m
2
Fz, rigid 4m 2 r cos 2 t
2
z
2 z cos
r
I z , tension c0 c2 sin 2 c4 sin 4 c6 sin 6
cos t 1 sin t 2
2
with
2
2 r 2 cos 2 t 2
cos
2
cos
4
t
and therefore the updated variable radius is given by: ra r
2
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
m
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:
and therefore the updated variable radius is given by: rb r rb
2
Fz Fz, rigid Fz,tension
and therefore the updated variable radius is given by: ra r ra
653
EA
2 z
m 2 9
2
r
2
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 rigidbody 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
0
z
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 timeinterval 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. Copyright © 2010 SciRes.
Figure 4. A digital record of the ground reaction (in dN or kgforce), 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 kgforce (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 semianalytical 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 powertoangular 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 apriori known. Furthermore, concerning the experimental part of this work, it is worthmentioning 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 welldesigned experimental device, a fact that presupposes high technical skills and hightech 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 figureeightshaped 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 nonvanishing net impulse. Intuitively, this claim is true because when the ratio of the spinning angular velocity over the first one (formation of the figureeightshaped path) is not an integer number, every mass does not repeat its initial position; in other words, when the mass completes the figureeightshaped 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 Copyright © 2010 SciRes.
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/NanoSystems, Springer, London, 2007, pp. 333334.
[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. 822827.
[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. 89, October 2003, pp. 887905.
[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. 733752.
[7]
I. I. Blekhman, A. L. Fradkov, H. Nijmeijer and A. Yu. Pogromsky, “On SelfSynchronization and Controlled Synchronization,” Systems and Control Letters, Vol. 31, No. 5, October 1997, pp. 299305.
[8]
G. Y. Stepanov, “Why is it Impossible to Have ‘Dean’s Apparatus’?” Journal Priroda (in Russian), Vol. 7, 1963, pp. 8591.
[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/TM2006214390, AIAA20064913, December 2006. http://gltrs.grc.nasa.gov/reports/2006/TM2006214390.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. 441461. [12] C. G. Provatidis, “Some Issues on Inertia Propulsion Mechanisms Using Two ContraRotating Masses,” Theory of Mechanisms and Machines, Vol. 8, No. 1, April ENG
C. G. PROVATIDIS
656 2010, pp. 3441.
[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 ScientificPractical 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
Copyright © 2010 SciRes.
ureEightShaped Closed Curves in the ThreeDimensional 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 WeightReduction of a Spinning Wheel,” Meccanica, Vol. 42, No. 4, August 2007, pp. 359364. [17] M. Tajmar, “Homopolar Artificial Gravity Generator Based on FrameDragging,” Acta Astronautica, Vol. 66, No. 910, MayJune 2010, pp. 12971301. [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 figureeightshaped curve For the purpose of completeness, details are provided here for the figureeightshaped 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 xaxis, 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
2
2
2
2
2
2
2
xa ya z a xb yb zb rsphere
, with rsphere r 2 R 2
Second, in the case of a nonvanishing 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
b
1a
1a
1a
Copyright © 2010 SciRes.
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, 658664 doi:10.4236/eng.2010.28084 Published Online August 2010 (http://www.SciRP.org/journal/eng).
ComputerAided 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 Email: {ezenwaogbonnaya, canjohnson2000}@yahoo.com,
[email protected] Received March 25, 2010; revised June 24, 2010; accepted June 28, 2010
2
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 flowinduced 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 1200°C 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
Copyright © 2010 SciRes.
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 builtup 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 7085% 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 Copyright © 2010 SciRes.
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
Cdamper
. .. 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
m
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
2
2 t
The general solution of Equation (7) can be obtained thus:
D F0cost
2
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: .. ΣF = 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)
2 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: Copyright © 2010 SciRes.
(5)
(6)
X
2 2 2 2 F0 4 n M 4 2 2 n2 2 2
which, Frequency, F =
2
2
cost
Hz and vibration displacement
ENG
E. A. OGBONNAYA ET AL.
amplitude.
661 Start
X
F0
M 4 2 2 n2 2
(8)
2
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
2 n
2
2
(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) µ = product of coefficient of damping (Nsm1). 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).
3. Results and Discussion 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
Copyright © 2010 SciRes.
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
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
m
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, proofreading 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, 812 June 2009. www.teurbexpo.org
[2]
R. Kurz, “Surge Control Design System Design,” Proceedings of ASME Turbo Expo, Orlando, Florida, 812 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. 34. T. Okada, A. Kawajiri, O. Yataka and O. Eisuke, “Stall Inception Process and Prospects for Active HubFlap Control in Three Stage Axial Flow Compressor,” Journal of Thermal Science, CoPublished with SpringerVarlag GmbH, 2008, pp. 48.
3040 3030 3020 3010
[3]
3000 6
6.2
6.4
6.6
6.8
7
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 Copyright © 2010 SciRes.
[4]
ENG
E. A. OGBONNAYA ET AL. [5]
[6]
[7]
[8]
[9]
E. A. Ogbonaya, “Modeling VibrationBased Faults in Rotor Shaft of Gas Turbine,” Ph.D. Dessertation, Department of Marine Engineering, Nigeria, 2004, pp. 2022, 9295, 180181. 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, 812 June 2009. www.Teurbexpo.org S. S. Rao, “Mechanical Vibration,” 4th Edition, Pvt Ltd., Dorling Kingsley, Licenses of Pearson Education in South Asia. C. B. MeherHomji, M. Chaker and A. E. Brouley, “The Fouling of Axial Flow CompressorCauses, Effects Susceptibility and Sensitivity,” Proceedings of ASME Turbo Expo, Orlando, Florida, 812 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 GT200850043, 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, 27 November 2003.
Copyright © 2010 SciRes.
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 90GT103, 1990. [12] C. B. MeherHomji, “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. 99111. [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. 142146. [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, PrenticeHall 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, McGrawHill 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)
Copyright © 2010 SciRes.
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) α = Inlet flow angle o (degree) Eo = Cascade collection efficiency (%) rp = Pressure ratio m = Mass flow rate (kg/s)
ENG
Engineering, 2010, 2, 665667 doi:10.4236/eng.2010.28085 Published Online August 2010 (http://www.SciRP.org/journal/eng).
Flipped Voltage Follower Design Technique for Maximised Linear Operation ChingMei Chen, Khaled Hayatleh, Bryan Hart, John Lidgey School of Technology, Oxford Brookes University, Oxford, UK Email:
[email protected] 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 preexisting, and new, low power/low voltage analogue circuits. A prototype FVF is a two transistor sourcefollower 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 sourcefollower with decreased output impedance and increased linearity in its voltage transfer characteristic. The socalled ‘Super SourceFollower’ [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 interconnected to form an Nchannel 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 loopgain 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],
Copyright © 2010 SciRes.
ID
βN 2 VGS VTN 2
(1)
2I D βN
(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(VGVS) = Gatesource voltage; VDS(VDVS) = Drainsource voltage; VTN = Threshold voltage; βN = μNCOX(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) βN and, VG VSS
2I X 2VTN βN
(4)
If ΔVG denotes the linear range then, from (3), (4), 2I X VG VTN (5) βN The problem, now, is that ΔVG 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: smallsignal 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 Widlartype 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 (IYIZ). 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 βP
2 IY I Z βN
(6)
in which subscript P refers to Pchannel 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 βP ‘high’ but IZ, ΙVTPΙ and βN ‘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).
Copyright © 2010 SciRes.
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) βP βN 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 loopgain 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 27ºC. All the mosfets, except M3, had L = 0.13 u, W = 10 u. It was assumed that for low voltage⁄low 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. RamírezAngulo, A. J. LópezMartin, A. Torralba, J. A. Gómez Galán, A. Carlosena and F. M. Chavero, “The Flipped Voltage Follower: A Useful Cell for LowVoltage LowPower Circuit Design,” IEEE Transactions on Circuits and SystemsI: Regular Papers, Vol. 52, No. 7, July 2005, pp. 12761291.
[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. RamírezAngulo, R. G. Carvajal and A. J. LópezMartin, “Highly Linear V/I Converter with Programmable Current Mirrors,” Circuits and Systems, ISCAS 2007, IEEE International Symposium, 2730 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).
Copyright © 2010 SciRes.
ENG
Engineering, 2010, 2, 668672 doi:10.4236/eng.2010.28086 Published Online August 2010 (http://www.SciRP.org/journal/eng).
Wire Bonding Using Offline Programming Method Yeong Lee Foo, Ah Heng You, Chee Wen Chin Faculty of Engineering and Technology, Multimedia University, Jln Ayer Keroh Lama, Melaka, Malaysia Email:
[email protected] 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 (DiOLP). Both versions utilize the bonding diagram and computer aided design data to speed up bonding program creation. However, the newly proposed DiOLP is more flexible as it can be used to create bonding program for multiple machine platforms in microelectronics industry. Some special features of DiOLP 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. DiOLP 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 preprogram 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. Copyright © 2010 SciRes.
They are material handling, bonding parameter and bonding path instruction. The material handling information such as magazine dimension, leadframeindexing 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 reuse 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 [68].
2. Vendor Specific Offline Programming Currently most machine vendors have their own version Copyright © 2010 SciRes.
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 addon 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 dieattached 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 (DiOLP) 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 DiOLP 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 DiOLP 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 DiOLP. 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 coordinates 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 DiOLP, 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 DiOLP. Just like vendor specific OLP, DiOLP 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 DiOLP 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. DiOLP 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. DiOLP 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 twodimensional 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 Copyright © 2010 SciRes.
671
o’ Poo’ Pno’
o
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 θ counterclockwise about the origin, the functional forms are x' = xcosθ − ysinθ and y' = xsinθ + ycosθ 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 θ in zaxis 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. DiOLP is a more suitable method to replace the time consuming manual method.
y P’ (x’, y’)
6. References P (x, y)
O
[1]
x
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 DiOLP 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 DiOLP as described. Both versions utilize the bonding diagram CAD data to speed up bonding program creation. However, the proposed DiOLP 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
Copyright © 2010 SciRes.
[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. 3036. L Nguyen, I. Singh, C. Murray, J. Jackson, J. DeRosa and D. Ho, “70 μm Fine Pitch Wire Bonding,” IEEE International Electronics Manufacturing Technology Symposium, Adelaide, 1998, pp. 394400. 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. 43014305. 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. 15. G. G. Harman, “Wire Bonding in Microelectronics Materials, Processes, Reliability and Yield,” 2nd Edition, McGrawHill, New York, 1997.
ENG
Call for Papers
ENGINEERING A Journal Published by Scientific Research Publishing, USA www.scirp.org/journal/eng 9 771947 393005
EditorinChief Prof. David L. Carroll
08
Wake Forest University, USA
Editorial Board Dr. Rehan Ahmed Prof. Moh’D A. M. AlNimr 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. ChuiChi Lee Prof. YongGang 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 tWa 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 ) C h o n g q i n g U n i v e r s i t y, 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 C h o n g q i n g U n i v e r s i t y, C h i n 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 peerreview 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 Reports—25 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 Reviews—Comments 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 EMail Hh ttp://www. scirp.org/journal/eng
[email protected]
Engineering, 2010, 2, 559672
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 Zircaloy4: Effects in the Dimensional Stability of Structural Components under Nuclear Reactor Operating Conditions P. Vizcaíno, C. P. Fagundez, A. D. Banchik.................................................................................... 573
Electrochemical Generation of ZnChitosan Composite Coating on Mild Steel and its Corrosion Studies K. Vathsala, T. V. Venkatesha, B. M. Praveen, K. O. Nayana............................................................ 580
Tunable ErbiumDoped Fiber Lasers Using Various Inline Fiber Filters S.K. Liaw, K. C. Hsu, N. K. Chen............................................................................................... 585
Behaviour of a Composite ConcreteTrapezoidal Steel Plate Slab in Fire T. Hozjan, M. Saje, I. Planinc, S. Srpcic, S. Bratina....................................................................... 594
The Effect of Initial Oxidation on LongTerm Oxidation of NiCoCrAlY Alloy C. Zhu, X. Y. Wu, Y. Wu, G. Y. Liang............................................................................................. 602
Highly Nonlinear BendingInsensitive Birefringent Photonic Crystal Fibres H. Ademgil, S. Haxha, F. AbdelMalek........................................................................................... 608
Progress in Antimonide Based IIIV 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. Özkol..................................................................................................................... 641
A Device that can Produce Net Impulse Using Rotating Masses C. G. Provatidis........................................................................................................................... 648
ComputerAided 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
Wire Bonding Using Offline Programming Method Y. L. Foo, A. H. You, C. W. Chin.................................................................................................... 668
Copyright
2010 SciRes.
Engineering, 2010, 2, 559672