ISSN 1905-7873 Â© 2012 - Maejo International Journal of Science ...

Journal Information 

Maejo International Journal of Science and Technology (ISSN 1905-7873 © 2012), 

the international journal for preliminary communications in Science and Technology is the 

first peerrefereed scientific journal of Maejo University (www.mju.ac.th). Intended as a 

medium for communication, discussion, and rapid dissemination of important issues in 

Science and Technology, articles are initially published online in an open access format, 

which thereby gives authors the chance to communicate with a wide range of readers in an 

international community. 

Publication Information 

MIJST is published triannually. Articles are available online and can be accessed free 

of charge at http://www.mijst.mju.ac.th. Printed and bound copies of each volume are 

produced and distributed to authors and selected groups or individuals. This journal and the 

individual contributions contained in it are protected under the copyright by Maejo 

University. 

Abstracting/Indexing Information 

MIJST is covered and cited by Science Citation Index Expanded, SCOPUS, Journal 

Citation Reports/Science Edition, Zoological Record, Chemical Abstracts Service (CAS), 

SciFinder Scholar, Directory of Open Access Journals (DOAJ), CAB Abstracts, ProQuest, 

and Google Scholar. 

Contact Information 

Editorial office: Maejo International Journal of Science and Technology (MIJST), 1st 

floor, Orchid Building, Maejo University, San Sai, Chiang Mai 50290, Thailand 

Tel: +66-53-87-3880 

E-mails: duang@mju.ac.th

MAEJO INTERNATIONAL JOURNAL 

OF SCIENCE AND TECHNOLOGY 

Editor 

Duang Buddhasukh, Maejo University, Thailand. 

Associate Editors 

Jatuphong Varith, Maejo University, Thailand. 

Wasin Charerntantanakul, Maejo University, Thailand. 

Morakot Sukchotiratana, Chiang Mai University, Thailand. 

Nakorn Tippayawong, Chiang Mai University, Thailand. 

Editorial Assistants 

James F. Maxwell, Chiang Mai University, Thailand. 

Jirawan Banditpuritat, Maejo University, Thailand. 

Prof. Dallas E. Alston 

Dr. Pei-Yi Chu 

Asst. Prof. Ekachai Chukeatirote 

Prof. Richard L. Deming 

Prof. Cynthia C. Divina 

Prof. Mary Garson 

Prof. Kate Grudpan 

Assoc. Prof. Duangrat Inthorn 

Prof. Minoru Isobe 

Prof. Kunimitsu Kaya 

Assoc. Prof. Margaret E. Kerr 

Dr. Ignacy Kitowski 

Asst. Prof. Andrzej Komosa 

Asst. Prof. Pradeep Kumar 

Asst. Prof. Ma. Elizabeth C. Leoveras 

Dr. Subhash C. Mandal 

Prof. Amarendra N. Misra 

Dr. Robert Molloy 

Prof. Mohammad A. Mottaleb 

Prof. Stephen G. Pyne 

Prof. Renato G. Reyes 

Dr. Waya Sengpracha 

Dr. Settha Siripin 

Prof. Paisarn Sithigorngul 

Prof. Anupam Srivastav 

Prof. Maitree Suttajit 

Assoc. Prof. Chatchai Tayapiwatana 

Emeritus Prof. Bela Ternai 

Asst. Prof. Narin Tongwittaya 

Asst. Prof. Jatuphong Varith 

Assoc. Prof. Niwoot Whangchai 

Editorial Board 

University of Puerto Rico, USA. 

Changhua Christian Hospital, Taiwan, R.O.C. 

Mae Fah Luang University, Thailand. 

California State University Fullerton, Fullerton CA 

Central Luzon State University, Philippines. 

The University of Queensland, Australia. 

Chiang Mai University, Thailand. 

Mahidol University, Thailand. 

Nagoya University, Japan. 

Tohoku University, Japan. 

Worcester State College, Worcester,MA 

University of Maria-Curie Sklodowska, Poland. 

University of Maria-Curie Sklodowska, Poland. 

Jaypee University of Information Technology, India. 


Jadavpur University, India. 

Fakir Mohan University, Orissa, India. 


Northwest Missouri State University, USA. 

University of Wollongong, Australia. 


Silpakorn University, Thailand. 

Maejo University, Thailand. 

Srinakharinwirot University, Thailand. 

College of Engineering and Technology, India. 

Naresuan University (Payao Campus), Thailand. 


La Trobe University, Australia. 



Maejo University,Thailand. 

Consultants 

Asst. Prof. Chamnian Yosraj, Ph.D., President of Maejo University 

Assoc. Prof. Thep Phongparnich, Ed. D., Former President of Maejo University 

Assoc. Prof. Chalermchai Panyadee, Ph.D., Vice-President in Research of Maejo University



The International Journal for the Publication of Preliminary 

Communications in Science and Technology 

http://www.mijst.mju.ac.th 

ISSN 1905-7873 © 2012 by Maejo University



Volume 6, Issue 1 (January - April 2012) 

CONTENTS 

Page 

Editor's Note 

Duang Buddhasukh, Editor-in-Chief .................…………………….……………………... i 

Enhancement of carotenoid and chlorophyll content of an edible freshwater alga (Kai: 

Cladophora sp.) by supplementary inorganic phosphate and investigation of its biomass 

production 

Taweesak Khuantrairong and Siripen Traichaiyaporn*………...……….…………….… 1-11 

Performance of self-excited induction generator with costeffective static compensator 

Yogesh K. Chauhan *, Sanjay K. Jain and Bhim Singh…..…………………………….. 12-27 

Benthic diatoms of Mekong River and its tributaries in northern and north-eastern Thailand 

and their application to water quality monitoring 

Sutthawan Suphan*, Yuwadee Peerapornpisal and Graham J. C. Underwood…………28-46 

Fourteen new records of cercosporoids from Thailand 

Pheng Phengsintham, Ekachai Chukeatirote, Eric H. C. McKenzie, Mohamed A. Moslem, 

Kevin D. Hyde * and Uwe Braun..………………………………………………..………47-61 

On the security of an anonymous roaming protocol in UMTS mobile networks 

Shuhua Wu *, Qiong Pu and Ji Fu………………..……………………………….……. 62-69 

Influence of MeV H+ ion beam flux on cross-linking and blister formation in PMMA resist 

Somrit Unai *, Nitipon Puttaraksa, Nirut Pussadee, Kanda Singkarat, Michael W. Rhodes, 

Harry J. Whitlow and Somsorn Singkarat……………………….……………...…….… 70-76 

Degradation of bisphenol A by ozonation: rate constants, influence of inorganic anions, and 

by-products 

Kheng Soo Tay *, Noorsaadah Abd. Rahman and Mhd. Radzi Bin Abas….………….… 77-94 

Chitinase production and antifungal potential of endophytic Streptomyces strain P4 

Julaluk Tang-um and Hataichanoke Niamsup*………………………………………... 95-104 

Water quality variation and algal succession in commercial hybrid catfish production ponds 

Chatree Wirasith and Siripen Traichaiyaporn*...………….….……………………….105-118 

Determination of production-shipment policy using a two-phase algebraic approach 

Yuan-Shyi Peter Chiu, Hong-Dar Lin and Huei-Hsin Chang *……..…………….… 119-129 

IIS-Mine: A new efficient method for mining frequent itemsets 

Supatra Sahaphong* and Veera Boonjing……………………………..……………... 130-151 

Lipase-catalysed sequential kinetic resolution of -lipoic acid 

Hong De Yan, Yin Jun Zhang, Li Jing Shen and Zhao Wang*…................................. 152-158 

http://www.mijst.mju.ac.th 

ISSN 1905-7873 © 2012 by Maejo University



Volume 6, Issue 1 (January - April 2012) 

Author Index 

Author 

Page 

Abas Mhd. R. B. 77 

Boonjing V. 130 

Braun U. 47 

Chang H.-H. 119 

Chauhan Y. K. 12 

Chiu Y.-S. P. 119 

Chukeatirote E. 47 

Fu J. 62 

Hyde K. D. 47 

Jain S. K. 12 

Khuantrairong T. 1 

Lin H.-D. 119 

McKenzie E. H. C. 47 

Moslem M. A. 47 

Niamsup H. 95 

Peerapornpisal Y. 28 

Phengsintham P. 47 

Pu Q. 62 

Pussadee N. 70 

Puttaraksa N. 70 

Rahman N. Abd. 77 

Author 

Page 

Rhodes M. W. 70 

Sahaphong S. 130 

Shen L. J. 152 

Singh B. 12 

Singkarat K. 70 

Singkarat S. 70 

Suphan S. 28 

Tang-um J. 95 

Tay K. S. 77 

Traichaiyaporn S. 1 

Traichaiyaporn S. 105 

Unai S. 70 

Underwood G. J. C. 28 

Wang Z. 152 

Whitlow H. J. 70 

Wirasith C. 105 

Wu S. 62 

Yan H. De 152 

Zhang Y. J. 152

Instructions for Authors 

A proper introductory e-mail page containing the title of the submitted article and certifying 

its originality should be sent to the editor (Duang Buddhasukh, e-mail : duang@mju.ac.th). The 

manuscript proper together with a list of suggested referees should be attached in separate files. The 

list should contain at least 5 referees with appropriate expertise. Three referees should be non-native 

from 3 different countries. Each referee's academic/professional position, scientific expertise, 

affiliation and e-mail address must be given. The referees should not be affiliated to the same 

university/institution as any of the authors, nor should any two referees come from the same 

university/institution. The editorial team, however, retain the sole right to decide whether or not the 

suggested referees are approached. 

Failure to conform to the above instructions will result in non-consideration of the 

submission. 

Please also ensure that English and style is properly edited before submission. UK style of 

spelling should be used. Authors who would like to consult a professional service can visit 

www.proof-reading-service.com, www.editage.com, www.enago.com, www.bioedit.co.uk 

(bioscience and medical papers), www.bioscienceeditingsolutions.com, www.scribendi.com, 

www.letpub.com, www.papersconsulting.com or www.sticklerediting.com. 

Important : Manuscript with substandard English and style will not be considered. 

Warning : Plagiarism (including self-plagiarism) may be checked for at the last stage of processing 

and, if detected, will result in a rejection and blacklisting. 

Manuscript Preparation 

Manuscripts must be prepared in English using a word processor. MS Word for Macintosh or 

Windows, and .doc or .rtf files are preferred. Manuscripts may be prepared with other software 

provided that the full document (with figures, schemes and tables inserted into the text) is exported to 

a MS Word format for submission. Times or Times New Roman font is preferred. The font size 

should be 12 pt and the line spacing 'at least 17 pt'. A4 paper size is used and margins must be 1.5 cm 

on top, 2.0 cm at the bottom and 2.0 cm on both left and right sides of the paper. Although our final 

output is in .pdf format, authors are asked NOT to send manuscripts in this format as editing them is 

much more complicated. Under the above settings, a manuscript submitted should not be longer than 

15 pages for a full paper or 20 pages for a review paper. 

A template file may be downloaded from the Maejo Int. J. Sci. Technol. homepage. 

(DOWNLOAD HERE) 

Authors' full mailing addresses, homepage addresses, phone and fax numbers, and e-mail 

addresses homepages can be included in the title page and these will be published in the manuscripts 

and the Table of Contents. The corresponding author should be clearly identified. It is the 

corresponding author's responsibility to ensure that all co-authors are aware of and approve of the 

contents of a submitted manuscript. 

A brief (200 word maximum) Abstract should be provided. The use in the Abstract of 

numbers to identify compounds should be avoided unless these compounds are also identified by 

names. 

A list of three to five keywords must be given and placed after the Abstract. Keywords may 

be single words or very short phrases.

Although variations in accord with contents of a manuscript are permissible, in general all papers 

should have the following sections: Introduction, Materials and Methods, Results and Discussion, 

Conclusions, Acknowledgments (if applicable) and References. 

Authors are encouraged to prepare Figures and Schemes in colour. Full colour graphics will be 

published free of charge. 

Tables and Figures should be inserted into the main text, and numbers and titles supplied for 

all Tables and Figures. All table columns should have an explanatory heading. To facilitate layout of 

large tables, smaller fonts may be used, but in no case should these be less than 10 pt in size. Authors 

should use the Table option of MS Word to create tables, rather than tabs, as tab-delimited columns 

are often difficult to format in .pdf for final output. 

Figures, tables and schemes should also be placed in numerical order in the appropriate place within 

the main text. Numbers, titles and legends should be provided for all tables, schemes and figures. 

Chemical structures and reaction schemes should be drawn using an appropriate software package 

designed for this purpose. As a guideline, these should be drawn to a scale such that all the details and 

text are clearly legible when placed in the manuscript (i.e. text should be no smaller than 8-9 pt). 

For bibliographic citations, the reference numbers should be placed in square brackets, i.e. [ ], 

and placed before the punctuation, for example [4] or [1-3], and all the references should be listed 

separately and as the last section at the end of the manuscript. 

Format for References 

Journal : 

1. D. Buddhasukh, J. R. Cannon, B. W. Metcalf and A. J. Power, “Synthesis of 5-n-alkylresorcinol 

dimethyl ethers and related compounds via substituted thiophens”, Aust. J. Chem., 1971, 24, 2655- 

2664. 

Text : 

2. A. I. Vogel, “A Textbook of Practical Organic Chemistry”, 3rd Edn., Longmans, London, 1956, pp. 

130-132. 

Chapter in an edited text : 

3. W. Leistritz, “Methods of bacterial reduction in spices”, in “Spices: Flavor Chemistry and 

Antioxidant Porperties” (Ed. S. J. Risch and C-T. Ito), American Chemical Society, Washington, DC, 

1997, Ch. 2. 

Thesis / Dissertation : 

4. W. phutdhawong, “Isolation of glycosides by electrolytic decolourisation and synthesis of 

pentinomycin”, PhD. Thesis, 2002, Chiang Mai University, Thailand. 

Patent : 

5. K. Miwa, S. Maeda and Y. Murata, “Purification of stevioside by electrolysis”, Jpn. Kokai Tokkyo 

Koho 79 89,066 (1979). 

Proceedings : 

6. P. M. Sears, J. Peele, M. Lassauzet and P. Blackburn, “Use of antimicrobial proteins in the 

treatment of bovine mastitis”, Proceedings of the 3rd International Mastitis Seminars, 1995, Tel-Aviv, 

Israel, pp. 17-18. 

(Download Submission Checklist) 

Manuscript Revision Time 

Authors who are instructed to revise their manuscript should do so within 45 days. Otherwise 

the revised manuscript will be regarded as a new submission.

Maejo Int. J. Sci. Technol. 2012, 6(01), i 

Maejo International 

Journal of Science and Technology 

ISSN 1905-7873 

Available online at www.mijst.mju.ac.th 

Editor’s Note 

The upkeep of the standard of a journal is a major and time-consuming job. At Maejo, 

however, we are willing to go to the extremes to do that kind of job. In trying to maintain the 

standards of an international journal, we subject every submitted article to a series of screening that 

may sometimes be inconvenient or plain unpleasant to a fair number of authors, to whom we 

sincerely apologise. As a matter of fact, some of their complaints have had a role in shaping several 

of our journal’s policies which will better serve future contributors. In so far as the language and 

academic standards of an article are not compromised, we thank them for their candid comments 

and always welcome more contributions from them. 

Duang Buddhasukh 

Editor-in-Chief

Communication 

Maejo Int. Maejo J. Sci. Int. Technol. J. Sci. Technol. 2012, 6(01), 2012, 1-11 6(01), 1-111 





Enhancement of carotenoid and chlorophyll content of an 

edible freshwater alga (Kai: Cladophora sp.) by supplementary 

inorganic phosphate and investigation of its biomass 

production 

Taweesak Khuantrairong and Siripen Traichaiyaporn* 

Algae and Water Quality Research Unit, Department of Biology, Faculty of Science, Chiang Mai 

University, Chiang Mai, 50200 Thailand 

* Corresponding author, e-mail: tsiripen@yahoo.com; tel +66 (0)53 943 346; fax +66 (0)53 852 259 

Received: 10 March 2011 / Accepted: 1 January 2012 / Published: 12 January 2012 

Abstract: Enhancement of the carotenoid and chlorophyll content of an edible freshwater alga (Kai: 

Cladophora sp.) by supplementary inorganic phosphate in canteen wastewater was investigated. The 

mass cultivation of the alga was conducted at ambient temperature and light intensity in diluted 

canteen wastewater enhanced with dipotassium hydrogen phosphate at 5-20 mg L -1 . An increase in 

total carotenoid and chlorophyll was observed. However, it had no effect on biomass production. As 

a result of the algal cultivation, there was a dramatic improvement in the quality of canteen wastewater. 

Keywords: freshwater alga, Cladophora, Kai 

________________________________________________________________________________ 

INTRODUCTION 

Algae are a significant source of human food, especially in Asia. In Thailand an edible 

freshwater alga, Cladophora (known locally as Kai), consists of two species, namely C. glomerata 

and a to-be-identified Cladophora sp. [1]. Kai is abundant in Nan and Mekong Rivers in the northern 

part of Thailand. The local people around these rivers collect it for domestic consumption and it is 

sold in the local markets. Many reports on the nutritional value of Cladophora spp. show that they 

contain a significant amount of carotenoids that are nutritionally essential for humans and animals [1- 

5]. Carotenoids and chlorophylls, which are found in plants and algae, are extremely important in 

photosynthesis and growth [6-8]. They are also powerful antioxidants that are beneficial to human

2 Maejo Int. J. Sci. Technol. 2012, 6(01), 1-11 

health and are now used for supplementary food and animal feed. Previous studies have suggested 

that they can prevent or delay cancer and degenerative diseases in humans and animals by 

contributing to antioxidative defenses against metabolic oxidative by-products [9-11]. 

Mass cultivation of algae in canteen wastewater for use as feed for Mekong giant catfish [2] 

and Tuptim tilapia [12] had been done. It was found that the algal biomass production strongly 

correlated with water quality and environmental factors, e.g. temperature, light intensity and nutrient 

concentration, results which were similarly obtained elsewhere [13-15]. However, several researches 

have indicated that phosphorus is a main factor related to growth and production of Cladophora 

[13-24]. Other researches have also shown that phosphorus is effective for algal carotenoid and 

chlorophyll production [25-30]. In contrast, stimulation of carotenoid and chlorophyll production 

with phosphorus starvation has been reported [31-35]. 

Culturing of algae in wastewater can also improve the water quality by decreaseing BOD, 

COD and nutrients of the wastewater [2,12,36]. In this research, mass cultivation of Cladophora sp. 

(Kai) in canteen wastewater with different phosphate concentrations was carried out to investigate 

the effects of supplementary phosphorus on carotenoid and chlorophyll content of this alga, its 

biomass production and physico-chemical characteristics of the culture water. 

MATERIALS AND METHODS 

Culture Media Preparation 

One thousand litres of canteen wastewater were collected from the canteen wastewater 

clarifier of Maejo University and left to settle in an open cement pond for 3 weeks to allow 

microorganisms to break down solid organic waste. The wastewater was then filtered through an 80- 

m plankton net filter [36]. The filtrate was diluted to 10% and analysed for pH, dissolved oxygen 

(DO), biochemical oxygen demand (BOD), total hardness, ammonia-nitrogen (NH 3 -N), nitratenitrogen 

(NO 3 - -N), nitrite-nitrogen (NO 2 - -N), total Kjeldal nitrogen (TKN) and orthophosphatephosphorus 

(PO 4 3- -P) by following the methods of APHA et al [37]. 

Culture Condition 

Cladophora sp. (Kai) was obtained from the Algae and Water Quality Research Unit, Chiang 

Mai University. The alga was cultured for 12 weeks at ambient air temperature (21-28 o C) and light 

intensity (12,333-39,267 lux) by attachment on plastic nets (60 g m -2 ) in cement raceway ponds (size 

1.2×2.3×0.5 m) each containing diluted canteen wastewater 20 cm deep (552 L per experiment) with 

continuous pump-driven circulation of 0.15 m s -1 [2,4]. A complete randomised design (CRD) was 

carried out with addition of dipotassium hydrogen orthophosphate (K 2 HPO 4 ) at concentrations of 5, 

10, 15 and 20 mg L -1 (treatment 1, 2, 3 and 4 respectively) with diluted canteen wastewater without 

K 2 HPO 4 as control. The experiment was done in triplicate. 

Carotenoid and Chlorophyll Analysis 

The alga at 12 weeks of culturing was harvested, washed with tap water, air dried and freezedried. 

The total carotenoid content was determined according to the method of Britton [38]. The 

freeze-dried sample (0.4 g) was homogenised with 20 mL of 95% ethanol in an extraction tube. Two

Maejo Int. J. Sci. Technol. 2012, 6(01), 1-11 

3 

mL of 5% KOH were then added, air was driven out of the tube with nitrogen gas and the tube was 

stored in the dark for at least 2 hours. Ether (3×5 ml) was added to extract the carotenoid portion. 

The ether supernatant was separated and its absorbance was read at 400-700 nm with a 

spectrophotometer. The total amount of carotenoid was calculated according to the following 

equation: total carotenoid = [(A max /0.25)×ether supernatant volume]/sample weight, where A max = 

maximum absorbance. 

Extraction of carotenoid components and chlorophylls was performed by a modified method 

of Yoshii et al. [3] and Dere et al. [39] as follows: the freeze-dried sample (0.5 g) was mixed with 

ice-cooled acetone (25 mL) in the dark, the mixture stored in the dark at -20 o C for 18 hours and the 

supernatant filtered. Carotene, xanthophyll and chlorophylls a and b (μg g -1 ) were determined 

spectrophotometrically at 470, 645 and 662 nm respectively by means of equations proposed by 

Lichtenthaler and Wellburn [40]: 

Chlorophyll a = 11.75A 662 – 2.35A 645 

Chlorophyll b = 18.61A 645 – 3.960A 662 

Carotene = (1000A 470 – 2.270C a – 81.4 C b ) / 227 (C a = chlorophyll a, C b = chlorophyll b) 

Xanthophyll = total carotenoid _ carotene 

Biomass Production 

The growth in terms of biomass (g m -2 wet weight) and specific growth rate (SGR) (% d −1 ) 

were measured every week following the method of Premila and Rao [41]. The SGR was calculated 

as SGR (% d −1 ) = {[ln(m 1 /m 0 )]/t}100, where m 0 = initial weight, m 1 = final weight, and t = time of 

culture in days. 

Water Quality Analysis 

Water samples from all the experimental ponds were collected once a week and analysed [37] 

for the physico-chemical properties, viz. temperature, pH, DO, BOD, total hardness, NH 3 -N, NO 3 - - 

N, NO 2 - -N, TKN and PO 4 3- -P. 

Statistical Analysis 

The data were presented as mean value ± standard deviation. Comparison of mean values was 

made by one-way analysis of variance (ANOVA), followed by Duncan's multiple range test (DMRT) 

at a significance level of p


Table 1. Carotenoid and chlorophyll content of Cladophora sp. (Kai) cultured at different phosphate 

concentrations 

Amount of pigments (g g -1 dry weight) 

Pigment 

Control Treatment 1 Treatment 2 Treatment 3 Treatment 4 

K 2HPO 4(mgL -1 )= 0 K 2HPO 4(mgL -1 )= 5 K 2HPO 4(mgL -1 )= 10 K 2HPO 4(mgL -1 )= 15 K 2HPO 4(mgL -1 )= 20 

Total carotenoid 889±157 a 1,084±253 a 1,130±147 a 1,151±151 a 1,729±212 b 

Carotene 44±4 a 86±9 b 87±6 b 96±7 bc 103±9 c 

Xanthophyll 779±42 a 997±254 a 1043±145 a 1055±154 a 1626±220 b 

Chlorophyll a 148±13 a 270±30 b 309±20 bc 305±52 bc 348±17 c 

Chlorophyll b 56±31 a 173±11 b 245±15 bc 200±49 c 249±16 c 

Note: Each value is mean ± SD 

Data in the same row with different superscripts are significantly different (p0.05). 

Carotenoids in green algae are produced in two different compartments and by two different 

pathways, i.e. the acetate-mevalonate pathway and the phosphoglyceraldehyde-pyruvate pathway, 

and they are further synthesised from isopentenyl diphosphate and its isomers [42-43]. In the present 

study it was observed that phosphate increased carotenoid and chlorophyll production in the alga, 

which agrees with the findings by Khuantrairong and Traichaiyaporn [4-5] that total carotenoid, β- 

carotene, lutein and zeaxanthin in Cladophora sp. positively correlate with phosphate level. Celekli 

et al. [29] reported that the phosphate supply increased biomass and carotenoid production in a bluegreen 

alga (Spirulina platensis) while Latasa and Berdalet [25] deserved that the synthesis of 

pigments in a dinoflagellate Heterocapsa sp. stopped upon phosphorus limitation. In contrast, 

Buapet et al. [30] suggested that phosphorus enrichment had no significant effect on chlorophyll a 

production by Ulva reticulate seaweed. Subudhi and Singh [35] reported that a high concentration 

of phosphate-phosphorus reduced the chlorophyll content of Azolla pinnata blue-green alga. 

Enhancement of astaxanthin in a green alga (Haematococcus pluvialis) [31,34] and that of β- 

carotene, zeaxanthin and violaxanthin in a marine microalga (Nannochloropsis gaditana) [33] were 

observed upon phosphorus limitation. Leonardos and Geider [32] stated that the nitrate-tophosphate 

supply ratio was related to carotenoid and chlorophyll-a production in the cryptophyte 

Rhinomonas reticulata. 

Biomass Production 

For 12-week cultures, the biomass production was similar among the treatments and ranged 

between 3,406-3,464 g m -2 (wet weight) (Figure 1). The highest value, 4,167 g m -2 , was observed in 

treatment 3 after culturing for 10 weeks. However, statistical analysis indicated that Cladophora sp. 

(Kai) cultured at different phosphorus concentrations showed no significant difference in growth 

(p>0.05). 

The specific growth rate (SGR) was similar among treatments and ranged between -2.7 _ 

38.9% d -1 . After one week of cultivation, the alga began to grow quickly and the highest SGR values


5 

were observed (Figure 2). Statistical analysis showed that the biomass production rate was not 

significantly different among the treatments (p>0.05). 

Figure 1. Biomass production of Cladophora sp. (Kai) cultured at different phosphorus 

concentrations (C = control group, T1 = treatment 1 (+5 mg L -1 K 2 HPO 4 ), T2 = treatment 2 (+10 

mg L -1 K 2 HPO 4 ), T3 = treatment 3 (+15 mg L -1 K 2 HPO 4 ), and T4 = treatment 4 (+20 mg L -1 

K 2 HPO 4 )) 

Figure 2. Specific growth rate (SGR) of Cladophora sp. (Kai) cultured at different phosphorus 

concentrations (C = control group, T1 = treatment 1 (+5 mg L -1 K 2 HPO 4 ), T2 = treatment 2 (+10 

mg L -1 K 2 HPO 4 ), T3 = treatment 3 (+15 mg L -1 K 2 HPO 4 ), and T4 = treatment 4 (+20 mg L -1 

K 2 HPO 4 ))


Thus, the biomass production and SGR of the alga cultured at different phosphorus 

concentrations showed no difference among all the treatments, indicating that the added phosphate 

had no effect on biomass production. Phosphorus is one of the main factors related to growth and 

production of freshwater Cladophora [13-24]. However, although Auer and Canale [24] observed 

that high dissolved phosphorus values in water induced an increase in stored phosphorus and growth 

rate of Cladophora, our results agreed with the findings that when phosphorus is above a certain 

concentration (0.01 mg L -1 ), it has no effect on the growth and biomass production of Cladophora 

[13,18,44]. The biomass production of Cladophora also strongly depends on environmental 

conditions, especially temperature and light intensity. High biomass production was observed in 

winter season and under high light intensity [2, 5, 21]. 

Water Quality 

For 12-week cultures, the physico-chemical properties of water of all treatments ranged as 

follows: temperature 19-28 o C, pH 8.3-8.9, DO 8.47-11.75 mg L -1 , BOD 1.00-10.27 mg L -1 , total 

hardness 38.91-68.21 mg L -1 (as CaCO 3 ), NH 3 -N 0.07-1.07 mg L -1 , NO 3 - -N 0.19-1.58 mg L -1 , NO 2 - - 

N 0.002-0.012 mg L -1 , TKN 0.19-4.16 mg L -1 and PO 4 3- -P 0.004-14.780 mg L -1 (Figure 3). It was 

noted that the PO 4 3- -P levels in all treatments were higher than the critical phosphorus level of 0.01 

mg L -1 for Cladophora growth in the natural ecosystem [20]. 

High DO values occurred in all experiments as a result of aeration by the air pump. Algae 

cultivation in canteen wastewater with high nutrients has been observed to dramatically reduce 

nitrogen and phosphorous levels and at the same time produce a useful product for animal feed [36]. 

In this work, it was noted that culturing of Cladophora sp. (Kai) showed similar improvement in 

water quality with decreasing BOD and nutrient levels (NH 3 -N, NO 3 - -N, TKN and PO 4 3- -P). 

CONCLUSIONS 

This study demonstrates that phosphate added to the culture of Cladophora sp. (Kai) 

enhances production of carotenoids and chlorophylls but has no effect on biomass production. In 

addition, cultivation of the alga in canteen wastewater improves the water quality by decreasing 

BOD and nutrients levels. 

ACKNOWLEDGEMENTS 

Financial support from Thailand Research Fund (TRF) through the Royal Golden Jubilee 

Ph.D. Program (Grant No. PHD/0130/2548) is acknowledged.


7 

A 

A 

B 

B 

C 

C 

D 

D 

E 

F F 

Figure 3. Water quality of Cladophora sp. (Kai) culture: water temperature (A), pH (B), DO (C), 

BOD (D), total hardness (E), and NH 3 -N (F) (C = control group, T1 = treatment 1, T2 = treatment 2, 

T3 = treatment 3, and T4 = treatment 4)


A 

A 

B 

B 

C 

D 

D 

Figure 3 (Continued). Water quality of Cladophora sp. (Kai) culture: NO 3 - -N (A), NO 2 - -N (B), TKN (C), 

and PO 3- -P (D) (C = control group, T1 = treatment 1, T2 = treatment 2, T3 = treatment 3, and T4 = 

treatment 4) 

REFERENCES 

1. Y. Peerapornpisal, D. Amornlertpison, C. Rujjanawate, K. Ruangrit and D. Kanjanapothi, “Two 

endemic species of macroalgae in Nan river, northern Thailand, as therapeutic agents”, 

ScienceAsia, 2006, 32 (supplement 1), 71-76. 

2. S. Traichaiyaporn, B. Waraegsiri and J. Promya, “Culture of a green alga genus Cladophora 

(Kai) as feed for the Mekong giant catfish (Pangasianodon gigas, Chevey)”, Final Report of 

The Thailand Research Fund, Bangkok, Thailand, 2011. 

3. Y. Yoshii, T. Hanyuda, I. Wakana, K. Miyaji, S. Arai, K. Ueda and I. Inouye, “Carotenoid 

compositions of Cladophora balls (Aegagropila linnaei) and some members of the 

Cladophorales (Ulvophyceae, Chlorophyta): Their taxonomic and evolutionary implication”, J. 

Phycol., 2004, 40, 1170-1177. 

4. T. Khuantrairong and S. Traichaiyaporn, “Production of biomass, carotenoid and nutritional 

values of Cladophora sp. (Kai) by cultivation in mass culture”, Phycologia, 2009, 48 

(supplement), 60-66.


9 

5. T. Khuantrairong and S. Traichaiyaporn, “The nutritional value of edible freshwater alga 

Cladophora sp. (Chlorophyta) grown under different phosphorus concentrations”, Int. J. Agric. 

Biol., 2011, 13, 297-300. 

6. T. W. Goodwin, “The Biochemistry of the Carotenoids”, 2nd Edn., Chapman and Hall, New 

York, 1980, pp.207-256. 

7. A. F. H. Marker, “The use of acetone and methanol in the estimation of chlorophyll in the 

presence of phaeophytin”, Freshwat. Biol., 1972, 2, 361-385. 

8. R. B. Woodward, “The total synthesis of chlorophyll”, Harvard University (no date), 

http://www.iupac.org/publications/pac/1961/pdf/0203x0383.pdf (Accessed: 29 Dec. 2011). 

9. P. Burtin, “Nutritional value of seaweeds”, Electron. J. Environ. Agric. Food Chem., 2003, 2, 

498-503. 

10. G. S. Omenn, G. E. Goodman, M. D. Thornquist, J. Balmes, M. R. Cullen, A. Glass, J. P. 

Keogh, F. L. Meyskens, B. Valanis, J. H. Williams, S. Barnhart and S. Hammar, “Effects of a 

combination of beta carotene and vitamin A on lung cancer and cardiovascular disease”, New 

Engl. J. Med., 1996, 334, 1150-1155. 

11. H. Tapiero, D. M. Townsend and K. D. Tew, “The role of carotenoids in the prevention of 

human pathologies”, Biomed. Pharmacother., 2004, 58, 100-110. 

12. J. Promya, “Assessment of immunity stimulating capacity and meat, egg qualities of hybrid 

Tumtim tilapia ND56 (Oreochromis sp.) fed on raw Spirulina”, PhD Thesis, 2008, Chiang Mai 

University, Thailand. 

13. B. A. Whitton, “Studies on the growth of riverain Cladophora in culture”, Arch. Microbiol., 

1967, 58, 21-29. 

14. C. E. R. Pitcairn and H. A. Hawkes, “The role of phosphorus in the growth of Cladophora”, 

Water Res., 1973, 7, 159-162. 

15. S. L. Wong and B. Clark, “Field determination of the critical nutrient concentrations for 

Cladophora in streams”, J. Fish. Res. Board Can., 1976, 33, 85-92. 

16. P. B. Birch, D. M. Gordon and A. J. McComb, “Nitrogen and phosphorus nutrition of 

Cladophora in the Peel-Hervey estuarine system of Western Australia”, Bot. Mar., 1981, 24, 

381-387. 

17. J. M. Graham, M. T. Auer, R. P. Canale and J. P., Hoffmann, “Ecological studies and 

mathematical modeling of Cladophora in Lake Huron: 4. Photosynthesis and respiration as 

functions of light and temperature”, J. Great Lakes Res., 1982, 8, 100-111. 

18. J. P. Hoffmann and L. E. Graham, “Effects of selected physicochemical factors on growth and 

zoosporogenesis of Cladophora glomerata (Chlorophyta)”, J. Phycol., 1984, 20, 1-7. 

19. J. R. Wharfe, K. S. Taylor and H. A. C. Montgomery, “The growth of Cladophora glomerata in 

a river receiving sewage effluent”, Water Res., 1984, 18, 971-979. 

20. D. S. Painter and G. Kamaitis, “Reduction of Cladophora biomass and tissue phosphorus in 

Lake Ontario, 1972-83”, Can. J. Fish. Aquat. Sci., 1987, 44, 2212-2215. 

21. W. K. Dodds and D. A. Gudder, “The ecology of Cladophora”, J. Phycol., 1992, 28, 415-427.


22. S. N. Higgins, R. E. Hecky and S. J. Guildford, “Environmental controls of Cladophora growth 

dynamics in eastern Lake Erie: Application of the Cladophora growth model (CGM)”, J. Great 

Lakes Res., 2006, 32, 629-644. 

23. S. N. Higgins, S. Y. Malkin, E. T. Howell, S. J. Guildford, L. Campbell, V. Hiriart-Baer and R. 

E. Hecky, “An ecological review of Cladophora glomerata (Chlorophyta) in the Laurentian 

Great Lakes”, J. Phycol., 2008, 44, 839-854. 

24. M. T. Auer and R. P. Canale, “Ecological studies and mathematical modeling of Cladophora in 

Lake Huron: 3. The dependence of growth rates on internal phosphorus pool size”, J. Great 

Lakes Res., 1982, 8, 93-99. 

25. M. Latasa and E. Berdalet, “Effect of nitrogen or phosphorus starvation on pigment 

composition of cultured Heterocapsa sp.”, J. Plankton Res., 1994, 16, 83-94. 

26. N. Leonardos and I. A. N. Lucas, “The nutritional values of algae grown under different 

culture conditions for Mytilus edulis L. larvae”, Aquaculture, 2000, 182, 301-315. 

27. M. Menéndez, J. Herrera and F. A. Comin, “Effect of nitrogen and phosphorus supply on 

growth, chlorophyll content and tissue composition of the macroalga Chaetomorpha linum 

(O.F. Müll.) Kütz, in a Mediterranean coastal lagoon”, Sci. Mar., 2002, 66, 355-364. 

28. M. Orosa, D. Franqueira, A. Cid and J. Abalde, “Analysis and enhancement of astaxanthin 

accumulation in Haematococcus pluvialis”, Bioresour. Technol., 2005, 96, 373-378. 

29. A. Celekli, M. Yavuzatmaca and H. Bozkurt, “Modeling of biomass production by Spirulina 

platensis as function of phosphate concentrations and pH regimes”, Bioresour. Technol., 2009, 

100, 3625-3629. 

30. P. Buapet, R. Hiranpan, R. J. Ritchie and A. Prathep, “Effect of nutrient inputs on growth, 

chlorophyll, and tissue nutrient concentration of Ulva reticulata from a tropical habitat”, 

ScienceAsia, 2008, 34, 245-252. 

31. B. R. Brinda, R. Sarada, B. S. Kamath and G. A. Ravishankar, “Accumulation of astaxanthin in 

flagellated cells of Haematococcus pluvialis - cultural and regulatory aspects”, Curr. Sci., 

2004, 87, 1290-1295. 

32. N. Leonardos and R. J. Geider, “Elemental and biochemical composition of Rhinomonas 

reticulata (Cryptophyta) in relation to light and nitrate-to-phosphate supply ratios”, J. Phycol., 

2005, 41, 567-576. 

33. E. Forján, I. Garbayo, C. Casal and C. Vílchez, “Enhancement of carotenoid production in 

Nannochloropsis by phosphate and sulphur limitation”, in “Communicating Current Research 

and Educational Topics and Trends in Applied Microbiology” (Ed. A. Méndez-Vilas), 

Formatex Research Center, Badajoz (Spain), 2007, pp.356-364. 

34. P. He, J. Duncan and J. Barber, “Astaxanthin accumulation in the green alga Haematococcus 

pluvialis: Effects of cultivation parameters”, J. Intgr. Plant Biol., 2007, 49, 447-451. 

35. B. P. R. Subudhi and P. K. Singh, “Effect of phosphorus and nitrogen on growth, chlorophyll, 

amino nitrogen, soluble sugar contents and algal heterocysts of water fern Azolla pinnata”, 

Biol. Plantarum, 1979, 21, 401-406.


11 

36. J. Promya, S. Traichaiyaporn and R. Deming, “Phytoremediation of kitchen wastewater by 

Spirulina platensis (Nordstedt) Geiteler: Pigment content, production variable cost and 

nutritional value”, Maejo Int. J. Sci. Technol., 2008, 2, 159-171. 

37. APHA, AWW and WPCF, “Standard Method for Examination of Water and Waste Water”, 

16 th Edn., American Public Health Association, Washington DC, 1985. 

38. G. Britton, “Workshop on carotenoid: The qualitative and quantitative analysis”, Prince of 

Songkla University, Thailand, 2005, pp.9-16. 

39. S. Dere, T. Günes and R. Sivaci, “Spectrophotometric determination of chlorophyll-a, b and 

total carotenoid contents of some algae species using different solvents”, Turk. J. Bot., 1998, 

22, 13-17. 

40. H. K. Lichtenthaler and A. R. Wellburn, “Determination of total carotenoids and chlorophyll a 

and b of leaf extract in different solvents”, Biochem. Soc. Trans., 1983, 11, 591-592. 

41. V. E. Premila and M. U. Rao, “Effect of crude oil on the growth and reproduction of some 

benthic marine algae of Visakhapatnam coastline”, Indian J. Mar. Sci., 1997, 26, 195-200. 

42. F. X. Cunningham, “Regulation of carotenoid synthesis and accumulation in plants”, Pure 

Appl. Chem., 2002, 74, 1409-1417. 

43. V. G. Ladygin, “Biosynthesis of carotenoids in the chloroplasts of algae and higher plants”, 

Russ. J. Plant Physiol., 2000, 47, 904-923. 

44. V. J. Bellis and D. A. McLarty, “Ecology of Cladophora glomerata (L.) Kütz. in Southern 

Ontario”, J. Phycol., 1967, 3, 57-63. 

© 2012 by Maejo University, San Sai, Chiang Mai, 50290 Thailand. Reproduction is permitted for 

noncommercial purposes.

12 Maejo Maejo Int. J. Sci. Int. Technol. J. Sci. Technol. 2012, 6(01), 2012, 12-27 6(01), 12-27 

Full Paper 





Performance of self-excited induction generator with costeffective 

static compensator 

Yogesh K. Chauhan 1, * , Sanjay K. Jain 2 and Bhim Singh 3 

1 

Department of Electrical Engineering, School of Engineering, Gautam Buddha University, G.B. 

Nagar-201310, India 

2 

Electrical and Instrumentation Engineering Department, Thapar University, Patiala-147004, India 

3 

Department of Electrical Engineering, Indian Institute of Technology, New Delhi-110016, India 

* Corresponding author, e-mail: chauhanyk@yahoo.com 

Received: 6 June 2010 / Accepted: 8 January 2012 / Published: 13 January 2012 

Abstract: The performance of a system consisting of a three-phase self-excited induction 

generator (SEIG) with static compensator (STATCOM) for feeding static resistive-inductive 

(R-L) and dynamic induction motor (IM) loads was investigated. The cost-effective 

STATCOM providing stable operation was designed by connecting additional shunt 

capacitance with the load. The STATCOM-controlled algorithm was realised by controlling 

the source current using two control loops with proportional-integral (PI) controller: one for 

controlling the SEIG terminal voltage and the other for maintaining the DC bus voltage. The 

SEIG-STATCOM performance was studied for two designs of STATCOM, namely costeffective 

STATCOM and full-rating STATCOM. The cost-effective SEIG-STATCOM 

system with the proposed control scheme exhibited improved performance with respect to 

starting time, voltage dip, generator current and total harmonic distortion under various 

transient conditions. 

Keywords: self-excited induction generator (SEIG), static compensator (STATCOM), 

voltage regulation, renewable energy source 

__________________________________________________________________________________ 

INTRODUCTION 

The fast rate of fossil fuel depletion is drawing attention to explore the alternative energy 

sources, e.g. small hydropower, wind and tidal power [1, 2]. The induction generator [3, 4] which 

operates in grid or self-excited mode is a strong candidate to harness electrical energy from these


13 

sources. An isolated electric supply using a self-excited induction generator (SEIG) is an economical 

option for such small-capacity applications as lighting and small motor loads in remote locations. 

The SEIG consists of a cage induction machine excited through an externally connected 

capacitor bank. The primary advantages of the SEIG in small capacity are the simple, brush-less and 

rugged construction, lower maintenance cost, small size and improved transient performance. The 

terminal voltage of SEIG is governed by parameters such as capacitance, prime mover speed and 

speed-torque characteristic and load. A poor voltage regulation results when the SEIG is loaded [5, 6] 

due to the increasing difference between the volt-ampere reactive (VAR) supplied by the capacitor 

bank and that demanded by the generator and load. The series capacitors [7-9] have been used with the 

SEIG for improved voltage regulation. The application of a thyristor as a static switch has resulted in 

such schemes as saturable core reactors and switching shunt capacitors for achieving improved voltage 

regulation of SEIG [10-12]. 

The modern power electronic converters are characterised by fast response, improved 

switching features and low cost. These converters are being applied as flexible alternating current 

transmission systems (FACTS) for various control and regulating purposes. The performance and cost 

of the shunt capacitor, static VAR compensator (SVC) and STATCOM were compared [13]. The 

STATCOM is normally a current controlled–voltage source inverter (CC-VSI) and has wide 

applications in the power system for improving power quality, harmonic elimination, reactive power 

compensation and load balancing [14,15]. The installation issue and capability of STATCOM have 

been demonstrated for the voltage limited feeder and industrial facility [16,17]. The concepts of static 

compensation for SEIG have been described for static loads [18,19]; the steps in designing a 

STATCOM for SEIG system have been summarised [20]. However, most of these attempts have been 

made to feed three-phase static loads and very little efforts are made for SEIG feeding the dynamic 

(motor) loads. 

In this paper, the performance of SEIG with STATCOM on feeding a static R-L load and 

induction motor is investigated. The dynamic model of the system is developed and a methodology to 

decide the ratings of STATCOM components such as the DC bus capacitor, AC side filter and 

insulated gate bipolar transistors (IGBT) is presented. The system performance is also studied for 

cost-effective STATCOM and full-rating STATCOM designs. 

SYSTEM DESCRIPTION AND CONTROL SCHEME 

The schematic description of the SEIG-STATCOM system is shown in Figure 1. A suitable 

capacitor bank is needed to obtain the rated voltage at no load. The STATCOM consists of an IGBTbased 

3- CC-VSI, an AC side-filter inductor, and a capacitor on the self-supporting DC bus. The 

scheme to control the STATCOM is depicted in Figure 2. In this scheme, two control loops are 

employed: one for controlling the terminal voltage of SEIG and other for maintaining the DC bus 

voltage. The proportional-integral (PI) controller is used with both control loops, which possess the 

advantage of being simple, effective and easy to tune.


Figure 1. SEIG-STATCOM system supplying the load 

Figure 2. Control scheme for STATCOM 

The control scheme is based on direct control of the source current, which consists of in-phase 

I sd 

and quadrature I 

sq 

components. I sd 

is the current drawn from SEIG to maintain the DC bus 

voltage by charging (or discharging) the DC bus capacitor. I 

sq 

is the reactive current required to 

maintain the SEIG terminal voltage. To regulate the DC bus voltage, the error signal for the PI 

controller is obtained from the sensed DC bus voltage V dc 

and its reference value V dc 

. The output of 

this PI controller is taken as I sd 

. The 3- reference in-phase current i sdabc 

is obtained by multiplying I 

sd 

with the sinusoidal in-phase unit voltage template u sdabc 

. To regulate the SEIG terminal voltage, the 

error signal for PI controller is obtained from the amplitude of SEIG voltage V sm 

and its reference 

value V sm 

. The output of this PI controller is taken as I sq 

. The 3- reference quadrature current i 

sqabc 

is 

obtained by multiplying I 

sq 

with the sinusoidal quadrature unit voltage template u sqabc 

. The total 

reference source current i sabc 

is taken as the sum of i sdabc 

and i 

sqabc 

. The sensed and reference source 

currents are processed in a rule based carrier less hysteresis band current controller to generate gating 

signals (S a , S b , S c ) for IGBT of CC-VSI of STATCOM.


15 

DESIGN METHODOLOGY 

The STATCOM is designed considering that the VAR rating of STATCOM is fixed, the source 

voltage is completely sinusoidal and the pulse width modulation (PWM) inverter operates in linear 

mode. The VAR rating of STATCOM, (VAR) STATCOM , is estimated using the capacitance needed for 

providing the rated voltage at no load (C NL ) and that needed for a stable operation of the induction 

motor while being loaded to its rated capacity (C FL ). These capacitances are computed using sequential 

unconstrained minimisation technique (SUMT) in conjunction with Rosenbroack’s direct search 

method [23]. The problem formulation is briefed in Appendix I. The (VAR) STATCOM is computed as 

 

2 1 1 

(VAR) STATCOM = 3V 

ab 

(1) 

 

XC 

X 

FL CNL 

 

and STATCOM line current I st is calculated as 

( VAR) 

STATCOM 

Ist 

(2) 

3V 

ab 

The reference DC bus voltage Vdcr 

of the voltage source converter of STATCOM depends upon 

the AC voltage. The STATCOM does not provide adequate compensation during the transient 

condition of low V dcr 

, whereas high V dcr 

may stress the devices. The V dcr 

is calculated using the peak 

supply line voltage V m [15] as 

V dcr = (1.2-2.0)V m , Vdcr 

Vm 

(3) 

The DC bus capacitor C dc stores energy and maintains the DC bus voltage with small ripple. 

The C dc is calculated using an energy balanced equation characterised by appropriate response time t 

and actual DC bus voltage V dca : 

C 

dc 

 

 

V 

3V ab 

I st 

t 

2 2 

dcr 

Vdca 

 

The AC filter inductance L f is decided by the allowable ripple in the compensation current [20]. 

L f is calculated by assuming a linear mode operation (modulation index m=1) and a switching 

frequency f s of 10 kHz: 

3 

Vdc 

2 

Lf 

 

 

(5) 

6af K 

s 

rp 

where the range of factor a (transient current) =1.2-2.0 and K rp (peak-to-peak ripple) =0.05-0.1. 

The ratings of IGBT suitable for medium rating and high frequency operation are decided as 

follows: 

Vdev (1 KL Kt ) Vm 

(6) 

I 2 K ( K 1) 

I 

dev s rp st 

where the ranges of K L (factor for filter inductor drop), K t (factor for transient voltage) and K s (factor 

of safety) are taken as 0.05-0.1, 0.1-0.2 and 1.25-1.50 respectively. 

(4)


SYSTEM MODEL 

The complete system as shown in Figure 1 consists of SEIG, STATCOM with associated 

control, and loads. The dynamic model of each component is presented herewith. 

SEIG Model 

The induction generator model is developed in a stationary q-d reference frame considering the 

effect of both main and cross flux saturation [21, 22]. The model, i.e. the q-d axis stator and rotor 

currents and the rotor speed (ω r ), in state space form is expressed using equations (7) and (8) 

respectively: 

p 

1 

i L v 

 

r 

i 

G 

i 

P 

p r 

TP Tem 

(8) 

2J 

3P 

Tem 

Lm 

iqsidr 

idsiqr 

, and v , i , r , L , G and T P are defined in Appendix-II. 

4 

where 

Shunt Capacitor Model 

The q-d axis stator currents i qs and i ds are converted into 3- stator currents i ga , i gb and i gc 

using 2- to 3- transformation [22]. Kirchhoff’s current law (KCL) [22] is applied to obtain the 

capacitor equations governing the SEIG voltage as 

pv 

pv 

ab 

bc 

 

 

iga ilda ica igb ildb icb 

 

3C 

sh 

iga ilda ica2igb ildb icb 

3C 

sh 

and v 

ab 

vbc 

vca 

0 

(10) 

where i ga 

, i lda 

and i ca 

are line ‘a’ currents for the generator, load and STATCOM respectively. 

STATCOM Model 

The charging (or discharging) of the DC bus capacitor C dc using hysteresis current controller 

switching functions (S a , S b , S c ) is expressed as 

pV 

 

i S i S i S 

 

ca a cb b cc c 

dc 

(11) 

Cdc 

The DC bus voltage V dc is reflected as voltages e a , e b and e c on the AC side of the PWM 

inverter as 

ea 2 1 1Sa 

Vdc 

e 

 

b 

1 2 1 

 

S 

 

 

 

b 

3 

 

 

e 

 

c 1 1 2 S 

 

c 

(7) 

(9) 

(12)


17 

The AC side filter equations in state space form are expressed as 

vbc 2vab ebc 2eab 3Rf ica 

 

pica 

 

3Lf 

vbc vab ebc eab 3Rf icb 

picb 

 

3L 

where eab ea eb 

and e bc 

e b 

e c 

. 

f 

(13) 

Static R-L Load 

The static load is considered as delta-connected. The phase currents for static R-L load (i prla , 

i prlb and i prlc ) are expressed as 

piprla ( vab Rai prla) 

La 

piprlb ( vbc Rbi prlb) 

Lb 

(14) 

pi ( v R i ) L 

prlc ca c prlc c 

Correspondingly, the line currents (i rla , i rlb and i rlc ) can be obtained as follows: 

i ( i i 

) 

rla prla prlc 

irlb ( iprlb iprla) 

i ( i i 

) 

rlc prlc prlb 

Induction Motor Load 

The values for v qs and v ds are obtained from SEIG voltages (v ab , v bc , v ca ) using 3- to 2- 

transformation [22]. The state space model of an induction motor dynamic load is similar to the 

induction generator model. It is expressed using motor parameters as 

1 

pi 

m 

Lm 

v 

m 

 

rm 

im 

 

G 

m 

im 

 

(16) 

Pm 

p rm 

Temm 

TL 

 

(17) 

2 J 

m 

The equations (7 _ 17) represent the model of the complete system. These equations are solved 

by fourth-order Runge-Kutta integration method [22] in MATLAB. 

(15) 

RESULTS AND DISCUSSION 

An investigation was carried out on a 3.7-kW induction machine operated as SEIG, which was 

loaded with a static R-L load of 0.8 pf and a 1.5-kW induction motor load. The parameters of this 

machine are given in Appendix III. The capacitances C NL and C FL were calculated as 16.1 F and 26.5 

F respectively for the rated SEIG voltage at no load and at the rated motor load under steady state. 

It was observed that the motor was unable to start even with C FL and resulted in a voltage 

collapse because the capacitor bank was unable to meet the excessive VAR requirements of SEIG and 

the motor load during starting. The starting of the motor was successful with a capacitance of 36F. 

The performance during voltage buildup, the starting of motor at 2.0 sec., and the subsequent loading 

on motor at 4.0 sec. are shown in Figure 3. During starting, the terminal voltage dropped to 0.25 p.u. 

and the induction motor took 0.7 sec. to stabilise the start-up transients. The excessive dip in voltage 

level and the longer duration for start-up were of serious quality concern. In addition, the SEIG 

exhibited poor voltage regulation even for the static load.


v (V) 

(Rad/s) 

ab 

T (Nm) i (A) i (A) 

rm emm ma gla 

1000 

0 

-1000 

20 

0 

-20 

10 

0 

-10 

50 

0 

-50 

400 

200 

0 

 

2 2.5 3 3.5 4 

2 2.5 3 3.5 4 

2 2.5 3 3.5 4 

2 2.5 3 3.5 4 

2 2.5 3 3.5 4 

Time(sec) 

Figure 3. SEIG feeding motor load with successful starting and consequent loading 

(Load conditions: 1.5-kW induction motor load at 2.0 sec. with C sh = 36 μF) 

The application of STATCOM was investigated for effective control and improved 

performance. The STATCOM was designed to give a cost-effective operation by selecting the C dc after 

considering the VAR requirement carefully. For a cost-effective STATCOM, the required VAR rating 

of STATCOM was calculated using the limiting capacitance of C NL and (C FL +C NL )/2 while an 

additional (C FL -C NL )/2 shunt capacitance was switched on with the load. For a full-rating STATCOM, 

the VAR rating of STATCOM was calculated with limiting capacitances C NL and C FL . For both the 

cost-effective and full-rating STATCOM designs, the shunt capacitance C NL needed to achieve the 

rated SEIG voltage at no load was taken as 16.1F. The different values of C FL were obtained for the 

static and motor loads. 

The capacitance needed for successful starting of the induction motor, which was 36 F, was 

taken as C FL for the motor load. C FL for the static load was considered as the capacitance needed to 

obtain the rated voltage at steady state, which was 30.4F. The STATCOM parameters are 

summarised in Table 1 for both cost-effective and full-rating STATCOM designs. The IGBT of 

reduced current ratings were needed for a cost-effective STATCOM, which reduced the cost, and 

therefore the operation should be cost-effective. 

Table 1. STATCOM design parameters 

Parameter Cost-effective STATCOM Full-rating STATCOM 

2.2kW static load 1.5 kW IM load 2.2kW static load 1.5 kW IM load 

(0.8 pf) 

(0.8 pf) 

STATCOM rating 1.21 kVAR 1.615 kVAR 2.42 kVAR 3.23 kVAR 

Current supplied by STATCOM 1.68 A 2.25 A 3.37 A 4.5 A 

Ref. DC bus voltage 700 V 700 V 700 V 700 V 

DC bus capacitance 86.3 F 118.1 F 172.6 F 236.3 F 

AC filter inductance 11.16 mH 8.16 mH 5.58 mH 4.08 mH 

Device selection IGBT IGBT IGBT IGBT 

Device voltage rating 1094 V 1094 V 1094 V 1094 V 

Device current rating 5.5 A 7.5 A 11.0 A 15.0 A


19 

On the basis of STATCOM design parameters, a comparison between the cost-effective and 

full rating STATCOM is summarised in Table 2. 

Table 2. Comparison of performance parameters for cost-effective and full-rating STATCOM 

Performance parameter Cost-effective STATCOM Full-rating STATCOM 

Performance (voltage regulation, 

starting time, THD, etc.) 

Energy consumption (mainly in 

IGBT, DC bus capacitor and filter 

inductor loss) 

Overall cost (IGBT, DC bus 

capacitor, filter inductor) 

Reasonably good 

Low (lower IGBT current rating, 

lower DC bus capacitance and 

higher filter inductor) 

Low 

Good 

High (higher IGBT current rating, 

higher DC bus capacitance and 

lower filter inductor) 

High 

Performance with Balanced Static R-L Load 

The performance characteristics of the systems with cost-effective and full-rating STATCOM 

are shown in Figure 4 and Figure 5 respectively for feeding balanced 0.8-pf static R-L load. A load of 

2.2 kW was applied at 2.5 sec., which was changed to 3.0 kW at 3.0 sec. and 2.2 kW at 3.5 sec. 

With cost-effective STATCOM (Figure 4), the visible transients lasting for about 0.2 sec were 

observed in v ab , i ga and V dc due to the application of 2.2-kW load. The V dc momentarily decreased to 

480V and varied between 480-800V before returning to the reference value. With full-rating 

STATCOM (Figure 5), the transients settled down after 5-6 cycles and the DC bus voltage decreased 

to 500V momentarily with the application of 2.2-kW load. The increase in static load to 3.0 kW at 3.0 

sec. did not result in any appreciable transients due to the choice of C dc corresponding to the motor 

load. In both cost-effective and full-rating STATCOM, V dc dropped momentarily to 630V and 

gradually returned to the reference value with the PI controller action. When the load was reduced to 

2.2 kW at 3.5 sec., the dynamics of the system changed accordingly. V dc momentarily rised to 760V 

before returning to the reference mark. A steady state was achieved within 0.12 sec. and 0.25 sec. with 

the full-rating and cost-effective STATCOM respectively. 

Performance with Unbalanced Static R-L Load 

The system performance with the cost-effective STATCOM was studied for an unbalanced R-L 

load and the corresponding characteristics for three-phase generator voltage v abc , generator current 

i gabc , load current i rlabc , compensation current i cabc , and DC bus voltage V dc are shown in Figure 6. 

Initially, the 2.2-kW, 3.0-kW and 1.0-kW loads of 0.8 pf were applied on ‘a’, ‘b’ and ‘c’ phases 

respectively at 3.0 sec. Further, an acute unbalance was made by unloading of ‘c’ phase at 3.5 sec. The 

STATCOM responded satisfactory during the unbalanced load condition and maintained the balanced 

condition at SEIG terminals.


1000 

v ab 

(V) 

0 

-1000 

2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 

20 

i ga 

(A) 

0 

-20 

2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 

10 

i rla 

(A) 

0 

-10 

2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 

10 

i ca 

(A) 

0 

-10 

2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 

V dc 

(V) 

800 

600 

400 

2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 

Time(sec) 

Figure 4. Performance characteristics of SEIG-STATCOM (cost-effective) system with 0.8-pf static R-L 

load (Load conditions : 2.2 kW at 2.5 sec., 3.0 kW at 3.0 sec. and 2.2 kW at 3.5 sec.) 

1000 

v ab 

(V) 

0 

-1000 

2.4 2.6 2.8 3 3.2 3.4 3.6 

20 

i ga 

(A) 

0 

-20 

2.4 2.6 2.8 3 3.2 3.4 3.6 

10 

i rla 

(A) 

0 

-10 

2.4 2.6 2.8 3 3.2 3.4 3.6 

10 

i ca 

(A) 

0 

-10 

2.4 2.6 2.8 3 3.2 3.4 3.6 

800 

V dc 

(V) 

600 

400 

2.4 2.6 2.8 3 3.2 3.4 3.6 

Time(sec) 

Figure 5. Performance characteristics of SEIG-STATCOM (full rating) with 0.8-pf static R-L load 

(Load conditions: 2.2 kW at 2.5 sec., 3.0 kW at 3.0 sec., and 2.2 kW at 3.5 sec.)


21 

1000 

v abc 

(V) 

0 

-1000 

20 

3 3.2 3.4 3.6 3.8 4 4.2 

i rlabc 

(A) 

i gabc 

(A) 

i cabc 

(A) 

V dc 

(V) 

0 

-20 

10 

0 

-10 

5 

0 

-5 

750 

700 

650 

3 3.2 3.4 3.6 3.8 4 4.2 

3 3.2 3.4 3.6 3.8 4 4.2 

3 3.2 3.4 3.6 3.8 4 4.2 

3 3.2 3.4 3.6 3.8 4 4.2 

Time(sec) 

Figure 6. Performance characteristics of SEIG-STATCOM (cost-effective) with unbalanced 0.8-pf R-L load 

(Load conditions: 3-φ load (2.2 kW, 3.0 kW, 1.0 kW) at 3.0 sec.; load (2.0 kW, 3.0 kW, 

unloading on ‘c’ phase) at 3.5 sec; and balanced 3-φ loading of 2.2.0 kW at 4.0 sec.) 

Performance with Motor Load 

The performance characteristics of the system during the starting and sudden loading of an 

induction motor are shown in Figures 7 and 8 for cost-effective and full-rating STATCOM 

respectively. The motor at no load was switched on at 3.0 sec. and was loaded at 4.0 sec. During the 

starting of the motor, excessive reactive VAR were drawn and therefore the V dc dropped to 300V level 

with both STATCOM. The V dc quickly rose above the reference 700V before returning to reference 

level after replenishing the charge on DC bus capacitor. At starting, the transients in v ab , i ga , i ma , T emm 

and V dc were more visible with cost-effective STATCOM due to its lower rating and the uncharged 

capacitor connection across the motor load terminals. Because of a lower compensation level, the i ca 

under steady state was lower compared to the corresponding value with full-rating STATCOM. With 

cost-effective STATCOM, the SEIG voltage decreased by 10% and the transients were settled in 0.5 

sec., whereas these values were 6% and 0.4 sec. respectively for full-rating STATCOM. Without 

STATCOM, as shown in Figure 3, the decrease in SEIG voltage was 30%. The motor was loaded with 

rated load torque at 4.0 sec. After small transients, the system response changed accordingly. The 

increase in i ga , i ma , and T emm and the decrease in rm were observed. The V dc decreased to 670V before 

returning to steady-state reference value of 700V. 

The performance of SEIG-IM configuration was compared in the presence and absence of 

STATCOM and the key parameter indices of the system are summarised in Table 3. The total 

harmonic distortion (THD) was obtained through fast Fourier transform (FFT) using discrete Fourier 

transform (DFT) algorithm of MATLAB. The generator-side THD values were within a permissible 

level, which demonstrated a satisfactory overall system performance.


1000 

v ab 

(V) 

0 

-1000 

20 

3 3.5 4 4.5 

i ga 

(A) 

0 

-20 

20 

3 3.5 4 4.5 

i (A) (Rad/sec) rm T (N ) 

i (A) 

ca emm m ma 

0 

-20 

50 

0 

500 

0 

20 

0 

-20 

800 

3 3.5 4 4.5 

3 3.5 4 4.5 

3 3.5 4 4.5 

3 3.5 4 4.5 

V dc 

(V) 

600 

400 

3 3.5 4 4.5 

Time(sec) 

Figure 7. Performance characteristics of SEIG-STATCOM (cost-effective) with induction motor load 

(Load conditions: T L = 0 at 3.0 sec. and T L = rated torque at 4.0 sec.) 

1000 

v ab 

(V) 

0 

-1000 

2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4 

20 

i ga 

(A) 

0 

-20 

2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4 

20 

(Rad/sec) 

i (A) rm T (Nm) i (A) 

ca emm 

ma 

0 

-20 

2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4 

50 

0 

2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4 

500 

0 

2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4 

50 

0 

-50 

2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4 

800 

V dc 

(V) 

600 

400 

2.8 3 3.2 3.4 3.6 3.8 4 4.2 4.4 

Time(sec) 

Figure 8. Performance characteristics of SEIG-STATCOM (full-rating) with induction motor load 

(Load conditions : T L = 0 at 3.0 sec. and T L = rated torque at 4.0 sec.)


23 

Table 3. Comparison of performance indices of the system with and without STATCOM 

Performance index 

Without 

STATCOM 

With STATCOM 

Cost-effective 

design 

Supply voltage dip at start-up 75% 10.0% 6.1% 

Start-up time 0.7 sec 0.18 0.16 sec 

Supply current THD at full motor load NA 1.4% 1.3% 

Rise/drop in DC bus voltage of 

STATCOM at the time of sudden IM 

loading 

Drop in DC bus voltage with full torque 

loading of IM 

Full-rating design 

NA 840V/300V 790V/330V 

NA 625V 650V 

CONCLUSIONS 

The investigations on SEIG-STATCOM system have been presented for feeding static R-L and 

induction motor loads. The dynamic model of the system has been developed in state space form. The 

STATCOM parameters have been calculated for two design cases, namely the cost-effective and fullrating 

designs using the proposed design procedure. The cost-effective STATCOM was designed to 

provide a stable operation by connecting additional shunt capacitance with the load. The system 

performance has been presented for both the cost-effective and full-rating STATCOM. With a 

controlling algorithm, the system exhibited improved performance in terms of parameters such as 

starting time, voltage dip, generator current and total harmonic distortion in the supply current under 

various transient conditions. The full-rating STATCOM could be considered for critical applications 

having stringent performance consideration while the cost-effective STATCOM should be for remote 

applications where the cost is important.


Appendix I 

The per phase equivalent circuit of SEIG feeding induction motor load at steady state is shown 

in Figure 9. In the equivalent circuit, the SEIG parameters R s and X ls are stator resistance and leakage 

reactance respectively, R r and X lr are rotor resistance and leakage reactance respectively, and X m is the 

magnetising reactance. The corresponding parameters are presented with subscript m for motor load. 

X csh is the reactance offered by the capacitor bank, and F and are per unit frequency and prime-mover 

speed respectively. 

jFX lr 

jFX m 

jFX ls 

R s 

-jX csh 

/F 

R sm 

jFX lsm 

jFX mm 

jFX lrm 

R rm 

/(F- m 

) 

R r 

/(F-) 

I s 

Figure 9. Per phase equivalent of SEIG feeding motor load 

Applying KVL on stator side loop of induction generator results in 

Z loop I s = 0 

Under steady state condition, I s cannot be zero and therefore Z loop 

should be zero. An 

optimisation problem has been formulated to obtain the unknown variables X csh and F. The objective 

function F n is expressed as 

Fn( X 

csh, F) abs( Zloop) 

The values of X csh and F should lie between the respective minimum and maximum limits: 

Fmn F Fmx , Xcmn Xcsh Xcmx 

 

The above optimisation problem is solved through SUMT in conjunction with the Rosenbroack 

method of direct search technique [23]. After the convergence, the capacitance is computed. 

Appendix II 

The induction machine model is developed in a stationary reference frame while incorporating 

the effects of both the main flux and cross-flux saturation. The forms of v, i, r, 

L 

and G are 

given as 

T 

T 

v v v v v i i i i i r diag r r r r 

qs ds qr dr ; qs ds qr dr ; 

L 

L L L L 

 

L L L L 

 

L L L L 

 

L L L L 

sq dq mq dq 

dq sq dq md 

mq dq rq dq 

dq md dq rd 

 

 

 

 

 

 

0 0 0 0 

 

0 0 0 0 

 

G 

 

0 

rLm 0 Lr 

 

 

 

rLm 0 Lr 

0 

 

s s r r 

The air gap voltage of SEIG does not remain constant during loading. Therefore, the 

magnetising inductance is calculated by calculating the magnetising current as 

 

2 2 

m 

 

qs 

 

qr 

 

ds 

 

dr 

i i i i i 

The inductances in [L] are evaluated [21] as


25 

L / i , Ld / di ,cos i / i , sin i / i 

m m m m m dm m qm m 

L ( LL 

)cossin 

dq 

m 

L Lcos L sin , L Lsin L 

cos 

2 2 2 2 

mq m md m 

L L L , L L L , L L L , L L L 

sq ls mq sd ls md rq lr mq rd lr md 

The prime mover torque driving the induction machine is expressed as 

T 6200 20 

P 

r 

Appendix III 

Generator parameters 

3.7 kW, 415 V, -connected, 7.6 A (line), 50 Hz, 4 pole, J=0.0842 kg-m 2 cage induction machine. 

R s =0.0585 pu, R r =0.06196 pu, X ls =X lr =0.1015 pu, X m(unsat) =2.858 pu. 

Magnetisation characteristic of SEIG 

2 

L K i K i K 

m 1 m 2 m 3 

where K 1 0.0091, K 2 2.0024 , K 3 348.35. 

Motor parameters 

1.5 kW, 415 V, -connected, 3.2 A (line), 50 Hz, 4 pole, J=0.0205 kg-m 2 cage induction machine. 

R sm =0.0832 pu, R rm =0.0853 pu, X lsm =X lrm =0.1101 pu, X mm(unsat) =1.83 pu. 

Magnetisation characteristic of motor 

5 4 3 2 

Lmm Km 1imm Km2imm Km3imm Km4imm Km5imm Km6 

where Km 

1 

0.0072, Km2 0.0849 , Km3 0.4298 , Km4 0.9924 , Km5 0.6847 , Km6 1.171.


REFERENCES 

1. S. Major, T. Commins, and A. Noppharatana, “Potential of wind power for Thailand: An 

assessment”, Maejo Int. J. Sci. Technol. 2008, 2, 255-266. 

2. F. F. Li, J. Kueck, T. Rizy and T. King, “A preliminary analysis of the economics of using 

distributed energy as a source of reactive power supply”, Report for the U.S. Department of 

Energy, 2006, www.localpower.org/documents/reporto_doe_reactivepowerandde.pdf. 

3. E. D. Bassett and F. M. Potter, “Capacitive excitation for induction generators”, AIEE Trans. 

(Elect. Eng.), 1935, 54, 540-545. 

4. J. M. Elder, J. T. Boys and J. L. Woodward, “Self-excited induction machine as a small low-cost 

generator”, IEE Proc. Gener. Transm. Distrib., 1984, 131, 33-41. 

5. S. S. Murthy, O. P. Malik and A. K. Tandon, “Analysis of self excited induction generators”, IEE 

Proc. Gener. Transm. Distrib., 1982, 129, 260-265. 

6. L. Shridhar, B. Singh, C. S. Jha and B. P. Singh, “Analysis of self excited induction generator 

feeding induction motor”, IEEE Trans. Ener. Convers., 1994, 9, 390-396. 

7. E. Bim, J. Szajner and Y. Burian, “Voltage compensation of an induction generator with longshunt 

connection”, IEEE Trans. Ener. Convers., 1989, 4, 526-530. 

8. M. H. Haque, “Selection of capacitors to regulate voltage of a short-shunt induction generator”, 

IET Gener. Transm. Distrib., 2009, 3, 257–265. 

9. L. Wang and J. Y. Su “Effects of long-shunt and short-shunt connections on voltage variations of a 

self-excited induction generator”, IEEE Trans. Ener. Convers., 1997, 12, 368-374. 

10. S. M. Alghuwainem, “Steady-state analysis of an induction generator self excited by a capacitor in 

parallel with saturable reactor”, Electr. Mach. Power Syst., 1998, 26, 617-625. 

11. S. C. Kuo and L. Wang, “Analysis of isolated self-excited induction generator feeding a rectifier 

load”, IEE Proc. Gener. Transm. Distrib., 2002, 149, 90-97. 

12. T. L. Maguire and A. M. Gole, “Apparatus for supplying an isolated DC load from a variablespeed 

self-excited induction generator”, IEEE Trans. Ener. Convers., 1993, 8, 468-475. 

13. A. S. Yome and N. Mithulananthan, “Comparison of shunt capacitor, SVC and STATCOM in 

static voltage stability margin enhancement”, Int. J. Electr. Eng. Edu., 2004, 41, 158-171. 

14. B. Singh, K. Al-Haddad and A. Chandra, “Harmonic elimination, reactive power compensation 

and load balancing in three-phase, four-wire electric distribution systems supplying non-linear 

loads”, Electr. Power Syst. Res., 1998, 44, 93-100. 

15. S. K. Jain, P. Agarwal and H. O. Gupta, “Fuzzy logic controlled shunt active power filter for power 

quality improvement”, IEE Proc. Electr. Power Appl., 2002, 149, 317-328.


27 

16. S. M. Ramsay, P. E. Cronin, R. J. Nelson, J. Bian and F. E. Menendez, “Using distribution static 

compensators (D-STATCOMs) to extend the capability of voltage-limited distribution feeders”, 

Proceedings of IEEE Rural Electric Power Conference, 1996, Fort Worth, TX, USA, pp. A4:1- 

A4:7. 

17. K. N. Choma and M. Etezadi-Amoli, “The application of a DSTATCOM to an industrial facility”, 

Proceedings of IEEE Power Engineering Society Winter Meeting, 2002, New York, USA, Vol. 2, 

pp. 725-728. 

18. B. Singh and L. B. Shilpakar, “Analysis of a novel solid state voltage regulator for a self-excited 

induction generator”, IEE Proc. Gener. Transm. Distrib., 1998, 145, 647-655. 

19. S. C. Kuo and L. Wang, “Analysis of voltage control for a self-excited induction generator using a 

current-controlled voltage source inverter (CC-VSI)”, IEE Proc. Gener. Transm. Distrib., 2001, 

148, 431-438. 

20. B. Singh, S. S. Murthy and S. Gupta, “Analysis and design of STATCOM-based voltage regulator 

for self-excited induction generators”, IEEE Trans. Ener. Convers., 2004, 19, 783-790. 

21. E. Levi, “Applications of the current state space model in analyses of saturated induction 

machines”, Electr. Power Syst. Res., 1994, 31, 203-216. 

22. P. C. Krause, “Analysis of electrical machinery”, McGraw Hill, New York, 1986. 

23. S. S. Rao, “Engineering optimization, theory and practice”, New Age International, New Delhi, 

1998. 



28Maejo Maejo Int. J. Int. Sci. J. Technol. Sci. Technol. 2012, 2012, 6(01), 6(01), 28-46 28-46 

Full Paper 





Benthic diatoms of Mekong River and its tributaries in 

northern and north-eastern Thailand and their application to 

water quality monitoring 

Sutthawan Suphan 1,* , Yuwadee Peerapornpisal 1 and Graham J. C. Underwood 2 

1 Department of Biology, Faculty of Science, Chiang Mai University, 50200, Thailand 

2 

Department of Biological Sciences, University of Essex, Colchester, Essex, United Kingdom CO4 

3SQ 

* Corresponding author, e-mail: suttawan@hotmail.com 

Received: 7 February 2011 / Accepted: 6 May 2011 / Published: 26 January 2012 

Abstract: Biomonitoring of benthic diatoms was performed to assess the water quality of 

Mekong River and its tributaries in northern and north-eastern Thailand. Fourteen 

sampling sites along the river and its tributaries were investigated. Two hundred and 

fifty-two species in 53 genera of diatoms were recorded. Each sampling site had distinct 

water chemistry and other physical variables. Cluster analysis identified 11 groups at 

80% similarity. The relationship between diatom community composition and water 

quality variables was determined by statistical techniques. A number of diatom species 

were found to be useful as indicators of some physico-chemical properties of water. 

Keywords: benthic diatoms, water quality, biomonitoring, Mekong River 

_______________________________________________________________________________________ 

INTRODUCTION 

One of the most important natural resources for human life is water resource. Human use 

including household, industrial, agricultural and recreational activities affects water quality. There 

is an urgent need to develop a more sustainable practice for the management and efficient use of 

water resources as well as the need to protect the ecosystems where these resources are located. 

Therefore, it is important to monitor the quality of our limited water supplies. Presently, there are 

several methods to monitor water quality. One of the methods that is successfully used for 

monitoring aquatic environments around the world is biological assessment of water quality. It is 

considered an essential part in the assessment of the ecological quality of running waters apart from


29 

the data obtained from other sources such as hydrology, eco-morphology, physico-chemical water 

analysis and eco-toxicological analysis [1]. 

In rivers and streams, benthic diatoms, the most common and diverse primary producers [2], 

are regarded as bioindicators due to their sensitivity and strong response to many physical, chemical 

and biological changes [3]. Diatoms are successfully used for monitoring aquatic environments 

around the world, especially in Europe, USA and Japan [4-5]. 

In Thailand, Mekong River flows through Chiang Rai province in the north. As it continues 

along its path in Laos, it once again flows back into Loei, Nong Khai, Nakhon Panom, Mukdaharn, 

Amnaj Charoen and Ubon Ratchadhani provinces in the north-east of Thailand. There are many 

smaller rivers which are tributaries that may affect the Mekong River. Thus, the water quality in the 

Mekong River and its tributaries should be monitored continuously. 

There have been few studies of diatoms in the Mekong River in the past [6-8]. In this study, 

the diversity and distribution of benthic diatoms in the Mekong River and its tributaries in northern 

and north-eastern Thailand was investigated. In addition the relationship between diatoms species 

and some water physico-chemical properties was studied. The basic ecological data that could be 

applied to develop sustainable water resources were also obtained. Furthermore, the study should 

provide more information on diatoms in the South-east Asian region where very few reports on 

benthic diatoms were documented. 


Study Area 

Fourteen sampling sites along the Mekong River and its tributaries in northern and northeastern 

Thailand were selected based on the distance and environmental impact. Nine sites were 

located along Mekong River and five sites along its tributaries. The detail of each sampling site is 

shown in Figure 1. Diatom samples and physico-chemical water quality were determined 3 times 

per year in each season during July 2005 – April 2007. 

Benthic Diatom Collection and Identification 

Ten replicates of benthic diatoms were collected at each sampling site. Diatoms were taken 

from stone surfaces using a toothbrush and a 10-cm 2 plastic sheet. The samples were put in plastic 

boxes and fixed with Lugol’s solution on site. Diatom samples were then taken to the laboratory and 

cleaned by concentrated acid digestion method [9]. Briefly, each sample was centrifuged at 3,500 

rpm for 15 minutes. The diatom cells were placed in an 18-cm core tube, added with concentrated 

nitric acid, heated in a boiler (70-80 o C) for 30-45 minutes, and rinsed 4-5 times with deionised 

water. 

Each cleaned diatom sample was mounted on a microscope slide with Naphrax ® , a mountant 

with a high refractive index [9-10]. Up to 300 diatom valves were counted and identified with an 

Olympus CH30 microscope at ×1000 magnification. Taxonomy and nomenclature was determined 

according to the relevant references [8, 11-17].


Sampling sites along Mekong River 

Sampling sites along tributaries 

Site 1 Golden Triangle, Chiang Rai province (GT) 

Site 2 Kok River, Chiang Rai province (KO) 

Site 3 Ban Had Krai, Chiang Rai province (HK) 

Site 4 Hueng River, Loei province (HG) 

Site 5 Kaeng Khood Khoo, Loei province (KK) 

Site 6 Ponpisai, Nong Khai province (PS) 

Site 7 Laung River, Nong Khai province (LG) 

Site 8 Mueng, Nakhon Phanom province (NP) 

Site 9 Songkram River, Sakonnakron province (SK) 

Site 10 Kaeng Ka Bao, Nakhon Phanom province (KB) 

Site 11 Had Hin Win Chai, Amnat Charoen province (HW) 

Site 12 Kaeng Hin Kan, Mukdahan province (KH) 

Site 13 Kaeng Sapue, Ubon Ratchathani province (KP) 

Site 14 Khong Jium, Ubon Ratchathani province (KJ) 

Figure 1. Map of Thailand showing the location of 14 sampling sites along Mekong River and its 

tributaries 

Physico-Chemical Data 

Water samples were collected in triplicate at each sampling site. The samples were put in 

polyethylene bottles and kept in a cool box at 5-7 o C. Water chemical and physical properties were 

determined by established methods [18]; soluble reactive phosphorus (SRP) was determined by 

ascorbic acid method, nitrate nitrogen (NO 3 - -N) by cadmium reduction method, ammonia nitrogen 

(NH 4 + -N) by Nesslerisation method, alkalinity (as mg/L CaCO 3 ) by phenolpthalein methyl orange 

indicator method, dissolved oxygen (DO) by azide modification of the Winkler method, and 

biochemical oxygen demand (BOD) by 5-day incubation and azide modification of the Winkler 

method. The pH was measured with a pH meter, conductivity with a conductivity meter, turbidity 

with a turbidity meter, water temperature with a thermometer, and water velocity with a velocity 

meter (Aquaflow Probe - Model 6900, Ricky Hydrological Company). 

Data Analysis 

ANOVA single factor was performed on water quality and data sets on benthic diatoms 

assemblage to determine any significant differences between groups. Pearson’s correlation was also 

calculated for certain variables. Water quality data and diatom species were analysed with canonical 

correspondence analysis (CCA), a multivariate direct gradient analysis method widely used in 

ecology, to determine the relationship between physico-chemical water quality and diatom species 

using multivariate statistical package (MVSP) software. Cluster analysis was performed on the log 

of transformed water quality data using MVSP software with unweighted pair-group method of 

arithmetic averages (UPGMA) cluster method to show similarity percentage between samples.


31 

Environmental variables of the clusters were then tested for significance between groups with 

ANOVA single factor. The number of benthic diatoms in each group was calculated by Minitab 

program to find significant differences and correlations between groups. 


Diatom Diversity and Distribution 

A total of 135,859 benthic diatom cells were counted. Two hundred and fifty-two species of 

benthic diatoms were found and classified into 3 classes, 6 subclasses, 14 orders, 27 families and 53 

genera. The majority (88.5%) was in Bacillariophyceae class with the remaining 6.0% in 

Fragilariophyceae class and 5.5% in Coscinodiscineae class. Nitzschia was the genus with the 

highest number of species (30 species) followed by Navicula (25 species), Gomphonema (16 

species), Eunotia (14 species), Luticola (12 species) and Pinnularia (8 species). The number of 

diatom species recorded was similar to that reported for other river systems of comparable size. 

Holmes and Whitton [19] listed 230 diatom species collected from Tees River in northern England 

while Archibald [20] recorded 310 diatom species from Sundays and Great Fish Rivers in South 

Africa. Similarly, 267 species of diatoms found in La Trobe River and tributaries in Australia was 

reported [21]. The distribution of the diatoms is shown in Figure 2. The lowest number of diatoms 

was recorded at HK in Chiang Rai, northern Thailand, during the fourth sampling in July 2006 

(rainy season). 

Figure 2. Number of benthic diatoms in Mekong River and its tributaries between July 2005 –April 

2007 

Among the 252 species, 29 were common species as shown in Table 1 and Figures 3. 

Nitzschia palea showed the highest percentage (36.0%) followed by Mayamaea atomus (35.4%), 

Eolimna minima (25.3%), Navicula cryptotenelloides (24.9%), Cymbella sp.1 (21.3%) and 

Achnanthidium minutissimum (20.5%).


Table 1. Twenty nine common species of benthic diatoms in Mekong River and its tributaries. The 

percentage of relative abundance is shown in bracket. 

Taxa 

Nitzschia palea (Kützing) Smith (36.0%) 

Planothidium frequentissimum (Lange-Bertalot) Round 

Mayamaea atomus (Kützing) Lange-Bertalot (35.4%) 

& Bukhtiyarova(11.9%) 

Navicula symmetrica Patrick (33.4%) Cymbella sumatrensis Hustedt (11.7%) 

Eolimna minima (Grunow) Lange-Bertalot (25.3%) Nitzschia filiformis (Smith) Hustedt (11.4%) 

Navicula cryptotenelloides Lange-Bertalot (24.9%) Ulnaria ulna (Nitzsch) Compère (10.8%) 

Gomphonema lagenula Kützing (24.6%) Eolimna subminuscula (Manguin) Gerd Moser (9.8%) 

Cymbella sp.1 (21.3%) Fragilaria bidens Heiberg (9.0%) 

Achnanthidium minutissimum (Kützing) Czarnecki (20.5%) Melosira varians Agardh (9.0%) 

Navicula cryptotenella Lange-Bertalot (19.9%) Navicula menisculus Schumann (6.9%) 

Nitzschia inconspicua Grunow (19.2%) Nitzschia dissipata (Kützing) Grunow (6.6%) 

Nitzschia supralitorea Lange-Bertalot (17.2%) Geissleria decussis (Østrup) Lange-Bertalot&Metzeltin (5.9%) 

Navicula rostellata Kützing (14.5%) Achnanthidium convergens (Kobayasi) Kobayasi (5.9%) 

Encyonema sp.1 (14.3%) Frustulia undosa Metzeltin & Lange-Bertalot (4.7%) 

Luticola goeppertiana (Bleisch) Mann in Round, Sellaphora pupula (Kützing) Mereschkovsky (3.1%) 

Crawford & Mann (13.7%) Nitzschia microcephala Grunow (2.9%) 

Nitzschia clausii Hantzsch (13.1%) 

According to Jüttner et al. [22], M. atomus, A. minutissimum and N. palea were common 

species in streams in agricultural catchments of Kathmandu valley, Nepal. Duong et al.[23] reported 

the impact of urban pollution in Hanoi area on benthic communities collected from Red, Nhue and 

Tolich Rivers in Vietnam. They also reported that the diatom assemblage at the Tolich site 

consisted mainly of Nitzschia umbonata, N. palea and Eolimma minima. Some of these diatoms 

were reported to have preference for tropical regions without being restricted to these latitudes [23- 

24]. Cymbella turgidula was recorded in many tropical countries such as Sri Lanka [25] and Río 

Savegre River in the central and southern parts of Costa Rica [26]. Cymbella tumida was also 

reported to be widespread but very often found in the tropics [11]. Patrick and Reimer [27] reported 

this species from New England. Diploneis subovalis was found in Iceland and Finland, but with 

higher frequency in tropical rivers [11]. Gomphonema parvulum var. lagenula was regarded as a 

tropical species [26]. 

Common species, especially Cymbella turgidula, C. tumida and Gomphonema parvulum 

var. lagenula (currently regarded as a synonym of Gomphonema lagenula) were found in many 

streams and rivers in Thailand [8, 28-38]. 

Water Physico-Chemical Properties 

The water physico-chemical properties of Mekong River and its tributaries between July 

2005 - April 2007 are shown in Table 2. Broad differences were apparent between sampling sites. 

The water temperature ranged between 25-32 o C, the temperatures at GT, KO and HK in the north 

being lower than those in the north-eastern areas. The water velocity was highest (8.30 m/s) at KB 

in the Mekong River. All sampling sites showed neutral pH with the highest (7.67) at GT and the 

lowest (6.75) at SK. Average alkalinity ranged between 23.3-67.8 mg/L. The highest value of 

conductivity was recorded at SK in the north-eastern area. The highest DO and BOD values were 

observed at KO in the north-eastern area, as were the highest NO 3 - -N (1.57 mg/L) and SRP (0.24 

mg/L). The highest NH 4 + -N (0.53 mg/L) was recorded at KP and the lowest (0.21 mg/L) at GT.


33 

Figure 3. Common species of benthic diatoms in Mekong River and its tributaries 

(scale bar = 10 m) 

(1) Nitzschia filiformis (W. Smith) Hustedt, (2) Nitzschia palea (Kützing) W. Smith, 

(3-4) Nitzschia inconspicua Grunow, (5-7) Nitzschia supralitorea Lange-Bertalot, 

(8-9) Nitzschia clausii Hantzsch, (10) Nitzschia dissipata (Kützing) Grunow, 

(11) Nitzschia microcephala Grunow, (12-13) Navicula menisculus Schumann, 

(14) Navicula symmetrica R.M. Patrick, (15) Navicula cryptotenelloides Lange- 

Bertalot, (16) Navicula cryptotenella Lange-Bertalot, (17) Navicula rostellata Kützing, 

(18-19) Fragilaria bidens Heiberg, (20) Ulnaria ulna (Nitzsch) P. Compère


Figure 3 (continued). Common species of benthic diatoms in Mekong River and its tributaries 

(scale bar = 10 m) 

(1) Frustulia undosa D. Metzeltin & H. Lange-Bertalot , (2-3) Geissleria decussis 

(Østrup) Lange-Bertalot&Metzeltin, (4-6) Gomphonema lagenula Kützing, 

(7-8) Sellaphora pupula (Kützing) Mereschkovsky, (9-10) Luticola goeppertiana 

(Bleisch) D.G.Mann in Round,Crawford&Mann, (11) Achnanthidium convergens 

(H. Kobayasi) H. Kobayasi, (12-14) Achnanthidium minutissimum (Kützing) Czarnecki, 

(15-16) Eolimna minima (Grunow) Lange-Bertalot, (17-18) Eolimna subminuscula 

(Manguin) Gerd Moser, (19-20) Mayamaea atomus (Kützing) H. Lange-Bertalot, 

(21-22) Cymbella sumatrensis Hustedt, (23) Cymbella sp.1, (24) Encyonema sp.1, 

(25-26) Planothidium frequentissimum (Lange-Bertalot) Round & L. Bukhtiyarova, 

(27) Melosira varians C. Agardh


35 

The water quality in the Mekong River and its tributaries was classified in the fourth 

category according to the standards for surface water quality of Thailand certified by the Pollution 

Control Department [39], and can be used for household consumption after a disinfection process 

and special water treatment. 

Table 2. Physicochemical properties of water at 14 sampling sites (average values and min-max 

values, n=14) 

Sampling 

Site 

Temperature 

( o C) 

Velocity 

(m/s) 

pH 

Conductivity 

(μS/cm) 

Turbidity 

(NTU) 

Alkalinity 

(mg/L as CaCO 3) 

DO 

(mg/L) 

BOD 

(mg/L) 

NO 3 - -N 

(mg/L) 

NH 4 + -N 

(mg/L) 

SRP 

(mg/L) 

GT 

25.6 

(18.1-30.8) 

5.4 

(1.8-9.2) 

7.7 

(7.3-8.0) 

218.8 

(153.0-312.0) 

137.1 

(54.0-301.0) 

55.70 

(9.90-103.00) 

8.00 

(5.40-10.00) 

2.60 

(1.80-4.10) 

0.78 

(0.10-1.90) 

0.21 

(0.04-0.47) 

0.14 

(0.06-0.55) 

KO 

26.0 

(20.5-29.2) 

5.8 

(4.0-7.8) 

7.2 

(6.9-7.8) 

138.5 

(78.9-198.2) 

199.1 

(57.0-305.0) 

41.05 

(6.90-70.20) 

8.20 

(5.20-11.00) 

3.10 

(2.00-5.00) 

1.57 

(0.80-3.00) 

0.37 

(0.29-0.46) 

0.24 

(0.10-1.19) 

HK 

25.3 

(18.5-29.8) 

7.8 

(3.0-12.3) 

7.6 

(7.2-8.1) 

221.0 

(162.0-270.0) 

212.6 

(48.0-419.0) 

60.80 

(17.60-97.00) 

7.80 

(5.20-9.80) 

1.50 

(0.20-3.10) 

1.55 

(0.20-3.40) 

0.31 

(0.13-0.67) 

0.15 

(0.02-0.67) 

HG 

30.2 

(23.3-35.8) 

6.7 

(0.5-14.3) 

7.6 

(7.2-8.1) 

180.4 

(64.0-389.0) 

82.5 

(3.0-438.0) 

57.80 

(19.00-101.00) 

7.00 

(4.20-8.40) 

2.70 

(0.40-6.40) 

0.88 

(0.00-1.60) 

0.42 

(0.15-0.72) 

0.22 

(0.10-0.61) 

KK 

28.9 

(23.9-32.8) 

3.5 

(0.1-5.6) 

7.6 

(7.0-8.7) 

192.8 

(49.0-325.0) 

205.2 

(105.0-453.0) 

67.80 

(35.00-104.00) 

5.90 

(4.80-8.60) 

2.30 

(0.00-5.60) 

1.20 

(0.20-2.00) 

0.28 

(0.14-0.43) 

0.19 

(0.14-0.29) 

PS 

30.0 

(26.0-33.0) 

2.74 

(0.0-6.2) 

7.4 

(7.0-7.8) 

249.7 

(147.0-349.0) 

183.4 

(96.0-367.0) 

65.80 

(25.00-107.00) 

5.60 

(4.40-8.20) 

1.20 

(0.20-4.40) 

1.12 

(0.60-1.80) 

0.38 

(0.16-0.58) 

0.19 

(0.07-0.26) 

LG 

31.5 

(26.5-34.1) 

1.3 

(0.0-6.3) 

7.2 

(6.4-8.1) 

292.0 

(146.0-401.0) 

53.3 

(12.0-125.0) 

36.00 

(8.00-85.00) 

4.40 

(2.60-5.80) 

1.30 

(0.10-3.00) 

0.79 

(0.50-1.10) 

0.41 

(0.20-0.74) 

0.20 

(0.07-0.40) 

NP 

30.6 

(25.1-32.9) 

1.4 

(0.0-4.4) 

7.4 

(6.7-7.8) 

195.7 

(91.0-263.0) 

127.4 

(28.0-266.0) 

61.20 

(20.50-95.00) 

5.70 

(4.00-8.20) 

1.70 

(0.20-4.90) 

0.76 

(0.00-1.30) 

0.36 

(0.17-0.55) 

0.13 

(0.00-0.33) 

SK 

31.4 

(27.0-34.8) 

3.7 

(0.1-10.4) 

6.8 

(6.0-7.5) 

341.1 

(127.0-759.0) 

70.1 

(10.0-288.0) 

23.30 

(4.10-45.00) 

4.60 

(2.20-7.00) 

1.60 

(0.00-3.60) 

0.99 

(0.20-1.80) 

0.40 

(0.18-0.73) 

0.09 

(0.01-0.20) 

KB 

30.4 

(24.5-32.6) 

8.3 

(2.0-16.5) 

7.5 

(6.8-8.0) 

193.2 

(133.5-255.0) 

134.1 

(23.0-271.0) 

51.40 

(17.00-92.00) 

6.00 

(4.60-8.40) 

1.90 

(0.30-4.60) 

0.68 

(0.10-1.90) 

0.30 

(0.14-0.54) 

0.15 

(0.04-0.29) 

HW 

31.0 

(27.1-34.8) 

6.9 

(0.1-14.9) 

7.4 

(6.8-7.9) 

166.3 

(66.0-255.0) 

136.0 

(16.0-303.0) 

49.20 

(18.50-87.00) 

6.10 

(4.60-8.60) 

1.10 

(0.00-4.90) 

0.55 

(0.10-1.20) 

0.31 

(0.10-0.60) 

0.15 

(0.00-0.37) 

KH 

30.3 

(22.4-36.4) 

2.7 

(0.0-7.1) 

7.5 

(6.8-7.8) 

134.9 

(43.0-250.0) 

144.1 

(22.0-325.0) 

52.50 

(20.00-90.00) 

5.89 

(4.00-8.40) 

1.40 

(0.10-4.80) 

0.55 

(0.10-1.10) 

0.31 

(0.16-0.61) 

0.17 

(0.00-0.70) 

KP 

32.1 

(28.0-36.0) 

5.2 

(0.1-8.8) 

6.9 

(6.2-7.9) 

175.7 

(80.0-265.0) 

72.6 

(13.0-119.0) 

43.90 

(10.00-73.20) 

4.86 

(3.60-7.20) 

1.44 

(0.00-5.20) 

0.65 

(0.00-1.90) 

0.53 

(0.17-0.85) 

0.10 

(0.02-0.30) 

KJ 

31.5 

(27.0-34.8) 

1.7 

(0.0-4.7) 

7.9 

(6.6-8.2) 

141.8 

(44.0-243.0) 

119.9 

(14.0-260.0) 

59.80 

(21.00-85.00) 

6.29 

(4.60-8.50) 

2.36 

(0.60-5.10) 

0.68 

(0.00-1.30) 

0.40 

(0.11-1.07) 

0.14 

(0.01-0.34) 

Correlation between Physico-Chemical Variables 

There were significant positive and negative correlations between some of the physicochemical 

variables of water as shown in Table 3. Significant positive correlation between SRP and 

NO 3 - -N was observed (P


correlation with NH 4 + -N (P


37 

Figure 4. CCA of the relationship between water quality and common species of diatom (% of 

relative abundance > 1) 

There are many environmental conditions which influence the growth of algae. In a lotic 

ecosystem, almost all algae live in benthic forms [40] and they usually grow on any surface of the 

substratum. Each type of substratum, either rock, mud, sand or silt, affects the benthic behaviour 

[41]. The substratum type is related to the current velocity and water volume [33]. The substratum 

characteristic has a direct impact on the distribution of benthic algae [42]. In this study, the lowest 

amounts of diatoms at all sampling times were recorded at HK sampling site (Figure 2). This 

observation was probably due to a high level of water in the wet season and the substrata along the 

bank being soft sediment of sand and silt, which is not suitable for diatom growth [43]. 

Furthermore, blooms of macroalgae, namely Cladophora spp. and Microspora spp., appeared in 

cool dry season and there were fewer suitable substrata for the attachment of benthic diatoms. 

Mpawenayo and Mathooko [44] published the structure of the diatom assemblage associated with 

Cladophora macroalgae and sediment in a highland stream of Kenya. The dominant species of 

diatoms were Nitzschia amphibia and Gomphonema parvulum. Similary, G. parvulum was found at 

HK where there was a full bloom of Cladophora spp. Jüttner et al. [45] reported that G. parvulum in 

particular preferred vegetation as its habitat. However, G. parvulum was found equally often on 

sediment because of transportation and contamination. Persistent stability and protection of diatoms 

attached on Cladophora could explain the existence of higher diatom species richness on 

Cladophora than in a more unstable sediment habitat.


In fast-flowing water, there are diatoms such as Achnanthes and Cocconeis that can attach 

on rocks and other hard surfaces [24]. Similarly, in this study, Cocconeis spp. were found as 

dominant species at HG during summer 2007 with high water velocity. In slower flowing water, 

Melosira varians and various species of Synedra, Gomphonema and Cymbella are found on hard 

substrata [24]. Throughout the sampling period, Gomphonema spp. were also observed at LG 

located about 100 metres downstream from a small dam, where the flow rate of water was 

extremely low. 

Species of Benthic Diatoms in Relation to Physico-Chemical Properties of Water by Cluster 

Analysis 

Twenty nine common diatom species were arranged in groups of sampling sites detected by 

cluster analysis of physico-chemical parameters of water quality at 80% similarity. The number of 

benthic diatoms in each group was calculated by Minitab program to find the significant difference 

and significant correlation between groups. The relationship between diatom community 

composition and water quality variables was determined using statistical techniques. Each sampling 

site had distinct water physico-chemical properties. At 80% similarity, the dendrogram divided 

sampling sites into eleven distinct clusters of characteristic water quality types: groups A to I 2 

(Figure 5 and Table 4). It was found that all sampling sites in groups A (n=3) and B (n=9) and some 

in group C (n=3) were those during December 2006 (cool dry season). Group E, the biggest group 

(n=54), was composed of sampling sites during May 2006 (summer) and some sites in two cool dry 

seasons. 

The values of physico-chemical parameters in each group were calculated by Minitab 

program to find significant differences and correlations between groups A to I 2 as shown in Table 5. 

It was found that groups A, C, G and H showed a non-significant correlation with NO 3 - -N 

concentration. Group F 1 showed higher concentrations of NO 3 - -N (average of 1.33 mg/L) than 

groups B, D, E, F 2 and I 1 (P


39 

UPGMA 

UPGMA 

40 50 60 70 80 90 100 

Per cent. Similarity 

Percent Similarity 

SK 5 

SK 5 

SK 5 

KJ 5 

KJ 5 

KJ 5 

KH5 

KH5 

KH5 

KK 5 

KK 5 

KK 5 

HG5 

HG5 

HG5 

SK 4 

SK 4 

SK 4 

LG4 

LG4 

LG4 

KP 5 

KP 5 

KP 5 

HG4 

HG4 

HG4 

KP 4 

KP 4 

KP 4 

KP 1 

KB 3 

KB 3 

KH3 

KH3 

KH3 

KB 3 

HG3 

HG3 

HG3 

KO2 

NP 5 

NP 5 

NP 5 

KP 3 

KP 3 

HK 3 

HK 3 

HK 3 

KJ 2 

KH2 

HW 2 

KB 2 

KK 2 

NP 2 

PS 2 

KK 3 

KO3 

KO3 

KO3 

KP 2 

KB 5 

KB 5 

KB 5 

SK 3 

SK 3 

SK 3 

KP 3 

NP 3 

NP 3 

LG2 

HW 5 

HW 5 

HW 5 

KJ 3 

KJ 3 

KJ 3 

NP 3 

GT 3 

GT 3 

GT 3 

LG3 

LG3 

LG3 

LG6 

LG6 

LG6 

HG6 

HG6 

HG6 

PS 5 

PS 5 

PS 5 

KK 3 

HG2 

PS 3 

PS 3 

PS 3 

KK 3 

HK 6 

HK 6 

HK 6 

PS 6 

PS 6 

PS 6 

KK 6 

KK 6 

KK 6 

GT 6 

GT 6 

GT 6 

SK 1 

LG1 

LG5 

LG5 

LG5 

KP 6 

KP 6 

KP 6 

SK 6 

SK 6 

SK 6 

NP 6 

NP 6 

KJ 6 

KJ 6 

KJ 6 

HW 6 

HW 6 

HW 6 

KH6 

KH6 

KH6 

KB 6 

KB 6 

KB 6 

NP 6 

HK 2 

GT 5 

GT 5 

GT 5 

GT 2 

SK 2 

KO5 

KO5 

KO5 

KO1 

KH4 

KH4 

KH4 

HW 4 

HW 4 

KJ 4 

HW 4 

NP 4 

NP 4 

KJ 4 

KJ 4 

NP 4 

KO4 

KO4 

KO4 

PS 1 

HG1 

KK 1 

HK 1 

KK 4 

KK 4 

KK 4 

PS 4 

PS 4 

PS 4 

HK 4 

HK 4 

HK 4 

HW 1 

KB 4 

KB 4 

KB 4 

GT 4 

GT 4 

GT 4 

KH1 

KB 1 

KJ 1 

NP 1 

HK 5 

HK 5 

HK 5 

KO6 

KO6 

KO6 

GT 1 

I2 

I1 

H G F2 

F1 

E 

D 

C 

B 

A 

Figure 5. Dendrogram of similarities between investigated sites according to physico-chemical 

parameters of water; 252 cases, 11 variables


Table 4. Groups of sampling sites detected by cluster analysis of physico-chemical parameters of 

water quality at 80% similarity 

Group Sampling site Description 

A SK5.1, SK5.2, SK5.3 Tributaries in north-eastern Thailand 

in cool dry season (second year) 

B KK5.1, KK5.2, KK5.3, KH5.1, KH5.2, KH5.3, KJ5.1, KJ5.2, 

KJ5.3 

Mekong River in north-eastern Thailand in cool 

dry season (second year) 

C HG5.1, HG5.2, HG5.3 Tributaries in north-eastern Thailand 

in cool dry season (second year) 

D HG4.1, HG4.2, HG4.3, LG4.1, LG4.2, LG4.3, SK4.1, SK4.2, 

SK4.3, KP5.1, KP5.2, KP5.3 

Tributaries in north-eastern Thailand 

in summer (second year) and cool dry season 

(second year) 

E GT3.1, GT3.2, GT3.3, KO2, KO3.1, KO3.2, KO3.3, HK3.1, 

HK3.2, HK3.3, HG3.1, HG3.2, HG3.3,KK2, KK3.2, PS2, LG2, 

Mekong River and its tributaries in summer (first 

year) and some of two cool dry seasons 

NP2, NP3.1, NP3.2,NP3.3, NP5.1, NP5.2, NP5.3, SK3.1, SK3.2, 

SK3.3, KB2, KB3.1, KB3.2, KB3.3, KB5.1, KB5.2, KB5.3, 

HW2, HW5.1, HW5.2, HW5.3, KH2, KH3.1, KH3.2, KH3.3, 

KP1, KP2, KP3.1, KP3.2, KP3.3, KP4.1, KP4.2, KP4.3, KJ2, 

KJ3.1, KJ3.2, KJ3.3 

F 1 GT6.1,GT6.2,GT6.3, HK6.1,HK6.2,HK6.3,HG2, HG6.1, HG6.2, 

HG6.3, KK3.1, KK3.3, K6.1, KK6.2, KK6.3, PS3.1, PS3.2, 

PS3.3,PS5.1,PS5.2,PS5.3, PS6.1, PS6.2, PS6.3,LG3.1, LG3.2, 

LG3.3,LG6.1, LG6.2, LG6.3 

Mekong River and its tributaries in two summers 

F 2 GT2, GT5.1, GT5.2, GT5.3, HK2, LG1, LG5.1,LG5.2, LG5.3, 

NP6.1, NP6.2, NP6.3, SK1, SK6.1, SK6.2, SK6.3, KB6.1, 

KB6.2, KB6.3, HW6.1, HW6.2, HW6.3, KH6.1, KH6.2, KH6.3, 

Mekong River and its tributaries in summer 

(second year) and some of cool dry season (second 

year) 

KP6.1, KP6.2, KP6.3, KJ6.1, KJ6.2,KJ6.3 

G SK2 Tributaries in north-eastern Thailand 

in cool dry season (first year) 

H KO1, KO5.1, KO5.2, KO5.3 Tributaries in northern Thailand 

in rainy season (first year) and cool dry season 

(second year) 

I 1 KO4.1, KO4.2, KO4.3, NP4.1, NP4.2, NP4.3,HW4.1, HW4.2, 

HW4.3, KH4.1, KH4.2, KH4.3, KJ4.1, KJ4.2, KJ4.3 

Mekong River and its tributaries in rainy season 

(second year) 

I 2 GT1, GT4.1, GT4.2, GT4.3, KO6.1, KO6.2, KO6.3, HK1, 

HK4.1, HK4.2, HK4.3, HK5.1, HK5.2, HK5.3, HG1, KK1, 

KK4.1, KK4.2, KK4.3, PS1, PS4.1, PS4.2, PS4.3, NP1,KB1, 

KB4.1, KB4.2, KB4.3, HW1, KH1, KJ1 

Mekong River and its tributaries in two rainy 

seasons and some of cool dry season (second year) 

and summer (second year) 

There were significant differences (P


41 

turbidity (average of 4.33 NTU) was recorded for group C while the highest (average of 314.16 

NTU) was recorded for group I 2 . 

_ 

Table 5. Values of average (X) and standard error (se) of physico-chemical parameters (P


Table 5. (continued) 

X = 0.82 

Parameter F 1 F 2 G H I 1 I 2 

NO — 3 N X = 1.33 

X = 1.80 

X =1.38 

X =0.68 

X =1.04 

cor F 1>B,D,E,F 2,I 1 cor F 2 < F 1 

cor ns 

cor ns 

cor I 1 < F 1 

cor I 2>D 

(mg/L) 

se = 0.08 

n = 30 

se = 0.08 

n = 31 

se = 0 

n = 1 

se = 0.54 

n=4 

se = 0.23 

n=15 

se=0.15 

n=31 

NH + 4 -N 

(mg/L) 

X = 0.32 

se = 0.03 

n = 30 

cor F 1=F 2


43 

CONCLUSIONS 

The number of benthic diatom species found (252 species from 53 genera) in Mekong River 

and its tributaries in north and north-eastern Thailand is similar to those reported for other river 

systems of comparable size. Determination of the relationship between diatom community 

composition and water quality variables shows that many species, notably Achnanthidium 

minutissimum, A. convergens, Navicula menisculus, Nitzschia clausii, Luticola goeppertiana, 

Eolimna minima, Ulnaria ulna and Mayamaea atomus, can act as indicators of some of the riverwater 

qualities, viz. conductivity, alkalinity, NH 4 + -N, NO 3 

- 

-N, soluble reactive phosphorus, and 

BOD. 


The authors would like to thank Thailand Research Fund (TRF) for its support in the form of 

a research grant (Grant No. PHD/0129/2547) through the Royal Golden Jubilee PhD Program. They 

are also grateful to Dr. Ingrid Jüttner from National Museum Cardiff, U.K. for her suggestions and 

assistance in diatom identification. 

REFERENCES 

1. A. Chovanec, P. Jäger, M. Jungwirth, V. Koller-Kreimel, O. Moog, S. Muhar and S. Schmutz, 

“The Austrian way of assessing the ecological integrity of running waters – a contribution to 

the EU water framework directive”, Hydrobiol., 2000, 422/423, 445-452. 

2. F. E. Round, “Use of diatoms for monitoring rivers”, in “Use of Algae for Monitoring Rivers” 

(Ed. B. A. Whitton, E. Rott and G. Friedrich), Institut fur Botanik, Universitat, Innsbruck, 

1991, pp.25-32. 

3. B. A. Whitton, E. Rott and G. Friedrich, “Use of algae for monitoring rivers”, Proceedings of 

International Symposium on Use of Algae for Monitoring Rivers I, 1991, Düsseldorf, 

Germany, pp.10-18. 

4. E. Rott, E. Pipp and P. Pfister, “Diatom methods developed for river quality assessment in 

Austria and a cross-check against numerical trophic indication methods used in Europe”, Algol. 

Stud., 2003, 110, 91-115. 

5. J. Prygiel, M. Coste and J. Bukowska, “Review of the major diatom-based techniques for the 

quality assessment of river-state of the art in Europe”, Proceedings of International Symposium 

on Use of Algae for Monitoring Rivers III, 1999, Douai, France, pp.224-238. 

6. M. Ohno, H. Fukushima and T. Ko-Bayashi, “Diatom flora of the Mekong water system, 

Cambodia”, Nat. Sci., 1972, 20, 1-11. 

7. Mekong River Commission, “Biomonitoring of the lower Mekong River and selected 

tributaries”, MRC technical paper No.13, Mekong River Comission, Vientiane, Lao PDR, 

2006. 

8. S. Pruethiworanon, “Diversity of macroalgae and benthic diatoms in Mekong River passing 

Thailand and their application for water quality”, MS Thesis, 2008, Chiang Mai University, 

Thailand.


9. I. Renberg, “A procedure for preparing large sets of diatom slides from sediment cores”, J. 

Paleolimnol., 1990, 4, 87-90. 

10. M. G. Kelly, A. Cazaubon, E. Coring, A. Dell’Uomo, L. Ector, B. Goldsmit, H. Guasch, J. 

Hürlimann, A. Jarlman, B. Kaweeka, J. Kwandrans, R. Laugaste, E.- A. Lindstrom, M. Leitao, 

P. Marvan, J. Padisak, E. Pipp, J. Prygiel, E. Rott, S. Sabater, H. van Dam and J. Vizinet, 

“Recommendations for the routine sampling of diatoms for water quality assessments in 

Europe”, J. Appl. Phycol., 1998, 10, 215-224. 

11. K. Krammer and H. Lange-Bertalot, “Bacillariophyceae Teil 1: Naviculaceae, Süsswasserflora 

von Mitteleuropa Band 2/1”, Gustav Fisher Verlag, Stuttgart, 1986. 

12. K. Krammer and H. Lange-Bertalot, “Bacillariophyceae Teil 2: Epithemiaceae Surirellaceae, 

Süsswasserflora von Mitteleuropa Band 2/2”, Gustav Fisher Verlag, Stuttgart, 1988. 

13. K. Krammer and H. Lange-Bertalot, “Bacillariophyceae Teil 3: Centrales Fragilariaceae 

Eunotiaceae, Süsswasserflora von Mitteleuropa Band 2/3”, Gustav Fisher Verlag, Stuttgart, 

1991a. 

14. K. Krammer and H. Lange-Bertalot, “Bacillariophyceae Teil 4: Achnanthaceae, 

Süsswasserflora von Mitteleuropa Band 2/4”, Gustav Fisher Verlag, Stuttgart, 1991b. 

15. H. Lange-Bertalot, “Diatoms of Europe”, A.R.G. Gantner Verlag K.G., Ruggell, 2001. 

16. T. Watanabe, T. Ohtsuka, A. Tuji and A. Houki, “Picture Book and Ecology of the Freshwater 

Diatoms”, Uchida-rokakuho, Tokyo, 2005. 

17. H. Lange-Bertalot, “Iconographia Diatomologica: Annotated Diatom Micrographs”, Vol.13, 

A.R.G. Gantner Verlag K.G., Ruggell, 2004. 

18. A. D. Eaton, L. S. Clesceri, E. W. Rice and A. E. Greenberg, “Standard Methods for 

Examination of Water and Wastewater: Centennial Edition”, 21 st Edn., American Public Health 

Association, Washington D. C., 2005. 

19. N. T. H. Holmes and B. A. Whitton, “Phytobenthos of the River Tees and its tributaries”, 

Freshwater Biol., 1981, 11, 139-163. 

20. R. E. M. Archibald, “The diatoms of the Sundays and Great Fish Rivers in the Eastern Cape 

Province of South Africa”, A.R.G. Gantner Verlag K.G., Ruggell, 1983. 

21. B. C. Chessman, “Diatom flora of an Australian river system: Spatial patterns and environment 

relationships”, Freshwater Biol., 1986, 16, 805-819. 

22. I. Jüttner, S. Sharma, B. M. Dahal, S. J. Ormerod, P. J. Chimonides and E. J. Cox, “Diatoms as 

indicators of stream quality in the Kathmandu Valley and Middle Hills of Nepal and India”, 


23. T. T. Duong, M. Coste, A. Feurtet-Mazel, D. K. Dang, C. Gold, Y. S. Park and A. Boudou, 

“Impact of urban pollution from the Hanoi area on benthic diatom communities collected from 

the Red, Nhue and Tolich Rivers (Vietnam)”, Hydrobiol., 2006, 563, 201-216. 

24. R. Patrick, “Ecology of freshwater diatoms and diatom communities”, in “The Biology of 

Diatoms” (Ed. D. Werner), Blackwell Scientific Publications, Oxford, 1977, pp.284-322. 

25. N. Foged, “Freshwater Diatoms in SriLanka”, Odense Publisher, Odense (Denmark), 1976. 

26. A. M. Silva-Benavides, “The epilithic diatom flora of a pristine and a polluted river in Costa 

Rica, Central America”, Diatom Res., 1996, 11, 105-142.


45 

27. R. Patrick and C. W. Reimer, “The Diatoms of the United States”, The Academy of Natural 

Science of Philadelphia, Philadelphia, 1966. 

28. T. Pekthong, “Diversity of phytoplankton and benthic algae in Mae Sa stream, Doi Suthep-Pui 

National Park, altitude 330-550 meters”, MS Thesis, 1998, Chiang Mai University, Thailand. 

29. T. Pekthong, “Fifty one new record species of freshwater diatoms in Thailand”, Chiang Mai J. 

Sci., 2002, 28, 97-112. 

30. P. Waiyaka, “Diversity of phytoplankton and benthic algae in Mae Sa stream, Doi Suthep-Pui 

National Park, altitude 550-1050 meters”, MS Thesis, 1998, Chiang Mai University, Thailand. 

31. T. Kunpradid, “Biodiversity of phytoplankton and macroalgae in Mae Sa stream, Doi Suthep- 

Pui National Park, Chiang Mai province”, MS Thesis, 2000, Chiang Mai University, Thailand. 

32. T. Kunpradid, “Diversity of macroalgae and benthic diatoms and their relationship with 

nutrient compounds in Ping and Nan Rivers”, PhD Thesis, 2005, Chiang Mai University, 

Thailand. 

33. Y. Peerapornpisal, T. Pekthong, P. Waiyaka and S. Promkutkaew, “Diversity of phytoplankton 

and benthic algae in Mae Sa stream, Doi Suthep-Pui National Park, Chiang Mai”, Siam Soc., 

2000, 48, 193-211. 

34. S. Suphan, “Diversity of macroalgae and benthic diatoms in the area of Golden Jubilee Thong 

Pha Phum Project, Thong Pha Phum district, Kanchanaburi province”, MS Thesis, 2004, 


35. T. Inthasotti, “Seasonal distribution of macroalgae and benthic diatoms in Mae Tang River, 

Wiang Haeng district, Chiang Mai province between March 2004-January 2005”, Independent 

Study for Master Degree, 2006, Chiang Mai University, Thailand. 

36. T. Inthasotti, “Diversity of macroalgae and benthic diatoms in Kham watershed, Chiang Rai 

province”, MS Thesis, 2006, Chiang Mai University, Thailand. 

37. P. Leelahakriengkrai, “Usage of macroalgae and benthic diatoms to monitor the water quality 

of Kok River and tributaries”, Independent study for Master Degree, 2006, Chiang Mai 


38. P. Leelahakriengkrai, “Diversity and usage of macroalgae and benthic diatoms to monitor water 

quality of Ping River, 2004-2005”, MS Thesis, 2006, Chiang Mai University, Thailand. 

39. National Environmental Board, “Surface water quality standards” (in Thai), the Royal 

Government Gazette, Vol. 111, Part 16 (February 24, 1994). 

40. H. B. N. Hynes, “The Ecology of Running Waters”, Liverpool University Press, Liverpool, 

1970, pp.53-77. 

41. S. Panha, “A new species of Amphidromus from Thailand (Stylommatophora: Camaenidae)”, 

Malacol. Rev., 1996, 29, 131-132. 

42. V. J. Chapman and D. J. Chapman, “The Algae”, Macmillan, London, 1973, pp.388-395. 

43. F. E. Round, “The Biology of the Algae”, 2 nd Edn., Edward Arnold Ltd., London, 1973, pp.64- 

69. 

44. B. Mpawenayo and J. M. Mathooko, “The structure of diatom assemblages associated with 

Cladophora and sediments in a highland stream in Kenya”, Hydrobiol., 2005, 544, 55-67.


45. I. Jüttner, H. Rothfritz and S. J. Ormerod, “Diatoms as indicators of river quality in the 

Nepalese Middle Hills with consideration of the effects of habitat-specific sampling”, 




Full Paper 

Maejo Maejo Int. J. Sci. Int. Technol. J. Sci. Technol. 2012, 6(01), 2012, 47-61 6(01), 47-61 





Fourteen new records of cercosporoids from Thailand 

Pheng Phengsintham 1,2 , Ekachai Chukeatirote 1 , Eric H. C. McKenzie 3 , Mohamed A. Moslem 4 , 

Kevin D. Hyde 1, 4, * and Uwe Braun 5 

1 School of Science, Mae Fah Luang University, Chiang Rai 57100, Thailand 

2 Department of Biology, Faculty of Sciences, National University of Laos, Laos 

3 

Landcare Research, Private Bag 92170, Auckland, New Zealand 

4 King Saud University, College of Science, Department of Botany and Microbiology, P.O. Box: 

2455, Riyadh 1145, Saudi Arabia 

5 Martin-Luther-Universität, Institut für Biologie, Bereich Geobotanik und Botanischer Garten, 

Herbarium, Neuwerk 21 D-06099 Halle/S, Germany 

* Corresponding author: kdhyde3@gmail.com 

Received: 24 July 2010 / Accepted: 27 January 2012 / Published: 1 February 2012 

Abstract: Comprehensive examination of cercosporoid leaf-spotting hyphomycetes was carried out 

in northern Thailand. Fourteen species assigned to the genera Cercospora (5), Passalora (3), 

Pseudocercospora (5) and Zasmidium (1) are new records for Thailand. Cercospora verniciferae 

and Zasmidium cassiicola are poorly known species and are fully described. 

Keywords: 

records 

anamorphic fungi, cercosporoid hyphomycetes, South-East Asia, taxonomy, new 

INTRODUCTION 

Cercospora sensu lato is one of the largest genera of hyphomycetes and is almost 

cosmopolitan in distribution. It causes leaf-spots and other lesions on a wide range of host plants. 

Species of this genus are important pathogens responsible for severe damage to beneficial plants such 

as maize, rice, grasses, vegetables, forest trees and ornamentals [1-3]. 

There have been several recent comprehensive accounts of the fungi of Thailand which are 

among the best documented in the region [4, and references therein]. In Thailand, the study of 

Cercospora and allied genera can be traced back to 1980 [5, 6]. Sixty cercosporoids were listed, 

including 13 unidentified species of Cercospora in the host index of plant diseases in Thailand [5],


whereas 21 species of Cercospora were specified as plant pathogens [6]. In 1989, 49 cercosporoid 

species were further identified in southern Thailand [7]. These reports, however, were based on the 

old generic characters, i.e. Cercospora sensu lato. Subsequently, 112 species of Cercospora as well 

as their synonyms were recorded in ‘The host index of plant diseases in Thailand’ [8]. It should also 

be noted that, according to the list, species names used were ambiguous since the criteria used for 

classification were based on both old and new criteria. In 2007, three new species of Cercospora 

were discovered; these included 11 species that were new to Thailand [9]. Forty-three cercosporoid 

species were included in an annotated list of cercosporoid fungi in northern Thailand [10], and two 

taxa associated with necrotic leaflets of an areca palm (Areca cathecu) were reported [11]. A PhD 

thesis on “Diversity and phylogeny of true cercosporoids fungi from northern Thailand” available in 

2009 encompassed 166 cercosporoids [12]. Recently, many new Cercospora species have been 

discovered, including Cercospora cristellae, a new cercosporoid species associated with the weed 

Cristella parasitica from northern Thailand [13], and three new records of cercosporoid fungi 

(Cercospora diplaziicola, Pseudocercospora duabangae and Pseudocercospora trematicola) from 

Thailand [14]. 

The genus Cercospora Fresen. s. lat., which is one of the largest genera of hyphomycetes, 

has been monographed with over 3000 names [15]. Similar to other fungal group, the classification 

and identification of Cercosporoid fungi are mainly based on morphological characteristics. 

Currently, an identification key of Cercosporoid proposed in 2003 has been widely accepted [16]. In 

this present study, we explored the diversity of Cercospora and allied genera by using this 

identification key. The objectives of this paper are to investigate the cercosporoid fungi of Thailand 

and Laos, and to provide data on Thailand’s fungi in comparison with the diversity of these fungal 

groups in neighbouring countries. 


Sample Collections and Examination of Fungal Structures 

Plants leaves with leaf spots or other lesions were collected during field trips in northern 

Thailand. Photographs of symptoms, including fungal colonies or fruiting bodies, were taken. 

Macroscopic characters were observed using a stereomicroscope to check lesions/leaf spots 

(shape, size, colour, margin) and colonies/caespituli (with details, i.e. amphigenous/epiphyllous, 

punctiform/pustulate/inconspicuous, effuse, loose, dense, brown/blackish, and others.) 

Microscopic examination, measurement, description and presentation of drawings followed 

the standard procedures [16,17]. In the illustrations, thin-walled structures were depicted by a single 

line, thick-walled ones by double lines, and stippling was used to accentuate shape and pigmentation. 

Identification of Fungi 

The species of cercosporoid hyphomycetes were determined using the key available in the 

current taxonomic publications cited in the list of references.


49 

Dried specimens were prepared and stored at the herbarium of the School of Science, Mae 

Fah Luang University (MFU). Duplicates were preserved at the herbarium of the Institute of 

Biology, Geobotany and Botanical Garden, Halle (Saale), Germany (HAL). 


Fourteen cercosporoid hyphomycetes were identified and assigned to species of the genera 

Cercospora (5), Passalora (3), Pseudocercospora (5) and Zasmidium (1). The cercosporoid species 

and their habitats are listed in Table 1. 

Table 1. Cercosporoid species examined in this study 

Fungal species F DD G U 

Cercospora malloti × 

Cercospora senecionicola × 

Cercospora sidicola × 

Cercospora verniciferae × 

Cercospora zizphigena × 

Passalora broussonetiae × 

Passalora fusimasculans × × 

Passalora graminis × 

Pseudocercospora cycleae × 

Pseudocercospora malloticola × 

Pseudocercospora olacicola × 

Pseudocercospora paederiae × 

Pseudocercospora polysciatis × 

Zasmidium cassiicola × 

Note: F = fallow land; DD = dry dipterocarp forest; G = garden; U = urban area 

Cercospora malloti [18] 

Notes: The collection no. MFLU10-0310 from Tadsak waterfall, Ching Rai province agrees 

well with Cercospora malloti as previously circumscribed [15,16] (conidiophores 10–50 × 3–5 μm 

and conidia 40–75 × 1.5–3 μm). C. malloti is part of the C. apii Fesen. complex from which it is 

morphologically barely distinguishable [16]. The collection from Thailand has conidiophores which 

are 32–140 × 5–6 μm, and conidia which are 20–146 × 2–4 μm. 

Known hosts: Mallotus apelta (Lour.) Müll. Arg., M. japonicus (L. f.) Müll. Arg., and M. 

repandus (Rotter) Müll. Arg. (Euphorbiaceae) [15,16]. 

Known distribution: Asia - China, Japan [15,16], Thailand (this paper); North America - 

USA (MS) [15,16].


Material examined: Phengsintham (MFLU10-0310) on leaf of Mallotus repandus 

(Euphorbiaceae) (Thailand: Chiang Rai province, Wiang Chiang Rung district, Tadsak waterfall), 23 

December 2009. 

Cercospora senecionicola [19] 

Notes: The collection no. MFLU10-0318 from Sri Pangsang village, Chiang Rai province 

agrees with Cercospora senecionicola as circumscribed in Chupp [15]. Brief description of the 

collection from Thailand: Leaf spots/Lesions circular to slightly irregular, 2–3 mm diam., at first 

dark green, later becoming brown to dark brown in the centre, dark brown margin. 

Caespituli/Colonies amphigenous, conspicuous, scattered, dark brown. Conidiophores single or 

fasciculate, arising from stromata (1–8 per fascicle), 0–5-geniculate, cylindrical, straight to curved, 

67–170 × 5–6 μm, 0–8-septate. Conidia acicular to obclavate, straight to curved, 17–82 × 4–7 μm, 

0–8-sepate, slightly constricted at the septa, hyaline to subhyaline, smooth, wall 0.3–0.5 μm thick, 

apex acute, based truncate to subtruncate, hila 2–3 μm wide, wall of the hila 0.5 μm wide, darkened. 

Known hosts: Senecio aureus L., S. aureus var. balamitea Toir. & Gray, and S. walkeri Arn. 

(Compositae = Asteraceae) [15,16]. 

Known distribution: Asia - China, Laos, Thailand [15]. 

Material examined: 1) Phengsintham (MFLU10-0318) on leaf of Senecio walkeri 

(Asteraceae) (Thailand: Chiang Rai province, Muang district, Sri Pangsang village), 11 August 2009; 

2) Phengsintham (P567) on leaf of Senecio walkeri (Asteraceae) (Laos: Bokeo province, Phimonsine 

village), 20 February 2010. 

Cercospora sidicola [20] 

Notes: The collection no. MFLU10-0312 from Tadsak waterfall, Chiang Rai province agrees 

with Cercospora sidicola as circumscribed by Chupp [15], but differs in formation of leaf spots. 

Known hosts: Sida acuta Burm. F., S. cordifolia (DC.) Fryxell , S. mysorensis Wight & Arn., 

S. rhombifolia L., S. spinosa L., and Sida sp. (Malvaceae) [15,16]. 

Known distribution: Asia - China, India, Thailand (this paper); North America - Cuba, 

Dominican Republic, Panama, Puerto Rico, USA (FL, LA, TX), Virgin Islands; South America - 

Argentina, Brazil [16]. 

Material examined: Phengsintham (MFLU10-0312) on leaf of Sida mysorensis (Malvaceae) 

(Thailand: Chiang Rai province, Wiang Chiang Rung district, Tadsak waterfall), 23 December 2009. 

Cercospora verniciferae [21] (Figure 1; Redescribed because this species is poorly known.) 

Leaf spots/Lesions small to medium, suborbicular to irregular, 1–4 mm in diam., brown in 

the centre, and with brown-yellow margin. Caespituli/Colonies hypophyllous, scattered, dark 

brown. Mycelium internal; hyphae branched, 3–5 μm wide (x = 3.57 μm, n = 7), septate, 

constricted at the septa, distance between septa 7–12 μm ( x = 8.42 μm, n = 7), brownish or greenhyaline, 

wall 0.5–0.8 μm wide (x = 0.54 μm, n = 7), smooth, forming plate-like plectenchymatous 

stromatic hyphal aggregations. Stromata developed, small to medium-sized, globular to subglobular, 

substomatal and intraepidermal, 16–33 μm in diam. (x = 23.6 μm, n = 8), dark brown to black in


51 

mass, composed of swollen hyphal cells, subglobose, rounded to angular in outline, 5–10 μm wide 

(x = 7.9 μm, n = 13), brown to dark brown, wall 0.5–0.8 μm wide (x = 0.68 μm, n = 13), smooth. 

Conidiophores fasciculate, arising from stromata (1–4 per fascicle), emerging through stomata, not 

branched, straight to curved, cylindrical, 45–89 × 5–7 μm ( x = 71.2 × 5.4 μm, n = 5), 2–5-septate, 

distance between septa 5–30 μm ( x = 16.2 μm, n = 16), medium brown, paler at the apex, wall 0.5– 

0.8 μm wide (x = 0.63 μm, n = 16), smooth, 0–2-times geniculate. Conidiogenous cells integrated, 

terminal, cylindrical, 16–30 × 4–5 μm (x = 24.5 × 4.5 μm, n = 4), pale brown; conidiogenous loci 

conspicuous, subcircular, 2–2.5 μm wide (x = 2.12 μm, n = 4), wall 0.5–0.8 μm thick (x = 0.57 

μm, n = 4), thickened and darkened. Conidia solitary, acicular to obclavate, straight to curved, 23– 

105 × 2–4 μm ( x = 64 × 2.8 μm, n = 5), 5–12-septate, hyaline to subhyaline, thin-walled 0.3–0.5 

μm (x = 0.31 μm, n = 5), smooth; tip acute; base truncate, hila thickened and darkened 1–1.5 μm 

wide (x = 1.1 μm, n = 5), wall of the hila 0.3–0.35 μm (x = 0.31 μm, n = 5) thick. 

Known hosts: Rhus vernicifera DC., Spondias dulcis Parkinson, and S. pinnata (L. F.) Kurz 

(Anacardiaceae) [15,16]. 

Known distribution: Asia - Thailand (this paper); Oceania - American Samoa; South 

America - Brazil [15,16]. 

Material examined: Phengsintham (MFLU10-0313) on leaf of Spondias pinnata 

(Anacardiaceae) (Thailand: Chiang Rai province, Muang district, Sri Pangsang village), 22 December 

2009. 

Notes: The collection from Thailand agrees well with Cercospora verniciferae as 

circumscribed by Chupp [15]. C. verniciferae has conidiophores that are 45–89 × 5–7 μm and 

conidia that are 23–105 × 2–4 μm. C. verniciferae is part of the C. apii Fesen. complex from which 

it is morphologically barely distinguishable [16]. 

Cercospora ziziphigena [22] 

Notes: The collection no. MFLU11-0019 from Mae Puem National Park, Pha Yao province 

(conidiophores 40–320 × 4–6 μm and conidia 163–195 × 2.5–3 μm) having a long conidia, differs 

from the Cercospora ziziphigena previously described [22] (conidiophores 13.8–92.5 × 3.5–6.3 μm 

and conidia 17.5–76.8 × 3.1–5.3 μm). A true Cercospora s. str. is quite distinct from Cercospora 

apii s. lat. [16]. 

Known hosts: Ziziphus incurva Roxb. and Ziziphus sp. (Rhamnaceae) [16, 22]. 

Known distribution: Asia - China [16, 22], Thailand (this paper). 

Material examined: Phengsintham (MFLU11-0019) on leaf of Ziziphus sp. (Rhamnaceae) 

(Thailand: Pha Yao province, Mae Jai district, Mae Puem National Park), 22 August 2010.


Figure 1(a). Cercospora verniciferae on Spondias pinnata from leaf spots: 1. Stroma with attached 

conidiophores; 2. Conidiophore; 3–7. Conidia. (Scale bar = 10 μm)


53 

Figure 1(b). Cercospora verniciferae on Spondias pinnata from leaf spots: 1–2. Lesions on host 

leaves (1. upper surface and 2. lower surface); 3. Caespituli; 4. Internal mycelium; 5. Stroma with 

attached conidiophores; 6. Stroma; 7–9. Conidia. (Scale bar: 1–2. = 10 mm; 3. = 3 mm; 4–9. = 10 

μm)


Passalora broussonetiae [16] 

Notes: The collection no. MFLU10-0314 from Tadsak waterfall, Chiang Rai province agrees 

well with Passalora broussonetiae [16], but the hyphae are smooth to distinctly verruculose 

(described to be smooth by Hsieh and Goh [3]). Brief description of Passalora broussonetiae from 

Thailand: Leaf spots/Lesions irregular, 1–9 mm diam., at first reddish brown, later becoming dark 

brown in the centre, gray to reddish brown margin. Caespituli/Colonies amphigenous, conspicuous. 

Conidiophores 170–390 × 2–5 μm, 5–17-septate. Conidia 6–28 × 4–6 μm, 0–3-septate. 

Known host: Broussonetia papyrifera (L.) L'Hér. ex Vent.[3, 23]. 

Known distribution: Asia - Taiwan [3,16], Thailand (this paper). 

Material examined: Phengsintham (MFLU10-0314) on leaf of Broussonetia papyrifera 

(Moraceae) (Thailand: Chiang Rai province, Wiang Chiang Rung district, Tadsak waterfall), 23 


Passalora fusimaculans [16] Cercospora fusimaculans [24] 

Notes: The collection no. MFLU10-0315 from Sri Pangsang village garden and no. 

MFLU10-0316 from Huay Kang Pah National Park, Chiang Rai province agree well with Passalora 

fusimaculans as circumscribed previously [3, 15, 25], but differ in having rather long conidiophores 

(up to 130 μm) and shorter conidia (up to 85 μm). 

Known hosts: Agrostis, Brachiaria, Brachia, Beckeropsis, Chasmopodium, Digitaria, 

Echinochloa, Eleusine, Entolasia, Ichnanthus, Leptoloma, Oplismenus, Panicum, Paspalidium, 

Pennisetum, Rottboellia, Setaria, Sorghum, Stenotaphrum, Urochloa, and Zea (Poaceae) [16]. 

Known distribution: Africa - Botswana, Ethiopia, Ghana, Guinea, Ivory Coast, Kenya, 

Malawi, Nigeria, Rwanda, Sierra Leone, South Africa, Sudan, Tanzania, Togo, Uganda, Zambia, 

Zimbabwe; Asia - Brunei, China, India, Japan, Malaysia, Philippines, Taiwan, Thailand (this paper); 

Europe - Azerbaijan, France, Georgia; North America - Costa Rica, Cuba, Dominican Rep., El 

Salvador, Guatemala, Honduras, Jamaica, Mexico, Nicaragua, Panama, Trinidad & Tobago, USA 

(AL, PL, IA, ID, KS, NC, ND, OK, OR, TX, VA, WI); Oceania - Australia, Fiji, New Zealand, 

Palau, Papua New Guinea, Samoa, Solomon Islands, Vanuatu; South America - Bolivia, Brazil, 

Colombia, Ecuador, Guyana, Peru, Venezuela [16]. 

Material examined: 1) Phengsintham (MFLU10-0315) on leaf of Echinochloa esculenta 

(Poaceae) (Thailand: Chiang Rai province, Muang district, Sri Pangsang village), 15 September 

2009; 2) Phengsintham (MFLU10-0316) on leaf of Echinochloa esculenta (Poaceae) (Thailand: 

Chiang Rai province, Huay Kang Pah National Park), 4 December 2009. 

Passalora graminis [26] 

Notes: The collection no. MFLU10-0317 from Doi Tung National Park, Chiang Rai province 

agrees well with Passalora graminis as circumscribed previously [15, 23]. Passalora graminis has 

conidiophores of 10–52 × 3–5 μm and conidia of 18–38 × 1.5–2 μm. 

Known hosts: Agrobardeum, Agropyron, Agrositanion, Agrostis, Alopecurus, Ammophila, 

Anthoxanthum, Arctagrostis, Arrhenatherum, Arundinaria, Avena, Beckmannia, Bromus,


55 

Calamagrostis, Cinna, Cynodon, Cynosurus, Dactylis, Danthonia, Deschampsia, Digitaria, Elymus, 

Elysitanion, Elytrigia, Egagrostis, Festuca, Glyceria, Hierochloe, Hordeum, Hystrix, Koeleria, 

Leersia, Leucopoa, Lolium, Melica, Milium, Miscanthus, Muhlenbergia, Oryzopsis, Panicum, 

Pennisetum, Phleum, Phragmities, Poa, Puccinella, Roegneria, Secale, Sitanion, Sparlina, 

Stenotaphrum, Stipa, Trisetum, and Zea (Poaceae) [16]. 

Known distribution: Asia - Thailand (this paper) and worldwide [16]. 

Material examined: Phengsintham (MFLU10-0317) on leaf of Agrostis sp. (Poaceae) 

(Thailand: Chiang Rai province, Doi Tung National Park), 18 August 2009. 

Pseudocercospora cycleae [23] 

Notes: The collection no. MFLU10-0319 from Khun Korn waterfall, Chiang Rai province 

agrees with the previous description of this species [27], but differs in having longer conidiophores. 

Known hosts: Cyclea fissicalyx Dunn, C. peltata Hook. f. & Thomson, and Cyclea sp. 

(Menispermaceae) [27]. 

Known distribution: Asia - China, India [27], Thailand (this paper). 

Material examined: Phengsintham (MFLU10-0319) on leaf of Cyclea peltata 

(Menispermaceae) (Thailand: Chiang Rai province, Khun Korn waterfall), 18 December 2009. 

Pseudocercospora malloticola [3] 


(Thailand) and no. NUOL P588 from Naloumai village, Savannakhet province (Laos) are similar to 

Pseudocercospora malloticola from Taiwan previously described [3]. Brief description of the 

collection from Thailand: Leaf spots/Lesions discoid to irregular, 1–6 mm diam., at first yellowish, 

later becoming brown or dark brown, and with yellowish margin. Conidiophores fasciculate, arising 

from stromata (2–11 per fascicle), emerging through stomata, nearly straight, cylindrical, 

unbranched, 10–40 × 3–5 μm. Conidia formed singly, cylindric, straight to slightly curved, 33–75 × 

3–4 μm. 

Known hosts: Mallotus barbatus (Wall.) Müll. Arg., M. japonicus (L. F.) Müll. Arg., and M. 

thorelii Gagnep [3]. 

Known distribution: Asia - Laos (this paper), Taiwan [3], Thailand (this paper). 

Material examined: 1) Phengsintham (MFLU10-0320) on leaf of Mallotus barbatus 

(Euphorbiaceae) (Thailand: Chiang Rai province, Muang district, Sri Pangsang village), 30 August 

2009; 2) Phengsintham (P588) on leaf of Mallotus thorelii (Euphorbiaceae) (Laos: Savannakhet 

province, Vilaboury district, Naloumai village), 23 June 2010. 

Pseudocercospora olacicola [28] 

Notes: The collection no. MFLU11-0020 from Mae Puem National Park, Pha Yao province 

(conidiophores 12–35 × 3–5 μm and conidia 11–58 × 2–3 μm) agrees with Pseudocercospora 

olacicola previously described [28] (conidiophores 10.5–40.5 × 2.5–5 μm and conidia 16–30.5 × 

2.5–4 μm).


Known hosts: Olax scandens Roxb., Olax wightiana Wall. ex Wight & Arn., O. zeylanica L., 

Olax sp., and Ximenia sp. (Olacaceae) [16, 28]. 

Known distribution: Asia - India [16, 28], Thailand (this paper). 

Material examined: Phengsintham (MFLU11-0020) on leaf of Olax scandens (Olacaceae) 

(Thailand: Pha Yao province, Mae Jai district, Mae Puem National Park), 22 August 2010. 

Pseudocercospora paederiae [3] 

Notes: The collection no. MFLU10-0321 from Tadsak waterfall, Chiang Rai province 

(conidiophores 5–20 × 3–5 μm and conidia 42–75 × 2–3 μm) is similar to the original description of 

this species, based on material from Taiwan, but there are slight differences in the size of the 

conidiophores and conidia. The collection from Taiwan has conidiophores that are densely 

fasciculate, 20–120 × 3–4 μm, subhyaline to pale brown and conidia that are obclavate, straight to 

moderately curved, 30–80 × 3.5–5 μm and medium olivaceous [3, 15, 27]. 

Known hosts: Paederia chinensis Hance, P. foetida L., P. scandens (Lour.) Merr., and P. 

tomentosa Blume (Rubiaceae) [3,27]. 

Known distribution: Asia - China, Japan, Korea, Taiwan [3, 27], Thailand (this paper). 

Material examined: Phengsintham (MFLU10-0321) on leaf of Paederia tomentosa 

(Rubiaceae) (Thailand: Chiang Rai province, Wiang Chiang Rung district, Tadsak waterfall), 23 


Pseudocercospora polysciatis [29] 


agrees well with Pseudocercospora polysciatis described previously [3, 27] but differs in having 

distinct constriction at the septa. 

Known hosts: Polyscias balfouriana (André) L.H. Bailey, P. guilfoylei (W. Bull) L.H. 

Bailey, and Polyscias sp. (Araliaceae) [3, 16, 27]. 

Known distribution: Africa - Mauritius, Ivory Coast; Asia - Brunei, Philippines, Taiwan, 

Thailand (this paper); North America - Cuba; Oceania - American Samoa, Cook Islands, Fiji, 

Kiribati, Marshall Islands, Micronesia, Niue, Samoa, Solomon Islands, Tonga [16]. 

Material examined: Phengsintham (MFLU10-0322) on leaf of Polyscias balfouriana 

(Araliaceae) (Thailand: Chiang Rai province, Muang district, Sri Pangsang village), 16 January 2010. 

Zasmidium cassiicola [30] Stenella cassiicola [31] (Figure 2; Redescribed because this species 

is poorly known.) 

Leaf spots/Lesions variable, more or less irregularly orbicular, 1–8 mm diam., typically deep 

brown. Caespituli/Colonies hypophyllous, conspicuous. Mycelium external; hyphae branched, 2–4 

μm wide (x = 3.2 μm, n = 9), septate, constricted at the septa, distance between septa 8–30 μm ( x 

= 18.56 μm, n = 9), pale olivaceous-brown, thin-walled 0.3–0.5 μm wide (x = 0.41 μm, n = 9), 

verruculose. Stromata absent. Conidiophores borne on external mycelial hyphae, unbranched, 

cylindrical, 30 – 117 × 3 – 4 μm ( x = 65.9 × 3.17 μm, n = 18), 3 – 8-septate, distance between


57 

Figure 2(a). Zasmidium cassiicola on Cassia fistula: 1–3. External mycelia with attached 

conidiophores; 4–7. Conidia. (Scale bar = 10 μm)


Figure 2(b). Zasmidium cassiicola on Cassia fistula from leaf spot: 1–2. Lesions on host leaves (1. 

upper surface and 2. lower surface); 3–6. Conidiophores; 7. Apex with attached conidium; 8. 

External mycelium; 9–11. Conidia; 12. Living culture. (Scale bars: 1–2. = 10 mm, 3–11. = 10 μm) 

septa 7–25 μm ( x = 15.8 μm, n = 30), mid pale golden brown, wall 0.5–0.8 μm ( x = 0.48 μm, n = 

30), smooth. Conidiogenous cells integrated, terminal or intercalary, 7–20 × 2–4 μm ( x = 13 × 

2.63 μm, n = 8), cylindrical, swollen and curved at the apex; conidiogenous loci forming minute, 

dark or refractive scars on lateral and terminal denticles, 1–2 μm diam. (x = 1.4 μm, n = 7), giving 

rise to branched conidial chains, wall 0.3–0.5 μm wide (x = 0.4 μm, n = 7), thickened, darkened.


59 

Conidia solitary or catenate, sometimes ellipsoidal-ovoid or subcylindrical, but mostly slightly 

obclavate, straight or slightly curved or sinuous, 11–70 × 2–4 μm ( x = 34.16 × 2.9 μm, n = 24), 1– 

5-septate, pale olivaceous, wall 0.3–0.5 μm wide (x = 0.33 μm, n = 24), smooth or finely 

verruculose; apex rounded or subtruncate, 1–1.5 μm wide, wall 0.3–0.5 μm wide; base short tapered 

at the base to the hilum, 1–2 μm wide (x = 1.3 μm, n = 9), wall 0.3–0.5 μm wide (x = 0.41 μm, n = 

9), thickened and darkened. 

Known host: Cassia fistula L. (Leguminosae) [31]. 

Known distribution: Asia - India [31], Thailand (this paper). 

Material examined: Phengsintham (MFLU10-0324) on leaf of Cassia fistula (Leguminosae) 

(Thailand: Chiang Rai province, Muang district, Sri Pangsang village), 16 January 2010. 

Cultural characteristics: Colonies on potato dextrose agar after three weeks at 25°C with 

spreading mycelium, surface ridged, black and wavy in the centre and gray margin, reaching 5–15 

mm diam.; hyphae often constricted at the septa, distances between septa 6–16 × 3–5 μm ( x = 10.5 

× 3.6 μm, n = 30), thin-walled, 0.3–0.5 μm wide (x = 0.45 μm, n = 30), hyaline, smooth or 

verruculose; Conidiophores and conidia not formed in culture. 

Notes: The collection from Thailand agrees well with the Indian Zasmidium cassiicola 

(Stenella cassiicola [13, 31]. This is the first record outside India. Zasmidium cassiae-fistulae is a 

similar species but differs in forming its conidia consistently singly [32]. 

CONCLUSIONS 

Cercosporoid fungi are one of the largest groups of pathogenic hyphomycetes causing leaf 

spots on a wide range of crops, fruit trees and other plants. The damage to living leaves and fruits 

may cause reduced yield. Fourteen species assigned to the genera Cercospora (5), Passalora (3), 

Pseudocercospora (5) and Zasmidium (1) are new records for Thailand. Cercospora verniciferae 

and Zasmidium cassiicola are poorly known species and are fully described. 


The authors would like to thank the Global Research Network for Fungal Biology and King 

Saud University and the Mushroom Research Foundation (MRF) for supporting this research. 

Special thanks also go to the Mushroom Research Foundation. 

REFERENCES 

1. G. N. Agrios, “Plant Pathology”, 5 th Edn., Academic Press, San Diego, 2005. 

2. P. W. Crous, “Taxonomy and phylogeny of the genus Mycosphaerella and its anamorphs”, 

Fungal Divers., 2009, 38, 1-24. 

3. W. H. Hsieh and T. K. Goh, “Cercospora and Similar Fungi from Taiwan”, Maw Chang Book 

Co., Taipei, 1990. 

4. E. B. G. Jones, M. Tantichareon and K. D. Hyde (Ed.), “Thai Fungal Diversity”, Biotec, 

Bangkok, 2005.


5. P. Giatgong, “Host Index of Plant Diseases in Thailand”, Ministry of Agriculture and 

Cooperatives, Bangkok, 1980. 

6. P. Sontirat, P. Pitakpraiwan, W. Choonbamroong and U. Kueprakone, “Plant Pathogenic 

Cercosporoids in Thailand”, Ministry of Agriculture and Cooperatives, Bangkok, 1980. 

7. V. Petcharat and M. Kajanamaneesathian, “Species of plant pathogenic Cercospora in southern 

Thailand”, Thai J. Phythopathol., 1989, 9, 23-27. 

8. P. Sontirat, P. Pitakpraivan, T. Khamhangridthrirong, W. Choonbamroong and U. Kueprakone, 

“Host Index of Plant Diseases in Thailand”, Ministry of Agriculture and Cooperatives, Bangkok, 

1994. 

9. C. Nakashima, K. Motohashi, J. Meeboon and C. To-anun, “Studies on Cercospora and allied 

genera in northern Thailand”, Fungal Divers., 2007, 26, 257-270. 

10. J. Meeboon, I. Hidayat and C. To-anun, “An annotated list of cercosporoid fungi in northern 

Thailand”, J. Agric. Technol., 2007, 3, 51-63. 

11. C. To-anun, J. Nguenhom, J. Meeboon and I. Hidayat, “Two fungi associated with necrotic 

leaflets of areca palms (Areca catechu)”, Mycol. Prog., 2009, 8, 115-121. 

12. J. Meeboon, “Diversity and phylogeny of true cercosporoid fungi from northern Thailand”, PhD 

Thesis, 2009, Chiang Mai University, Thailand. 

13. C. To-anun, I. Hidayat and J. Meeboon, “Cercospora cristellae, a new cercosporoid fungus 

associated with weed Cristella parasitica from northern Thailand”, J. Agric. Technol., 2010, 6, 

331-339. 

14. P. Phengsintham, E. Chukeatirote, K. A. Abdelsalam, K. D. Hyde and U. Braun, “Cercospora 

and allied genera from Laos 3”, Cryptogamie Mycol., 2010, 31, 305-322. 

15. C. Chupp, “A Monograph of the Fungus Genus Cercospora”, Charles Chupp, Ithaca (New 

York), 1954. 

16. P. W. Crous and U. Braun, “Mycosphaerella and Its Anamorphs: 1. Names Published in 

Cercospora and Passalora”, Centraalbureau voor Schimmelcultures, Utrecht, 2003. 

17. U. Braun, “A Monograph of Cercosporella, Ramularia and Allied Genera (Phytopathogenic 

Hyphomycetes)”, Vol. 1, IHW Verlag, Eching (Germany), 1995. 

18. J. B. Ellis and B. M. Everhart, “New species of fungi from various localities”, J. Mycol., 1888, 

4, 113-118. 

19. J. J. Davis, “Notes on parasitic fungi in Wisconsin XX”, Trans. Wisc. Acad. Sci., 1937, 30, 1- 

16. 

20. J. B. Ellis and B. M. Everhart, “New species of hyphomycetes fungi”, J. Mycol., 1889, 5, 68-72. 

21. A. P. Viegas, “Alguns fungos do Brasil, Cercosporae”, Bol. Soc. Bras. Agron., 1945, 8, 1-160. 

22. L. Xu and Y. L. Guo, “Studies on Cercospora and allied genera in China XIII”, Mycosystema, 

2003, 22, 6-8. 

23. F. C. Deighton, “Studies on Cercospora and allied genera. VI. Pseudocercospora Speg., 

Pantospora Cif. and Cercoseptoria Petr.”, Mycol. Papers, 1976, 140, 1-168. 

24. G. F. Atkinson, “Some Cercosporae from Alabama”, J. Elisha Mitchel Sci. Soc., 1891, 8, 33- 

67.


61 

25. M. B. Ellis, “More Dematiaceous Hyphomycetes”, Commonwealth Mycological Institute, Kew, 

Surrey, 1976. 

26. P. Srivastava, “Recombinations in genus Passalora Fries”, J. Living World, 1994, 1, 112-119. 

27. Y. L. Guo and W. H. Hsieh, “The Genus Pseudocercospora in China”, International Academic 

Publishers, Beijing, 1995. 

28. Kamal, M. K. Khan and R. K. Verma, “New species and combinations in Pseudocercospora 

from India”, Mycol. Res., 1990, 94, 240-242. 

29. U. Braun and A. Sivapalan, “Cercosporoid hyphomycetes from Brunei”, Fungal Divers., 1999, 

3, 1-27. 

30. Kamal, “Cercosporoid Fungi of India”, Bishen Singh Mahendra Pal Singh, Dehradun (India), 

2010. 

31. S. Misra, A. K. Srivastava and Kamal, “Further additions to Stenella from India and Nepal”, 

Mycol. Res., 1999, 103, 268-270. 

32. U. Braun, P. Crous and Kamal, “New species of Pseudocercospora, Pseudocercosporella, 

Ramularia and Stenella (cercosporoid hyphomycetes)”, Mycol. Prog., 2003, 2, 197-208. 




Short Communication 





On the security of an anonymous roaming protocol in UMTS 

mobile networks 

Shuhua Wu 1, 2, * , Qiong Pu 2, 3 and Ji Fu 1 

1 Department of Network Engineering, Information Engineering University, Zhengzhou, China 

2 State Key Laboratory of Information Security, Graduate University of Chinese Academy of 

Sciences, Beijing, China 

3 CIMS Research Centre, Tongji Univerity, Shanghai, China 

* Corresponding author, e-mail: pqwsh@yahoo.com.cn 

Received: 21 April 2011 / Accepted: 4 February 2012 / Published: 10 February 2012 

Abstract: In this communication, we first show that the privacy-preserving roaming 

protocol recently proposed for mobile networks cannot achieve the claimed security level. 

Then we suggest an improved protocol to remedy its security problems. 

Keywords: cryptanalysis, anonymous roaming, authentication, UMTS 

INTRODUCTION 

With the advancement and tremendous development of computer networks and 

telecommunications, user mobility has become a highly desirable network feature nowadays, 

especially in wireless networks (e.g. cellular networks [1-3]). This technology enables users to access 

services universally and without geographical limitations. In other words, they can go outside the 

coverage zone of their home networks, travel to foreign networks and access services provided by 

the latter as a visiting user or a guest. This capability is usually called roaming. Security is one of the 

major requirement in roaming networks. In addition to authentication, user’s privacy is equally 

important in such networks. To preserve this feature, not only should the user’s identity be protected 

(anonymity requirement), but also his location and the relation between his activities should be kept 

secret (untraceability requirement). The violation of either of the mentioned requisites can seriously 

endanger the user’s privacy. Samfat et al. [1] have proposed a comprehensive classification for 

different levels of privacy protection according to the knowledge of different entities about the user’s 

identification information. The classification is as follows:


63 

• C1: Each user is anonymous to eavesdroppers and his activities are unlinkable to them. 

• C2: In addition to C1, each user is anonymous to the foreign servers and his activities are 

unlinkable to them. 

• C3: In addition to C2, the relationship between the user and servers (the home server and the 

foreign servers) is anonymous for eavesdroppers. 

• C4: In addition to C3, the home server of the user is anonymous to the foreign servers. 

• C5: In addition to C4, each user is anonymous and his activities are unlinkable to his home 

server. 

In the standard universal mobile telecommunication system (UMTS) [2], the home server 

must be always aware of the mobile user’s location in order to route the incoming calls towards the 

user. Moreover, the foreign server should know the identity of the home server for billing purpose. 

Therefore, it seems that the admissible level of privacy protection in this scenario is C3. In the last 

decades, several schemes addressed the privacy of users in mobile networks [3-15]. However, the 

most perfect and practical scheme that has been proposed so far only achieves the C2 class of 

anonymity and the possible C3 class has not been provided in UMTS yet. To fill this blank, Fatemi et 

al. [2] recently proposed a privacy-preserving roaming protocol based on hierarchical identity-based 

encryption (IBE) [16] for mobile networks. This protocol was claimed to achieve the acceptable C3 

level of privacy. In this communication, we first show that it has some security weakness and thus 

the claimed security level is not achieved. Finally, we propose an enhanced protocol to remedy the 

existing security loopholes. 

PRELIMINARY 

In this section we recall the concept of Identity-Based Encryption (IBE) and hierarchical IBE 

(HIBE) schemes, upon which Fatemi et al.’s scheme builds. Here we just follow their description [2]. 

At first, we introduce the concept of a bilinear map between two groups, which will be used in the 

IBE scheme. Let G 1 

be an additive group and G 2 

be a multiplicative group, both of order q ( q 

should be some large prime, e.g. 160 bits). We say that a map eG1G1 G2 

is an admissible 

bilinear map if all the three following conditions are satisfied: 

1) eaPbQ ( ) ePQ 

( ) ab for all ab Zq 

and PQ G1 

(bilinear condition); 

2) the map does not send all elements of G 1 

G 1 

to the identity element of G 2 

(non-degeneracy 

condition); and 

3) there is an efficient algorithm to compute ePQ ( ) for all PQ G1 

(computability condition). 

Throughout this communication, the Bilinear Diffie-Hellman (BDH) in G 1 

G 2 

e is 

believed to be hard (i.e. it is hard to compute ePP ( ) abc G2 

, given PaPbPcP 

for some 

abc Z q 

). Since the BDH problem is not harder than the computational Diffie-Hellman (CDH) 

problem in G 

1 

or G 

2 

, the CDH problem in G 1 

or G 

2 

is also believed to be hard. The CDH problem 

in G 

1 

is as follows: given random PaPbP 

for ab Zq 

, compute abP ; the CDH problem in 

G 

2 

is defined similarly. In addition, for a point Q G 

1 

, the isomorphism fQ 

G1 G2 

by 

f ( ) ( ) 

Q 

P e P Q is considered as a one-way function ( P cannot be inferred from ePQ ( ) and Q )


since an efficient algorithm for inverting f Q 

for some Q results in an efficient algorithm for solving 

CDH problem in G 

2 

. 

Now we begin to introduce the IBE system. An IBE is a public key cryptosystem in which 

the public key takes any arbitrary string such as a name or an e-mail address, and the private key 

generator (PKG) can produce a private key corresponding to each string. Hence, one can encrypt a 

message by a public key even if the public key’s owner has not yet set up his private key. An efficient 

IBE is presented [17], which is called a Boneh-Franklin scheme. Let P be a generator of G 1 

and 

s Z 

q 

be the PKG’s master key. Then in the Boneh-Franklin scheme, each user’s identity-based 

private key should be computed as ( ) 

 

kU 

sH1 U , where H 1 

{01} 

G 1 

is a cryptographic hash 

function and U is the user’s identity. Then one can encrypt a message using the public key U , and 

U can decrypt the ciphertext using the private key k U 

. The BF scheme is resistant to the chosen 

ciphertext attack, assuming the hardness of the BDH problem [17]. 

Similar to the public key cryptosystems, a hierarchy of PKGs is desirable in an IBE system to 

reduce the workload of the master servers. A two-level HIBE (2-HIBE) is presented [16]. There are 

three entities involved in a 2-HIBE scheme: a root PKG which possesses a master key s , the domain 

PKGs which gain their domain keys from the root PKG, and the users with private keys generated by 

 

their domain PKGs. The 2-HIBE scheme benefits from a linear one-way function hG1Zq 

G1 

with the following properties: 

 

1) For all PG1 axZ h( aPx) ah( P x) 

, 

q 

 

2) Given xx 

iZqP G1 

and xih( aPxi) 

for i 1 n, haPx ( ) cannot be computed with 

any probabilistic polynomial-time algorithm. 

The function h defined above is a one-way function with respect to its first argument, i.e. P 

cannot be inferred from hPx ( ) and x . Then the key for domain S is kS 

sH1( S) 

G1 

and the key 

 

for user U in domain S is kU 

h( kSH2( S U)) 

G1, where H 2 

{01} Zq 

is a cryptographic 

hash function and denotes concatenation. Finally, one can encrypt a message by a public key 

SU 

and U can decrypt the ciphertext using k . U 

FATEMI ET AL.’ S ROAMING PROTOCOL 

Review of Protocol 

Here we just follow the description of Fatemi et al [2]. Like Wan et al.’s scheme [18], they 

also assume that a 2-HIBE is implemented in the system and the domain servers have received their 

private keys { KS 

sH1( S )} 

i 

i 

from a root server. Also, they suppose that the user U obtains his 

private key KU 

h( KHS H2( HS U)) 

during the registration at his home domain server HS . In 

addition, a temporary key K e( h( h( H1( HS) H2( HS Nym)) 

H2( HS)) sH1( HS)) 

corresponding to 

a pseudonym Nym will be computed by the user during the roaming protocol and will be used for 

the authentication and key agreement purposes when he enters a foreign network domain. 

As shown in Figure 1, the protocol is as follows:


65 

Step 1. When the foreign server ( FS ) detects a new user in his domain, it generates a nonce 

a random number r s 

(both from 

Z q 

) and computes rP. s 

Then it stores the values 

N 

s 

and 

N 

s 

and rP s 

in 

his database and sends the first message including his identity IDFS Ns 

and rP s 

to the user. 

Figure. 1. Fatemi et al.’s roaming protocol [2] 

Step 2. Similarly, the user U generates a nonce N u 

and a random number r u 

and computes 

k u rrP 

u s 

. Then he fetches the only unused pair of ( Nym K) 

from his memory and computes 

the session key to be shared with the foreign server as sk ( 1) 

u 

H4 K k 

u FS Nym Nu Ns 

and a verifier mac ( u 

H 

4 

K k u FS Nym Nu 

N 

s 

0) , where H 4 

is a hash function which 

* 

maps {0 1} to {0 1} l for some security parameter l . After that, he selects an arbitrary Nym 

next 

to 

be used in the next execution of the roaming protocol (either in the current FS or another FS ). 

In order to compute the corresponding key K 

next 

with the help of FS and HS , the user selects a 

 

random number a Zq 

and computes the following values: 

 

 

 

 

x1 hhaH ( ( 

1( HS) H2( HS Nymnext 

)) H2( HS)) 

, x2 hhaH ( ( 

1( HS) H2( HS U)) H2( HS)) 

. 

Also, he chooses random numbers ba1 a2 Z 

 

 

q 

and computes ax 

1 1 ax 

2 2 

and 

EHS ( bUE( KU U) IDFS 

), where ES 

( M ) denotes the ID-based encryption of message M with 

the public key S (e.g. HS or FS ), and EK ( 

U 

U) 

denotes the symmetric encryption of U with 

 

the key K U 

. Next, he sends the values EFS ( NymIDHS ) Nuru PNsrs Pmacu 

x1a1x1 

ax 

2 2 

and 


) to the foreign server. 

Step 3. Upon receiving the above values, the foreign server checks if N s 

and rP 

s 

exist in its 

database and aborts the connection if it does not find such values. Otherwise, it decrypts 

EFS 

( Nym IDHS 

) with its private key sH 1 

( FS ) and obtains the Nym and ID 

HS 

. Then it 

generates a random number c Z 

q 

and computes z h( h( cH1( HS) H 2( HS Nym)) H 

2( HS)) 

. 

Subsequently, the FS sends z and E ( bUE( K U) ID ) to the HS . 

HS U FS


Step 4. The home server decrypts the message E ( bUE( K U) ID ) with its private key 

HS U FS 

sH ( ) 1 

HS and checks whether it has received the messages from the server with the identity 

ID 

FS 

. Then it authenticates the user U by verifying the correctness of EK ( 

U 

U) 

. The home 

server terminates the connection if any of these verifications fails. Otherwise, it computes 

y e( zsH1( HS)) 

( sb) H1( HS) sH1( HS) 

bH ( HS) 

1 

and sends them back to the FS . 

1 

c 

Step 5. The FS computes k s rrP 

s u 

and the values K 

y 

 

, ( 

macu 

H4 K k 

s FS Nym 

Nu 

Ns 0) . The FS rejects the connection if the equality macu mac 

u 

does not hold. 

Otherwise, it accepts K as the user’s key corresponding to Nym and authenticates the user. 

The computed K together with the message EHS ( bUE( KU U) IDFS 

) are credentials by 

which the foreign server will be able to request the user’s home server for service charge. 

Indeed, these values become a proof for payment request. In the next step the foreign server 

computes the session key ( 

sk H 1) 

4 

K k s FS Nym N Ns and the authenticator 

s 

( 

4 s 

s 

mac H K k FS Nym N Ns 2) . Moreover, the foreign server calculates 

u 

 

 

y1 e( x1( s b) H1( HS)) 

and y2 e( a1x1 a2x2( s b) H1( HS)) 

to make the computation of 

K 

next 

feasible for the user. Finally, it returns macs y1 

and H4( y 

2) 

to the user. 

Step 6. When the user receives the messages from the foreign server, he computes 

mac ( 2) 

s 

H4 K k U FS Nym Nu 

Ns and checks the equality macs mac 

s 

. If it 

does not hold, the user aborts the connection. Otherwise, he authenticates the foreign server 

 

 

and computes the following values: y1 y1e( x1 bH1( HS)) 

, 

 

1 

2 

( 

1) a 

 

y y [ e( h( a K 

U 

H 

2( HS )) H 1( HS )) e( 

x2 

( 1 

1 ))]a 2 

( ) 

bH HS , ( 

1 

) a 

 

Knext 

y . Afterwards, the 

user considers whether H4( y2) H4( y 

2) 

. If the equation holds, he accepts K next 

as the key 

corresponding to Nym 

next 

. If not, he rejects the connection. 

At the end of the protocol, sku sks 

is the key that the user and the home server have agreed 

upon to be used for security purpose. 

Weakness of Fatemi et al.’s Protocol 

We assume the adversary has totally controlled a mobile user U or equivalently he has 

revealed the secret keys K through side channel attacks [19]. We further assume the adversary has 

U 

corrupted one of foreign networks, e.g. FN . Let HS be the home server of U and FS be the 

server of FN . The adversary impersonated U to visit FN and initiated an execution of Fatemi et 

al.’s roaming protocol. He obtained from the corrupted network the message transmitted from HA 

to FS in Step 4: ( s 

b) H1( HS) 

, where b is the random number chosen by the adversary 

in Step 2. He could then compute HS ’s secret key K through 

HS 

K sH1( HS) ( s b) H1( HS) bH1( HS) 

. When the adversary knows K , the problem of 

HS 

HS 

Fatemi et al.’s roaming protocol becomes evident: 

• Firstly, the adversary can reveal the real identity of any other subscriber of HS . When a mobile 

user U ( U ) performs the authentication procedure with a foreign server FS ( FS ), the 

adversary eavesdrops their communication and can easily get the message transmitted between 

u


67 

U and FS : ID 

FS 

in Step 1 and EHS ( bUE( KU U) IDFS 

) in Step 2. Then the adversary just 

guesses that U is a subscriber of HS and attempts to decrypt the message 


) with K sH ( ) 

HS 1 

HS . If he can retrieve ID 

FS 

from the decrypted 

message, he confirms his guess is correct, i.e. HS HS , because otherwise the probability that 

he gets any meaningful results from decryption for verification is next to zero. Then he further 

retrieves item U from the decrypted message and thus knows the user’s real identity U . This 

even contradicts the C1 security requirements. However, if U is not a subscriber of HS , his 

attack cannot succeed, but he will always succeed for any subscriber of HS . 

• Secondly, the adversary can impersonate any other subscriber of HS (e.g. U ) because he may 

derive K 

U 

from K as follows: K ( 

HS 

U 

h K H2( HS U)) 

. In other words, the authentication 

HS 

mechanism of the protocol is completely compromised. 

IMPROVED ROAMING PROTOCOL 

The above demonstrated attacks show that Fatemi et al.’s protocol does not seem to achieve 

authentication or anonymity. In this section we present an enhanced protocol to remedy the security 

loopholes. As shown in Figure 2, our protocol is based on that of Fatemi et al. and it has the 

following changes: 

Figure 2. Improved roaming protocol 

 

• In Step 2 the computation of ax 1 1 

a 2 

x 2 

is not needed any longer and finally the user U sends 

the values EFS ( NymIDHS ) Nuru PNsrs Pmacu 

x 

1 

and EHS 

( bUE( KUU) IDFS 

) to the 

foreign server. 

• In Step 3 the FS also forwards x 1 

to the HS. That is, the FS sends z, 

E ( bUE( K U) ID ) and x 1 

to the HS. 

HS U FS


• In Step 4 the computation of ( sb) H1( HS) sH1( HS) bH1( HS) 

is not needed. Instead, the 

 

 

home server computes a new item w E( KU 

e( x1 sH1( HS)) x1) 

to make the computation of 

K 

next 

feasible for the user and finally sends w along with y back to the FS . 

• In Step 5 the computation of y 1 

and y 

2 

is not needed and the foreign server returns mac 

s 

and 

w to the user. 

• In Step 6 the computation of y 1 

and y 2 

is not needed. After the verification of mac 

s 

is passed 

and the foreign server is authenticated, the user U decrypts w using his own secret key K 

U 

to 

retrieve the two items ex ( 

 

1 sH1( HS)) 

x1. If the decrypted x 1 

is the same as x 1 

computed in 

1 

( ) 

Step 1, he proceeds to compute ( ( 

1 1( ))) a 

 

Knext 

e x sH HS and accepts K 

next 

as the key 

corresponding to Nym 

next 

. If not, he rejects the connection. 

In our improved protocol, the item ex ( 1 

sH1( HS)) 

is used to make the computation of K 

next 

feasible for the user. Given ex ( 1 

sH1( HS)) 

, it is impossible for the adversary to compute sH ( HS ) 1 

since the isomorphism f Q 

(here Q x 

1 

) is a one-way function. Therefore, the attacks described 

previously will not work any more. Although the changes introduce some computation overhead on 

the side of HS due to the computation of w , the computation cost of FS or U is significantly 

reduced since both FS and U omit several costly operations (including bilinear map and 

exponentiation). In practice the device of the mobile user is much less powerful than the servers’. 

Our protocol, therefore, would be more practical. 


This work was supported in part by the National Natural Science Foundation of China 

(Project No. 61101112) and China Postdoctoral Science Foundation (Project No. 2011M500775). 

REFERENCES 

1. D. Samfat , R. Movla and N. Asokan , “Untraceability in mobile networks”, Proceedings of the 

1st International Conference on Mobile Computing, 1995, Santa Barbara, California, USA, 

pp.26-36. 

2. M. Fatemi, S. Salimi and A. Salahi, “Anonymous roaming in universal mobile telecommunication 

system mobile networks”, IET Inf. Secur., 2010, 4, 93-103. 

3. W.-S. Juang and J.-L. Wu, “Efficient 3GPP authentication and key agreement with robust user 

privacy protection”, Proceedings of IEEE Wireless Communications and Networking 

Conference, 2007, Hong Kong, China, pp.2720-2725. 

4. B. Sattarzadeh, M. Asadpour and R. Jalili, “Improved user identity confidentiality for UMTS 

mobile networks”, Proceedings of 4th European Conference on Universal Multiservice 

Networks, 2007, Toulouse, France, pp.401-409. 

5. Y. Jiang, C. Lin, X. Shen and M. Shi, “Mutual authentication and key exchange protocols for 

roaming services in wireless mobile networks”, IEEE Trans. Wireless Comm., 2006, 5, 2569- 

2577.


69 

6. G. Yang, D. S. Wong and X. Deng, “Efficient anonymous roaming and its security analysis”, 

Proceedings of the 3rd International Conference on Applied Cryptography and Network 

Security, 2005, New York, USA, pp.334-349. 

7. C.-T. Li and C.-C Lee, “A novel user authentication and privacy preserving scheme with smart 

cards for wireless communications”, Math. Comp. Model., 2012, 55, 35-44. 

8. S. Wu, Y. Zhu and Q. Pu, “Security analysis of a cocktail protocol with the authentication and 

key agreement on the UMTS”, IEEE Comm. Lett., 2010, 14, 366-369. 

9. S. Wu,. Y. Zhu and Q. Pu, “A novel lightweight authentication scheme with anonymity for 

roaming service in global mobility networks”, Int. J. Network Manage., 2011, 21, 384-401. 

10. C.-C. Chang and H.-C. Tsai, “An anonymous and self-verified mobile authentication with 

authenticated key agreement for large-scale wireless networks”, IEEE Trans. Wireless Comm., 

2010, 9, 3346-3353. 

11. T.-Y. Youn and J. Lim, “Improved delegation-based authentication protocol for secure roaming 

service with unlinkability”, IEEE Comm. Lett., 2010, 14, 791-793. 

12. D. He, J. Bu, S. Chan, C. Chen and M. Yin, “Privacy-preserving universal authentication 

protocol for wireless communications”, IEEE Trans. Wireless Comm., 2011, 10, 431-436. 

13. D. He, M. Ma, Y. Zhang, C. Chen and J. Bu, “A strong user authentication scheme with smart 

cards for wireless communications”, Comp. Comm. , 2011, 34, 367-374. 

14. J. Xu, W-T. Zhu and D-G. Feng, “An efficient mutual authentication and key agreement 

protocol preserving user anonymity in mobile networks”, Comp. Comm., 2011, 34, 319-325. 

15. C. Chen, D. He, S. Chan, J. Bu, Y. Gao and R. Fan, “Lightweight and provably secure user 

authentication with anonymity for the global mobility network”, Int. J. Comm. Syst., 2011, 24, 

347-362. 

16. J. Horwitz and B. Lynn, “Toward hierarchical identity-based encryption”, Proceedings of the 

21st International Conference on the Theory and Applications of Cryptographic Techniques, 

2002, Amsterdam, Netherlands, pp.466-481 

17. D. Boneh and M. Franklin, “Identity-based encryption from the weil pairing”, Proceedings of 

the 21 st International Cryptology Conference on Advances in Cryptology, 2001, Santa Barbara, 

California, USA, pp.213-229 

18. Z. Wan, K. Ren and B. Preneel, “A secure privacy-preserving roaming protocol based on 

hierarchical identity-based encryption for mobile networks”, Proceedings of the 1st ACM 

Conference on Wireless Network Security, 2008, New York, USA, pp.62-67. 

19. P. Kocher, J. Jaffe and B. Jun, “Differential power analysis”, Proceedings of the 19th Annual 

International Cryptology Conference, 1999, Santa Barbara, California, USA, pp.388-397. 




Full Paper 





Influence of MeV H + ion beam flux on cross-linking and blister 

formation in PMMA resist 

Somrit Unai 1, 2, * , Nitipon Puttaraksa 1, 3 , Nirut Pussadee 1, 2 , Kanda Singkarat 4 , 

Michael W. Rhodes 5 , Harry J. Whitlow 3 and Somsorn Singkarat 1, 2 

1 Plasma and Beam Physics Research Facility, Department of Physics and Materials Science, 

Faculty of Science, Chiang Mai University, Chiang Mai 50200, Thailand 

2 

Thailand Centre of Excellence in Physics, CHE, 328 Si Ayutthaya Road, Bangkok 10400, Thailand 

3 Department of Physics, University of Jyväskylä, P.O. Box 35 (YFL), FIN – 40014, Finland 

4 Department of Physics and Materials Science, Faculty of Science, Chiang Mai University, 

Chiang Mai 50200, Thailand 

5 Science and Technology Research Institute, Chiang Mai University, Chiang Mai 50200, Thailand 

* Corresponding author, e-mail: somrit@fnrf.science.cmu.ac.th 

Received: 31 March 2011 / Accepted: 13 February 2012 / Published: 14 February 2012 

Abstract: In soft lithography, a pattern is produced in poly(dimethylsiloxane) (PDMS) 

elastomer by casting from a master mould. The mould can be made of poly(methylmethacrylate) 

(PMMA) resist by utilising either its positive or negative tone induced by an ion beam. Here we 

have investigated the irradiation conditions for achieving complete cross-linking and absence of 

blister formation in PMMA so that its negative characteristic can be used in making master 

moulds. PMMA thin films approximately 9 μm thick on Si were deposited by spin coating. The 

2-MeV H + ion beam was generated using a 1.7-MV tandem Tandetron accelerator. The beam 

was collimated to a 500×500 μm 2 cross section using programmable proximity aperture 

lithography system with a real-time ion beam monitoring system and a high precision current 

integrator. The irradiated areas were investigated by a standard scanning electron microscope 

and a profilometer. It was found that both the ion beam flux and the stopping power of the ions 

in the polymer have a critical influence on the blister formation. 

Keywords: soft lithography, H + ion irradiation, PMMA resist, blister formation


71 

INTRODUCTION 

Numerous studies of ion-beam-induced modification of polymers have been performed in the 

last three decades because of its potential for technological applications [e.g. 1-3]. In recent years, ion 

beam lithography on a resist for the development of microfluidic devices has been extensively 

investigated [4-6]. The resist that has the tendency to form cross-linkage of the main polymer chains or 

pendant side chains during irradiation is of the negative type (also called negative tone). This is 

because the irradiated area becomes less soluble in developing solutions compared to an unirradiated 

region. On the other hand, the predominant modification in a positive resist (also called positive tone) 

is chain scissioning of the main and side polymer chains, which enhances the solubility rate in the 

exposed region [7]. Typically, the master mould for soft lithography is fabricated from either an ionirradiated 

negative resist such as SU-8 [8] or a positive resist such as poly(methyl methacrylate) 

(PMMA) [5]. In addition, Licciardello et al. [9] found that the PMMA would undergo a changeover 

from a positive to a negative tone by the action of high fluence irradiation, but the utilisation of 

PMMA as a negative tone resist for master mould fabrication has been little used [10]. For both 

positive and negative resists, a good master mould used in a direct replica moulding technique must 

have smooth surfaces. However, blisters or craters have been observed on irradiated polymer surfaces 

[11-12]. These defects can spoil the mould since they may introduce blockages into 

poly(dimethylsiloxane) (PDMS) replicas for microfluidic networks. Consequently, blisters and craters 

must be avoided. This paper reports on an experimental study to control defects from 2-MeV H + ion 

irradiation until the dominance of cross-linking in PMMA has been achieved. Some effects of 1-MeV 

H + ion beam are also included for comparison. 

PRINCIPLE 

According to the stopping and range of ions in matter (SRIM) simulation program, version 

SRIM-2008 [13], for the projected range of 65 μm, the tracks of 2-MeV H + ions in PMMA is straight, 

especially in the first 10 μm. An individual ion loses about 200 keV of kinetic energy during its 

passage through the first 10 μm within 0.5 ps. For light incident ions with a velocity much greater than 

the Bohr velocity (2.2×10 6 m/s) in thin polymer targets, the energy deposited by the ion mainly results 

in the excitation and ionisation of atoms and molecules of the polymer in the ion-track zone. These 

initial physical processes of ion-solid interaction can cause chain scission, cross-linking and gas 

evolution [14]. In the case of PMMA, several simple low-mass gas products such as H 2 , CO, CO 2 and 

CH 4 have been detected, with carbon monoxide as the dominant product [15-16]. This is indicative of 

bond breaking, which is a precursor of both the chain scission and cross-linking. There were reports 

that the gas yield from the polymers is strongly dependent on the linear energy transfer (LET) of the 

ions [17] and on the incident ion fluence [18]. Although the fundamental mechanism is still not fully 

understood, the proposed models [11-12] agree that blisters and craters on the surface of a polymer are 

formed by the gas products as they pass into the surrounding vacuum.



The PMMA used in this work had a molecular weight of 950 kDa (950 PMMA A11, 

MicroChem). All PMMA films were spin-coated on clean 1×1-cm 2 silicon substrates at a spin speed of 

2,500 rpm for 45 sec. by using a self-made spin coater. Subsequently, the films were soft-baked on a 

hot-plate at 160 o C for 2 min. This process was repeated three times to attain a total film thickness of 

8.8±0.1 μm as measured with a stylus profilometer (P-15 Profiler, KLA-Tencor, USA). The 1.7-MV 

tandem accelerator (1.7 MV high current Tandetron, High Voltage Engineering Europa B. V., the 

Netherlands) was used to irradiate the polymer with 1- and 2-MeV H + ion beams. The pressure during 

irradiation was about 5×10 -6 mbar. The irradiated area was 500×500 μm 2 for all experiments and was 

defined by two computerised L-shaped blade apertures of the programmable proximity aperture 

lithography (PPAL) system [10]. The experimental set-up when utilising the PPAL technique is shown 

in Figure 1. The two L-shaped blades were made of copper plates 100 μm thick with well-polished 

edges. Each of the L-shaped blades was mounted on a computerised micro-stepper motor that 

independently moved with high precision in either vertical or horizontal direction. In this manner, the 

PPAL system could produce any rectangular pattern with dimensions between 1-1000 μm 2 . The 

aperture was located at ~2 mm in front of the sample. The PMMA film on silicon substrate was 

mounted to the translation stage holder, which could move in both x and y directions with a resolution 

of 1 μm. 

x 

Electron 

suppressor 

Aperture 

y 

Sample 

Faraday cup 

A 

Pico-ammeter 

Sample holder 

~3mm 

2-MeV H + 

ion beam 

-200 V 

Alumina 

screen 

A 

Pico-ammeter 

Figure 1. Schematic illustration of an experimental set-up using the PPAL technique for H + ion beam 

irradiation (not to scale) 

The key parameters in this study were the values of ion beam fluence (ions/cm 2 ) and ion beam 

flux (ions/cm 2·s) at the irradiated areas. To ensure the accuracy of these measurements, the ion beam 

current at two different positions, as shown in Figure 1, was measured with high precision. A 

picoammeter with 4.8-pC resolution was specially designed for this purpose. To assure correctness in 

measuring, an electron suppressor with a 5-mm-diameter hole and a -200V potential was placed in 

front of the aperture to prevent secondary electrons from leaving the copper blades. 

The picoammeter was used to measure the ion beam current detected by an 8-mm-innerdiameter, 

65-mm-long Faraday cup. The Faraday cup was isolated and buried in the sample holder. 

The ion beam current was collected every second from start to stop. Together with the known hole area


73 

at the aperture, the ion beam fluence and flux could be easily calculated. Moreover, the major part of 

the incident ion beam blocked by the aperture was monitored also; this was used to fine-tune the 

accelerator for a stable ion beam current. 

After irradiation, the surface morphology of the irradiated areas was investigated by an optical 

microscope and a scanning electron microscope (SEM) (JSM-5410LV, JEOL, Japan), while the 

shrinkage of the irradiated areas was measured by the profilometer. 


It was found that for 2-MeV H + ions, the cross-linking of PMMA within the entire pattern 

occurred when the fluence exceeded 3.5×10 14 ions/cm 2 , which is consistent with a previous report [19]. 

Therefore, all the incident fluences used in this work were above this value. Accordingly, in Figure 2 

both a and d, b and e, and c and f patterns were irradiated by the 2-MeV H + ion beam with the same 

fluence, viz. 1.0×10 15 , 1.25×10 15 and 1.75×10 15 ions/cm 2 respectively, while the ion beam flux used 

for patterns a-c and d-f was 4.7×10 11 and 3.0×10 12 ions/cm 2·s respectively. It was clearly seen that 

flaws on the PMMA surface were strongly dependent on the ion beam flux. For the same irradiation 

fluence but with small values of ion current, the flaws were absent. Figure 3(a) shows that the flaws 

are blisters and not craters. Each isolated blister has a common appearance of a round shape with about 

the same diameter. 

Figure 2. Optical microscope images (13×) of the six 500×500 μm 2 patterns of irradiated areas on the 

PMMA thin film 

The surface profiles of patterns a, b and c in Figure 2 were measured by the profilometer. As 

shown in Figure 4, the surface regions modified by the 2-MeV H + ion beam are remarkably lower than 

those in the unirradiated regions. Also, the compaction or shrinking occurs even at the smooth patterns 

and increases with ion fluence. Hnatowicz and Fink [20] reported that the compaction of a polymer is 

initiated by cross-linking and tends to increase the material density. Experimental evidence from this 

study suggests that the gas evolution always occurs during irradiation even though no blister is


observed, as shown in Figure 2 (a-c). However, the gas-release mechanisms for low-flux and high-flux 

irradiation may be different. More information can be found in a comparison of the smooth pattern a 

with the blister-filled pattern d, for example. Both of them experienced the same fluence but the 

irradiation duration of pattern a was 1.72 times longer than that of pattern d. 

Figure 3. Tilted SEM images of the blisters created by (a) 2-MeV H + ions and (b) 1-MeV H + ions. 

The average blister diameters are (a) 61.0±5.8 μm and (b) 92.8±2.5 μm. 

Figure 4. The surface profile of the 3 irradiated areas in the top row of Figure 2 

It is reasonable to draw a conclusion that each ion track can be considered isolated in the case 

of low flux. Degassing due to the pressure gradient is thought to be the main mechanism. Moreover, 

the very thin PMMA film would allow the gases to leave the polymer easily without much build-up. 

On the other hand, in the high ion flux case, the temporal interval between the closely-spaced incident 

ion tracks would allow a build-up of dense low-molecular-weight fragments from intense main-chain 

scissions. This might nucleate gas bubbles in the whole irradiated volume of the polymer and 

accumulate gases which eventually converted it into a ‘foamed’ structure [21]. In this case, the gas 

bubbles that reached the polymer surface would form blisters.


75 

It was also found that a 1-MeV H + ion beam creates blisters at lower ion beam flux than does a 

2-MeV H + ion beam. By comparing Figure 3(b) with Figure 3(a), we can see that although each 

isolated blister has the same shape, the blisters from 1-MeV irradiation are significantly larger in 

diameter than those from 2-MeV irradiation. From the SRIM code, the stopping power is 1.5 times and 

the displacement damage 2.4 times greater in 1-MeV than in 2-MeV proton irradiation. 

CONCLUSIONS 

At an ion irradiation fluence of 10 14 ions/cm 2 of 2-MeV H + ions, the cross-linking process in 

PMMA started to overcome the chain scission process. The full cross-linking began at an ion fluence 

of 6.6×10 14 ions/cm 2 . The formation of blisters was strongly dependent on the ion beam flux and 

stopping power of the polymer for the ions. A 2-MeV proton flux of less than 4.7×10 11 ions/cm 2·s 

achieved a blister-free condition for the PMMA film ~9 μm thick. This work has confirmed that 

blisters are created by gas evolution due to chain scissions induced by the ion beams, especially in the 

early stage of irradiation. 


This work was partly supported by the CoE Program of Chiang Mai University and the 

International Atomic Energy Agency (IAEA, Vienna). S.U. gratefully acknowledges a scholarship 

from the ThEP Centre. N. P. (N. Puttaraksa) gratefully acknowledges financial support from Thailand 

Research Fund (TRF) in the form of an RGJ scholarship. H. J. W.’s and N. P.’s work was carried out 

under the Academy of Finland Centre of Excellence in Nuclear and Accelerator Based Physics (Ref. 

213503). H. J. W. is grateful for a senior researcher grant from the Academy of Finland (Ref. 129999). 

The authors also wish to thank Mr. Chome Thongleurm and Mr. Witoon Ginamoon for their technical 

support. 

REFERENCES 

1. T. M. Hall, A. Wagner and L. F. Thompson, "Ion beam exposure characteristics of resists: 

Experimental results", J. Appl. Phys., 1982, 53, 3997-4010. 

2. S. V. Springham, T. Osipowicz, J. L. Sanchez, L. H. Gan and F. Watt, "Micromachining using 

deep ion beam lithography", Nucl. Instr. Meth. Phys. Res. B, 1997, 130, 155-159. 

3. F. Menzel, D. Spemann, S. Petriconi, J. Lenzner and T. Butz, "Proton beam writing of 

submicrometer structures at LIPSION", Nucl. Instr. Meth. Phys. Res. B, 2007, 206, 419-425. 

4. P. G. Shao, J. A. van Kan, K. Ansari, A. A. Bettiol and F. Watt, "Poly(dimethylsiloxane) 

micro/nanostructure replication using proton beam written masters", Nucl. Instr. Meth. Phys. 

Res. B, 2007, 260, 479-482. 

5. S. Gorelick, N. Puttaraksa, T. Sajavaara, M. Laitinen, S. Singkarat and H. J. Whitlow, 

"Fabrication of microfluidic devices using MeV ion beam Programmable Proximity Aperture 

Lithography (PPAL)", Nucl. Instr. Meth. Phys. Res. B, 2008, 266, 2461-2465. 

6. C. Udalagama, E. J. Teo, S. F. Chan, V. S. Kumar, A. A. Bettiol and F. Watt, "Proton beam 

writing of long, arbitrary structures for micro/nano photonics and fluidics applications", Nucl. 

Instr. Meth. Phys. Res. B, 2011, 269, 2417-2421.


7. M. J. Madou, "Fundamentals of Microfabrication: The Science of Miniaturization", 2nd Edn., 

CRC Press, Boca Raton, 2002, pp.6-7. 

8. V. Auzelyte, M. Elfman, P. Kristiansson, C. Nilsson, J. Pallon, N. A. Marrero and M. Wegdén, 

"Exposure parameters for MeV proton beam writing on SU-8", Microelectron. Eng., 2006, 83, 

2015-2020. 

9. A. Licciardello, M. E. Fragalà, G. Foti, G. Compagnini and O. Puglisi, "Ion beam effects on the 

surface and on the bulk of thin films of polymethylmethacrylate", Nucl. Instr. Meth. Phys. B, 

1996, 116, 168-172. 

10. N. Puttaraksa, S. Unai, M. W. Rhodes, K. Singkarat, H. J. Whitlow and S. Singkarat, 

"Fabrication of a negative PMMA master mold for soft-lithography by MeV ion beam 

lithography", Nucl. Instr. Meth. Phys. B, 2012, 272, 149-152. 

11. J. Kaur, S. Singh, D. Kanjilal and S. K. Chakarvarti, "Nano/micro surface structures by swift 

heavy ion irradiation of polymeric thin films on GaAs", Digest J. Nanomater. Biostruct., 2009, 

4, 729-737. 

12. D. He and M. N. Bassim, "Atomic force microscope study of crater formation in ion 

bombarded polymer", J. Mater. Sci., 1998, 33, 3525-3528. 

13. J. F. Ziegler and J. P. Biersack, “The stopping and range of ions in matter”, 2008, 

http://www.srim.org. 

14. M. E. Fragalà, G. Compagnini, A. Licciardello and O. Puglisi, "Track overlap regime in ionirradiated 

PMMA", J. Polym. Sci. B, 1998, 36, 655-664. 

15. Z. Chang and J. A. LaVerne, "The gases produced in gamma and heavy-ion radiolysis of 

poly(methyl methacrylate)", Radiat. Phys. Chem., 2001, 62, 19-24. 

16. M. E. Fragalà, G. Compagnini, L. Torrisi and O. Puglisi, "Ion beam assisted unzipping of 

PMMA", Nucl. Instr. Meth. Phys. B, 1998, 141, 169-173. 

17. M. B. Lewis and E. H. Lee, "G-values for gas production from ion-irradiated polystyrene", J. 

Nucl. Mater., 1993, 203, 224-232. 

18. F. Schrempel and W. Witthuhn, "Deep light ion lithography in PMMA __ a parameter study", 

Nucl. Instr. Meth. Phys. B, 1997, 132, 430-438. 

19. N. Puttaraksa, R. Norarat, M. Laitinen, T. Sajavaara, S. Singkarat and H. J. Whitlow, 

"Lithography exposure characteristics of poly(methyl methacrylate) (PMMA) for carbon, 

helium and hydrogen ions", Nucl. Instr. Meth. Phys. B, 2012, 272, 162-164. 

20. V. Hnatowicz and D. Fink, "Macroscopic changes in ion-irradiated polymers", in 

"Fundamentals of Ion-Irradiated Polymers" (Ed. D. Fink), Springer-Verlag, Berlin, 2004, 

pp.349-375. 

21. A. Chapiro, "Chemical modifications in irradiated polymers", Nucl. Instr. Meth. Phys. B, 1988, 

32, 111-114. 



Maejo Maejo Int. J. Int. Sci. J. Technol. Sci. Technol. 2012, 6(01), 2012, 6(01), 77-94 77-9477 



Full Paper 



Degradation of bisphenol A by ozonation: rate constants, 

influence of inorganic anions, and by-products 

Kheng Soo Tay * , Noorsaadah Abd. Rahman and Mhd. Radzi Bin Abas 

Environmental Research Group, Department of Chemistry, Faculty of Science, University of Malaya, 

50603 Kuala Lumpur, Malaysia 

* Corresponding author, e-mail: khengsoo@um.edu.my 

Received: 8 February 2011 / Accepted: 23 February 2012 / Published: 27 February 2012 

Abstract: The second-order rate constants for the reaction between bisphenol A (BPA) and ozone 

were evaluated over the pH range of 2-12. The rate constants showed minimum values (×10 4 M -1 s -1 ) 

under acidic condition (pH < 4) and were of maximum values (×10 9 M -1 s -1 ) under basic condition 

(pH >10). From pH 4 to 7, the second-order rate constants were found to increase by a magnitude 

of almost 10 2 and this was due to the increase in anionic BPA species in the solution. The rate 

constants increased almost twofold when pH increased from 9.6 to 10.2. The presence of common 

inorganic anions at levels commonly found in the environment did not affect the rate of degradation 

of BPA. 

The degradation by-products from the ozonation of BPA were identified as 4-(prop-1-en-2- 

yl)phenol, hydroquinone, 4-hydroxyacetophenone, 2-(2-(4-hydroxyphenyl)propan-2-yl)succinaldehyde, 

2-(1-(4-hydroxyphenyl)vinyl)pent-2-enal, 3-formyl-4-(4-hydroxyphenyl)-4-methylpent-2- 

enoic acid, monohydroxy-BPA and dihydroxy-BPA. In conclusion, ozonation was found to be an 

effective method for the removal of BPA even in the presence of common inorganic anions at 

environmental concentrations. However, incomplete treatment of BPA might produce a variety of 

degradation by-products. 

Keywords: bisphenol A, ozonation, rate constant, anions, competitive kinetics, OH radical 

________________________________________________________________________________ 

INTRODUCTION 

Bisphenol-A (BPA; 2,2-bis(4-hydroxylphenyl)propane) is a commonly used chemical in the 

synthesis of polymers, especially for food and beverage packages. BPA has been reported to have 

developmental toxicity, carcinogenicity, possible neurotoxicity [1] and estrogenic effects [2]. This


chemical can leach from plastic products under normal and high temperatures [2-3]. Thus, the 

leaching of BPA into the environment from disposed plastic materials is expected. So far, legislation 

on the use of BPA and its discharge into the environment is sorely lacking. Consequently, the 

occurrences of BPA in the environment have been widely reported [e.g. 4-6]. The presence of BPA 

in coastal waters and supermarket seafood from Singapore and mussels from South and South-east 

Asia [7-8] indicates that BPA pollution is severe not only in developed countries but also in the 

Asian region. 

Since BPA has the potential to cause undesirable ecological and human health effects, various 

treatment technologies have been developed for its removal from water. The oxidative degradation 

methods for BPA such as ozonation [9-14], photo-Fenton reaction [15], photocatalytic reaction by 

TiO 2 [16] and ultrasound-UV-iron(II) treatment [17] have been reported. Among the treatment 

methods, ozonation has been projected to be one of the fastest growing water disinfection 

technologies in the market [18-19]. It has been shown to be effective in the removal of organic 

pollutants from water and wastewater [20-24]. During ozonation, organic pollutants undergo a series 

of oxidation processes by ozone (O 3 ) and hydroxyl radical (•OH) formed by the decomposition of O 3 

in water [25]. In some cases, toxic by-products may be generated [22]. Therefore, evaluation and 

determination of degradation by-products from the ozonation of organic pollutants is an important 

consideration. 

Although the ozonation of BPA has been widely reported, based on our literature review the 

influence of inorganic anions on the removal of BPA by ozonation has not been evaluated. The 

influence of inorganic anions is important since chloride and phosphate ions, for example, can react 

with O 3 and •OH, thus affecting the rate of removal of the organic pollutants in water [26-27]. The 

increase in salt concentration has also been reported to affect O 3 solubility in water [28]. The main 

objective of this study, therefore, is to evaluate the effect of inorganic anions, namely phosphate, 

nitrate, sulphate and chloride ions. In addition, the second-order rate constants for the reaction 

between BPA and O 3 at pH 2-12 were determined. The variation in the rate constant at different pH 

values, especially within the two pK a values (9.6 and 10.2) of BPA, was studied in detail in order to 

compare the reactivity of anionic and dianionic species of BPA in aqueous solution towards O 3 . The 

degradation by-products (DBPs) of BPA were also identified. Some DBPs of BPA have already been 

determined by previous studies [11,13]. However, in this study, we managed to identify a few 

additional compounds and degradation pathways of BPA during ozonation were proposed. 


Chemicals 

BPA (>99% purity) was obtained from Aldrich and used without further purification. Its 

stock solution (200 mg/L) was prepared by dissolving in boiling ultrapure deionised water (Elcagan, 

UK). All chemicals were used without further purification. Sodium phosphate (96%) and sodium 

dihydrogen phosphate (99%) was obtained from Aldrich. Sodium chloride (≥99.5%), sodium nitrate 

(99%) sodium sulphate decahydrate (>99%), sodium phosphate monobasic (99%) and tert-butyl 

alcohol (t-BuOH) (99.5%) were purchased from Sigma. Sodium phosphate dibasic (99%) and 

disodium hydrogen phosphate (99%) were purchased from Riedel-de-Haën. A mixture of BSTFA


79 

(N,O-bis(trimethylsilyl)trifluoroacetamide) and TMCS (trimethylchlorosilane) in a ratio of 99:1 was 

obtained from Supelco. All solvents (HPLC grade) and phosphoric acid (85%) were obtained from 

Merck. Sodium hydroxide (>98%) was purchased from Fluka. Purified oxygen (99.8%) was 

obtained from MOX-linde (Malaysia). Phosphate buffer (0.5 mol/L) was prepared using sodium 

dihydrogen phosphate and/or disodium hydrogen phosphate and the pH was adjusted using either 

phosphoric acid or sodium hydroxide solutions. O 3 was produced from purified oxygen by an ozone 

generator (model OZX03K, Enaly Trade Co. Ltd., Canada). All tubing from the ozone generator 

was of ozone-inert silicone material. 

Ozonation of BPA 

Determination of second-order rate constant for reaction of BPA with O 3 

A detailed description of the rate constant determination was given in our previous study 

[29]. Briefly, a competitive kinetic method was applied using phenol as a reference compound. 

Experiments were performed at room temperature (27-30ºC) in 20-mL vials with solution containing 

equal concentration of BPA and phenol (4 μmol/L) as well as 20 mmol/L of t-BuOH. The pH of the 

solution was adjusted using 20 mmol/L of phosphate buffer. Ozone solutions with concentrations 

ranging from 1.5 to 7.5 μmol/L were added. The aqueous O 3 stock solution was prepared by 

sparging O 3 at a rate of 0.70 g/hr into deionised water placed inside a water-jacketed beaker at 2ºC. 

The concentration of O 3 was measured by the indigo method [30]. The final volume of the mixtures 

was 20 mL, which also minimised the headspace. Each vial was then shaken vigorously. The time of 

total ozone consumption was estimated from the half-life of ozone in the BPA solution under 

different pH conditions. The half-lives for ozone in BPA solution were found to be 28, 9 and


hr. The silylated extract was dried with nitrogen stream and redissolved in 30 μL of 

dichloromethane,1.2 μL of which was analysed by gas chromatography-mass spectrometry (GC- 

MS). 

Analytical methods 

The concentration of BPA in the reaction mixture was determined using HPLC (Thermo 

Separation Product HPLC System P2000, HiTech Trader, USA) equipped with a UV detector and a 

degasser. A 250×4.6mm RP-8 (5μm) Lichrospher-100 analytical column (Merck) was used for 

separation. Acetonitrile (65%) in deionised water with 0.1% trifluoroacetic acid was used as the 

mobile phase. The separation was carried out under isocratic condition. The separated components 

were detected at 230 nm. The flow rate was maintained at 1.0 mL/min for all runs and the sample 

volume for HPLC analysis was 20 μL. 

The analysis of the degradation by-products was carried out using a Hewlett-Packard HP 

6890 gas chromatograph coupled with HP5972 mass spectrometer. The column was HP-5 (5% 

phenylmethylpolysiloxane) column with the dimension of 30 m 0.25 mm and 0.25 μm of film 

thickness. Helium was used as the carrier gas with an average flow rate of 40 cm/sec., and the GC 

oven temperature was initially 60°C (maintained for 2 min.) and increased to 280°C at the rate of 

6°C/min and maintained at this temperature for 2 min. The temperatures of the injection port and the 

transfer line were set at 290°C and 300°C respectively. The data for quantitative analysis was 

acquired in the electron impact mode (70 eV) with scanning in the range of 50-600 amu at 1.5 

sec./scan. 


Kinetics of Degradation of BPA by Ozonation 

The reaction of ozone with an organic compound has been reported to be of first order with 

respect to both ozone and the organic compound. Thus, the kinetics of ozonation can be expressed 

as a second-order reaction [32]. The determination of its rate constants was performed using a 

competitive kinetics method with phenol as reference compound [20, 29, 33-34]. Phenol was 

selected because it was expected to have a similar decomposition pathway and rate constant as BPA 

in the ozonation [34]. For phenol, the second-order rate constants for the reaction with O 3 at 

different pH values ( k 

app,Phenol-O 

) were calculated using equation 1: 

3 

-pH 

-pKa 

10 10 

kapp,Phenol-O k 

3 O 3,Phenol -pKa -pH 

k 

O 3,Phenolate 

 

-pKa -pH 

10 +10 10 +10 

(1) 

where pK a of phenol was 9.9 and the intrinsic rate constants for phenol and phenolate were 1.3 ×10 3 

M -1 s -1 ( k 

O 3 ,Phenol) and 1.4 ×10 9 M -1 s -1 ( k 

O 3 ,Phenolate 

) respectively [9, 34]. The rate constants for the 

reaction between O 3 and BPA ( ) were estimated using competitive kinetics method derived 

from equation 2 [35, 36]: 

kO3 

BPA


81 

[BPA] 

k 

n O3 

BPA 

[Phenol] 

 

n 

ln 

ln 

[BPA] 

k [Phenol] 

 

0 app,O3 

-Phenol 0 

(2) 

where [BPA] 

0 

and [Phenol] 

0 

represent the initial concentration of BPA and phenol and [BPA] n 

and 

[Phenol] n 

represent the concentration of BPA and phenol after the ozonation reaction at different 

ozone dose, n. 

Ozonation of BPA was carried out at pH 2.0-12.0 in the presence of excess t-BuOH 

([t-BuOH] / [O 3 ] > 300). t-BuOH is a radical scavenger added to scavenge the •OH radical, thus 

allowing BPA to react only with O 3 during ozonation. Lee et al. [9] and Deborde et al. [34] reported 

k at 20°C = 1.68×10 4 and 1.30×10 4 M -1 s -1 at pH 2, and 1.11×10 9 and 1.60×10 9 M -1 s -1 at pH 12 

o3 

-BPA 

respectively. In this work, somewhat higher values of ko3 

-BPA 

were obtained, viz. (1.7±0.5)×10 4 and 

(9.0±0.5)×10 9 M -1 s -1 at pH 2 and 12 respectively. This difference might be due to a large error that 

can occur in the competitive kinetics method [20] and also to a higher reaction temperature that was 

selected in this study. 

According to Staples et al. [37], the two dissociation constants of BPA are 9.6 (pK a 1) and 

10.2 (pK a 2). BPA is therefore a weak organic acid which can dissociate in solution as either an 

anionic or dianionic species. Generally, when pH = pK a , the undissociated and ionic species exist at 

equal concentration in solution. When pH < pK a , the undissociated species is predominant and when 

pH > pKa, the ionic species is predominant [38]. Therefore, for BPA, when pH < pK a 1, 

undissociated BPA exists predominantly in water. When pK a 1 < pH < pK a 2, anionic BPA is the 

predominant species and when pH > pK a 2, dianionic BPA are predominant (Figure 1). 

Figure 1. Dissociation of BPA 

It has been previously shown that in the presence of the •OH radical, the rate of BPA 

degradation by ozone increases from pH 2 to 7 and then decreases at pH 10 due to •OH scavenging 

by carbonate and bicarbonate ions [14]. In this study, t-BuOH was added to scavenge the •OH 

radical from the beginning in order to study the rate constants for BPA degradation by O 3 only. The 

results are shown in Figure 2, indicating that the reactivity increases as the pH increases from 2 to 

12. Under acidic condition (pH < 7), the reaction would be mainly between O 3 and undissociated 

BPA. Between pH 4-7, the rate constant was observed to increase by a magnitude of almost 10 2 and 

this would be due to the increase of anionic species of BPA in the solution. The rate constant 

increased almost twofold when the pH increased from 9.6 (= pK a 1) to 10.2 (= pK a 2).


10 9 

10 8 

Rate constants (M -1 s -1 ) 

10 7 

10 6 

10 5 

10 4 

10 3 

pH 

2 4 6 8 10 12 14 

pH 

Rate constants (M -1 s -1 ) 

1.0x10 9 

8.0x10 8 

pK a 

2 

6.0x10 8 

4.0x10 8 

2.0x10 8 

0.0 

pK a 

1 

6 8 10 12 14 

Figure 2. Variation of second-order rate constant of BPA-ozone reaction at different pH values 

In order to study the effect of the ratio of [BPA 2- ] / [BPA - ] on the rate constant, the obtained 

second-order rate constants were plotted against [BPA 2- ] / [BPA - ] values at pH between 9.6-10.3 

(Figure 3). The ratio of [BPA 2- ] / [BPA - ] was estimated using Henderson-Hasselbach equation as 

follows [39]: 

2 

[BPA ] 

pH pKa 

2 log [BPA 

] 

 

(3) 

 

9x10 8 

Second-order rate constnats (M -1 s -1 ) 

8x10 8 

7x10 8 

6x10 8 

5x10 8 

4x10 8 

R 2 = 0.9973 

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 

Estimated [ BPA 2- ]/[ BPA - ] 

Figure 3. Effect of [BPA 2- ] / [BPA - ] on second-order rate constant of BPA-ozone reaction at the 

pH 9.6-10.3 

From Figure 3, the rate constant increases proportionally with [BPA 2- ] / [BPA - ], which 

implies that the dianionic species of BPA (BPA 2- ) is more easily oxidised by O 3 compared to its 

anionic counterpart (BPA - ). From Figure 2, it can be clearly observed that when the pH value


83 

increases further from 10.2 to 12.0, the reactivity of BPA with O 3 remains almost constant. This is 

most likely due to the fact that the maximum fraction of BPA 2- has been reached at pH 10.2 and 

further increasing of pH does not influence the amount of BPA 2- anymore. 

Influence of Inorganic Anions 

Most wastewaters normally contain inorganic anions coexisting with organic pollutants. The 

effects of the anions on the rate of BPA degradation was studied at concentrations found in 

wastewaters [40-42]. Figure 4 shows plots of BPA degradation in the presence of phosphate, nitrate, 

chloride and sulphate ions. The results indicate that the presence of common inorganic anions at 

concentration levels in the wastewaters does not significantly affect the rate of BPA degradation. 

This might be due to the fast reaction between O 3 and BPA as indicated by the determined secondorder 

rate constants. 

(a) 

(b) 

1.0 

0.8 

without phosphate 

0.05 ppm phosphate 



1.0 

0.8 

without nitrate 

0.1ppm nitrate 

1.0 ppm nitrate 

10 ppm nitrate 

[BPA]/[BPA] 0 

0.6 

0.4 

[BPA]/[BPA] 0 

0.6 

0.4 

0.2 

0.2 

0.0 

0.0 

0 2 4 6 8 10 

Time (min) 

0 2 4 6 8 10 

Time (min) 

(c) 

(d) 

[BPA]/[BPA] 0 

1.0 

0.8 

0.6 

0.4 

without chloride 

0.5 ppm chloride 

5 ppm chloride 



[BPA]/[BPA] 0 

1.0 

0.8 

0.6 

0.4 

without sulfate 

0.2 ppm sulfate 

2 ppm sulfate 

20 ppm sulfate 

200 ppm sulfate 

0.2 

0.2 

0.0 

0.0 

0 2 4 6 8 10 

Time (min) 

0 2 4 6 8 10 

Time (min) 

Figure 4. Effect of (a) phosphate, (b) nitrate, (c) chloride and (d) sulphate ions on the degradation 

of BPA (temperature = 25ºC; O 3 dose = 0.69 g/h; pH = 6.5; [BPA] 0 = 100 mg/L) 

Degradation By-Products (DBPs) of BPA 

GCMS analyses were performed by comparing the chromatogram of BPA with those of the 

aliquots taken at different ozonation times. All samples were subjected to similar derivatisation


procedure as mentioned in the experimental section. A chromatogram showing the distribution of 

DBPs of BPA is presented in Figure 5 and the proposed DBPs are presented in Table 1. All 

derivatised compounds occurred as trimethylsilyl derivatives characterised by the peak at m/z 73 in 

the mass spectrum. Identification of DBPs was carried out based on fragmentation patterns in the 

mass spectrum and/or by comparing the mass spectrum with the library available in the instrument 

database. 

BPA 

Relative abundance 

DBP 7 

DBP 6 

DBP 2 DBP 4 DBP 5 

DBP 1 DBP3 

DBP 8 

Retention time (min) 

Figure 5. Gas chromatogram of BPA after 4 min. of ozonation time ([BPA] 0 = 100 mg/L, pH = 6.5, 

temperature = 25 ºC, O 3 dose = 0.70 g/hr) 

The mass spectrum of BPA shows the molecular ion peak at m/z 372 (Figure 6a). Besides the 

peak at m/z 357 representing the loss of methyl group from trimethylsilyl group ((M-CH 3 ) + ), the 

other significant peak is at m/z 207 representing the loss of trimethyl(phenoxy)silane from silylated 

BPA. DBP 1 , DBP 2 and DBP 3 represent the breakdown products of BPA and their mass spectra (as 

silylated derivatives) are presented in Figures 6 (b-d). The formation of DBP 1 , DBP 2 and DBP 3 has 

also been detected in previous studies [13,15,16]. 

The mass spectra of DBP 4 and DBP 5 , formed by aromatic-ring opening during the ozonation 

process, are presented in Figures 7a and 7b respectively. Silylated DBP 5 with a molecular weight of 

292 amu shows a peak representing (M−1) +· at m/z 291 (Figure 7b). DBP 6 is a BPA degradation byproduct 

proposed by Deborde et al. [11]. The structure of this compound was derived based on its 

fragmentation pattern in the mass spectrum (Figure 7c). The peak at m/z 378 corresponds to the 

molecular ion peak of silylated DBP 6 . Other major peaks are at m/z 212 and 217. The peak at m/z 

212 represents the radical cation of silylated 3-formyl-4-methylpenta-2,4-dienoic acid. The peak at 

m/z 217 represents the loss of silylated carboxylic acid and aldehyde functional groups from the 

parent compound. 

Monohydroxylated (DBP 7 ) and dihydroxylated (DBP 8 ) BPA were also detected during the 

ozonation process. The molecular ion peaks of DBP 7 and DBP 8 are at m/z 460 (Figure 7d) and m/z 

548 (Figure 7e) respectively. Additional 88 amu and 176 amu over the molecular ion peak of


85 

silylated BPA (m/z 372) indicated the addition of one and two [(CH 3 ) 3 SiO] + groups to silylated BPA 

respectively. 

Table 1. Degradation by-products of BPA 

Compound identified 

Retention 

time (min.) 

(Label) 

Proposed structure of 

compound 

(Molecular weight) 

Name 

(m/z 372) 

25.01 

BPA 

(228.3) 

Bisphenol A 

(m/z 206) 

15.71 

(DBP 1 ) 

(134.1) 

4-(Prop-1-en-2-yl)phenol 

(m/z 254) 

16.23 

(DBP 2 ) 

(110.0) 

Hydroquinone 

(m/z 208) 

17.15 

(DBP 3 ) 

(136.1) 

4-Hydroxyacetophenone 

(m/z 292) 

19.23 

(DBP 4 ) 

HO O O 

(220.1) 

2-(2-(4- 

Hydroxyphenyl)propan- 

2-yl)succinaldehyde 

(m/z 274) 

20.24 

(DBP 5 ) 

(202.1) 

2-(1-(4- 

Hydroxyphenyl)vinyl)- 

pent-2-enal 

TMSO 

O O OTMS 

(m/z 378) 

25.09 

(DBP 6 ) 

(234.1) 

3-Formyl-4-(4- 

hydroxyphenyl)-4- 

methylpent-2-enoic acid 

(m/z 460) 

26.20 

(DBP 7 ) 

(244.1) 

Monohydroxy-BPA 

(m/z 548) 

27.01 

(DBP 8 ) 

(260.1) 

Dihydroxy-BPA


(a) 

(b) 

m/z 

m/z 

(c) 

(d) 

m/z 

m/z 

Figure 6. Mass spectra of DBP 1 , DBP 2 , DBP 3 and BPA 

Figure 8 shows the time profiles of the major DBPs, most of which were successfully 

removed after 10 min. of ozonation, with the exception of DBP 3 , DBP 6 and DBP 8 . These three byproducts 

thus seemed to be more resistant to ozonation compared to others. In the degradation 

experiment, ozonation was performed without a radical scavenger, so BPA could react with both O 3 

and •OH. As compared to the reaction between O 3 and BPA [11], the results in this study show that 

the presence of •OH seemed to produce more species of the breakdown products such as DBP 1 , 

DBP 3 and DBP 4 , which was most likely due to the non-selective behaviour of •OH, which can also 

react at the aliphatic chain and aromatic ring of BPA [33, 43].


87 

Mass spectrum 

Fragmentation pathway in mass 

spectrum 

(a) DBP 4 

m/z 

(b) DBP 5 

m/z 

(c) DBP 6 

m/z 

Figure 7. Mass spectra and fragmentation patterns of DBP 4 , DBP 5 , DBP 6 , DBP 7 and DBP 8


(d) DBP 7 

(e) DBP 8 

m/z 

m/z 

Figure 7. (continued) 

BPA 

DBP 1 

1.0 

DBP 2 

DBP 3 

Relative abundance 

0.8 

0.6 

0.4 

0.2 

DBP 4 

DBP 5 

DBP 6 

DBP 7 

DBP 8 

0.0 

0 2 4 6 8 10 

Ozonation time (min) 

Figure 8. Time profiles of major BPA degradation by-products ([BPA] 0 = 100 mg/L, pH = 6.5, 

temperature = 25ºC, and O 3 dose = 0.69 g/h). Relative abundance of the degradation byproducts 

was calculated by normalising the peak areas at the defined ozonation time to the 

highest peak areas.


89 

Formation of Degradation By-Products 

Formation of DBP 1 , DBP 2 , DBP 3 , DBP 7 and DBP 8 

Formation of DBP 7 and DBP 8 can occur through hydroxylation of BPA or via a direct 

reaction between BPA and O 3 (Figure 9). Hydroxylated BPA is proposed to be an important 

intermediate of by-products resulting from ring opening, especially DBP 4 , DBP 5 and DBP 6 . 

Formation of polyhydroxylated BPA through the reaction between BPA and •OH has also been 

reported during the degradation of BPA by ultrasound-UV-iron (II) treatment [17]. 

The formation of DBP 1 , DBP 2 and DBP 3 are presented in Figure 9b. Since O 3 with its 

eletrophilic nature reacts selectively with an electron-rich reaction site, it is proposed that the 

reaction begins at the side chain of BPA with an initial attack of •OH via hydrogen abstraction and 

leads to the formation of radical A. Intramolecular rearrangement of radical A then leads to the 

cleavage of C-C bond to form DBP 1 (4-isopropenylphenol) and a phenol radical, B. B will then react 

with •OH to form DBP 2 (hydroquinone). DBP 1 can further react with either O 3 or •OH, leading to 

the formation of dihydroxylated DBP 1 (C). C would then react with •OH via hydrogen abstraction to 

form radical D, which undergoes a fragmentation to form DBP 3 (4-hydroxyacetophenone). 

(a) 

(b) 

Figure 9. Proposed pathways for the formation of: (a) DBP 1 , DBP 2 and DBP 3 ; (b) DBP 7 and DBP 8 

from BPA during ozonation


Formation of DBP 4 , DBP 5 and DBP 6 

The formation pathway of DBP 4 is presented in Figure 10. The formation of DBP 6 is 

proposed to begin with the initial attack of •OH on the monohydroxylated BPA giving a radical 

intermediate (F) (Figure 11a). Intra-molecular rearrangement of F and cleavage of C-C bond lead to 

the opening of an aromatic ring, resulting in the formation of radical G, which then rearranges to 

form radical H, which reacts with water to form I. Attack of •OH at the enol site of I affords radical 

J, which gives radical K on cleavage of a C-C bond. Radical K further reacts with •OH to form L, 

which upon hydrogen abstraction gives radical M. M then further reacts with •OH leading to the 

formation of DBP 6 . The formation pathway of DBP 5 is presented in Figure 11b. 

- 

H 

HO 

BPA 

OH 

OH 

HO 

H 

OH 

OH 

HO 

OH 

OH 

OH 

+ OH 

O 

OH 

OH 

OH 

HO 

HO 

H 

OH 

HO 

H 

OH 

O 

HO 

OH 

O 

H 

+H 

O 

O 

O 

OH 

- 

CH 2 OH 

O 

H 

- H 

O 

HO 

OH 

HO 

OH 

OH 

HO 

OH 

OH 

- CO 2 

O 

OH 

CH 

OH 

HO 

O 

HO 

OH 

HO 

OH 

DBP 4 

Figure 10. Proposed pathway for the formation of DBP 4


91 

(a) 

(b) 

O 

OH 

O 

OH 

O 

OH 

HO 

G 

O H 

HO 

O 

OH 

HO 

O 

OH 

O 

O 

O 

H 

O 

OH 

- HOO 

O 

OH 

HO 

- H 

HO 

+ H 

HO 

CH 2 

H OH 

H 

OH 

OH 

CH 2 

CH 

- CO 2 

- CH 3 

OH 

O 

HO 

HO 

HO 

- H 

CH 2 

O 

O 

HO 

DBP 5 

HO 

Figure 11. Proposed pathways for the formation of: (a) DBP 5 and (b) DBP 6 

CONCLUSIONS 

The rate constant for the reaction between BPA and ozone in aqueous solution in the 

presence of t-BuOH increased with increase in pH between pH 2-10, beyond which it was fairly 

constant between pH 10-12. The reactivity of BPA increased in the order: undissociated form < 

anionic form < dianionic form. The presence of common inorganic anions (chloride, sulphate, nitrate


and phosphate ions) at environmental levels did not significantly affect the rate of degradation of 

BPA by ozone. 

The degradation products of BPA during ozonation were identified to be 4-(prop-1-en-2- 

yl)phenol, hydroquinone, 4-hydroxyacetophenone, 2-(2-(4-hydroxyphenyl)propan-2-yl)succinaldehyde, 

2-(1-(4-hydroxyphenyl)vinyl)pent-2-enal, 3-formyl-4-(4-hydroxyphenyl)-4-methylpent-2- 

enoic acid, monohydroxy-BPA and dihydroxy-BPA. 


This research was financially supported by Malaysia Toray Science Foundation (MTSF), 

University of Malaya (Grant No. P0266-2007A, SF025-2007A, PS153-2007B and FS305-2007C) 

and Ministry of Higher Education Malaysia (FRGS). 

REFERENCES 

1. Y. M. Lee, M. J. Seong, J. W. Lee, J. K. Lee, T. M. Kim, S. Y. Nam, D. J. Kim, Y. W. Yun, T. 

S. Kim, S.Y. Han and J. T. Hong, “Estrogen receptor independent neurotoxic mechanism of 

bisphenol A, an environmental estrogen”, J. Vet. Sci., 2007, 8, 27-38. 

2. J. A. Brotons, M. F. Olea-Serrano, M. Villalobos, V. Pedraza and N. Olea, “Xenoestrogens 

released from lacquer coatings in food cans”, Environ. Health Perspect., 1995, 103, 608-612. 

3. C. Dash, M. Marcus and P. D. Terry, “Bisphenol A: Do recent studies of health effects among 

humans inform the long-standing debate”, Mutat. Res., 2006, 613, 68-75. 

4. D. de A. Azevedo, S. Lacorte, P. Viana and D. Barcelό, “Occurrence of nonylphenol and 

bisphenol-A in surface waters from Portugal”, J. Braz. Chem. Soc., 2001, 12, 532-537. 

5. E. Z. Harrisson, S. R. Oakes, M. Hysell and A. Hay, “Organic chemicals in sewage sludges”, 

Sci. Total Environ., 2006, 367, 481-497. 

6. S. Zhang, Q. Zhang, S. Darisaw, O. Ehie and G. Wang, “Simultaneous quantification of 

polycyclic aromatic hydrocarbons (PAHs), polychlorinated biphenyls (PCBs), and 

pharmaceuticals and personal care products (PPCPs) in Mississippi river water, in New Orleans, 

Louisiana, USA”, Chemosphere, 2007, 66, 1057-1069. 

7. C. Basheer, H. K. Lee and K. S. Tan, “Endocrine disrupting alkylphenols and bisphenol-A in 

coastal waters and supermaket seafood from Singapore”, Mar. Pollut. Bull., 2004, 48, 1161- 

1167. 

8. T. Isobe, H. Takada, M. Kanai, S. Tsutsumi, K. O. Isobe, R. Boonyatumanond and M. P. 

Zakaria, “Distribution of polycyclic aromatic hydrocarbons (PAHs) and phenolic endocrine 

disrupting chemicals in South and Southeast Asian mussels”, Environ. Monit. Assess., 2007, 

135, 423-440. 

9. J. Lee, H. Park and J. Yoon, “Ozonation characteristics of bisphenol A in water”, Environ. 

Technol., 2003, 24, 241-248. 

10. B. Xu, N. Gao, M. Rui, H. Wang and H. Wu, “Degradation of endocrine disruptor bisphenol A 

in drinking water by ozone oxidation”, Front. Environ. Sci. Eng. China, 2007, 1, 350-356. 

11. M. Deborde, S. Rabouan, P. Mazellier, J. P. Duguet and B. Legube, “Oxidation of bisphenol A 

by ozone in aqueous solution”, Water Res., 2008, 42, 4299-4308.


93 

12. F. J. Rivas, Á. Encinas, B. Acedo and F. J. Beltrán, “Mineralization of bisphenol A by advanced 

oxidation processes”, J. Chem. Technol. Biotechnol., 2009, 84, 589-594. 

13. I. Gultekin, M. Valko and H. I. Nilsun, “Degradation of bisphenol-A by ozonation”, J. Adv. 

Oxidat. Technol., 2009, 12, 242-248. 

14. T. Garoma and S. Matsumoto, “Ozonation of aqueous solution containing bisphenol A: Effect 

of operational parameters”, J. Hazard Mater., 2009, 167, 1185-1191. 

15. H. Katsumata, S. Kawabe, S. Kaneco, T. Suzuki and K. Ohta, “Degradation of bisphenol A in 

water by the photo-Fenton reaction”, J. Photochem. Photobiol. A, 2004, 162, 297-305. 

16. K. Nomiyama, T. Tanizaki, T. Koga, K. Arizono and R. Shinohara, “Oxidative degradation of 

BPA using TiO 2 in water, and transition of estrogenic activity in the degradation pathways”, 

Arch. Environ. Contam. Toxicol., 2007, 52, 8-15. 

17. R. A. Torres, C. Pétrier, E. Combet, F. Moulet and C. Pulgarin, “Bisphenol A mineralization by 

integrated ultrasound-UV-iron (II) treatment”, Environ. Sci. Technol., 2007, 41, 297-302. 

18. J. Sowmya, “European emerging trends and technologies include UV, Ozone”, Water and 

Waste Water International Magazine (April, 2008), PennWell Publishing Corporation. 

(http://www.waterworld.com/index/display/article-display/329727/articles/water-wastewaterinternational/volume-23/issue-2/regional-focus/middle-east-north-africa/european-emergingtrends-amp-technologies-include-uv-ozone.html) 

(Accessed on 28 th December 2011). 

19. A. Conner, “Reducing cooling tower costs with ozone technology”, CTI Bibliography of 

Technical Papers __ Ozone, Cooling Technology Institute, 2005, http://www.cti.org/tech_papers/ 

ozone.php (Accessed on 28 th December 2011). 

20. M. M. Huber, S. Canonica, G. Y. Park and U. von Gunten, “Oxidation of pharmaceuticals 

during ozonation and advanced oxidation processes”, Environ. Sci. Technol., 2003, 37, 1016- 

1024. 

21. F. J. Beltrán, “Ozone Reaction Kinetics for Water and Wastewater Systems”, Lewis Publishers, 

Boca Raton, 2004, pp.1-4. 

22. K. Ikehata, N. J. Naghashkar and M. G. El-Din, “Degradation of aqueous pharmaceuticals by 

ozonation and advanced oxidation process: A review”, Ozone Sci. Eng., 2006, 28, 353-414. 

23. S. A. Snyder, E. C. Wert, D. J. Rexing, R. E. Zegers and D. D. Drury, “Ozone oxidation of 

endocrine disruptors and pharmaceuticals in surface water and wastewater”, Ozone Sci. Eng., 

2006, 28, 445-460. 

24. S. Esplugas, D. M. Bila, L. G. Krause and M. Dezotti, “Ozonation and advanced oxidation 

technologies to remove endocrine disrupting chemicals (EDCs) and pharmaceuticals and 

personal care products (PPCPs) in water effluents”, J. Hazard Mater., 2007, 149, 631-642. 

25. M. D. H. Guil, M. Petrović, J. Radjenovic, A. R. Fernández-Alba, A. R. Fernández-Alba and D. 

Barceló, “Removal of pharmaceuticals by advanced treatment technologies”, in “Comprehensive 

Analytical Chemistry, Volume 50: Analysis, Fate and Removal of Pharmaceuticals in the Water 

Cycle”, (Ed. M. Petrović and D. Barceló), Elsevier, Amsterdam, 2007, pp.451-474. 

26. J. Staehelin and J. Hoigné, “Decomposition of ozone in water in the presence of organic solutes 

acting as promoters and inhibitors of radical chain reactions”, Environ. Sci. Technol., 1985, 19, 

1206-1213. 

27. M. S. Siddiqui, “Chlorine-ozone interactions: Formation of chlorate”, Water Res., 1996, 30, 

2160-2170.


28. M. A. Boncz, H. Bruning, W. H. Rulkens, H. Zuilhof and E. J. R. Sudhölter, “The effect of salts 

on ozone oxidation processes”, Ozone Sci. Eng., 2005, 27, 287-292. 

29. K. S. Tay, N. A. Rahman and M. R. B. Abas, “Ozonation of parabens in aqueous solution: 

Kinetics and mechanism of degradation”, Chemosphere, 2010, 81, 1446-1453. 

30. H. Bader and J. Hoigné, “Determination of ozone in water by the indigo method”, Water Res., 

1981, 15, 449-456. 

31. K. S. Tay, N. A. Rahman and M. R. B. Abas, “Degradation of DEET by ozonation in aqueous 

solution”, Chemosphere, 2009, 76, 1296-1302. 

32. J. Hoigné and H. Bader, “Rate constants of reactions of ozone with organic and inorganic 

compounds in water – II: Dissociating organic compounds”, Water Res., 17, 185-194. 

33. K. S. Tay, N. A. Rahman and M. R. B. Abas, “Removal of selected endocrine disrupting 

chemicals and personal care products in surface waters and secondary wastewater by ozonation”, 

Water Environ. Res., 2011, 83, 684-691. 

34. M. Deborde, S. Rabouan, J. P. Duguet and B. Legube, “Kinetics of aqueous ozone-induced 

oxidation of some endocrine disruptor”, Environ. Sci. Technol., 2005, 39, 6086-6092. 

35. J. Benner, E. Salhi, T. Ternes and U. von Gunten, “Ozonation of reverse osmosis concentrate: 

Kinetics and efficiency of beta blocker oxidation”, Water Res., 2008, 42, 3003-3012. 

36. F. J. Benitez, J. L. Acero, F. J. Real and G. Roldán, “Ozonation of pharmaceutical compounds: 

Rate constants and elimination in various water matrices”, Chemosphere, 2009, 77, 53-59. 

37. C. A. Staples, P. B. Dorn, G. M. Klecka, S. T. Oblock and L. R. Harris, “A review of the 

environmental fate, effects, and exposures of bisphenol A”, Chemosphere, 1998, 36, 2149-2173. 

38. C. Tian, J. T. Wang and X. L. Song, “Sediment-water interactions of bisphenol A under 

simulated marine conditions”, Water Air Soil Pollut., 2009, 199, 301-310. 

39. J. C. Kotz, P. M. Treichel and J. R. Townsend "Chemistry and Chemical Reactivity", 7th Edn., 

Brooks/Cole, Belmont, 2010, pp.810-859. 

40. S. V. Karmarkar, “Analysis of wastewater for anionic and cationic nutrients by ion 

chromatography in a single run with sequential flow injection analysis”, J. Chromatogr. A, 1999, 

850, 303-309. 

41. I. R. Santos, R. C. Costa, U. Freitas and G. Fillmann, “Influence of effluents from a wastewater 

treatment plant on nutrient distribution in a coastal creek from southern Brazil”, Braz. Arch. 

Biol. Technol., 2008, 51, 153-162. 

42. O. A. Al-Khashman, “Chemical evaluation of Ma’an sewage effluents and its reuse in irrigation 

purpose”, Water Resour. Manage., 2009, 23, 1041-1053. 

43. K. S. Tay, N. A. Rahman and M. R. B. Abas, “Characterization of atenolol transformation 

products in ozonation by using rapid resolution high-performance liquid chromatography/ 

quadrupole-time-of-flight mass spectrometry”, Microchem. J., 2011, 99, 312-326. 



Full Paper 

Maejo Maejo Int. J. Int. Sci. J. Technol. Sci. Technol. 2012, 6(01), 2012, 6(01), 95-104 95-10495 





Chitinase production and antifungal potential of endophytic 

Streptomyces strain P4 

Julaluk Tang-um and Hataichanoke Niamsup* 

Department of Chemistry and Centre of Excellence for Innovation in Chemistry, Faculty of Science, 

Chiang Mai University, Chiang Mai 50200, Thailand 

* Corresponding author, e-mail: hataichanoke.n@cmu.ac.th 

Received: 2 August 2011 / Accepted: 11 February 2012 / Published: 29 February 2012 

Abstract: The endophytic actinomycete P4 strain, previously isolated from sweet pea root, was 

identified as Streptomyces sp. by full 16S rRNA sequencing. It is mostly related to Streptomyces 

griseoflavus with a 99.7% identity score. The Streptomyces sp. P4 was tested for its hydrolytic 

activities by plate method. The result showed the presence of chitinase. The extent of chitinase activity 

was assessed by spectrophotometric method along with growth monitoring. Chitinase production was 

growth-associated and showed the highest activity on the fifth day. The dual culture method revealed 

that the strain was effective in restricting the radial growth of Fusarium oxysporum f.sp. lycopersici, an 

important phytopathogen of tomato. Scanning electronic microscopic analysis showed that the rupture 

of the F. oxysporum mycelial cell wall occurred at the area of interaction between F. oxysporum and 

Streptomyces sp. P4. This was possibly due to the chitinolytic activity of the P4. Thus, this 

actinomycete has the potential for being used as a biocontrol agent, thereby reducing the use of 

chemical fungicides. 

Keywords: endophyte, streptomycete, chitinase, fusarium, wilt, antimicrobial activity 

__________________________________________________________________________________ 

INTRODUCTION 

As worldwide concern for the natural environment and human health has increased, so has 

interest in organic farming. A research institute of organic agriculture (FiBL) and the international 

federation of organic agriculture movements (IFOAM) reported that organic agricultural lands have 

expanded globally from 11.0 million hectares in 1999 to 37.2 million hectares in 2009, accounting for 

0.85% of the total agricultural lands [1]. One of the criteria for organic farming is the avoidance of 

chemical usage. Soilborne plant diseases can be controlled through agronomic practices and microbial


biocontrol agents instead [2]. Biological controlling agents can replace chemical agents in controlling 

pathogenic insects, microbials and weeds. Several biofungicides are based on antibiotic metabolites and 

hydrolytic enzymes. For example, Streptomyces griseoviridis strain K61, a soilborne fungal antagonist 

which produces aromatic antibiotics with characteristic 7-membered rings in the molecules was 

commercialised as Mycostop by Verdera Oy, a Finnish company [3], and Streptomyces sp. Di-944 was 

formulated to suppress Rhizoctonia damping-off [4]. 

Streptomyces is a major genus of actinimycetes, the Gram-positive terrestrial or marine bacteria 

found in both colony and mycelium forms. Although Streptomyces species with a characteristic earthy 

smell may be thought of as pathogens, the antibiotics that they produce have been profitably exploited 

[5]. For example, S. clavuligerus produces the -lactam cephamycin C and clavulanic acid, a - 

lactamase inhibitor. A new stereoisomeric anthracyclin with anticancer activity was isolated from 

Sreptomyces sp. Eg23 [6]. Various hydrolytic enzymes, e.g. proteases/peptidases, chitinases/chitosanases, 

cellulases/endoglucanases, amylases, and pectate lyases, are produced by S. coelicolor [7]. 

In this study, the hydrolytic enzyme production of the endophytic actinomycete P4 strain, 

previously isolated from sweet pea root [8], is investigated. Additionally, the potential use of this 

endophyte as a biocontrol agent is studied by assessing its antagonism against pathogenic fungi. 


Microorganism Strains and Growth Conditions 

The bacterial strain P4, generously provided by Asst. Prof. Dr. Ampan Bhromsiri (Department 

of Soil Science and Conservation, Chiang Mai University) was previously isolated from sweet pea root. 

The bacteria was maintained at 30C on an IMA-2 agar medium consisting of 5 g glucose, 5 g soluble 

starch, 1 g beef extract, 1 g yeast extract, 2 g N-Z-case ® , 2 g NaCl and 1 g CaCO 3 per litre [9]. For the 

production of chitinase, the culture was transferred into a colloidal chitin medium, 1 litre of which 

consisted of 20 g colloidal chitin, 0.5 g yeast extract, 1 g (NH 4 ) 2 SO 4 , 0.3 g MgSO 4 . 7H 2 O and 1.36 g 

KH 2 PO 4 , with an adjusted pH of 7.0 [10]. Colloidal chitin was prepared according to the method 

described by Souza et al. [11]. The liquid culture was incubated at 30C with agitation at 160 revs/min. 

The pathogenic fungi Fusarium oxysporum f.sp. lycopersici, Corynespora cassiicola and 

Rhizoctonia solani were obtained from the Department of Agriculture, the Ministry of Agriculture and 

Cooperatives (Thailand). They were maintained on potato-dextrose agar (PDA). 

Identification by 16s rRNA Sequencing 

An approximately 1.5-kb polymerase chain reaction (PCR) product was amplified from genomic 

DNA of the bacterial strain P4 using two primers, 20F (5’-GAG TTT GAT CCT GGC TCA G-3’) and 

1500R (5’-GTT ACC TTG TTA CGA CTT-3’), directed to the 16S rRNA region. The following 

conditions were used: an initial denaturation step at 94C for 3 min., 25 cycles at 94C (1 min.), 50C 

(1 min.) and 72C (2 min.), followed by a final extension at 72C (3 min.). The purified PCR product 

was sequenced by an ABI PRISM® BigDye Terminator Ready Reaction Cycle Sequencing Kit 

(Applied Biosystems, USA) on an ABI Prism® 3730XL DNA Sequence (Applied Biosystems, USA). 

Homology was analysed using the BLAST program from the GenBank database [12].


97 

Plate Screening of Hydrolytic Enzymes 

The bacterial strain P4 (identified as Streptomyces) was screened for its capacity to produce 

hydrolytic enzymes using the plate method. The bacterium was allowed to grow at 30C on nutrient 

agar plates supplemented with different substrates, i.e. colloidal chitin, gelatin, sodium carboxymethylcellulose 

and Tween20, for detection of chitinase, protease, cellulase and lipase respectively. The clear 

zone around the colonies observed after 7-14 days is an indication for enzyme production. In the cases 

of cellulase and protease, the plates were reacted with 0.2% Congo red and saturated (NH 4 ) 2 SO 4 

respectively prior to the observation of growth [13-14]. 

Quantification of Extracellular Chitinase Activity 

The strain P4 was grown in a colloidal chitin medium broth at 30ºC with continuous shaking at 

160 rpm. The supernatant fluid was harvested every two days for 17 days by filtration through 

Whatman no.1 filter. Chitinase activity in the supernatant was assayed using 0.6% colloidal chitin as a 

substrate and was based on a procedure by Taechowisan et al. [13]. The supernatant fluid (700 μl) was 

added with 2% colloidal chitin (300 μl) in 0.1M potassium phosphate buffer at pH 7.0, and the mixture 

was incubated in a water bath at 40ºC for 3 hr. One mL of Somogyi’s reagent [15] was added and the 

reaction mixture was boiled at 100ºC for 10 min. and cooled to room temperature. Then Nelson’s 

reagent (1 mL) was added and the mixture cooled to room temperature for 20 min. After centrifugation 

of the reaction mixture, the amount of N-acetyl glucosamine (GlcNAc) released in the supernatant was 

spectrophotometrically measured by the method of Somogyi-Nelson [15]. The method is based on the 

520-nm absorbance given by a coloured complex formed between a copper-oxidised sugar and 

arsenomolybdate. One unit (U) of chitinase activity was defined as the amount of enzyme required to 

produce 1 mol of reducing sugar per min. under the conditions of the experiment. 

The cell growth was followed by measurement of the cells’ dry weight. After removing the 

supernatant for chitinase assay, the cell pellet was dried in an oven at 80C until a constant weight was 

obtained. All measurements were performed in triplicate. 

In Vitro Antagonism Tests against Fungi 

The identified Streptomyces P4 was evaluated for antagonistic activity towards three fungal 

phytopathogens, i.e. F. oxysporum, C. cassiicola and R. solani, by the dual culture method. An agar 

plug of 6 mm in diameter taken from a 7-day-old colony of each test fungus and that taken from a 14- 

day-old P4 bacterial strain were placed on PDA plates with 4-cm spacing (3 replicates each). The 

cultures were incubated at room temperature (30-32C) and the diameters of the fungal colonies in the 

direction of the actinomycete were measured every 2 days for 14 days. The test fungi were grown 

alone to serve as control. Data were statistically analysed for significance (p


osmium tetroxide. After dehydration in ethanol, the samples were dried to their critical points with 

carbon dioxide [16], mounted on slides and coated with gold for observation by SEM. 


Identification of Bacterial Strain 

A sequence of 1438-bp in length was obtained from the 1.5-kb PCR product. The DNA 

sequence was then submitted to GenBank database under accession no. JN102356 and was blasted 

against non-redundant DNA sequences in the database [12]. The 16S ribosomal DNA sequence showed 

the highest similarity to Streptomyces griseoflavus gene (accession no. EU741217) with a 99.7% 

identity score, followed by 99.5% score when aligned with S. variabilis (DQ442551, AB184884, 

AB184763), S. vinaceus (AB184186) and S. griseoincarnatus (AB184207, AJ781321). The P4 strain 

could not be definitely identified down to species level. It is worth noting that the 16S ribosomal RNA 

sequences are not highly variable among Streptomyces species and that Streptomyces systematics are 

rather complex. Lanoot et al. [17] suggested that 16S-ITS RFLP fingerprinting had a higher taxonomic 

resolution than 16S rDNA sequencing. A transformation reaction of progesterone has also been 

proposed for Streptomyces taxonomic classification [18]. 

S. griseoflavus was reported to produce many potent secondary metabolites, e.g. hormaomycin, 

a peptide lactone and a bacterial signalling metabolite and narrow-spectrum antibiotic [19]; 

okilactomycin, a polyketide antibiotic against gram-positive bacteria [20]; bicozamycin, a cyclic peptide 

antibiotic [21]; desferrioxamine, a precursor of iron chelator [22]; and an alkaline protease inhibitor 

[23]. However, there has been no report on the hydrolytic enzyme production in S. griseoflavus. 

Chitinase Activity 

From a preliminary screening of enzymes by the plant method (Figure 1), a clear zone 

surrounding the bacterial colonies was observed in the plate containing colloidal chitin as shown in 

Figure 1(A), indicating that the Streptomyces sp. P4 produced chitinase, whose activity was assayed 

during cell growth. Figure 2 shows that the activity was at a maximum on the fifth day, followed by a 

decrease upon approaching a stationary phase of growth. The findings imply that chitinase production 

is growth-associated. It should be noted that the Streptomyces sp. P4 strain had been pre-cultured in 

colloidal chitin medium prior to this experiment. As a consequence, the lag phase of growth was not 

observed, while logarithmic phase continued until the 5 th day before reaching a stationary phase. 

This result of chitinase production of Streptomyces sp. P4, induced in a colloidal chitincontaining 

environment as previously reported [10, 24] and indicating growth-associated behaviour, is 

similar to that obtained from the study on S. hygroscopicus [25]. However chitinase production in S. 

hygroscopicus occurred 1-2 days before cell growth, while in our case chitinase production of 

Streptomyces sp. P4 was closely associated with cell growth, which should be because the pre-cultured 

and pre-induced bacteria were used in our experiment. Closely paralleling growth, the chitinolytic 

enzyme production by Streptomyces may be for the purpose of hydrolysing chitin into monosaccharides 

to be used as carbon and nitrogen sources [26]. However, the chitinase activity of 0.00093 U/mL in 

Streptomyces sp. P4, which corresponds to a specific activity of 0.050 U/mg protein, was 4 times lower 

than that in S. viridificans (0.0038 U/mL) [10]. In nature chitinase is produced by actinomycetes in 

order to degrade complex nutrients from the soil. As fungal cell walls and insect structures largely 

contain chitin, chitinase produced from endophytes can be deleterious to pathogens and pests [27-28].


99 

A 

B 

C 

D 

Figure 1. Plate screening tests for hydrolytic enzyme production of Streptomyces sp. P4. Agar plates 

contain the corresponding substrates for chitinase (A), protease (B), cellulase (C) and lipase (D). Each 

plate represents a duplicate experiment. 

Antifungal Activity 

The results in Figures 3-4 show that Streptomyces sp. P4 could effectively suppress the growth 

of Fusarium oxysporum f.sp. lycopersici, a fungus causing Fusarium wilt, a severe disease in tomato 

[29]. Maximal inhibition was observed on the 9th day with 12.50% inhibition (Figure 3). The radial 

growth diameter of F. oxysporum when grown on the same plate as Streptomyces sp. P4 (5.370.23 

cm) was statistically smaller than that when cultured alone (6.130.38 cm). However, Streptomyces sp. 

P4 did not suppress the growth of the other two tested fungi, i.e. Corynesopra cassiicola (which causes 

leaf spot [30]) and Rhizoctonia solani (which causes root rot [31]) during the 14 days of observation 

(Figure 3). Anitha and Rabeeth [28] also reported the different responses of fungi to S. griseus 

chitinase. They suggested that the protein composition in the cell walls of different pathogenic fungi 

might make some fungal cell walls more resistant to chitinolytic degradation. Thus, only the co-culture 

containing F. oxysporum and Streptomyces sp. P4 was selected for SEM experiment.


Figure 2. Chitinase activity and cell dry weight during the growth of Streptomyces sp. P4 . Error bars 

represent the standard deviation of 3 replicates. 

* 

Figure 3. In vitro inhibitory activity of Streptomyces sp. P4 against three fungi: Fusarium oxysporum 

f.sp. lycopersici, Corynesopra cassiicola and Rhizoctonia solani. Dark blue and light blue bars 

correspond to radial growth diameters of the fungi cultured alone (control) and co-cultured with 

Streptomyces sp. P4 (dual culture) respectively on the 9 th day of growth on PDA plates. Error bars 

represent standard deviation of 3 replicates. The asterisk indicates that the value differs significantly 

from control (p


101 

A 

B 

Figure 4. Growth of Fusarium oxysporum f.sp. lycopersici grown alone (A) and co-cultured with 

Streptomyces sp. P4 (B) on PDA for 14 days 

Results obtained from SEM showed the breakage of the cell walls of F. oxysporum mycelia 

growing towards Streptomyces sp. P4 (Figure 5B) as compared to a control region (Figure 5A). The 

findings suggest that extracellular secondary metabolites and/or hydrolytic enzymes including chitinase 

play a crucial role in fungal growth inhibition. Prapagdee et al. [25] reported that the antifungal activity 

of S. hygroscopicus during exponential growth was mainly due to hydrolytic enzymes, while in the 

stationary phase it was due to secondary thermostable compound(s). In addition, there was a report on 

a positive correlation between chitinolytic and antagonistic activities of Streptomyces against the fungi 

Collectotrichum sublineolum, Guignardia citricarpa, Rhizoctonia solani and Fusarium oxysporum, but 

not in the oomycetes Pythium sp. and Phytophthora parasitica, which contain cellulose as a major cell 

wall component [16]. 

A 

B 

Figure 5. Scanning electron microscopic analysis of Fusarium oxysporum f.sp. lycopersici grown 

alone (A) and co-cultured with Streptomyces sp. P4 (B). Bars indicate 1 m.


CONCLUSIONS 

The present study provides background information for the potential use of the endophytic 

Streptomyces sp. P4 strain as a biocontrol agent antagonistic to specific fungi. The chitinase production 

and its association with the growth of Streptomyces sp. P4 were demonstrated. The fungal growth 

inhibition of F. oxysporum f.sp. lycopersici by Streptomyces sp. P4 was observed and demonstrated to 

result from the disruption of the fungal cell walls. 


This work was financially supported by the Centre of Excellence for Innovation in Chemistry 

(PERCH-CIC), Lanna Products Company and the Graduate School of Chiang Mai University. We 

thank S. Chantal E. Stieber (Department of Chemistry and Chemical Biology, Cornell University) for 

critical reading of the manuscript. 

REFERENCES 

1. H. Willer and L. Kilcher (Eds.), “The world of organic agriculture __ statistics and emerging trends 

2011”, IFOAM and FiBL, 2011, http://www.organic-world.net/ (Accessed: 1 August 2011). 

2. M. Shoda, “Bacterial control of plant diseases”, J. Biosci. Bioeng., 2000, 89, 515-521. 

3. D. R. Fravel, “Commercialization and implementation of biocontrol”, Annu. Rev. Phytopathol., 

2005, 43, 337-359. 

4. S. Sabaratnam and J. A. Traquair, “Formulation of a Streptomyces biocontrol agent for the 

suppression of Rhizoctonia damping-off in tomato transplants”, Biol. Control, 2002, 23, 245-253. 

5. G. Sykes and F. A. Skinner, “Actinomycetales: Characteristics and Practical Importance”, Academic 

Press, London, 1973, pp.64-75. 

6. M. S. Abdelfattah, “Screening of terrestrial Streptomyces leading to identification of new 

stereoisomeric anthracyclines”, World J. Microbiol. Biotechnol., 2008, 24, 2619-2625. 

7. P. Dyson, “Streptomyces”, in “Encyclopedia of Microbiology” (Ed. M. Schaechter), 3 rd Edn., 

Academic press, London, 2009, pp.318-332. 

8. P. Thapanapongworakul, “Characterization of endophytic actinomycetes capable of controlling 

sweet pea root rot diseases and effect on root nodule bacteria”, MS Thesis, 2003, Chiang Mai 


9. M. Shimizu, Y. Nakagawa, Y. Sato, T. Furumai, Y. Igarashi, H. Onaka, R. Yoshida and H. Kunoh, 

“Studies on endophytic Actinomycetes (I) Streptomyces sp. isolated from rhododendron and its 

antifungal activity”, J. Gen. Plant Pathol., 2000, 66, 360-366. 

10. R. Gupta, R. K. Saxena, P. Chaturvedi and J. S. Virdi, “Chitinase production by Streptomyces 

viridificans: Its potential in fungal cell wall lysis”, J. Appl. Bacteriol., 1995, 78, 378-383. 

11. C. P. Souza, E. M. Burbano-Rosero, B. C. Almeida, G. G. Martins, L. S. Albertini and I. N. G. 

Rivera, “Culture medium for isolating chitinolytic bacteria from seawater and plankton”, World J. 

Microbiol. Biotechnol., 2009, 25, 2079-2082. 

12. D. A. Benson. I. Karsch-Mizrachi, D. J. Lipman, J. Ostell and E.W. Sayers, “GenBank”, Nucleic 

Acids Res., 2011, 39 (Database issue), D32-37.


103 

13. T. Taechowisan, J. F. Peberdy and S. Lumyong, “Chitinase production by endophytic Streptomyces 

aureofaciens CMU Ac 130 and its antagonism against phytopathogenic fungi”, Ann. Microbiol., 

2003, 53, 447-461. 

14. G. L. Maria, K. R. Sridhar and N. S. Raviraja, “Antimicrobial and enzyme activity of mangrove 

endophytic fungi of southwest coast of India”, J. Agric. Technol., 2005, 1, 67-80. 

15. F. Green, C. A. Clausen and T. L. Highley, “Adaptation of the Nelson-Somogyi reducing-sugar 

assay to a microassay using microtiter plates”, Anal. Biochem., 1989, 182, 197-199. 

16. M. C. Quecine, W. L. Araujo, J. Marcon, C. S. Gai, J. L. Azevedo and A. A. Pizzirani-Kleiner, 

“Chitinolytic activity of endophytic Streptomyces and potential for biocontrol”, Lett. Appl. 

Microbiol., 2008, 47, 486-491. 

17. B. Lanoot, M. Vancanneyt, B. Hoste, K. Vandemeulebroecke, M. C. Cnockaert, P. Dawyndt, Z. 

Liu, Y. Huang and J. Swings, “Grouping of Streptomycetes using 16S-ITS RFLP fingerprinting”, 

Res. Microbiol., 2005, 156, 755-762. 

18. F. M. Atta and A. A. Zohri, “Transformation reactions of progesterone by different species of 

Streptomyces”, J. Basic Microbiol., 1995, 35, 3-7. 

19. I. Höfer, M. Crüsemann, M. Radzom, B. Geers, D. Flachshaar, X. Cai, A. Zeeck and J. Piel, 

“Insights into the biosynthesis of hormaomycin, an exceptionally complex bacterial signaling 

metabolite”, Chem. Biol., 2011, 18, 381-391. 

20. H. Imai, A. Nakagawa and S. Omura, “Biosynthesis of the antibiotic okilactomycin”, J. Antibiot., 

1989, 42, 1321-1323. 

21. K. Ochi, Y. Tsurumi, N. Shigematsu, M. Iwami, K. Umeharaand and M. Okuhara, “Physiological 

analysis of bicozamycin high-producing Streptomyces griseoflavus used at industrial level”, J. 

Antibiot., 1988, 41, 1106-1115. 

22. R. Nazari, A. Akbarzadeh, D. Norouzian, B. Farahmand, J. Vaez, A. Sadegi, F. Hormozi, M. Kiani- 

Rad and B. Zarbaksh, “Applying intra-specific protoplast fusion in Streptomyces griseoflavus to 

increase the production of desferrioxamine B”, Curr. Sci., 2005, 88, 1815-1820. 

23. A. Kuramoto, A. Lezhava, S. Taguchi, H. Momose and H. Kinashi, “The location and deletion of 

the genes which code for SSI-like protease inhibitors in Streptomyces species”, FEMS Microbiol. 

Lett., 1996, 139, 37-42. 

24. B. Mahadevan and D. L. Crawford, “Properties of the chitinase of the antifungal biocontrol agent 

Streptomyces lydicus WYEC108”, Enzyme Microb. Technol., 1997, 20, 489-493. 

25. B. Prapagdee, C. Kuekulvong and S. Mongkolsuk, “Antifungal potential of extracellular 

metabolites produced by Streptomyces hygroscopicus against phytopathogenic fungi”, Int. J. Biol. 

Sci., 2008, 4, 330-337. 

26. M. Charpentier and F. Percheron, “The chitin-degrading enzyme system of a Streptomyces species”, 

Int. J. Biochem., 1983, 15, 289-292. 

27. N. Dahiya, R. Tewari and G. S. Hoondal, “Biotechnological aspects of chitinolytic enzyme: A 

review”, Appl. Microbiol. Biotechnol., 2006, 71, 773-782. 

28. A. Anitha and M. Rabeeth, “Degradation of fungal cell walls of phytopathogenic fungi by lytic 

enzyme of Streptomyces griseus”, Afr. J. Plant Sci., 2010, 4, 61-66. 

29. A. Anitha and M. Rabeeth, “Control of fusarium wilt of tomato by bioformulation of Streptomyces 

griseus in green house condition”, Afr. J. Basic Appl. Sci., 2009, 1, 9-10.


30. S. Kurt, “Genetic variation in Corynesopra cassiicola, the target leaf spot pathogen”, Pakistan J. 

Biol. Sci., 2005, 8, 618-621. 

31. K. L. Olmos, S. H. Delgado and N. M. Pérez, “AFLP fingerprinting for identification of 

anastomosis groups of Rhizoctonia solani Kühn from common bean (Phaseolus vulgaris L.) in 

Mexico”, Rev. Mexicana Fitopatol., 2005, 23, 147-151. 



Full Paper 

Maejo Maejo Int. J. Sci. Int. Technol. J. Sci. Technol. 2012, 6(01), 2012, 105-118 6(01), 105-118 105 





Water quality variation and algal succession in commercial 

hybrid catfish production ponds 

Chatree Wirasith 1,2 and Siripen Traichaiyaporn 3,* 

1 

Program of Environmental Science, Department of Biology, Faculty of Science, Chiang Mai 

University, Chiang Mai 50200, Thailand 

2 

Program of Fisheries, Faculty of Agricultural Technology and Agro-Industry, Rajamangala 

University of Technology, Suvarnabhumi, Ayutthaya 13000, Thailand 

3 

Department of Biology, Faculty of Science, Chiang Mai University, Chiang Mai 50200, Thailand 

* Corresponding author, e-mail: tsiripen@yahoo.com 

Received: 8 October 2011 / Accepted: 23 March 2012 / Published: 30 March 2012 

Abstract: This study on water quality variation and algal succession in commercial hybrid catfish 

production ponds was conducted in 2007 in Bang Pa-In district, Ayutthaya province, Thailand. The 

study covered two fish crops, May-August and September-December. The physico-chemical water 

quality in the catfish ponds changed dramatically over the study period due to the practices of water 

changing, lime application and the culture duration before harvesting. Samples of algae collected 

during the first crop period contained 83 species belonging to the following divisions: Chlorophyta 

(34 species), Cyanophyta (28 species), Euglenophyta (12 species), Bacillariophyta (6 species), 

Chrysophyta (1 species), Pyrrhophyta (1 species) and Cryptophyta (1 species). Samples collected 

during the second crop contained 60 species of the following divisions: Chlorophyta (28 species), 

Cyanophyta (16 species), Euglenophyta (10 species) and Bacillariophyta (6 species). Cyanophyta 

was the most abundant in both crops, followed by Chlorophyta, Euglenophyta, Bacillariophyta, 

Chrysophyta, Cryptophyta and Pyrrhophyta. The blue-green algae Microcystis increasingly 

dominated the algal population during the course of the culture period. Pseudanabaena spp. were 

succeeded by Oscillatoria spp. and then Microcystis spp. in the first crop. Microcystis spp. 

dominated during the first two months of the second crop, and then was succeeded by 

Planktolyngbya spp. and Nitzschia spp. in the third and fourth months. In summary, water quality 

may account for algal proliferation resulting in algal blooms and influence algal succession in 

commercial catfish production ponds. 

Keywords: water quality, algal succession, commercial production pond, hybrid catfish


INTRODUCTION 

Hybrid catfish, the offspring of Clarias macrocephalus crossed with Clarias gariepinus, is 

among the most popular freshwater fish cultured commercially in South-east Asia, especially in 

Thailand [1]. In 2006, the production of this hybrid catfish in Thailand was estimated at 149,000 tons 

and valued at about 4,998.9 million Baht [2]. Since they are air breathers, hybrid catfish can be pondcultured 

at extremely high density, up to 100 fish/m 2 , with production reaching up to 100 tons/ha 

[1]. However, the off-flavour in the flesh of cultured catfish can be a problem, leading to market 

value reduction and/or making the fish unmarketable for a certain period of time, from a few days to 

weeks [3]. The off-flavour problem in cultured catfish is caused by compounds produced by certain 

kinds of blue-green algae, which are absorbed by the catfish and impart a bad flavour to the flesh if 

the harvest is delayed [4]. These blue-green algae can be found growing in catfish production ponds 

where an excessive amount of waste nutrients are generated. High density of the fish stocks and 

intensive feed input can result in extreme quantities of waste nutrients entering the production pond, 

which may account for the algal proliferation and resulting algal blooms [5]. Catfish cultured entirely 

and intensively in production ponds are commonly fed with pellet feed, trash fish and ground chicken 

skeletons and offal. Although the water in the production pond is normally changed completely 

during each production cycle of about 120-150 days, such feeding nevertheless causes a general 

deterioration of water quality and a decrease in dissolved oxygen in the pond water [6]. Low water 

quality also influences fish growth. In addition, the wastewater effluent from hybrid catfish 

production ponds can contain concentrated algal compounds and nutrients with a high nitrogen 

content, making it unsuitable for other profitable uses such as the culturing of other aquatic animals 

[7–9]. 

The seasonal succession of nutrients and phytoplankton populations in temperate and 

tropical systems has been extensively documented [10,11]. In tropical shallow water systems the 

roles of wet/dry seasons and wind typically have a greater impact on phytoplankton biomass 

production than inter-seasonal variations [12]. However, many other different types of algae 

prevailing in other climates also exhibit these wide swings in population densities. Some possible 

causes of these fluctuations include changes in temperature, pH, carbon dioxide, light intensity, 

nutrient concentration and the release of toxins by other organisms including competing algae [3]. 

Inthamjit et al. [9] reported a significant change in water quality during intensive culturing of hybrid 

catfish, where the value ranges of the parameters contributing to water quality were: dissolved 

oxygen (DO), 4.8-30.8 mg/L; biochemical oxygen demand (BOD), 24-90 mg/L; chemical oxygen 

demand (COD), 62-330 mg/L; chlorophyll a, 218-1,908 μg/L; total suspended solids (TSS), 378- 

1,490 mg/L; ammonia-nitrogen (NH 3 -N), 0.003-0.270 mg/L; nitrate-nitrogen (NO 3 - -N), 0.00-0.06 

mg/L; nitrite-nitrogen (NO 2 - -N), 0.01-0.03 mg/L; total phosphorus (TP), 1.50-5.81 mg/L; 

orthophosphate phosphorus (PO 4 3- -P), 0.00-2.11 mg/L; alkalinity, 91-388 mg/L; hardness (as 

CaCO 3 ), 300-580 mg/L; electrical conductivity, 800-1,900 μS/cm; and pH, 6.7-7.8. Stephens and 

Farris [13] compared the water quality from two channel catfish farms near Paragould, Arkansas 

(USA) during the summer of 2001 and found the following values: DO, 8.3 and 9.6 mg/L; 

chlorophyll a, 62 and 143 mg/L; TSS, 102 and 81 mg/L; NH 3 -N, 0.16 and 0.16 mg/L; NO 3 - -N, 


107 

phosphorus (SRP)), 1.188 and 2.38 mg/L; alkalinity, 118 and 167 mg/L; hardness, 93 and 192 mg/L; 

conductivity, 303 and 354 μS/cm; pH, 9.0 and 8.9; water temperature, 23 and 29 ºC; and fecal 

coliform bacteria, 603 and 433 CFU/100 mL. 

In Thailand, Ayutthaya province has many commercial hybrid catfish farms. Geographically, 

the province is mainly a lowland plain situated in the Chao Phraya River basin of central Thailand, 

where the soil is highly fertile and water is readily available year-round. Because of these natural 

advantages, the province is an important farming area for many other types of fish in addition to 

hybrid catfish. Based on Thailand’s fisheries statistics in 2005 [2], Ayutthaya, with a total land area 

of 2,412.8 ha, has a total of 3,623 farms engaged in pisciculture with a total yield of 2,176 tons. In 

Bang Pa-In district of the province (Figure 1) where this study was conducted, various forms of fish 

culture are practiced, including extensive, semi-intensive, intensive and integrated fish culture. For 

commercial hybrid catfish farming in this area, most of the culture are intensive systems. 

This study investigated the variations of water quality in terms of physical, chemical and 

biological aspects, as well as the succession of algae in commercial hybrid catfish production ponds. 

Sampling site 

Map of Ayutthaya 

Figure 1. Map of Thailand and location of the study area in Phra Nakhon, Sri Ayutthaya province 


Study Area 

The study area selected, Bang Pa-In district of Ayutthaya province (Figure 1), has a total land 

area of 488.8 ha, where 873 farmers were involved in fish culture [14]. Early in 2007, a field survey 

was conducted in the district. Three hybrid catfish farms, which were located near to one another and


which practiced similar fish culture systems, were selected as the sampling sites for this study. The 

study duration covered two fish crops in 2007, with the first crop from May to August and the 

second from September to December. Three replicates of water samples were collected monthly 

from the hybrid catfish production ponds at the three selected fish farms. In keeping with standard 

industrial practice for the culture of hybrid catfish, farmers released fingerlings into the production 

ponds (each 0.08 ha in size) at a density of 50 fingerlings/m2. The fingerlings were fed twice a day 

between 8-9 a.m. and 5-6 p.m. with pellet feed during the first and second months. Then during the 

third and fourth months they were fed chicken offal mixed with cassava chips in a ratio of 95:5. 

Water changing and lime application were carried out occasionally to manage water quality in the 

production ponds. The fish were cultured for 120-130 days before being harvested. The average fish 

yields were 59.38 tons/ha and 57.50 tons/ha for the first and second crops respectively, with the food 

conversion rate reaching 3.8-4.2. 

Physico-Chemical Water Quality Analysis 

Three replicates of water samples were collected monthly from the production ponds during 

both fish crops: May-August and September-December, 2007. All water samples were collected at a 

depth of 0.3-0.4 m and then preserved in an icebox until further processing. Water temperature, pH 

and DO were measured in situ using a portable hand-held meter (Multi 350i; WTW, Germany). 

Water transparency and water depth were measured using a Secchi disk and a measuring tape 

respectively. The analyses of chemical parameters were then carried out using suitable methods [15- 

16]: BOD by azide modification method; NH 3 -N by Nesslerisation method; NO 3 - -N by 

phenoldisulphonic acid method; total Kjeldahl nitrogen (TKN) by macro-Kjeldahl method; total 

phosphate by persulphate digestion/stannous chloride method; and orthophosphate phosphorus 

(PO 4 3- -P) by stannous chloride method. 

Algal Analysis 

For the algae count, the water sample (500 mL) from each production pond was transferred 

to a 500-mL cylinder and fixed with 5 mL of Lugol’s solution (20 g glacial acetic acid, 20 g 

potassium iodide and 20 g iodine dissolved in 200 mL distilled water). The preserved sample was left 

to stand in the dark for 10 days to allow concentration by decantation. A 20-25 mL sample from the 

lower layer of the 500-mL cylinder, containing the sedimented algae, was obtained and transferred to 

a 50-mL cylinder. A second decantation was conducted after another 7 days in the dark; a 10-mL 

sample from the lower layer of the 50-mL cylinder, containing the sedimented algae, was put into a 

glass vial and stored in a dark cupboard [17]. This concentrated sample of algae was used for their 

identification [18-22] and counting under a compound light microscope [17]. 

Statistical Analysis 

Collected data were statistically analysed using SPSS software program, version 14. 

Differences in means of water quality and algal population were established using analysis of variance 

(ANOVA), and relationships between algae and water quality parameters were tested using Pearson 

product-moment correlation. The level of significance was set at 0.05.


109 


Water Quality 

Water quality based on physico-chemical and biological parameters from production ponds at 

all three sampling sites and for both fish crops is presented in Table 1. The temperature optimum for 

aquaculture in Thailand is between 25-33ºC, depending on the species of fish being cultured; at 

temperatures above or below the optimum, fish growth is reduced [23]. There was a significant 

difference in temperature between the two study crops and a marked decline during the 4 th month 

(December) for the second crop. The optimal water transparency for aquaculture is 30 cm [24]. 

Water transparency in the study ponds ranged between 2.3-25.5 cm for the first fish crop and 1.1- 

10.3 cm for the second crop __ a significant difference between the two crops. There was no 

significant difference in water depth between the two crops. 

The optimal pH range for water used in aquaculture is between 6.5-8.5; however this will 

vary slightly depending on the cultured species [25]. Ingthamjit et al.[9] reported a pH range of 6.8- 

7.9 in hybrid catfish production ponds. The pH in the study ponds ranged between 6.6-7.1 for the 

first fish crop and 6.7-7.4 for the second crop. Generally, it is recommended that alkalinity be 

maintained within 50-300 mg/L to provide a sufficient buffering (stabilising) effect against pH swings 

that occur in ponds due to the respiration of the aquatic flora [25, 26]. Alkalinity in the study ponds 

ranged between 115.7-145.7 mg/L and 114.4-160.0 mg/L in the first and second fish crops 

respectively, thus displaying a significant difference. 

DO is probably the most critical water quality variable in freshwater aquaculture ponds. To 

achieve optimal growth, a good rule of thumb is to maintain the DO level at saturation or at least 5 

mg/L [25, 26]. The ranges of DO values in the study ponds were between 0.8-4.8 mg/L during the 

first fish crop and 1.5-4.3 mg/L during the second crop, with a significant difference in the 4 th -month 

values of the two study crops. Also, in the 4 th month, BOD reached 78.3 and 85.8 mg/L for the first 

and second fish crops respectively. These values were significantly different from the 1 st - and 2 nd - 

month values for both fish crops. 

According to a study by Boyd [26] on unfertilised woodland ponds in Alabama, the average 

total NH 3 -N (NH 4 + plus NH 3 expressed in terms of N) was 0.052 mg/L and and that for NO 3 - -N was 

0.075 mg/L. In intensive fish culture ponds, much higher concentrations of inorganic N are common. 

Channel catfish culture ponds can contain up to 0.5 mg/L of total NH 3 -N and 0.25 mg/L of NO 3 - -N 

[26]. NH 3 -N concentrations in the study ponds showed no significant difference between the first and 

second fish crops (0.06-1.77 mg/L). NO 3 - -N concentrations varied between 0.01-0.24 mg/L and 

0.02-0.03 mg/L for the first and second fish crops respectively, a significant difference being found 

between the 1 st and 4 th months of the first crop. Nitrogen is also present as soluble organic 

compounds and as constituents of living and dead particulate organic matter. Concentrations of 

organic nitrogen are usually well below 1 mg/L in unpolluted natural water [26]. In fish production 

ponds, phytoplankton blooms are normally heavy and the concentration of organic nitrogen may 

exceed 2-3 mg/L. In the study ponds, the TKN concentration reached a maximum level (5.6 mg/L) in 

the 4 th month of the second fish crop.


Table 1. Monthly means and standard deviations (mean ± SD) of physico-chemical characteristics of water samples from commercial hybrid catfish 

production ponds, Ayutthaya province, Thailand, 2007 

Parameter 

Month 

Crop 1 (mean ± SD) Crop 2 (mean ± SD) 

May June July August September October November December 

Water temperature (ºC) 29.4 a ± 0.1 31.8 a ± 0.1 31.0 a ± 0.1 30.8 a ± 0.1 29.0 b ± 0.1 28.8 b ± 0.1 28.4 b ± 0.1 27.5 b ± 0.3 

Water transparency (cm) 25.5 a ± 0.1 10.4 a ± 0.9 5.7 a ± 0.8 2.3 a ± 0.2 10.3 b ± 0.3 5.5 b ± 0.9 1.7 b ± 0.2 1.1 b ± 0.1 

Water depth (m) 1.25 ± 0.1 1.25 ± 0.1 1.25 ± 0.1 1.25 ± 0.1 1.25 ± 0.1 1.25 ± 0.1 1.25 ± 0.1 1.00 ± 0.2 

pH 6.7 ± 0.1 7.1 a ± 0.2 6.6 ± 0.1 7.0 b ± 0.1 6.7 ± 0.1 6.7 b ± 0.1 6.7 ± 0.1 7.4 a ± 0.2 

Alkalinity as CaCO3 (mg/L) 145.7 a ± 5.1 140.0 a ± 3.0 123.7 b ± 3.5 115.7 b ± 2.1 121.1 b ± 3.9 114.4 b ± 1.9 151.0 a ± 3.6 160.0 a ± 1.7 

DO (mg/L) 4.8 ± 0.6 3.2± 0.9 2.6 ± 0.4 0.8 b ± 0.2 4.3 ± 0.4 3.5 ± 0.3 2.4 ± 0.5 1.5 a ± 0.3 

BOD (mg/L) 24.2 b ± 1.4 38.2 b ± 2.8 65.1 ± 6.8 78.3 ± 6.0 49.7 a ± 3.2 61.1 a ± 4.6 75.0 ± 2.5 85.8 ± 5.2 

NH3-N (mg/L) 0.2 ± 0.1 0.06 ± 0.02 0.7 ± 0.1 1.7 ± 0.1 0.06 ± 0.04 0.1 ± 0.05 0.6 ± 0.2 1.8 ± 0.3 

NO3 - -N (mg/L) 0.01 b ± 0.01 0.02 ± 0.01 0.03 ± 0.02 0.2 a ± 0.04 0.03 a ± 0.01 0.02 ± 0.01 0.03 ± 0.01 0.03 b ± 0.01 

TKN (mg/L) 1.2 ± 0.1 1.0 ± 0.1 2.4 ± 0.4 2.9 b ± 0.3 1.3 ± 0.05 1.27 ± 0.2 2.0 ± 0.2 5.6 a ± 0.7 

Total P (μg/L) 4.1 ± 3.4 11.8 ± 2.9 11.9 ± 2.4 22.1 ± 4.3 6.0 ± 1.0 17.0 ± 3.8 12.7 ± 2.5 14.9 ± 3.5 

PO4 3- -P (μg/L) 0.1b ± 0.1 0.8 b ± 0.5 9.5 ± 1.8 16.6 a ± 0.9 4.5 a ± 1.2 8.6 a ± 1.2 8.6 ± 1.3 10.6 b ± 0.5 

Note: Values in the same row followed by different superscripts indicate significant difference (p


111 

The phosphorus concentration in water is usually quite low; dissolved orthophosphate 

concentration lies within 5-20 μg/L and seldom exceeds 100 μg/L even in highly eutrophic waters, 

while the concentration of total phosphorus seldom exceeds 1,000 μg/L [26]. Total phosphorus 

concentrations in the study ponds varied between 4.1-22.1 μg/L with no significant difference 

between the two crops, while PO 4 3- -P concentrations were significantly different between the two 

crops in the 1 st , 2 nd and 4 th months. 

The waste effluent from intensive fish culture as a source of pollution of natural bodies of 

water has been a major concern [27]. In the present study, water quality deteriorated as the farming 

season progressed, an occurrence shared by several other findings [9, 13, 28-30]. However, there 

was a difference in the amount of nutrients added to the water in the fish production ponds owing to 

difference in the feed applied. In this study, hybrid catfish were fed chicken offal mixed with cassava 

chips as fresh feed, which actually caused the water quality to deteriorate more rapidly. This was 

similar to the findings of Yi et al. [6], who showed that using trash fish and chicken offal as feed for 

fish culture could lead to a rapid deterioration of water quality. 

Algal Succession 

Algae found in the water samples of the first fish crop were categorised into 83 species of 7 

divisions, namely Chlorophyta (34 species), Cyanophyta (28 species), Euglenophyta (12 species), 

Bacillariophyta (6 species), Chrysophyta (1 species), Pyrrhophyta (1 species) and Cryptophyta (1 

species), whereas those of the second fish crop were categorised into 60 species of 4 divisions, 

namely Chlorophyta (28 species), Cyanophyta (16 species), Euglenophyta (10 species) and 

Bacillariophyta (6 species) (Table 2). 

Abundance percentages of the algal divisions are shown in Table 3. There was no significant 

difference between the two fish crops in the number of algae of each division. Chlorophyta was most 

abundant in both fish crops, followed by Cyanophyta, Euglenophyta, Bacillariophyta, Chrysophyta, 

Cryptophyta and Pyrrhophyta. This confirmed the results obtained by Ingthamjit et al. [9] and Boyd 

[26] as well as several other reports [28-33]. Phytoplankton occurring in fish production ponds 

includes members of the following taxonomic divisions: green algae (Chlorophyta), blue-green algae 

(Cyanophyta), euglenophytes (Euglenophyta), yellow-green and golden-brown algae, diatoms 

(Chrysophyta) and dinoflagellates (Pyrrhophyta) [26, 28-33]. The dominant algae in the first fish 

crop were Microcystis spp., followed by Pseudanabaena spp., Monoraphidium spp., Oscillatoria 

spp., Scenedesmus spp., Euglena spp., Merismopedia spp., Cyclotella spp., Coelastrum spp., 

Tetrastrum sp. and Spirulina sp.(Table 4). The second fish crop was dominated by Microcystis spp., 

followed by Planktolyngbya sp., Spirulina sp., Cyclotella spp., Pseudanabaena spp., Merismopedia 

spp., Nitzschia spp., Monoraphidium spp., Scenedesmus spp., Tetrastrum sp. and Phacus spp. 

(Table 5). Algae of the division Cyanophyta (Microcystis) grew densely and were the dominant 

division in both fish crops. As reported by Welker et al. [34], Microcystis is commonly present in 

eutrophic temperate lakes during summer. These findings are supported by the results of the present 

study: Microcystis species demonstrated better growth in summer and under eutrophic conditions 

with high concentrations of nutrients in the fish production pond water. Lin [32] and Chowdhury and 

Mamun[33] also reported that Cyanophyta dominated nutrient-rich channel catfish ponds.


Table 2. Diversity and classification of algae occurring in commercial hybrid catfish production 

ponds, Ayutthaya province, Thailand, 2007 

DIVISION CHLOROPHYTA 

Actinastrum hantzschii, Ankistrodesmus sp., Closteriopsis sp., Closterium sp. 1, Coelastrum astroideum, Coelastrum 

pseudomicroporum, Coelastrum sp.1, Cosmarium sp.1, Crucigenia crucifera, Crucigeniella rectangularis, 

Dictyosphaerium granulatum, Dictyosphaerium sp., Elakatothrix sp., Kirchneriella sp, Monoraphidium arcuatum, 

Monoraphidium caribeum, Monoraphidium contortum, Monoraphidium griffithii, Monoraphidium minutum, Oocystis 

sp., Pediastrum duplex, Pediastrum simplex, Scenedesmus bernardii, Scenedesmus disciformis, Scenedesmus 

microspina, Scenedesmus opoliensis, Scenedesmus pannonicus, Scenedesmus perforates, Scenedesmus velitaris, 

Scenedesmus sp. 1, Scenedesmus sp. 2, Staurastrum cingulum, Tetraedron caudatum, Tetrastrum heteracanthum 

DIVISION CYANOPHYTA 

Anabaena catenula, Aphanothece sp., Arthrospira sp., Chloroflexus sp., Chroococcus minutes, Chroococcus sp, 

Cylindrospermopsis curvispora, Cylindrospermopsis helicoidea, Cylindrospermopsis raciborskii, Gomphosphaeria sp., 

Komvophoron sp., Lyngbya sp., Merismopedia convulata, Merismopedia glauca, Merismopedia punctata, Microcystis 

aeruginosa, Microcystis wesenbergii, Microcystis sp., Oscillatoria agardhii, Oscillatoria limosa, Oscillatoria redekei, 

Planktolyngbya limnetica, Pseudanabaena catenata, Pseudanabaena sp.1, Pseudanabaena sp.2, Raphidiopsis sp., 

Romeria sp., Spirulina sp. 

DIVISION EUGLENOPHYTA 

Euglena acus, Phacus orbicularis, Phacus triqueter, Phacus sp.1, Phacus sp.2, Phacus sp.3, Strombomonas sp., 

Trachelomonas acanthostoma, Trachelomonas caudata, Trachelomonas cylindrical, Trachelomonas volvocina, 

Trachelomonas sp. 

DIVISION BACILLARIOPHYTA 

Aulacoseira granulata, Cyclotella sp., Fragilaria sp., Melosira sp., Nitzschia sp.1, Nitzschia sp.2 

DIVISION CHRYSOPHYTA 

Isthmochloron sp. 

DIVISION PYRRHOPHYTA 

Peridinium sp. 

DIVISION CRYPTOPHYTA 

Cryptomonas sp.


113 

Table 3. Means and standard deviations of the percentages of abundance of each algal division in hybrid catfish ponds by month 

Division 

Abundance (%) 

Crop 1 (mean ± SD) Crop 2 (mean ± SD) 

May June July August September October November December 

Chlorophyta 16.26 ± 8.43 34.67 ± 2.61 19.42 ± 12.52 29.79 ± 10.30 12.29 ± 10.88 9.24 ± 7.15 10.56 ± 4.11 21.23 ± 7.18 

Cyanophyta 68.40 ± 14.84 41.26 ± 8.59 73.23 ± 5.91 53.57 ± 1.9 74.77 ± 15.68 76.03 ± 12.46 69.02 ± 16.76 49.40 ± 12.98 

Bacillariophyta 3.77 ± 0.52 9.51 ± 10.69 2.69 ± 1.29 15.59 ± 11.85 7.33 ± 4.96 9.51 ± 5.16 19.88 ± 14.21 28.36 ± 7.73 

Euglenophyta 8.61 ± 4.83 12.10 ± 11.52 4.66 ± 0.65 0.75 ± 0.07 5.60 ± 0.38 5.19 ± 1.6 0.54 ± 0.21 1.02 ± 0.5 

Chrysophyta 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.31 ± 0.03 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 

Pyrrhophyta 1.07 ± 0.38 2.47 ± 1.80 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 

Cryptophyta 1.89 ± 1.63 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00


Table 4. Means and standard deviations of the percentages of abundance of each algal genus in hybrid 

catfish ponds by month (Crop 1) 

Genus (Division) Abundance (%) 

May June July August 

Microcystis spp. (Cyanophyta) 0.00 ± 0.00 1.45 ± 0.37 74.66 ± 7.60 28.89 ± 3.27 

Pseudanabaena spp. (Cyanophyta) 47.96 ± 21.62 4.14 ± 0.91 3.38 ± 1.01 4.39 ± 3.80 

Monoraphidium spp. (Chlorophyta) 12.62 ± 8.20 27.43 ± 8.60 3.17 ± 0.34 11.50 ± 9.15 

Oscillatoria spp. (Cyanophyta) 4.92 ± 0.76 37.37 ± 3.89 3.59 ± 1.37 1.30 ± 0.71 

Scenedesmus spp. (Chlorophyta) 11.41 ± 0.24 6.86 ± 3.99 4.83 ± 1.63 9.10 ± 4.64 

Euglena spp. (Euglenophyta) 12.58 ± 6.01 10.75 ± 9.08 1.53 ± 0.42 0.62 ± 0.04 

Merismopedia spp. (Cyanophyta) 3.73 ± 6.46 0.68 ± 1.18 3.57 ± 6.18 13.18 ± 4.99 

Cyclotella spp. (Bacillariophyta) 0.00 ± 0.00 3.47 ± 2.81 2.15 ± 2.33 14.35 ± 10.03 

Coelastrum spp. (Chlorophyta) 6.79 ± 5.96 2.07 ± 0.46 1.48 ± 1.18 2.33 ± 1.99 

Tetrastrum sp. (Chlorophyta) 0.00 ± 0.00 4.76 ± 5.02 1.29 ± 0.64 5.14 ± 4.06 

Spirulina sp. (Cyanophyta) 0.00 ± 0.00 1.00 ± 0.57 0.67 ± 0.46 9.18 ± 9.54 

Table 5. Means and standard deviations of the percentages of abundance of each algal genus in hybrid 

catfish ponds by month (Crop 2) 

Genus (Division) 

Abundance (%) 

September October November December 

Microcystis spp. (Cyanophyta) 32.93 ± 4.57 30.03 ± 5.45 23.31 ± 11.12 16.22 ± 9.73 

Planktolyngbya sp. (Cyanophyta) 14.21 ± 4.32 9.56 ± 6.30 28.90 ± 23.35 6.51 ± 4.89 

Spirulina sp. (Cyanophyta) 17.68 ± 8.68 29.84 ± 11.36 3.37 ± 0.63 0.95 ± 0.63 

Cyclotella spp. (Bacillariophyta) 3.80 ± 2.04 7.04 ± 7.20 19.21 ± 18.01 16.65 ± 3.49 

Pseudanabaena spp. (Cyanophyta) 14.33 ± 7.13 10.83 ± 1.20 12.77 ± 20.85 6.51 ± 4.89 

Merismopedia spp. (Cyanophyta) 3.91 ± 4.15 5.17 ± 0.68 5.50 ± 4.33 15.93 ± 6.86 

Nitzschia spp. (Bacillariophyta) 1.20 ± 0.38 0.00 ± 0.00 0.52 ± 0.13 19.48 ± 2.33 

Monoraphidium spp. (Chlorophyta) 3.91 ± 4.15 1.42 ± 1.11 1.87 ± 1.49 8.94 ± 6.15 

Scenedesmus spp. (Chlorophyta) 3.09 ± 0.77 2.57 ± 3.28 2.14 ± 0.76 6.09 ± 5.99 

Tetrastrum sp. (Chlorophyta) 2.34 ± 3.01 0.22 ± 0.38 1.85 ± 0.80 1.25 ± 1.92 

Phacus spp. (Euglenophyta) 4.37 ± 4.42 1.91 ± 0.72 0.20 ± 0.20 0.18 ± 0.13


115 

Figure 2 shows the dominant species among the algal populations during the different months 

of the study. Pseudanabaena spp. were the dominant species during the 1 st month of the first fish 

crop, followed by Oscillatoria spp. and Microcystis spp. in the 2 nd and 3 rd months respectively. For 

the 2 nd fish crop, Microcystis spp. dominated during the first two months, followed by 

Planktolyngbya sp. in the 3 rd month and Nitzschia spp. in the 4 th month. The succession of algae 

seemed to be associated with nutrient accumulation and water changing. This could be determined 

from the results of the correlation analysis of algal populations and water quality parameters. The 

prevalence of Pseudanabaena spp. was found to be significantly associated with water transparency, 

which was at the highest level in the first month when these species were the dominant algae. 

Microcystis spp. showed a high correlation with nitrate-nitrogen levels, which increased in the 3 th 

and 4 th months of the first fish crop and in the 1 st and 2 nd months of the second fish crop. The 

decrease in nitrate-nitrogen level also coincided with a decline in the population of Microcystis and 

an increase of Planktolyngbya and Nitzschia in the succeeding months. Planktolyngbya sp. showed 

no significant correlation with any of the studied water quality parameters, but tended to grow better 

in water with less transparency and lower temperature. Nitzschia spp. were found to significantly 

correspond to TKN level. An increase in TKN during the last month of the second fish crop might 

contribute to the succession of Nitzschia spp., as also indicated in studies by Ingthmjit et al.[9], 

Brunson et al. [35], and Zimba et al [36]. However, the succession of algae in the catfish production 

ponds differs from that occurring in natural lakes, owing to the dense stock of fish in the ponds and 

the daily feeding which provides abundant nutrients. As reported by Stephens and Farris [13], algal 

density in commercial channel catfish production ponds is limited more by nutrient availability than 

by light. Many other different types of algae also exhibit these wide swings in population density. 

Possible causes of these fluctuations include changes in temperature, pH, carbon dioxide 

concentration, light intensity and nutrient concentration, and the release of toxins by other organisms 

including competing algae [3]. 

CONCLUSIONS 

The physico-chemical water quality in commercial hybrid catfish production ponds located in 

Ayutthaya changed dramatically over the culture period. Certain water quality parameters influenced 

algal dominance and succession. While Microcystis of the division Cyanophyta dominated 

throughout, Pseudanabaena was succeeded by Oscillatoria, followed by Microcystis in the first fish 

crop period, whereas Microcystis was succeeded by Planktolyngbya and Nitzschia in the second fish 

crop period.


Figure 2. Percentages of abundance of algae in hybrid catfish ponds by month 


Our sincere gratitude is extended to Rajamangala University of Technology Suvarnabhumi 

for financial support. 

REFERENCES 

1. S. Areerat, “Clarias culture in Thailand”, Aquaculture, 1987, 63, 355-362.


117 

2. Fisheries Information Technology Center, “Fisheries Statistics of Thailand 2005”, Department 

of Fisheries, Ministry of Agriculture and Cooperatives, Bangkok, 2007. 

3. H. S. Killian, “Phytoplankton in catfish ponds”, Cooperative Extension Program, University of 

Arkansas, http://www.uaex.edu/aquaculture2/FSA/FSA9070.htm (Accessed: 2011). 

4. P. V. Zimba, C. S. Tucker, C. C. Mischke and C. C. Grimm, “Short-term effects of diuron on 

catfish pond ecology”, N. Am. J. Aquaculture, 2002, 64, 16-23. 

5. P. V. Zimba, L. Khoo, P. S. Gaunt, S. Brittain and W. W. Carmichael, “Confirmation of catfish, 

Ictalurus punctatus (Rafinesque), mortality from Microcystis toxins”, J. Fish Diseas., 2001, 24, 

41-47. 

6. Y. Yi, C. K. Lin and J. S. Diana, “Hybrid catfish (Clarias macrocephalus x C. gariepinus) and 

Nile tilapia (Oreochromis niloticus) culture in an integrated pen-cum-pond system: Growth 

performance and nutrient budgets”, Aquaculture, 2003, 217, 395-408. 

7. P. R. Adler, F. Takeda, D. M. Glenn and S. T. Summerfelt, “Enhancing aquaculture 

sustainability through utilizing byproducts”, World Aquaculture, 1996, 27, 24-26. 

8. T. V. R. Pillay, “The challenges of sustainable aquaculture”, World Aquaculture, 1996, 27, 7-9. 

9. S. Ingthamjit, S. Areerat, P. Tientong and S. Wallie, “Variation of water quality, phytoplankton 

and bacteria in catfish ponds”, Academic document No. 129, 1992, Freshwater Fisheries 

Research Institute, Department of Fisheries, Ministry of Agriculture and Co-operatives, 

Bangkok. 

10. G. E. Hutchinson, “A Treatise on Limnology”, Vol. 1, Wiley, New York, 1957. 

11. W. S. Rast, V. H. Smith and J. A. Thornton, “Characteristics of eutrophication”, in “The 

Control of Eutrophication in Lakes and Reservoirs” (Ed. S. O. Ryding and W. Rast (Eds.), 

UNESCO and Parthenon Publishers, London, 1989, pp.37-64. 

12. P. V. Zimba, “The use of nutrient enrichment bioassays to test for limiting factors affecting 

epiphytic growth in Lake Okeechobee, Florida: Confirmation of nitrogen and silica limitation”, 

Archiv. Hydrobiol., 1998, 141, 459-468. 

13. W. W. Stephens and J. L. Farris, “Instream community assessment of aquaculture effluents”, 

Aquaculture, 2004, 231, 149-162. 

14. The Center for Fisheries Development of Phranakorn-Sri Ayutthaya, “Fisheries statistics of 

Phranakorn-Sri Ayutthaya 2007” (unpublished document), Department of Fisheries, Ministry of 

Agriculture and Co-operatives, Bangkok. 

15. APHA, AWWA and WPCF, “Standard Methods for Examination of Water and Wastewater”, 

20 th Edn., American Public Health Association, Washington, DC, 1998. 

16. S. Traichaiyaporn, “Water Quality Analysis”, Department of Biology, Faculty of Science, 

Chiang Mai University, Chiang Mai, 2000 (Textbook in Thai). 

17. S. Traichaiyaporn, “Applied Phycology”, Department of Biology, Faculty of Science, Chiang 

Mai University, Chiang Mai, 2000 (Textbook in Thai). 

18. L. Wongrat, “Phytoplankton”, Department of Fishery Biology, Faculty of Fisheries, Kasetsart 

University, Bangkok, 1999 (Textbook in Thai). 

19. G. W. Prescott, “How to Know the Freshwater Algae”, 3 rd Edn., W. C. Brown, Dubuque, 

1978. 

20. C. B. Harrold and J. W. Michaell, “Introduction to the Algae”, 2 nd Edn., Prentice-Hall, 

Englewood Cliffs, 1985.


21. A. Shirota, “The Plankton of South Vietnam: Fresh Water and Marine Plankton”, Overseas 

Technical Cooperation Agency, Tokyo, 1966. 

22. X. N. Verlencar and S. R. Desai, “Phytoplankton Identification Manual”, National Institute of 

Oceanography, Dona Panla, 2004. 

23. M. Duangsawasdi and J. Somsiri, “Water Quality and Analytical Methods for Fisheries 

Research”, Department of Fisheries, Ministry of Agriculture and Cooperatives, Bangkok, 1986 

(Textbook in Thai). 

24. Y. Musit, “Water Quality for Aquaculture”, Department of Aquaculture, Faculty of Fisheries, 

Kasetsart University, Bangkok, 1992 (Textbook in Thai). 

25. J. K. Buttner, R. W. Soderberg and D. E. Terlizzi, “An introduction to water chemistry in 

freshwater aquaculture”, Fact Sheet No.170, 1993, Northeastern Regional Aquaculture Center, 

Dartmouth, Massachusetts, USA. 

26. C. E. Boyd, “Water Quality in Ponds for Aquaculture”, Department of Fisheries and Allied 

Aquacultures, Auburn University, Auburn (USA), 1990. 

27. H. S. Egna and C. E. Boyd, “Dynamics of Pond Aquaculture”, CRC Press, Boca Raton, 1997. 

28. B. Sen and F. Sonmez, “A study on the algae in fish ponds and their seasonal variations”, Int. J. 

Sci. Technol., 2006, 1, 25-33. 

29. M. A. Burford and D. C. Pearson, “Effect of different nitrogen sources on phytoplankton 

composition in aquaculture ponds”, Aquat. Microb. Ecol., 1998, 15, 277-284. 

30. C. S. Tucker and M. van der Ploeg, “Seasonal changes in water quality in commercial channel 

catfish ponds in Mississippi”, J. World Aquaculture Soc., 1993, 24, 473-481. 

31. M. M. Littler and J. H. Graffius, “The annual distribution of phytoplankton communities in a 

south-eastern Ohio pond”, Ohio J. Sci., 1974, 74, 313-324. 

32. C. K. Lin, “Biological principles of pond culture: Phytoplankton and macrophytes”, in 

“Principles and Practices of Pond Aquaculture” (Ed. J. E. Lamann, R. O. Smitherman and G. 

Tchobanoglous), Oregon State University, Corvallis, 1983. 

33. A. H. Chowdhury and A. A. Mamun, “Physio-chemical conditions and plankton population of 

two fishponds in Khulna”, Univ. J. Zool. Rajshahi Univ., 2006, 25, 41-44. 

34. M. Welker, M. Brunke, K. Preussel, I. Lippert and H. von Döhren, “Diversity and distribution 

of Microcystis (Cyanobacteria) oligopeptide chemotypes from natural communities studied by 

single-colony mass spectrometry”, Microbiology, 2004, 150, 1785-1796. 

35. M. W. Brunson, C. G. Lutz and R. M. Durborow, “Algae blooms in commercial fish production 

ponds”, Publication No. 466, 1994, Southern Regional Aquaculture Center, Stoneville, 

Mississippi, USA. 

36. P. V. Zimba, C. C. Mischke and S. S. Brashear, “Pond age-water column trophic relationships in 

channel catfish Ictalurus punctatus production ponds”, Aquaculture, 2003, 219, 291-301. 



Communication 

Maejo Maejo Int. J. Sci. Int. Technol. J. Sci. Technol. 2012, 6(01), 2012, 119-129 6(01), 119-129 119 





Determination of production-shipment policy using a two-phase 

algebraic approach 

Yuan-Shyi Peter Chiu 1 , Hong-Dar Lin 1 and Huei-Hsin Chang 2,* 

1 

Department of Industrial Engineering and Management, Chaoyang University of Technology, 

Wufong, Taichung 413, Taiwan 

2 

Department of Finance, Chaoyang University of Technology, Wufong, Taichung 413, Taiwan 

*Corresponding author, e-mail: chs@cyut.edu.tw 

Received: 31 March 2011 / Accepted: 31 March 2012 / Published: 7 April 2012 

Abstract: The optimal production-shipment policy for end products using mathematical 

modeling and a two-phase algebraic approach is investigated. A manufacturing system 

with a random defective rate, a rework process, and multiple deliveries is studied with the 

purpose of deriving the optimal replenishment lot size and shipment policy that minimises 

total production-delivery costs. The conventional method uses differential calculus on the 

system cost function to determine the economic lot size and optimal number of shipments 

for such an integrated vendor-buyer system, whereas the proposed two-phase algebraic 

approach is a straightforward method that enables practitioners who may not have 

sufficient knowledge of calculus to manage real-world systems more effectively. 

Keywords: manufacturing system, replenishment lot size, delivery, two-phase algebraic 

approach, random defective rate 

________________________________________________________________________________ 

INTRODUCTION 

With the purpose of minimising total set-up and holding costs, inventory controllers in most 

companies need to address two basic issues for items they routinely stock: when to start 

replenishment and how much to refill. For items made in-house by manufacturing firms, production 

planners must, without exception, decide when to initiate a production run and how many items to 

produce in a run [1]. An inventory model that uses mathematical techniques to derive the most 

economical production lot was first proposed by Taft [2] several decades ago. This is also known as 

the economic production quantity (EPQ) model [3].


The classic EPQ model assumes a continuous inventory issuing policy to satisfy product 

demand. However, in real-world vendor-buyer systems, multiple or periodic deliveries of end items 

are commonly adopted. Hence, the determination of the optimal number of shipments for a finished 

lot becomes a critical issue to such a vendor-buyer system in terms of production-delivery cost 

minimisation. Schwarz [4] examined a one-warehouse N-retailer deterministic inventory system with 

the objective of determining the stock policy that minimises the long-run average system cost per 

unit time. He derived optimal solutions along with a few necessary properties for a one-retailer and 

N-identical-retailer problems. Heuristic solutions for the general problem were also suggested. Goyal 

[5] studied an integrated single-supplier-single-customer problem and presented a method that is 

typically applicable to those inventory problems where a product is procured by a single customer 

from a single supplier using examples to demonstrate the proposed model. Studies related to various 

aspects of supply chain optimisation have since been extensively carried out [e.g. 6-13]. The classic 

EPQ model also assumes that all items produced are of perfect quality. However, in a real-life 

manufacturing environment, the generation of nonconforming items is almost inevitable due to 

process deterioration or various other factors. In the past decades, many studies have attempted to 

address the issues of defective products and quality assurance in production systems [e.g. 14-24]. 

Shih [14] extended two inventory models to the case where the proportion of defective 

units in the accepted lot is a random variable with known probability distribution. Optimal solutions 

to the amended systems were developed and comparisons with the traditional models were also 

presented via numerical examples. Moinzadeh and Aggarwal [17] studied a production-inventory 

system that was subjected to random disruptions. They assumed that the time between breakdowns is 

exponential, the restoration times are constant, and excess demand is back-ordered. An (s, S) policy 

was proposed and the policy parameters that minimise the expected total cost per unit time were 

investigated. A procedure for finding the optimal values of the policy was also developed. Makis 

[18] investigated the optimal lot sizing and inspection policy for an economic manufacturing quantity 

(EMQ) model with imperfect inspections and assumed that the process could be monitored through 

inspections and that both the lot size and the inspection schedule were subjected to control. It was 

assumed that the in-control periods were generally distributed and the inspections imperfect. Using 

Lagrange's method and solving a non-linear equation, a two-dimensional search procedure was 

proposed for finding the optimal lot sizing and inspection policy. Rahim and Ben-Daya [19] studied 

the simultaneous effects of deteriorating product items and deteriorating production processes on the 

economic production quantity, inspection schedules, and economic design of control charts. 

Deterioration times for both product and process were assumed to follow an arbitrary distribution, 

and the product quality characteristic was assumed to be normally distributed. Numerical examples 

were provided to demonstrate the usage of their models. Chiu et al. [21] studied the optimal 

replenishment policy for the EMQ model with rework failure, backlogging and random breakdowns. 

Mathematical modelling and cost analysis were employed in their study, along with a renewal reward 

theorem for dealing with variable cycle length. They derived a long-run average cost function for 

their proposed model and proved that it was a convex function. Finally, they obtained an optimal 

replenishment policy for such an imperfect EMQ model.


121 

Recently, an algebraic method of determining the economic order quantity (EOQ) model with 

backlogging was introduced by Grubbström and Erdem [25]. They used algebraic derivation to find 

the optimal order quantity without reference to the first- or second-order derivatives. Similar 

methodologies have been applied to solve various aspects of supply chain optimisation [26-28]. This 

paper extends such an approach in order to re-examine a manufacturing system with a random 

defective rate, a rework process, and multiple deliveries of its end product [13]. 

METHODS 

We present a two-phase algebraic approach [13] in order to re-examine a manufacturing 

system with a random defective rate, a rework process, and multiple shipments of finished items. 

Such a specific model is described as follows. Assume a production system has an annual production 

rate P and randomly produces a proportion x of defective items during its uptime at a production rate 

d. All manufactured items are screened and the inspection cost is included in the unit production cost 

C. Non-conforming products fall into two groups: the scrap (a proportion of θ) and the repairable 

(1-θ). The rework process has a rate of P 1 units per year and commences immediately after regular 

production in each cycle. A proportion θ 1 of reworked items fails during rework and is treated as 

scrap. Under regular supply, the constant production rate P must be larger than the sum of the 

demand rate λ and the production rate of defective items d, i.e. (P-d-λ) > 0, where d can be 

expressed as d = Px. Let d 1 denote the production rate of scrap during rework; d 1 can then be 

expressed as d 1 = P 1 θ 1 . Furthermore, the proposed system considers a multi-delivery policy for the 

end items with quality assurance. That is, the finished items can only be delivered to the customers if 

the whole lot is quality assured at the end of the reworking process. A fixed quantity of n 

installments of the finished batch is delivered to customers at fixed interval of time during production 

downtime t 3 (Figure 1). Other notations used in the proposed system are listed below. 

t 1 = regular production time in the proposed model, 

t 2 = time required to rework defective items, 

t 3 = time required to deliver all perfect-quality end products, 

t n = fixed interval of time between each installment of finished end products delivered 

during production downtime t 3 , 

T = cycle length, 

Q = manufacturing batch size __ the decision variable, 

n = number of fixed-quantity installments of finished batch to be delivered to 

customers __ the decision variable, 

H 1 = maximum level of on-hand inventory when regular production ends, 

H = maximum level of on-hand inventory when the rework process finishes, 

I(t) = on-hand inventory of perfect quality end items at manufacturer’s end at time t, 

TC(Q,n) = total production-inventory-delivery costs per cycle, 

K = set-up cost per cycle, 

C = unit production cost, 

h = unit holding cost, 

C R = unit rework cost, 

= holding cost for each reworked item, 

h 1


C S = disposal cost per scrap item, 

K 1 = fixed delivery cost per shipment, 

C T = delivery cost per item shipped to customers, 

φ = overall scrap rate per cycle (sum of scrap rates in periods t 1 and t 2 ), 

h 2 = holding cost for each item kept by customer, 

E[TCU(Q,n)] = long-run average cost per unit time. 

Figure 1. On-hand inventory of perfect end items in the proposed model with random defective 

rate, reworking and multi-delivery policy [13] 

With reference to Figure 1, the total production-inventory-delivery cost per cycle, TC(Q, n), 

consists of the following. (a) set-up cost and variable manufacturing costs per cycle; (b) total quality 

costs including variable repairing costs, holding costs for reworked items, and disposal costs for 

scrap items per cycle; (c) fixed and variable delivery costs per cycle; (d) total holding costs at the 

manufacturer’s end for all items produced in the periods t 1 , t 2 and t 3 ; and (e) total holding costs at the 

customer’s end for all items stocked in t 3 : 

P1 

t2 

TC Q, n K CQ CR x 1Qh 1 t2CSxQnK1 

2 

H1 dt1 H1 

H n 

1 

 

(1) 

CT Q1 

xh 

t1 t2 Ht3 

2 2 2n 

 

 

 

h2 

H 

t 

3 

T H 

t 

3 

2 

n 

 

 

With further derivations, the long-run average cost per unit time E[TCU(Q,n)] for the 

proposed system can be written as follows (see mathematical modelling section in Chiu et al. [13]):


 

, 

C 

 

1 

CT 

ET 1 

Ex Q1 

 

Ex 

2 2 

1 

 

h1 

Ex Q1 

 

CSEx 

1Ex 

2P1 

1Ex 

1Ex 

hQ 

hQ 

2 2 

 

 

 

2E x E x E x 1 

 

 

1 

 

 

1 hQ 1 

E x hQ 

hQEx1 

 

E TC Q n K nK 

ETCUQ, 

n 

 

C E x 

R 

 

 

2P 1 E x 2P 1 E x 

 

 

1 n 

 

 

 

2 2P 2P1 

 

2 2 2 

 

1 

 

1hQ 1hQ 

hQ 1E x 

1 Ex 

1 1 

 

n 2 n 2P 2 n 

P1 

 

 

 

123 

(2) 

Derivation of Optimal Policy using Two-phase Algebraic Approach 

Unlike the conventional method, which uses differential calculus on the cost function 

E[TCU(Q, n)] to find the optimal point [13], a straightforward two-phase algebraic approach to 

determining the optimal production-shipment policy for the proposed model is adopted here. 

Phase 1: Derivation of n* 

It can be seen that Eq.2 has two decision variables, namely Q and n. Moreover, there are 

several different forms of these decision variables in the right-hand side of Eq.2, e.g., Q, Q -1 , nQ -1 

and Qn -1 . Therefore, Eq.2 can be rearranged as 

or 

 

 

 

1 

 

 

1 

 

 

 

 

C 

CRE x CSE x 

ETCUQ, 

n 

CT 

 

1 E x 1E x 1E x 

2 

 

2 

 

h1 

E x h h 

 

 

2 2 

 

 

 

2E x E x E x 1 

 

 

Q 

1 E x 2P1 2P 2P 

 

 

 

 

 

1 

 

 

 

 

 

K 

 

Q 

P P1 

 

1 

Ex 

h 1 E x E x 1 

 

hh2 

 

 

2 2 2 

 

K1 

1 

 

nQ 

1 E x 

 

1 

Ex Ex1 

 

 

 

+ 

 

 

 

 

 

1 

h h2 

n Q 

2 2P 

2P1 

 

 

1 1 1 

, 

1 2 3 4 5 

 

 

ETCU Q n Q Q nQ n Q 

(4) 

 

Q 

1 

(3) 

where α 1 , α 2 , α 3 , α 4 and α 5 denote: 

 

1 

C 

CREx1 

CSE x 

CT 

 

(5) 

1Ex 

1Ex 

1Ex


 

 

2 

3 

 

4 

 

 

 

 

 

1 

E x 

 

1 

 

 

2 2 

h1 

E x h h 

 

 

2P1 2P 2P 

 

1 

 

 

h 1E x E x 1 

 

hh2 

 

 

 

 

2 2P 

2P1 

 

K 

 

2 2 

2Ex 

Ex Ex 

1 

 

 

1 

Ex 

(7) 

5 

K 

1 

1 

Ex 

(8) 

1 

Ex Ex1 

 

hh 

 

 

(9) 

2 

 

2 2P 

2P 

With further rearrangements, Eq.4 becomes 

1 

 

 

 

 

 

(6) 

1 2 1 1 

ETCUQ, 

n 2 

1 

Q 2 Q 3 n Q 4 

nQ 

 

 

5 

 

 

2 2 

1 1 1 

, 

1 2 3 4 

ETCU Q n Q Q n Q nQ 

 

5 

 

2 2 

 

2 3 4 5 

Eq.11 will be minimised if its second and third terms in it equal zero. That is: 

(10) 

(11) 

Q 

n 

 

3 

(12) 

2 

 

5 

Q 

(13) 

4 

Substituting Eq.6 and 7 into Eq.12, and then substituting Eq.8, 9 and 12 into Eq.13, the optimal 

number of shipments n* is 

n 

 

E x 1 

K 

hh2 

1 

Ex 

5 

P P1 K 

3 

1 

 

 

2 2 

4 2 

 

h1 

Ex 1 

 

h h 

 

2 2 

 

 

 

 

2E x E x 

E x 1 

1 E x P1 P P 

 

1 

 

 

 

 

 

 

E x 

h1 

Exhh2 

 

 

 

P 

 

 

1 

 

P1 

 

 

(14) 

It is noted that Eq.14 is identical to that obtained using the conventional differential calculus 

method [13]. We can also see that although in real-world situation the number of deliveries takes 

integer values only, Eq. 14 results in a real number. In order to locate the integer value of n* that 

minimises the long-run average cost for the proposed system, the two adjacent integers to n must be 

examined respectively for cost minimisation [11]. Let n + denote the smallest integer greater than or


125 

equal to n (derived from Eq. 14) and n - denote the largest integer less than or equal to n. Because n* 

is either n + or n - , we can first treat E[TCU(Q,n)] (Eq. 4) as a cost function with a single-decision 

variable Q, and perform the following rearrangements. 

Phase 2: Searching for Q* 

First, the long-run cost function E[TCU(Q, n)] (i.e. Eq.4) can be rearranged as the following 

single-decision-variable function: 

where α 6 and α 7 denote: 

With further rearrangements, Eq.15 becomes 

1 

, 

 

E TCU Q n Q Q 

1 6 

 

7 

(15) 

 

(16) 

1 

6 

2 n 5 

n 

(17) 

7 3 4 

 

 

2 

1 

 

 

1 6 7 6 7 

ETCU Q, n Q 

Q 

 

2 

 

(18) 

Upon derivation of Eq.18, it can be noted that E[TCU(Q,n)] will be minimised if the second term in 

it equals zero. That is: 

Q* 

 

7 

(19) 

 

6 

Substituting Eq.16 and 17 into Eq. 19, the optimal production lot size is 

Q* 

 

 

1 

 

 

2 2 

1 

2 2 

2 

 

2Ex 

Ex Ex 

1 

1 h1 

Ex 

 

1 

 

2 1 

Exh h2 

h2 

Ex 

n 

P P1 

n 

2 

K nK 

h E x 1 

h h 

1 

 

P1 P P 

 

1 

n 

 

 

 

E x 1 

1 1 1 

 

 

 

(20) 

It is noted that Eq.20 is identical to that obtained using the conventional differential calculus 

method [13]. 

To find the optimal production-shipment (Q*, n*) policy, we substitute all related system 

parameters, along with n + and n - , into Eq.20. Then, applying the resulting (Q, n + ) and (Q, n - ) 

respectively in Eq. 4, we choose the one that gives the minimum long-run average cost as the optimal 

production-shipment policy (Q*, n*). A numerical example to demonstrate the practical usage of this 

method is provided in the next section.


NUMERICAL EXAMPLE 

The aforementioned two-phase algebraic approach and its resulting Eq.14, 20 and 4 are 

verified in this section using the same numerical example [13]. Suppose an end product can be 

produced at a rate of 60,000 units per year, its annual demand being estimated to be 3,400 units, and 

during the production process there is a random defective rate x that follows a uniform distribution 

over a range of [0, 0.3]. A proportion θ = 0.1 of the imperfect items is considered to be scrap and 

the other portion is assumed to be repairable with a rework rate of P 1 = 2,100 units per year. It is 

further estimated that there is a proportion θ 1 = 0.1 of reworked items that fail (become scrap) during 

the rework period. As a quality assurance policy, the finished items can only be delivered to 

customers if the whole lot is quality-assured after reworking. A fixed quantity of n installments of the 

perfect end items are shipped to customers at a fixed interval of time during delivery time t 3 (Figure 

1). Other selected parameter values in this example are as follows: 

C = $100 per item, 

C R = $60 for each reworked item, 

C S = $20 for each scrap item, 

K = $20,000 per production run, 

h = $20 per item per year, 

h 1 = $40 per reworked item per unit time, 

K 1 = $2,400 per shipment, 

C T = $0.1 per item delivered, 

h 2 = $80 per item kept at the customer’s end per unit time. 

Applying Eq.14, we obtain n=2.736. Because the number of deliveries has to be an integer, 

we have n + =3 and n - =2. Substituting all system parameters, along with n + and n - respectively, into 

Eq.20, we find two possible policies, namely (Q, n + )=(1735, 3) and (Q, n - )=(1579, 2). We then apply 

(Q, n + ) and (Q, n - ) in Eq.4 to obtain E[TCU(1735,3)]=$485,541 and E[TCU(1579,2)]=$487,071. 

Selecting that with the minimum cost, we find that the optimal policy (Q*, n*)=(1735, 3) and 

the long-run average cost E[TCU(Q*, n*)]=$485,541. The results are noted to be identical to those 

obtained by Chiu et al. [13]. 

The effect of varying the lot-size Q on the long-run average cost function E[TCU(Q, n)] and 

on the components of E[TCU(Q, n)], for n* = 3, is depicted in Figure 2.


127 

Figure 2. Effect of varying lot size Q on the long-run average cost function 

E[TCU(Q, n)] and on the components of E[TCU(Q, n)] for n* = 3 

CONCLUSIONS 

This paper proposes a two-phase algebraic approach to determining the optimal productionshipment 

policy for an end product in an integrated supplier-customer system with quality assurance. 

Unlike the conventional method, which uses differential calculus on the system cost function 

to find the economic lot size and optimal number of deliveries, the proposed two-phase algebraic 

approach is a straightforward method that may enable practitioners with little or no knowledge of 

differential calculus to understand and manage real-world systems more effectively. The research 

results were confirmed to be identical to those obtained by the traditional method 


The authors would like to express their appreciation of the support of this research project by 

the National Science Council of Taiwan under grant number: NSC 99-2221-E-324-017. 

REFERENCES 

1. F. S. Hillier and G. J. Lieberman, “Introduction to Operations Research”, McGraw Hill, New 

York, 2001. 

2. E. W. Taft, “The most economical production lot”, Iron Age, 1918, 101, 1410-1412. 

3. S. Nahmias, “Production and Operations Analysis”, McGraw Hill, New York, 2005, pp.183- 

205. 

4. L. B. Schwarz, “A simple continuous review deterministic one-warehouse N-retailer inventory 

problem”, Manage. Sci., 1973, 19, 555-566. 

5. S. K. Goyal, “An integrated inventory model for a single supplier-single customer problem”, Int. 

J. Prod. Res., 1977, 15, 107-111.


6. L. B. Schwarz, B. L. Deuermeyer and R. D. Badinelli, “Fill-rate optimization in a onewarehouse 

N-identical retailer distribution system”, Manage. Sci., 1985, 31, 488-498. 

7. A. Banerjee and S. Banerjee, “Coordinated order-less inventory replenishment for a vendor and 

multiple buyers”, Int. J. Tech. Manage., 1992, 7, 328-336. 

8. R. M. Hill, “The optimal production and shipment policy for the single-vendor singlebuyer 

integrated production-inventory problem”, Int. J. Prod. Res., 1999, 37, 2463-2475. 

9. D. J. Thomas and S. T. Hackman, “A committed delivery strategy with fixed frequency and 

quantity”, Eur. J. Oper. Res., 2003, 148, 363-373. 

10. K. Ertogral, M. Darwish and M. Ben-Daya, “Production and shipment lot sizing in a vendorbuyer 

supply chain with transportation cost”, Eur. J. Oper. Res., 2007, 176, 1592-1606. 

11. Y-S. P. Chiu, F-T. Cheng and H-H. Chang, “Remarks on optimization process of 

manufacturing system with stochastic breakdown and rework”, Appl. Math. Lett., 2010, 10, 

1152-1155. 

12. C-C. Chern and I-C. Yang, “A heuristic master planning algorithm for supply chains that 

consider substitutions and commonalities”, Expert Syst. Appl., 2011, 38, 14918-14934. 

13. S. W. Chiu, H-D. Lin, M-F. Wu and J-C. Yang, “Determining replenishment lot size and 

shipment policy for an extended EPQ model with delivery and quality assurance issues”, Sci. 

Iran., 2011, 18, 1537-1544. 

14. W. Shih, “Optimal inventory policies when stockouts result from defective products”, Int. J. 

Prod. Res., 1980, 18, 677-686. 

15. M. J. Rosenblatt and H. L. Lee, “Economic production cycles with imperfect production 

processes”, IIE Trans., 1986, 18, 48-55. 

16. X. Zhang and Y. Gerchak, “Joint lot sizing and inspection policy in and EOQ model with 

random yield”, IIE Trans., 1990, 22, 41-47. 

17. K. Moinzadeh and P. Aggarwal, “Analysis of a production/inventory system subject to random 

disruptions”, Manage. Sci., 1997, 43, 1577-1588. 

18. V. Makis, “Optimal lot sizing and inspection policy for an EMQ model with imperfect 

inspections”, Naval Res. Log., 1998, 45, 165-186. 

19. M. A. Rahim and M. Ben-Daya, “Joint determination of production quantity, inspection 

schedule, and quality control for an imperfect process with deteriorating products”, J. Oper. 

Res. Soc., 2001, 52, 1370-1378. 

20. S. W. Chiu, D-C. Gong and H-M. Wee, “Effects of random defective rate and imperfect rework 

process on economic production quantity model”, Japan J. Ind. Appl. Math., 2004, 21, 375- 

389. 

21. S. W. Chiu, K-K. Chen and J-C. Yang, “Optimal replenishment policy for manufacturing 

systems with failure in rework, backlogging and random breakdown”, Math. Comput. Model. 

Dyn. Sys., 2009, 15, 255-274. 

22. A. Amirteimoori and A. Emrouznejad, “Input/output deterioration in production processes”, 

Expert Syst. Appl., 2011, 38, 5822-5825. 

23. Y-S. P. Chiu, H-D. Lin and H-H. Chang, “Mathematical modeling for solving manufacturing 

run time problem with defective rate and random machine breakdown”, Comput. Ind. Eng., 

2011, 60, 576-584.


129 

24. B. Sarkar, S. S. Sana and K. Chaudhuri, “An imperfect production process for time varying 

demand with inflation and time value of money - An EMQ model”, Expert Syst. Appl., 2011, 38, 

13543-13548. 

25. R. W. Grubbström and A. Erdem, “The EOQ with backlogging derived without derivatives”, 

Int. J. Prod. Econ., 1999, 59, 529-530. 

26. S. W. Chiu, “Production lot size problem with failure in repair and backlogging derived without 

derivatives”, Eur. J. Oper. Res., 2008, 188, 610-615. 

27. H-D. Lin, Y-S. P. Chiu and C-K. Ting, “A note on optimal replenishment policy for imperfect 

quality EMQ model with rework and backlogging”, Comput. Math. Appl., 2008, 56, 2819- 

2824. 

28. Y-Y. Xiao, R-Q. Zhang and I. Kaku, “A new approach of inventory classification based on loss 

profit”, Expert Syst. Appl., 2011, 38, 9382-9391. 



130 Maejo Maejo Int. J. Int. Sci. J. Technol. Sci. Technol. 2012, 2012, 6(01), 6(01), 130-151 130-151 

Full Paper 





IIS-Mine: A new efficient method for mining frequent itemsets 

Supatra Sahaphong * and Veera Boonjing 

Department of Mathematics and Computer Science, Faculty of Science, King Mongkut’s Institute of 

Technology Ladkrabang, Bangkok 10520, Thailand 

* Corresponding author, e-mail: supatra@ru.ac.th 

Received: 11 September 2011 / Accepted: 4 March 2012 / Published: 23 April 2012 

Abstract: A new approach to mine all frequent itemsets from a transaction database is 

proposed. The main features of this paper are as follows: (1) the proposed algorithm 

performs database scanning only once to construct a data structure called an inverted 

index structure (IIS); (2) the change in the minimum support threshold is not affected by 

this structure, and as a result, a rescan of the database is not required; and (3) the 

proposed mining algorithm, IIS-Mine, uses an efficient property of an extendable itemset, 

which reduces the recursiveness of mining steps without generating candidate itemsets, 

allowing frequent itemsets to be found quickly. We have provided definitions, examples, 

and a theorem, the completeness and correctness of which is shown by mathematical 

proof. We present experiments in which the run time, memory consumption and scalability 

are tested in comparison with a frequent-pattern (FP) growth algorithm when the 

minimum support threshold is varied. Both algorithms are evaluated by applying them to 

synthetics and real-world datasets. The experimental results demonstrate that IIS-Mine 

provides better performance than FP-growth in terms of run time and space consumption 

and is effective when used on dense datasets. 

Keywords: association rule mining, data mining, frequent itemsets mining, frequent 

patterns mining, knowledge discovering 

________________________________________________________________________________ 

INTRODUCTION 

The objective of frequent itemset mining is to identify all frequently occurring itemsets using 

a support threshold. Decision-makers are interested in all itemsets associated with high frequencies. 

Association rule mining algorithms can be broken down into two major phases. The first phase finds 

all of the itemsets that satisfy the minimum support threshold, which are the frequent itemsets. The


131 

second phase is rule generation, in which all the high confidence rules from the frequent itemsets 

found in the previous phase are extracted [1]. Many previous investigations focused on the first 

phase. Early algorithms based on generated and tested candidate itemsets have two major defects. 

First, the database must be scanned multiple times to generate candidate itemsets, which increases 

the I/O load and is time-consuming. The search space of itemsets that must be explored grows 

exponentially. Second, enormous candidate itemsets are generated and calculated from their 

supports, which consumes a large amount of CPU time [2]. 

To overcome the above-mentioned problems, a next generation of algorithms using a 

compact tree structure was proposed, called a frequent-pattern (FP) tree [3], which finds frequent 

itemsets directly from the data structure. However, most of the FP-tree-based algorithms have the 

following weaknesses. First, the mining of frequent itemsets from the FP-tree to generate a huge 

conditional FP-tree requires a large amount of run time and space. The best case is when a database 

has the same set of transactions; an FP-tree then contains only a single branch of nodes. The worst 

case is when a database has a unique set of transactions [3]. Second, when the users change to a new 

minimum support threshold for their new decision, the algorithm restarts the whole operation and 

scans the database twice. 

Many researchers have tried to solve the above problems using a vertical data layout. 

However, most of the algorithms have the drawback of increasing the run time and space 

consumption due to the following reasons. First, when the users change to a new minimum support 

threshold for their new decision, the algorithm restarts the whole operation more than one time to 

scan a database and construct their data structure. Rescanning the database for a new minimum 

support threshold wastes both run time and space. Second, all of the FP-tree-based algorithms 

generate a huge conditional FP-tree, which has a large number of recursive processing steps and 

requires a large amount of run time and space consumption. 

In this paper we present a new, efficient method to solve the above-mentioned problems by 

proposing both a data structure and a mining algorithm for decreasing the consumption of run time 

and space. First, the proposed method performs database scanning to construct a data structure 

called an inverted index structure (IIS) only once. In addition, changing the minimum support 

threshold does not affect the IIS; therefore, database rescanning is not required. Second, IIS-Mine is 

a new algorithm that mines all of the frequent itemsets without generating candidate itemsets and 

uses a new tree structure called the IIS item Tree. IIS-Mine employs an efficient property of the 

extendable itemset, which decreases the number of recursive processing steps when mining frequent 

itemsets. The completeness and correctness of the algorithm is proved using a mathematical proof. 

Last, the efficiency of IIS-Mine is compared with that of FP-growth in terms of run time and space 

consumption through simulation experiments. Our experiments show that IIS-Mine is more efficient 

than FP-growth in run time and space consumption for dense datasets. 

RELATED WORK 

The first algorithm to generate all frequent itemsets is the AIS algorithm, which was first 

introduced by Agrawal et al [4]. However, this algorithm constructs a list of all of the possible


itemsets at each level of traversal, so infrequent itemsets that are not needed are also generated. 

Later, the algorithm was improved upon and renamed the Apriori algorithm by Agrawal et al [5]. The 

Apriori algorithm uses a level-wise and breadth-first search approach for generating association 

rules. Many efficient association mining techniques have been developed based on the Apriori 

algorithm. Vu et al. [6] proposed a rule-based location prediction technique to predict the user’s 

featured location, but this proposal generates more candidate itemsets than are required. These 

algorithms are also expensive in terms of I/O load and run time when the database must be scanned 

multiple times to generate candidate itemsets. 

The above-mentioned problems can be improved upon by using a compact tree structure and 

finding frequent itemsets directly from the data structure. The algorithms scan a database twice. The 

first scan of the database is to discard infrequent itemsets; the second is to construct a tree. The FPgrowth 

algorithm, developed by Han et al. [7], is the most popular method. It performs a depth-first 

search approach in a search space. It encodes a dataset using a compact data structure called an FPtree 

or prefix tree and extracts frequent patterns directly from the FP-tree. Many approaches have 

been proposed to extend and improve upon this algorithm. Pei et al. [8] developed the H-mine 

algorithm using array- and tree-based data structures to improve the main memory cost. The 

PatriciaMine algorithm [9] compressed Patricia tries to store datasets, which is space efficient for 

both dense and sparse datasets. The FP-growth algorithm [10] reduces the FP-tree traversal time 

using an array technique. Zhu [11] proposed a new method to compress a large database into an FPtree 

with a children table but not a header table, and applied a depth-first search with this tree for the 

mining step, which reduces both the run time and the space consumption. Sahaphong and Boonjing 

[12] proposed a new algorithm which constructs a pattern base using a new method that is different 

from the pattern base in the FP-growth and mined frequent itemsets using a new combination method 

without the recursive construction of a conditional FP-tree. An approach based on the FP-tree and 

co-occurrence frequent items (COFI) was proposed to find frequent items in multilevel concept 

hierarchy by using a non-recursive mining process [13]. A new data structure called improved FP 

tree was proposed, which can reduce space consumption and enhance the efficiency of an attribute 

reduction algorithm [14]. To maintain the anti-monotone property of approximate weighted frequent 

patterns, a robust concept was proposed to relax the requirement for exact equality between the 

weighted supports of patterns and a minimum threshold [15]. However, most of the FP-tree-based 

algorithms require a large amount of run time and space to generate the huge conditional FP-trees. 

Moreover, the algorithm restarts the whole operation and requires that a database be scanned twice 

when the minimum support threshold is changed. 

As mentioned above, most of the algorithms that mine frequent itemsets use a horizontal data 

layout. However, many researchers use a vertical data layout. The Eclat algorithm was proposed 

[16] to generate all frequent itemsets in a breadth-first search using the joining step from the Apriori 

property when no candidate items can be found. The Eclat algorithm is very efficient for large 

itemsets but is less efficient for small ones. The diffset technique [17] was introduced to improve the 

memory requirement. Chai et al. [18] detailed a data structure called large-item bipartite graph to 

accommodate the data when a database is scanned. Similar to the FP-growth algorithm, this method


133 

mines frequent patterns using the recursive conditional FP-tree. The BitTableFI algorithm [19] uses 

horizontal and vertical data layouts to compress a database. Yen [20] presented an algorithm based 

on an undirected itemset graph that finds frequent itemsets by searching undirected graphs. When the 

database and minimum support change, this algorithm requires that the graph structure be researched 

to generate new frequent itemsets. The Index-BitTableFI [21] was developed to reduce the 

cost of candidate generation and to support counting. Sahaphong and Boonjing [22] proposed a new 

algorithm that reduces the run time. The drawback of this algorithm is its large memory consumption 

from generation of many repeated nodes. The JoinFI-Mine algorithm [23] uses a sorted-list structure 

constructed from the vertical data layout and finds all frequent itemsets using a depth-first search for 

joining frequent itemsets. Therefore, this algorithm consumes time and space in its joining step. 

METHODS 

Frequent-Itemsets Mining Problem 

We introduce the basic concepts of mining frequent itemsets. All terminologies in this section 

are proposed by Han et al [2]. 

Let I = { x 1, 

x 2 ,..., x m} 

be a set of items and DB = { T1 , T2 

,..., Tn} 

be a transaction database, where 

T 1 , T2 

,..., T n are transactions that contain items in I. The support, or supp (occurrence frequency), of a 

pattern A, where A is a set of items, is the number of transactions containing A in DB. A pattern A is 

frequent if A’s support is no less than a predefined minimum support threshold, minsup. 

Given a DB and a minimum support threshold minsup, the problem of finding a complete set 

of frequent itemsets is called the frequent-itemsets mining problem. 

For a greater understanding, we provide an example to illustrate the above definitions. 

Example 1. An example of the database by Han et al. [2] is used here. Table 1 is a DB. It 

consists of 5 transactions (T 1 , T 2 , T 3 , T 4 , and T 5 ) and 17 items (a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, 

p, and s). For example, the first transaction is T 1 , which contains f, a, c, d, g, i, m, and p. 

Table 1. A transaction database 

Transaction 

T 1 

T 2 

T 3 

T 4 

T 5 

Item 

f, a, c, d, g, i, m, p 

a, b, c, f, l, m, o 

b, f, h, j, o 

b, c, k, s, p 

a, f, c, e, l, p, m, n 

IIS: Design and Construction 

We present a data structure that contains transaction data called an inverted index structure 

(IIS). The IIS is a structure that holds a relationship between items and the transactions included 

within. The IIS is constructed from one scan of the DB. This original IIS can support every minsup;


therefore, it does not need to rescan the DB when the minsup is changed. According to the 

definitions in the previous section, we present a new definition with an example and an algorithm to 

demonstrate how to construct this IIS. 

Definition 1 (IIS). Let DB be a transaction database and I be a non-empty finite set of all 

items in each transaction in the DB, where each transaction is a set of items in I associated with an 

identifier, and let S be a set of all non-empty subsets of DB. An IIS is the function f : I → S defined 

by f ( a) 

= S1 

if T contains a for eachT 

∈S1 

. This function can be identified as a table consisting of the 

attributes of the items in I and the corresponding transactions in the DB. That is, each row in the IIS 

contains an item in I as well as the transactions in the DB that contain that item. The set of 

transactions are written in the order of their ascending identification numbers. 

With the above definition, the IIS represents the relationship between each item in I and its 

corresponding transactions; therefore, the IIS can apply to all minimum support thresholds, and a 

rescan of the database is not required. We demonstrate the steps to construct the IIS through the 

following example. 

Example 2. We use the example of a DB in Table 1. The DB is scanned once to create the 

IIS. The scan of the first transaction is T 1 , which consists of items f, a, c, d, g, i, m and p. The 

transaction T 1 will be inserted for each corresponding item sorted in ascending order (a, c, d, f, g, i, 

m, p). T 1 will be the first transaction inserted in the transactions of item a. The second examined item 

is c, so we insert T 1 in item c. Next, we examine item d; we then subsequently insert T 1 in item d. The 

remaining items (f, g, i, m and p) in T 1 can be similarly inserted. The remaining transactions (T 2 , T 3 , 

T 4 and T 5 ) in the DB are performed in a similar manner. 

Algorithm 1 shows how to construct the IIS. Figure 1 shows all of the items of the IIS after 

scanning the DB once, and the bold items are all frequent items that have a support greater than or 

equal to the minsup, which is assumed to be 3. 

Algorithm 1 (IIS construction) 

Input: DB. 

Output: IIS. 

Method: The IIS is constructed as follows. 

1 Begin 

2 Create header that contains all items. 

3 For each transaction T in DB do // scanning DB once 

4 Sort items in T // ascending order 

5 Create transaction to each corresponding item 

6 End //For 

7 End //Begin


135 

Item Transaction 

a T 1 T 2 T 5 

b T 2 T 3 T 4 

c T 1 T 2 T 4 T 5 

d T 1 

e T 5 

f T 1 T 2 T 3 T 5 

g T 1 

h T 3 

i T 1 

j T 3 

k T 4 

l T 2 T 5 

m T 1 T 2 T 5 

n T 5 

o T 2 T 3 

p T 1 T 4 T 5 

s T 4 

Figure 1. An example of the IIS 

IIS-Mine Algorithm 

We present a new algorithm called IIS-Mine. This algorithm uses a new tree structure, called 

the IIS item Tree, to mine frequent itemsets. The main features of this algorithm are as follows: (1) 

every frequent itemset is found without generating candidate itemsets; (2) the algorithm reduces the 

recursion of mining steps using the property of extendable itemset; and (3) the algorithm supports 

the mining of frequent itemsets with any value of the minsup without needing to rescan the database. 

From the above features, we can quickly find the frequent itemsets and completely and correctly 

obtain them. We now introduce the terminologies of the IIS item Tree, its construct, the theorem, the 

examples and the algorithms to describe how to mine frequent itemsets. 

Definition 2 (Itemset-tree structure). An itemset-tree structure is a tree structure 

constructed from the IIS. It is a finite set of one or more nodes with the following structure: 

(i) It consists of the root which contains an item, a set of item subtrees as the children of the root, 

and a set of header tables. 

(ii) Each node in this tree comprises five fields: item-name, which registers which item this node 

represents; support, which registers the number of transactions represented by the portion of the path 

reaching this node; same-item, which represents a pointer that points to the node in the itemset-tree 

structure that carries the same item-name; parent, which represents a pointer that points to the 

previous node in the same path; and child, which represents a pointer that points to the child node. 

(iii) Each member of the header table consists of two fields, item-name and head of node link, 

where head of node link represents a pointer that points to the first node in the itemset-tree structure 

carrying the item-name.


Definition 3 (IIS item tree). Let x 0 be an arbitrary frequent item in a given transaction database 

and IIS be the inverted index structure of the transaction database. A tree T constructed from the 

IIS is called an inverted index structure- x 0 tree, denoted by IIS x tree , if it satisfies the following: 

0 

(i) Each node of T is of the form ( A : s), 

where A is a frequent itemset and s is its support. If 

( A : s) 

is a node of T and A = { a}, 

where a is a frequent item, then ( A : s) 

is simply written by ( a : s). 

(ii) Let ( x 0 : s0 

) be its root, where s 0 is the support of x 0 . 

(iii) Let x 0 , x1,..., 

xk 

be frequent items in the IIS, and s i = supp { x 0 , x1,..., 

xi} 

for all i = 0,1,..., k. 

In 

this case, P = (( x0 : s0 

), ( x1 

: s1 

),...,( x k : sk 

)) is a path from the root ( x 0 : s0 

) to a leaf ( x k : sk 

) of the tree T 

if and only if s 0 ≥s 

1 ... ≥s 

k > 0 , x 0 < l x1 

< l ... < l xk 

, where 

l 

is the lexicographic order. If a is a 

frequent item in the IIS and if ( a : s) 

is a node of T such that x k < a or x i < l a < l xi+ 

1 for all i = 0,1,..., k, 

then supp{ x0 , x1,..., 

xi , a} = 0 and ( a : s) 

is not a node of P. 

The header table of IIS x0 

tree is a set of all frequent items a of a node ( a : s) 

of this tree. 

Based on the above definition, we have the IIS item Tree construction algorithm, as shown in 

Algorithm 2. It is evident that if minsup>0 is a minimum support threshold, then every frequent 

itemset can be derived from an IIS item Tree. 

Algorithm 2 (IIS item Tree construction) 

Input: IIS. 

Output: IIS item Tree. 

Method: An IIS item Tree is constructed as follows. 

1Begin 

2 Create header table 

3 Read frequent item x in IIS 

4 Create root R and initial supp(R) to 1 

5 Link R to header table 

6 For each transaction T of root R where T = T1 

to Tn 

do 

7 While next frequent item≠(last frequent item)+1 do 

8 Read next frequent item (N) that has same T with R 

9 Call InsertTree (N,R) 

10 End//While 

11 End//For 

12End //Begin 

Procedure InsertTree (N,R) 

1Begin 

2 If IIS R Tree has a node C such that C.item-name = N.item- name then 

3 Increment supp(N) by 1 

4 Else 

5 Create new node N and initial supp(N) to 1 

6 Link N to N’s parent 

7 Link N to N’s header table 

8 Link N to same-item 

9 End //If 

10End //Begin 

If no confusion arises, then { x 1, 

x2 

,..., xn} 

and ({ x1 , x2 

,..., xn }: s) 

are replaced by x 1x2... 

xn 

, and 

x1 x2... 

xn : s respectively, where x 1 , x2 

,..., xn 

are items. 

Example 3. In this example, we describe the steps to construct all of the IIS item Trees except 

the last frequent item p using the IIS in Figure 1 and minsup = 3. The first frequent item in the IIS is


137 

a; therefore, we first construct the IIS a Tree, and item a is a root. We obtain two paths: (, 

, , , ) and (, , , , ). The first path consists of 

frequent items (a,c,f,m,p) that appear twice in the DB. Similarly, the second path indicates frequent 

items (a,b,c,f,m) that are contained in only one transaction in the DB. These two paths share the 

frequent item a; thus, a appears three times in the DB. Other IIS item Trees, except the IIS p Tree, can be 

similarly constructed. In Figure 2, all of the IIS item Trees, except the last IIS p Tree, are illustrated. 

a:3 

c:2 

b:1 

f:2 c:1 

m:2 

f:1 

p:2 m:1 

(a) (b) (c) (d) (e) 

Figure 2. IIS item Trees 

Definition 4 (A 1 -tree). Let DB be a transaction database; let T be a tree such that each node 

of the tree is of the form ( A : s) 

, where A is a frequent itemset in DB and s is the support of A; and let 

A 1 be a frequent itemset in the DB. T is called an A 1 -tree if, for any path P of T, P is of the form 

P = (( A1 : s1), 

( A2 

: s2 

),...,( A k : sk 

)), where k is a positive integer; A i and A j are pairwise disjoint 

frequent itemsets for all i, j = 1,2,…,k with i j 

i 

; s i is the support of 

m 

Am for i =1,2,…,k; (A 1 :s 1 ) 

= 1 

is the root of T; and ( A k : sk 

) is a leaf of T. 

Example 4. According to Figure 2 (a), the ac-tree is shown in Figure 3. 

Definition 5 (Prefix subpath). Let T be a tree, a 1 be the root of T , and P ( a 1 

,..., a m 

) be a 

path of T, where m is a positive integer. Every path Q = ( a 1 ,..., a k ) of T is then called a prefix subpath 

of P, where 1 ≤k ≤m 

. 

Example 5. According to Figure 2(a), the paths ((a:3),(c:2)) and ((a:3),(c:2),(f:2)) are prefix 

subpaths of ((a:3),(c:2),(f:2),(m:2)). 

Definition 6 (Subheader). Let T be an A-tree and (x:s(x)) be a node of T, where x is a 

frequent item in a given transaction database and s(x) is the support of x. Suppose that all of the 

nodes (of T) containing x are only in the paths P 1 ,…,P k of T from the root (A:s) to some leafs of T, 

and suppose that Q i is a prefix subpath (of P i ) from the root ( A : s) 

to ( x : s( 

Qi 

)) for all i = 1,..., k. 

The 

subheader of T, denoted by SH(A), is defined as the order set SH(A) = {x|x is a frequent item not 

k 

contained in A, ( x : s( 

Qi 

)) is a node in Q i for i = 1,…,k and ∑s ( Qi) 

≥minsup}. 

i = 1


Example 6. According to Figure 2(a), SH(a) = {c,f,m}, where SH(a ) is the subheader of the 

tree in Figure 2 (a), supp(c) = 3, supp(f) = 3, and supp(m) = 3. 

Definition 7 (Conditional itemset-tree). Let A 1 be a frequent itemset, T be an A 1 -tree, x 2 be 

a frequent item with x2 ∈SH( 

A1 

) - A 1, 

and A1 x2 

: = A1 

∪{ x2} 

be a frequent itemset. A conditional A 1 x 2 - 

tree, denoted by T ( A 1 x 2 ), is a tree that satisfies the following: 

(i) ( A 1x2 

: s2 

) is the root of T ( A 1 x 2 ), where s 2 is the support of A 1 x and 2 s 2 ≥ minsup. 

(ii) The number of items in SH ( A 1 ) > 1. 

(iii) All of the paths are derived from T in the following way: Q = (( A1 x2 

: s2 

), ( x3 

: s3 

),...,( x k : sk 

)) is a 

path of T ( A 1 x 2 ) if and only if a path P = (( A1 : s1 

), ( x2 

: s2 

),...,( x : l sl 

)) exists from the root ( A 1 : s1 

) to the 

leaf ( x l : sl 

) of T, where l ≥k 

and xk 

is in SH ( A 1 ); and if there is a node ( x r : sr 

) of P such that r > k 

and (( A1 : s1 

),( x2 

: s2 

),...,( x r 

: sr 

)) is a prefix subpath of P, then x r must not be in SH ( A 1 ). 

Example 7. According to Figure 2 (a), the conditional itemset-tree, or conditional ac-tree, is 

shown in Figure 4. 

Notably, every non-empty subset of a frequent itemset is also frequent. This fact leads to the 

following definition. 

Definition 8 (Frequent itemset * of length m derived from tree). Let A be a frequent 

itemset, x be a frequent item, and x A. Suppose that T is a conditional Ax-tree, m is a positive 

integer greater than 1, and Ax = A∪{ 

x} 

has m elements. Ax is called a frequent itemset * of length m 

derived from T, denoted by FS * m ( Ax ) , if T contains precisely two nodes and Ax is a frequent itemset, 

or if SH ( A) 

= { x}. 

Example 8. According to Figure 5, the frequent itemset * of length 4 derived from the 

* 

conditional acf-tree is FS 4 ( acfm) 

= acfm. 

. 

Figure 3. ac-tree Figure 4. Conditional ac-tree Figure 5. Conditional acf-tree 

Definition 9 (Extendable frequent itemset * ). Let m be a positive integer greater than 1, T 

be a T (Ax) 

defined as in definition 7 with A 1 = A and x 1 = x, 

and Ax be a frequent itemset* of length 

* 

m derived from T. Each itemset in FS m ( Ax) 

is said to be extendable if m ≥3 

. For every k = 2, 3,…, 

m, we let * ( Ax 

* 

Ext k ) denote the set of all itemsets containing exactly k items of FS m ( Ax), 

where m ≥3 

. 

Each element of Ext * ( Ax k ) is called an extendable frequent itemset * (derived from T) of length k for 

all k ≥2 

. The set of all extendable frequent itemsets* of length k that is denoted 

by Ext* k 

= ∪{ Ext* k ( Ay) 

| Ay is a frequent itemset* of length at least k derived from a conditional Aytree} 

for all k ≥2 

.


139 

* 

Example 9. According to the previous example, the length of FS 4 ( acfm ) is 4; we then find 

* 

that 2 ( acfm 

* 

Ext ) = {ac, af, am, cf, cm, fm} and Ext 3 ( acfm) 

= {acf, acm, afm, cfm}. Therefore, Ext * 2 = 

{ac, af, am, cf, cm, fm, and all members of other * * 

Ext 2 ( Ay) 

} and Ext 3 = {acf, acm, afm, cfm, and all 

members of other Ext * 3 ( Ay) 

}. 

Definition 10 (Frequent itemset * of length m). Let k be the maximum length of all frequent 

itemsets in a transaction database, FS * m ( Ax ) and Ext* k be given as in definitions 8 and 9 respectively; 

let Am = { FS * m ( Ax) 

| Ax being a frequent itemset * of length m derived from T } for m = 2, 3,…, k; 

* 

define FS 2 = A 2 ∪{ Ax | Ax being a frequent itemset of length 2}; and define FS * m = Am for m = 2, 3,…, 

k}. Let FI * * 

* * * 

m be defined by FI 1 = {{ x} 

| x being a frequent item} and FI m = FS m ∪Extm 

for m = 2, 3,…, 

k}. Any element of FI * m is called the frequent itemset * of length m. 

* 

Example 10. According to the previous example, we find that FI 2 = {ac, af, am, cf, cm, 

* 

* 

fm}, FI 3 = {acf, acm, afm, cfm}, and FI 4 = {acfm}. 

Definition 11 (Frequent itemset * ). Let FI * m be given as in definition 10, and let FI 

* denote 

k * 

∪m = 1 FI m , where k is the maximum length of all frequent itemsets in a transaction database. Any 

element of FI 

* is called a frequent itemset * . 

Example 11. According to the previous example, FI 

* is {ac, af, am, cf, cm, fm, acf, acm, 

afm, cfm, acfm}. 

On the basis of the above definitions and examples, Algorithm 3 presents the IIS-Mine 

algorithm to show how it can be used to mine all frequent itemsets. 

Algorithm 3: (IIS-Mine: Mining frequent itemsets using IIS item Tree) 

Input: IIS, IIS item Trees constructed according to Algorithm 2, and minsup 

Output: FI* 

Procedure AllFreqItemset (IIS, FI*) 

1 Begin 

2 For each frequent item x in the IIS do 

3 

* 

FI1 = {{ x | x ∈I 

, supp(x) ≥ minsup} 

4 Call IIS x Tree , which is constructed from Algorithm 2 

5 If Tree ≠{} 

6 Call IIS-Mine (Tree, x) 

7 End //If 

8 End //For 

9 

* 

Find FI ∪ 

FI * 

// FI* is given in definition 11 

= m=1 

m 

10 End //Begin 

Procedure IIS-Mine(Tree, x) 

1 Begin 

2 Call SubHeader (A-tree, subheaderA) 

3 Generate all Ax with its support 

//All Ax are the frequent itemsets where A is the root of A-tree and 

4 For each | λ |> 1 do // λ is Ax 

5 Flag=1 

6 While Flag = 1 do 

7 Call SkipFreqItemset (subheaderA, λ , δ ,FlagRepeat, FlagTree) 

8 If FlagRepeat=0 and FlagTree=0 then // ( δ * 

FI m ) 

9 Call CondItemsetTree (A-tree, δ , conditional δ -tree) 

x ∈subheaderA


10 Else 

11 Flag=0 

12 End // If 

13 If conditional Ax-tree contains greater than two nodes 

14 Call IIS-Mine(conditional δ -tree, x) 

15 Else Flag=0 

16 End //If 

17 End //While 

18 

* 

* 

If | FS (δ) |≥3 

and FS (δ) 

* 

FI then 

m 

m 

* 

m 

* 

k 

19 Call ExtFreqItemset( FS m (δ) 

, Ext ) // FS is given in definition 8 

20 

* 

* * 

* 

Save FS m (δ) 

and all Extk 

to FI m 

// FI m is given in definition 10 

21 Else 

22 

* 

If FSm 

(δ) 

* 

FI m then 

23 

* 

Save FS (δ) 

to 

m 

* 

FI m 

24 End //If 

25 End //If 

26 End //For 


Procedure SubHeader (A-tree, subheaderA) 

1 Begin 

2 Each frequent item x of A-tree // SH(A) is given in definition 6 

3 Find {(x:s(x))|x∈SH(A), s(x) is the support of x in A-tree} 

4 End // Begin 

Procedure CondItemsetTree(Tree, δ , conditional δ -tree) 

1 Begin 

2 Scan tree once to collect the paths that have an association with root δ 

3 For all paths are derived from tree do 

4 Connect all paths to δ 

5 End //For 

6 End //Begin 

Procedure SkipFreqItemset (SubheaderA, λ , δ , FlagRepeat, FlagTree) 

1 Begin // To skip the construction of conditional item tree 

2 // x is the frequent item in subheaderA // λ is the root of tree; or a frequent itemset 

3 // FlagRepeat=0 means δ * 

FI m // FlagTree=0 means the conditional item-tree is constructed 

4 // n is the maximum number of elements of itemsets in subheaderA 

5 FlagRepeat =1, FlagTree=1 

6 β = λ, α = β 

7 While( FlagRepeat=1 and (order of x ≤ n )) do 

8 If β * 

FI m then 

9 FlagRepeat=0, FlagTree=0 

10 If x is the last item 

11 If | δ |= 2 then FlagTree=1 

12 End //If 

13 δ = α 

14 End //If 

15 Else 

16 If x is the last item then 

17 Increment order of item x 

18 Else 

19 Increment order of item x 

20 β = δ∪x 

21 α = δ 

22 End// If 

* 

m


141 

23 δ = β 

24 End //If 

25 End //while 


* 

Procedure ExtFreqItemset( FS (δ) 

, 

m 

* 

Ext k ) 

1 Begin // Ext is given in definition 9 

* 

* 

2 Generate all subsets of FSm 

(δ) 

and save to Ext k 

3 End // Begin 

* 

k 

Example 12. This example is given to demonstrate how the proposed IIS-Mine algorithm 

can be used to mine all frequent itemsets. Assume minsup = 3 for all of the definitions above, and for 

Examples 1-3, Figure 2 and Algorithm 3. For simplicity, this example is divided into five main steps 

ordered by five frequent items in the IIS. The proposed mining algorithm proceeds as follows. 

Let step 1 be the first main step. According to Procedure AllFreqItemset, the frequent item a 

is the first frequent item in the IIS that is mined. The IIS a Tree is constructed, as illustrated in Figure 

2(a). The algorithm calls Procedure IIS-Mine, so Procedure Subheader is called in order. The SH(a) 

is (c,f,m), where supp(c) = 3, supp(f) = 3 and supp(m) = 3. After line 3 in the procedure, IIS-Mine 

generates all of the frequent itemsets with its support, i.e. ac, af and am; all of these frequent 

itemsets have support equal to three. Let steps 1.1, 1.2 and 1.3 represent each of the frequent 

itemsets. 

The step 1.1, the first frequent itemset is ac, which has support equal to three. The algorithm 

* 

checks |ac|>1 and then calls Procedure SkipFreqItem. At this procedure, ac is not in FI 2 , so the 

algorithm rolls back to line 9 of Procedure IIS-Mine. The frequent itemset ac with its support is 

defined to be the root; then, the conditional ac-tree, which has root “”, is constructed using 

the input IIS a Tree. The conditional ac-tree is illustrated in Figure 4. The algorithm is iterated by 

calling Procedure IIS-Mine again because the conditional ac-tree contains more than two nodes; let 

this call be step 1.1.1. 

In step 1.1.1, the Procedure IIS-Mine is called; SH(ac) = (f,m), where supp(f) and supp(m) 

are then equal to 3. At line 3, the algorithm generates frequent itemsets, which are acf and acm, 

where supp(acf) and supp(acm) are then equal to 3. The first frequent itemset in this step is acf and 

* 

| acf |> 1, so the procedure SkipFreqItem in line 7 is processed, and it finds that acf is not in FI 3 . 

Next, the algorithm rolls back to line 9 to construct a conditional acf-tree that has a conditional actree 

as an input tree, which is illustrated in Figure 5. The condition of line 13 is that the conditional 

* 

acf-tree contains only two nodes, so the algorithm obtains FS 4 ( acfm) 

= acfm . At line 18, the size of 

FS * 

4 ( acfm ) is greater than three and * * 

FS 4 ( acfm) 

FI 4 . Thus, at line 19, Procedure ExtFreqItemset is 

* 

called to find all of the subsets of 4 ( acfm 

* 

FS ) , which are Ext 2 ( acfm ) and Ext * 

3 ( acfm ) : * 

Ext 2 ( acfm) 

= {ac, 

af, am, cf, cm, fm} and Ext * 

3 ( acfm ) = {acf, acm, afm, cfm}; hence, * 

FI 2 = {ac, af, am, cf, cm, fm}, 

* 

3 

* 

FI = {acf, acm, afm, cfm}, and FI = {acfm}. 

4 

In step 1.1.2, the next frequent itemset generated together with step 1.1.1 is acm. At line 7 of 

Procedure IIS-Mine, the algorithm calls Procedure SkipFreqItem and obtains acm, which is already a


* 

member of FI 3 ; m is the last frequent item in SH(ac), so we exit from this step without the 

construction of a conditional acm-tree. 

In step 1.2, at line 4 of Procedure IIS-Mine, the next frequent itemset is af, and at line 7, 

* 

Procedure SkipFreqItem is called and obtains af, which is contained in FI 2 . In the loop in line 20 of 

Procedure SkipFreqItem, after af is combined with the next frequent item in SH(a), which is m, afm 

* 

is obtained, which is contained in FI 3 , and item m is the last item in SH(a). The algorithm rolls back 

to line 8 of Procedure IIS-Mine, so the conditional af-tree is not constructed. 

In step 1.3, at line 4 of Procedure IIS-Mine, the next frequent itemset is am. At line 7, 

* 

Procedure SkipFreqItem is called and obtains am, which is contained in FI 2 , and there is no frequent 

item in SH(a). Therefore, the conditional am-tree is not constructed. 

Let step 2 be the second main step. According to line 2 of Procedure AllFreqItemset, b is the 

second frequent item in the IIS that is mined. After line 4, the IIS b Tree is constructed, as illustrated 

in Figure 2(b). The algorithm calls Procedure IIS-Mine at line 6. At line 2 of Procedure IIS-Mine, 

Procedure Subheader is called and obtains an empty SH(b), so the size of the frequent itemset 

generated with b is one. The processing of this step is terminated, and we return to line 2 of 

Procedure AllFreqItemset. 

Let the third main step be step 3. According to line 2 of Procedure AllFreqItemset, c is the 

third frequent item in IIS that is mined. Line 4 is called to construct the IIS c Tree, as illustrated in 

Figure 2(c). At line 6, the algorithm calls Procedure IIS-Mine to mine frequent itemsets. At line 2 of 

Procedure IIS-Mine, Procedure Subheader is called to obtain the SH(c) that is (f,m,p). Next, at line 

3, the algorithm generates frequent itemsets, which are cf, cm and cp, where supp(cf), supp(cm) and 

supp(cp) = 3. Let steps 3.1, 3.2 and 3.3 represent each of the frequent itemsets. 

In step 3.1, at line 4 of Procedure IIS-Mine, the first frequent itemset is cf, where |cf| > 1; line 

7 then calls Procedure SkipFreqItemset. At Procedure SkipFreqItemset, cf combines with the next 

frequent item in SH(c), which is m, so cfm with a support of 3 is obtained after processing lines 19- 

* 

23. Next, line 8 is checked, and as cfm is already obtained in FI 3 , lines 19-23 are checked again, and 

* 

cfmp is obtained. The frequent itemset cfmp is not a member in FI 4 , and p is the last frequent item in 

SH(c), so cfm is set to be a root for a conditional cfm-tree. The algorithm goes back to line 8 of 

Procedure IIS-Mine to construct a conditional cfm-tree, where the IIS c Tree is an input tree, which is 

* 

illustrated in Figure 6. Supp(p) is less than minsup, hence FS 3 ( cfm) 

= cfm . Procedure ExtFreqItemset 

in line 18 is not called because FS * 

3 ( cfm ) is contained in * 

FI 3 . In this step, the algorithm skips the 

construction of the conditional cf-tree. 

In step 3.2, according to line 4 of Procedure IIS-Mine, the next frequent itemset is cm, and 

Procedure SkipFreqItemset in line 7 is called. Line 8 of Procedure SkipFreqItemset checks that cm is 

* 

a member in FI 2 . Next, lines 19-23 are checked, so cm combines with the next item in SH(c), which 

* 

is p, and we obtain cmp with a support of 3. The frequent itemset cmp is not in FI 3 , and p is the last 

frequent item in SH(c), so cm is set to be a root for a conditional cm-tree. The algorithm goes back 

to line 8 of Procedure IIS-Mine to construct a conditional cm-tree, where the IIS c Tree is an input 

* 

tree, which is illustrated in Figure 7. Supp(p) is less than minsup, hence FS 2 ( cm) 

= cm and FS * 

2 ( cm ) 

* 

are already members in FI 2 .


143 

Figure 6. Conditional cfm-tree 

Figure 7. Conditional cm-tree 

In step 3.3, according to line 4 of Procedure IIS-Mine, the next frequent itemset is cp. Next, 

* 

at line 5, Procedure SkipFreqItemset is called, and cp is not a member in FI 2 . There is no frequent 

item in SH(c), so this procedure is terminated, and we return to line 8 of Procedure IIS-Mine. After 

lines 22-23 are checked, the new answer is contained in FI 

* 2 , which is cp. This step skips the 

construction of a conditional cp-tree. 

The remaining steps such as the fourth main step, which constructs the IIS f Tree and is 

illustrated in Figure 2(d), and the fifth main step, which constructs the IIS m Tree and is illustrated in 

Figure 2(e), are performed in the same way in sequence. 

The complete frequent itemsets are shown by item in Table 2 and by length in Table 3. 

Table 2. Complete frequent itemsets by item 

Item 

a 

b 

c 

f 

m 

p 

Frequent itemset 

a, acfm, ac, af, am, cf, cm, fm, acf, acm, amf, cfm 

b 

c, cp 

f 

m 

p 

Table 3. Complete frequent itemsets by length 

m-Length 

Frequent itemset 

1 a, b, c, f, m, p 

2 ac, af, am, cf, cm, cp, fm 

3 acf, acm, afm, cfm 

4 acfm 

The advantages of our algorithm are as follows. First, according to step 1.1.1 of Example 12, 

the frequent itemsets acfm are obtained, which are derived from a conditional acf-tree. This step 

shows the properties of an extendable frequent itemset * , which are given in Definition 9 and 

Procedure ExtFreqItemset in Algorithm 3. This step then finds all of the subsets of acfm, so we 

derive ten frequent itemsets, which are ac, af, am, cf, cm, fm, acf, acm, amf and cfm, without 

contributing more trees or using recursion to mine. Therefore, our algorithm can reduce many


* 

subsequent steps in mining frequent itemsets. It can be noticed that if | FS m 

( Ax) 

| is large, then the 

property of an extendable frequent itemset * is frequently used. Second, the method is also good for 

reducing run time and space consumption, and its performance will be shown in the experimental 

section. Last, according to step 3.1 of Example 12, the conditional cfm-tree is obtained, which 

reduces one node in the tree because of the step that sets frequent itemsets as the root node in 

Procedure SkipFreqItemset in Algorithm 3. The algorithm reduces the number of nodes, levels and 

size of the tree, thus reducing space consumption. In general, users change many minimum support 

thresholds to make their decision. Our method, which uses an IIS and the restart algorithm shown in 

Algorithm 3, supports the ability to make these changes without rescanning the database. 

Correctness 

The following theorem and proof are given to demonstrate that the proposed IIS-Mine 

algorithm can mine frequent itemsets completely and correctly. 

Theorem: The set of all frequent itemsets * derived from IIS-Mine is the complete set of all of 

the frequent itemsets. 

Proof: Let I be the nonempty finite set of all items in a given transaction database, FI denote 

the set of all frequent itemsets in the transaction database, and minsup 0 . It must be proved that 

* 

FI = FI . 

First, we prove that FI ⊆ FI 

* . Let F = { a 1 ,..., ak } ∈FI 

with a 1 < l ... < l ak 

, and si 

= supp { a 1 ,..., a i } 

for all i 1,..., 

k. 

Then, s 1 ≥... ≥sk 

≥α 

, and an item b exists such that b a1 

, and IIS b tree contains 

a 1 ,...,a k . It can be assumed that b is the item in I having these properties because I is the nonempty 

' 

' 

finite set of all items. Because s i ≥... 

≥sk 

≥α 

for all i 1,..., 

k , there exists a path (( b 1 : s1 

),...( b l : sl 

)) of 

IIS b tree containing a : s ),...,( a k 

: s ); that is, for all i 1,..., 

k , there exists j 1,..., 

l such that 

( 

1 1 

k 

( a i : si 

) = ( b j : s j ). It is obvious from the IIS-Mine algorithm that FI FI . 

* 

* 

We also show that FI ⊆FI 

. Let F = { a1,..., 

ak } ∈FI 

with a 1 < l ... < l ak 

. Then, for some 

positive integer m, ∈ 

* * 

F FI m . It is evident from definition 10 that if m 1, F ∈FI 

m implies F ∈ FI . 

* 

Now, suppose m 2 ; then, F ∈FI 

m or F ∈ Ext 

* m. 

In the first case, F ∈ FS 

* m , and from Definition 8, 

SH{a 1 ,…,a k-1 }=(a k ) or the conditional { a 

1,..., 

a k 1} 

ak 

tree contains exactly two nodes, 

({ a 

1,..., 

a k 1}: 

sk 

1 

) and ( a 

k 

: sk 

) , where si 

= supp { a 1 ,..., a i } for i k 1, 

k . Then from definitions 7 and 

8, we obtain { a1,..., 

ak 

1} 

ak 

{ 

a1,..., 

ak 

} F FI. 

In the other case, suppose that F ∈ Ext* 

m for m 2 . 

Then from definition 9, an Ax exists such that F ∈Ext* 

m ( Ax) 

, and Ax is a frequent itemset * of length at 

least m derived from the conditional Ax-tree. Again, from definition 9, a positive integer k greater 

than 2 exists such that F has itemsets containing precisely m items of FS * k ( Ax ) , and from definitions 

6 and 8, FS * k ( Ax ) is a frequent itemset, hence F ∈FI. 

The proof is complete. 

⊆ * 


We have presented the experiments in which the run time, memory consumption and 

scalability are tested for the IIS-Mine algorithm and FP-growth algorithm with different datasets and 

varying minimum support thresholds. The experiments were performed on a Microsoft Windows XP


145 

Professional Version 2002 Service Pack 3 operating system on a personal computer with 1 GB of 

main memory and Pentium (R) CPU 3.00 GHz. All algorithms were coded using C language. Two 

groups of benchmark datasets, i.e. two synthetic datasets and two real datasets, were used. 

For the first group of datasets, we also presented the experimental results for two synthetic 

datasets generated by the IBM Almaden Quest research group [24-25]. The datasets serve as the 

FIMI repository, which is a result of the workshops on frequent itemset mining implementations [26, 

27]. The two original databases of synthetic datasets are T10I4D100K and T40I10D100K, which are 

sparse datasets. The notation TxIyDzK denotes a dataset where K is 1,000 transactions. Table 4 lists 

the parameters of the synthetic datasets, which vary in the number of transactions, i.e. 20%, 40%, 

60% and 80% of the original database. 

Table 4. Parameters of the synthetic datasets 

|T| 

|I| 

|D| 

Average number of items per transaction 

Average length of a frequent itemset 

Number of transactions 

For the second group of datasets, the real datasets from the UCI machine learning repository 

[28] were used to test the proposed method. The real datasets used in the experiment were Chess 

[29] and Mushroom [30], which are dense datasets with a great number of long frequent itemsets. 

The characteristics of the real datasets are shown in Table 5. 

Table 5. Characteristics of real datasets 

Real dataset 

Chess 

Mushroom 

Description 

Average number of items per transaction = 37, number of transactions = 3,196, and 

number of items = 75 

Average number of items per transaction = 23, number of transactions = 8,124, and 

number of items = 119 

Run Time 

Figures 8(a) and 8(b) show the performance of the algorithms on two synthetic datasets, 

T10I4D100K and T40I10D100K respectively. In Figure 8(a), IIS-Mine performs better than FPgrowth 

in every support threshold. The gap in the graph becomes larger as the support threshold 

decreases. In Figure 8(b), when the minimum support is set at 20%, 17.5%, 15% or 12.5%, the run 

time between the two algorithms is not very different. However, when the minimum support is set at 

10%, 7.5% or 5%, the run time of FP-growth increases significantly when compared to that of IIS- 

Mine, which confirms that IIS-Mine performs better than FP-growth. The results shown in Figures 

8(a) and 8(b) can be explained as follows. With sparse datasets, when the minimum support is high, 

the number of frequent itemsets is low. However, when the minimum support is low, many frequent 

itemsets are obtained. IIS-Mine is always faster than FP-growth method, especially when the


minimum support is low, because FP-growth constructs bushy and wide FP-trees when the minimum 

support is low. So FP-growth is computationally expensive for tree traversing the FP-trees. IIS-Mine 

has the step of finding the root node from previous frequent itemsets, which can reduce the number 

of nodes and levels of the conditional itemset-tree. Therefore, traversing in the conditional itemsettree 

is on a reduced tree, which can result in a low run time consumption. However, the run time of 

both algorithms relies on the length of the transaction (as observed in a comparison of the graphs in 

Figures 8(a) and 8(b) with a minimum support of 5%); when the transaction is long, so it the run 

time of both algorithms. 

Run time (s) 

80 

60 

40 

20 

0 

T10I4D100K 

IIS-Mine 

FP-growth 

Run time (s) 

4000 

3000 

2000 

1000 

0 

T40I10D100K 

IIS-Mine 

FP-growth 

5 3 2 1 

Minimum support (%) 

20 

17.5 

15 

12.5 

10 

7.5 

5 


(a) 

(b) 

Run time (s) 

50 

40 

30 

20 

10 

0 

90 80 70 60 50 


Chess 

IIS-Mine 

FP-growth 

Run time (s) 

8 

6 

4 

2 

0 

40 30 20 10 


Mushroom 

IIS-Mine 

FP-growth 

(c) 

(d) 

Figure 8. Run time of mining on: a) T10I4D100K; b) T10I4D100K; c) Chess; d) Mushroom 

Figures 8(c) and 8(d) show the performance of algorithms on two dense datasets: chess and 

mushroom. Figure 8(c) shows that the run time of IIS-Mine is better than that of FP-growth in every 

support threshold. The run time of both algorithms increases when the minimum support threshold is 

reduced to 50%. In Figure 8(d), IIS-Mine again performs better than FP-growth in every minimum 

support. The run time of FP-growth increases significantly compared with IIS-Mine when the 

minimum support is less than 30%. The results shown in Figures 8(c) and 8(d) can be explained as 

follows. In the two Figures, IIS-Mine is faster than FP-growth for dense datasets. The main work in 

FP-growth is traversing FP-trees and constructing new conditional FP-trees after the first FP-tree is 

constructed from the original database. For dense datasets, we have found from numerous 

experiments that the time spent on traversing FP-trees is very long. IIS-Mine improves this problem 

using the property of extendable itemsets to reduce the number of recursive mining steps so that the 

size and number of constructing and traversing the trees are reduced. The run time of IIS-Mine is 

then less than that of FP-growth.


147 

Memory Consumption 

Figures 9(a) and 9(b) show the memory consumption of the algorithms on the synthetic 

datasets. In Figure 9(a), FP-growth consumes more memory than IIS-Mine. The graphs clearly 

separate out when the minimum support is less than 3%. In Figure 9(b), there is no difference in 

memory consumption until the minimum support is less than 15%. Thus, we can see that IIS-Mine 

consumes less memory than FP-growth on synthetic datasets. The large memory consumption of FPgrowth 

when running on synthetic datasets can be explained by the fairly low minimum support and 

the presence of many single items in the datasets; therefore, FP-growth constructs wide and bushy 

trees to mine all frequent itemsets. However, IIS-Mine uses the property of extendable itemsets, 

which reduces the construction of conditional itemset-trees, and uses the step of finding the root 

node from previous frequent itemsets. Therefore, the node construction and tree sizes are reduced, 

resulting in a reduction in memory consumption. 

Main memory (K) 

Main memory(K) 

4000000 

3000000 

2000000 

1000000 

0 

10000000 

8000000 

6000000 

4000000 

2000000 

0 

5 3 2 1 


(a) 

90 80 70 60 50 


(c) 

T10I4D100K 

IIS-Mine 

FP-growth 

Chess 

IIS-Mine 

FP-growth 

Main memory (K) 


15000000 

10000000 

5000000 

500000 

400000 

300000 

200000 

100000 

0 

0 

20 

15 

10 

5 


(b) 

40 30 20 10 


(d) 

T40I10D100K 

IIS-Mine 

FP-growth 

Mushroom 

IIS-Mine 

FP-growth 

Figure 9. Memory consumption of mining on: a) T10I4D100K; b) T40I10D100K; c) Chess; 

d) Mushroom 

Figures 9(c) and 9(d) show that the memory consumption of IIS-Mine is less than that of FPgrowth 

on dense datasets. Figure 9(c) shows that when the minimum support is less than 80%, the 

memory consumption of FP-growth increases significantly compared with that of IIS-Mine. Figure 

9(d) also shows that when the minimum support is less than 30%, the gap of the graphs clearly 

widens, which confirms that FP-growth consumes more memory than IIS-Mine. In both figures, FPgrowth 

consumes a great deal more memory when the minimum support is low because FP-growth 

has constructed large FP-trees for mining all frequent itemsets, whereas IIS-Mine uses the property 

of extendable itemsets, which perform better for dense datasets. Consequently, the recursion of


mining frequent itemsets in the next loops is reduced. Therefore, the construction of nodes and the 

sizes of the conditional item-trees are reduced. 

Scalability 

The scalability of the algorithms was tested by running them on datasets generated from 

T10I4 and T40I10. The number of transactions in the datasets ranged from 20K to 100K, where K is 

1000 transactions. In Figure 10(a) and Figure 11(a), the algorithms were run on all of the datasets 

generated from T10I4 at a minimum support of 1%. Both run time and memory consumption were 

recorded. Figure 10(a) shows the speed scalability, which means that the number of transactions 

increases as the run time increases. Figure 11(a) shows the memory scalability of the algorithms; the 

curve of FP-growth is over that of IIS-Mine, which means that FP-growth consumes more memory 

than IIS-Mine. The figure also shows that the memory consumption of the algorithms increases 

linearly with the size of the datasets. 

In Figures 10(b) and 11(b), the algorithms were run on all of the datasets generated from 

T40I10 at a minimum support of 5%. Both run time and memory consumption were recorded. 

Figure 10(b) confirms that the run time of both algorithms relies on the length and number of 

transactions: if the length or number of transactions increases, so does the run time of both 

algorithms. However, the run time of IIS-Mine was better than that of FP-growth for every number 

of transactions. Figure 11(b) confirms that the memory consumption of the algorithms increases with 

the length and number of transactions. However, the memory consumption of IIS-Mine was better 

than that of FP-growth for every number of transactions. 

Run time (s) 

80 

60 

40 

20 

0 

T10I4 

IIS-Mine 

FP-growth 

Run time (s) 

4000 

3000 

2000 

1000 

0 

T40I10 

IIS-Mine 

FP-Growth 

20 40 60 80 100 

20 40 60 80 100 

Number of transactions (K) 


(a) 

(b) 

Figure 10. Scalability of runtime on a) T10I4; b) T40I10 


4000000 

3000000 

2000000 

1000000 

0 

20 40 60 80 100 


T10I4 

IIS-Mine 

FP-growth 


15000000 

10000000 

5000000 

0 

20 40 60 80 100 


T40I10 

IIS-Mine 

FP-growth 

(a) 

(b) 

Figure 11. Scalability of memory consumption on: a) T10I4; b) T40I10


149 

CONCLUSIONS 

A data structure called inverted index structure (IIS) can store transaction data by scanning a 

database only once. Changing the minimum support does not affect the IIS and rescanning of the 

database is not required. A new algorithm called IIS-Mine can find frequent itemsets without 

generating candidate itemsets. It employs a more efficient use of the extendable-itemset property to 

reduce the number of recursive steps of mining. The node construction and the size of trees are then 

reduced, thereby reducing the run time and memory consumption. Although the proposed method 

accesses the IIS structure multiple times, experimental results demonstrated that for dense datasets 

the IIS-Mine algorithm is better than FP-growth algorithm in run time and space consumption. 


This work was supported by the National Centre of Excellence in Mathematics, PERDO, 

Office of the Higher Education Commission, Thailand. 

REFERENCES 

1. J. Han and M. Kamber, “Data Mining: Concepts and Techniques”, Elsevier, Maryland Heights 

(MO), 2006, pp.227-231. 

2. J. Han, J. Pei, Y. Yin and R. Mao, “Mining frequent patterns without candidate generation: A 

frequent-pattern tree approach”, Data Mining Knowl. Discov., 2004, 8, 53-87. 

3. G. Grahne and J. Zhu, “Fast algorithms for frequent itemset mining using FP-Trees”, IEEE 

Trans. Knowl. Data Eng., 2005, 17, 1347-1362. 

4. R. Agrawal, T. Imielinski and A. Swami, “Mining association rules between sets of items in 

large databases”, Proceedings of ACM SIGMOD International Conference on Management of 

Data, 1993, Washington, DC, USA, pp.207-216. 

5. R. Agrawal and R. Srikant, “Fast algorithms for mining association rules”, Proceedings of 20th 

International Conference on Very Large Data Bases, 1994, Santiago de Chile, Chile, pp.487- 

499. 

6. T. H. N. Vu, J. W. Lee and K. H. Ryu, “Spatiotemporal pattern mining technique for locationbased 

service system”, ETRI J., 2008, 30, 421-431. 

7. J. Han, J. Pei and Y. Yin, “Mining frequent patterns without candidate generation”, Proceedings 

of ACM SIGMOD International Conference on Management of Data, 2000, Dallas (TX), USA, 

pp.1-12. 

8. J. Pei, J. Han, H. Lu, S. Nishio, S. Tang and D. Yang, “H-mine: Hyper-structure mining of 

frequent patterns in large databases”, Proceedings of IEEE International Conference on Data 

Mining, 2001, San Jose (CA), USA, pp.441-448. 

9. A. Pietracaprina and D. Zandolin, “Mining frequent itemsets using Patricia tries”, Proceedings 

of 3rd IEEE International Conference on Data Mining, 2003, Melbourne (FL), USA. 

10. G. Grahne and J. Zhu, “Efficiently using Prefix-Trees in mining frequent itemsets”, Proceedings 

of 3rd IEEE International Conference on Data Mining, 2003, Melbourne (FL), USA.


11. Q. Zhu and X. Lin, “Depth first generation of frequent patterns without candidate generation”, 

Proceedings of 11th Pacific-Asia Conference on Knowledge Discovery and Data Mining, 2007, 

Nanjing, China, pp.378-388. 

12. S. Sahaphong and V. Boonjing, “The combination approach to frequent itemsets mining”, 

Proceedings of 3rd International Conference on Convergence and Hybrid Information 

Technology, 2008, Busan, Korea, pp.565-570. 

13. V. K. Shrivastava, P. Kumar and K. R. Padasani, “FP-tree and COFI based approach for mining 

of multiple level association rules in large database”, Int. J. Comp. Sci. Inf. Secur., 2010, 7, 

273-279. 

14. L. Huang, J. Z. Liang, Y. Pan and Y. Xian, “A complete attribute reduction algorithm based on 

improved FP tree”, Proceedings of International Conference on Circuits, Communications and 

System, 2010, Beijing, China, pp.1-4. 

15. U. Yun and K. H. Ryu, “Approximate weighted frequent pattern mining with/without noisy 

environments”, Knowl.-Based Syst., 2011, 24, 73-82. 

16. M. J. Zaki, “Scalable algorithms for association mining”, IEEE Trans. Knowl. Data Eng., 2000, 

12, 372-390. 

17. M. J. Zaki and K. Gouda, “Fast vertical mining using diffsets”, Proceedings of 9th ACM 

SIGKDD International Conference on Knowledge Discovery and Data Mining, 2003, 

Washington, DC, USA, pp.236-355. 

18. D. J. Chai, L. Jin, B. Hwang and K. H. Ryu, “Frequent pattern mining using bipartite graph”, 

Proceedings of 18th International Conference on Database and Expert Systems Applications, 

2007, Regensburg, Germany, pp.182-186. 

19. J. Dong and M. Han, “BitTableFI: An efficient mining frequent itemsets algorithm”, Knowl.- 

Based Syst., 2007, 20, 329-335. 

20. W. Yen, “A new mining algorithm based on frequent item sets”, Proceedings of International 

Workshop on Knowledge Discovery and Data Mining, 2008, Adelaide, Australia, pp.410-413. 

21. W. Song, B. Yang and Z. Xu, “Index-BitTableFI: An improved algorithm for mining frequent 

itemsets”, Knowl.-Based Syst., 2008, 21, 507-513. 

22. S. Sahaphong and V. Boonjing, “Mining of frequent itemsets by using the property of 

extendable-itemset”, Proceedings of 7th International Joint Conference on Computer Science 

and Software Engineering, 2010, Bangkok, Thailand, pp.168-173. 

23. S. Sahaphong and G. Sritanratana, “Mining of frequent itemsets with JoinFI-Mine algorithm”, 

Proceedings of 10th WSEAS International Conference on Artificial Intelligence, Knowledge 

Engineering and Database, 2011, Cambridge, United Kingdom, pp.73-78. 

24. Frequent Itemset Mining Dataset Repository, “T10I4D100K”, http://fimi.cs.helsinki.fi/data/, 

2003 (Accessed: January 11, 2010). 

25. Frequent Itemset Mining Dataset Repository, “T40I10D100K”, http://fimi.cs.helsinki.fi/data/, 

2003 (Accessed: January 11, 2010). 

26. Workshop on Frequent Itemset Mining Implementations, http://fimi.ua.ac.be/fimi03/, 2003 

(Accessed: January 2, 2010).


151 

27. Workshop on Frequent Itemset Mining Implementations, http://fimi.ua.ac.be/fimi04/, 2004 

(Accessed: January 2, 2010). 

28. UCI Machine Learning Repository, http://archive.ics.uci.edu/ml/, 2007 (Accessed: March 15, 

2010). 

29. UCI Machine Learning Repository, “Chess”, http://archive.ics.uci.edu/ml/datasets.html, 1989 

(Accessed: July 19, 2010). 

30. UCI Machine Learning Repository, “Mushroom”, http://archive.ics.uci.edu/ml/datasets.html, 

1987 (Accessed: July 19, 2010). 



152 Maejo Maejo Int. J. Int. Sci. J. Technol. Sci. Technol. 2012, 2012, 6(01), 6(01), 152-158 152-158 

Communication 





Lipase-catalysed sequential kinetic resolution of α-lipoic acid 

Hong De Yan 1 , Yin Jun Zhang 1 , Li Jing Shen 2 and Zhao Wang 1, * 

1 College of Biological and Environmental Engineering, Zhejiang University of Technology, Hang Zhou 

310014, People’s Republic of China 

2 Jiaxing Vocational and Technical College, Jiaxing 314036, Zhejiang, People’s Republic of China 

* Corresponding author, e-mail: hzwangzhao@163.com 

Received: 16 April 2011 / Accepted: 18 April 2012 / Published: 30 April 2012 

Abstract: Lipase from Aspergillus sp. WZ002 was employed to kinetically resolve racemic 

α-lipoic acid by a sequential esterification process. Though the remoteness of this substrate’s 

stereocentre from the reaction centre provided a significant challenge, the introduction of 

sequential kinetic resolution dramatically enhanced the lipase’s enantioselectivity for 

esterification at the terminal carbonyl group, producing the desired (R)-enantiomer virtually 

enantiomerically pure. The enantiomeric excess of the (R)-enantiomer increased from 52% in 

the first step to 92% for in second step. 

Keywords: esterification, lipase-catalysed reactions, α-lipoic acid, sequential kinetic 

resolution 

__________________________________________________________________________________ 

INTRODUCTION 

(R)-α-Lipoic acid is a naturally occurring cofactor of several α-keto acid dehydrogenases and a 

growth factor for a variety of microorganisms [1-3]. It has also been reported that α-lipoic acid and its 

derivatives are highly active as an anti-oxidant [4], anti-inflammatory agent [5], anti-HIV [6] and antitumour 

[7]. Generally, the (R)-enantiomer is much more active than the (S)-enantiomer [8], which has 

fostered significant interest in stereoselective synthesis of the pure enantiomers. Chemical synthesis of 

(R)- and (S)-α-lipoic acid has been achieved either from a ‘chiral pool’ starting material [9] or by 

asymmetric synthesis [10-11]. Alternative methods involving enzyme catalysis include Bakers’ yeast 

reduction [12], mono-oxygenase catalysis [13] and lipase-catalysed kinetic resolution [14-15]. α-Lipoic 

acid provides a significant challenge due to the remoteness of stereocentre located four carbon atoms 

away from the reaction centre (carboxylic group). Lipases from Aspergillus oryzae WZ007 [14] and


153 

Candida rugosa [15] show enantionselectivity towards the (S)-enantiomer, leaving the target (R)-αlipoic 

acid in unreacted form. In this study, the tested lipases show opposite enantionselectivity towards 

the (R)-enantiomer (Table 1) although with a low enantiomeric excess. In order to improve the 

enantiomeric purity, a sequential kinetic resolution was undertaken (Scheme 1). Herein we report on a 

successful application of sequential biocatalytic resolution of (R)-α-lipoic acid with high enantiomeric 

purity. 

(CH 2 ) 4 

COOH 

S 

S 

(CH 2 ) 4 

First esterification 

COOH 

n-Octanol / Lipase 

S S 

(S) -1 

(CH 2 ) 4 

COOOct 

Alkaline hydrolysis 

(CH 2 ) 4 

COOH 

(RS) -1 

S S 

S S 

(R) -2 (R) -1 

S 

S 

(CH 2 ) 4 

COOH 

Alkaline hydrolysis 

S 

S 

(CH 2 ) 4 

COOOct 

Second esterification 

n-Octanol / Lipase 

(R) -1 

(R) -2 

Scheme 1. Sequential kinetic resolution of α-lipoic acid ((RS)-1) with lipase 


Chemicals and Reagents 

α-Lipoic acid (purity≥98.0%) was purchased from Fluka BioChemika (Switzerland). (R)-αlipoic 

acid (purity≥98.0%) was purchased from Aladdin Reagent (China). Porcine pancreas lipase (Type 

II) was purchased from Sigma (USA). Nov 435 (an immobilised lipase from Candida antartica) and 

Lip TL (an immobilised lipase from Thermomyces lanuginosus) were purchased from Novozymes 

(Denmark). Lipase from Penicillium expansum was purchased from Shenzhen Leveking Bioengineering 

(China). All other chemicals were obtained from commercial sources and were of analytical 

reagent grade. 

Production of Lipase from Aspergillus sp. WZ002 

The strain WZ002 of Aspergillus sp., isolated from carrion and conserved in our laboratory, was 

maintained on potato dextrose agar (PDA) medium. A sequence analysis revealed that the internal 

transcribed spacer (ITS) DNA sequence of the strain WZ002 (GenBank accession no. JQ670919) 

showed high similarity (100% homology) to 47 strains of Aspergillus sp. The strain WZ002 was 

therefore primarily identified as a strain of Aspergillus sp. The culture was grown aerobically at 30°C 

and 200 rpm for 48 hr in cell growth medium consisting of glucose (10 g/L), peptone (5 g/L), KH 2 PO 4 

(1 g/L), MgSO 4·7H 2 O (0.5 g/L), FeSO 4·7H 2 O (0.01 g/L), KCl (0.5 g/L) and olive oil (10 mL/L). After 

harvesting by filtration, cells were washed with 100 mM Tris-HCl buffer (pH 7.3) and then freeze-dried. 

The lyophilised microbial cells (intracellular lipase) were used to catalyse the esterification reaction.


Screening of Lipase 

Lipase from Aspergillus sp. WZ002 (ASL), porcine pancreas lipase (PPL), lipase from 

Penicillium expansum (PEL), lipase from Candida antartica (Nov 435) and lipase from Thermomyces 

lanuginosus (Lip TL) were investigated in the catalysis of the esterification of α-lipoic acid with n- 

octanol [14]. The reaction mixture was made of α-lipoic acid (206 mg, 1 mmol), n-octanol (0.79 mL, 5 

mmol), heptane (20 mL) and an appropriate amount of lipase. The reaction mixture was shaken at 200 

rpm and after a specified time the reaction was quenched by removing enzyme particles through 

centrifugation. Unreacted α-lipoic acid was extracted with 0.5% NaHCO 3 and recovered with 

dichloromethane after acidification with 20% HCl. Dichloromethane was removed by vacuum 

distillation and the recovered α-lipoic acid was analysed by high performance liquid chromatography 

(HPLC). 

Effect of Time on ASL-catalysed Esterifying Reaction 

To investigate the effect of time on the conversion ratio and enantiomeric excess of (R)-1 at the 

first step of the esterification reaction (Scheme 1), the reaction mixture consisting of α-lipoic acid (206 

mg, 1 mmol), n-octanol (0.79 mL, 5 mmol), heptane (20 mL) and ASL (400 mg) were shaken at 200 

rpm and 40°C for 48, 60, 72 and 84 hr. Unreacted α-lipoic acid was extracted with 0.5% NaHCO 3 , and 

the solvent of the resulting upper organic phase was removed by vacuum distillation to enrich ester (R)- 

2, which was hydrolysed to the corresponding (R)-1 by alkaline hydrolysis. The resulting enriched (R)-1 

was subjected to HPLC analysis. 

Sequential Kinetic Resolution 

The esterification reaction was carried out at 200 rpm and 40°C on a shaker. The reaction 

mixture was made up of α-lipoic acid (206 mg, 1 mmol), n-octanol (0.79 mL, 5 mmol), heptane (20 

mL) and ASL (400 mg). The reaction was quenched after 84 hr. Unreacted α-lipoic acid was extracted 

with 0.5% NaHCO 3 and the solvent of the resulting upper organic phase was removed by vacuum 

distillation to enrich ester (R)-2. The ester was dissolved in 20 mL of 95% ethanol and mixed with 150 

mg of NaOH. The mixture was stirred for 6 hr at room temperature. The solvent was removed by 

vacuum distillation and the residue was partitioned with 10 mL of heptane and 20 mL of distilled water. 

The resulting aqueous phase was acidified by 20% HCl. The enriched (R)-1 was obtained by extraction 

with dichloromethane from the acidified solution. 

(R)-1 was then subjected to a second enzymatic resolution by the same procedure. 

Analysis 

An Agilent 1100 HPLC with a Chiralpak AS-H column (250 mm×4.6 mm, 5 μm, Daicel) was 

used to analyse the conversion ratio and enantiomeric excess of α-lipoic acid. Hexane:2- 

propanol:trifluoroacetic acid (97:3:0.1) was used as eluent at a flow-rate of 0.8 mL/min. Absorbance of 

column effluent was monitored at 220 nm [14]. The two enantiomers of α-lipoic acid were identified in 

the HPLC chromatogram by their different retention times using optically pure (R)-α-lipoic acid as 

reference compound. The enantiomeric ratios (E) were calculated using the following equations [16]: E 

= ln[(1−c)(1−ee S )]/ln[(1−c)(1+ee S )], c = [(c 0 -c e )/c 0 ]×100%, ee S = [([S]−[R])/([S]+[R])]×100%, ee R =


155 

[([R]−[S])/([S]+[R])]×100%, where c is the conversion ratio of reaction, c 0 is the initial amount of 

racemic α-lipoic acid, c e is the amount of residual α-lipoic acid at the end of reaction, ee S is the 

enantiomeric excess of the residual α-lipoic acid, and [R] and [S] are the peak areas for (R)-α-lipoic acid 

and (S)-α-lipoic acid respectively. 


Screening of Lipase 

The results were summarised in Table 1. All lipases tested showed (R)-stereopreference and 

considerable conversion ratio, but the enantiomeric ratios (E) of the transformations were very low. 

After 84 hr of transformation with ASL (59.2% conversion), the ester (R)-2 was separated from the 

unreacted substrate (S)-1; then the ester (R)-2 was hydrolysed under alkaline conditions to yield the 

corresponding (R)-1 enantiomer (ee R 52.4%) (Scheme 1, Figure 1). PPL and PEL afforded 64.3% and 

57% ee S of the unreacted substrate ((S)-1) at 90.1% and 79.8% conversion respectively, and their E- 

values were close to 2.0. When immobilised Nov 435 and Lip TL were used, a high activity (>72% 

conversion after 1 hr) was observed, but the unreacted enantiomer ((S)-1) was obtained in poor 

enantiomeric excess, which was also shown in their E-values, suggesting that both enantiomers might 

easily enter into the catalytic active site of the enzymes, leading to little enantioselectivity by the 

enzymes [17]. Thus, ASL with the highest enantiomer selectivity (E = 3.4-3.6) was chosen as the 

biocatalyst in sequential esterification to resolve α-lipoic acid. 

Table 1. Kinetic resolution of α-lipoic acid by different lipases 

Enzyme 

Temp. 

(°C) 

Time 

(hr) 

c (%) ee S (%) Preferred 

enantiomer 

ASL (400mg) 40 60 47.6 38.0 R 3.5 

40 72 52.7 43.6 3.4 

40 84 59.2 54.4 3.6 

PPL (400 mg) 37 4 63.6 32.8 R 1.9 

37 5 71.0 43.9 2.1 

37 6 90.1 64.3 1.8 

PEL (1000 mg) 37 3 39.6 17.0 R 2.0 

37 4 59.2 33.1 2.1 

37 5 79.8 57.0 2.1 

Nov 435 (50 mg) 26 0.5 54.8 9.3 R 1.3 

26 1 77.6 20.0 1.3 

Lip TL (50 mg) 26 0.5 46.7 8.7 R 1.3 

26 1 72.5 17.2 1.3 

E


Effect of Time on ASL-catalysed Esterifying Reaction 

Figure 1 plotted the time course for esterification of racemic α-lipoic acid ((RS)-1) by ASL. The 

conversion ratio increased significantly with increase in reaction time. On the contrary, enantiomeric 

excess of the hydrolysate of ester (R)-2 (ee R ), namely enriched (R)-1, decreased a little and ranged 

between 52.4-57%. So the reaction time of 84 hr, with 59.2% c and 52.4% ee R , was selected in order 

to obtain a higher yield of the target (R)-enantiomer. 

Conversion ratio /% 

65 

60 

55 

50 

45 

60 

56 

52 

48 

44 

ee R 

/% 

40 

40 

48 56 64 72 80 88 

Time /hr 

Figure 1. Time course for ASL-catalysed esterification of α-lipoic acid ((RS)-1) at the first resolution 

step: □ = conversion ratio; Δ = ee R of (R)-1) 

Sequential Kinetic Resolution of α-Lipoic Acid 

As shown in Scheme 1, racemic α-lipoic acid ((RS)-1) was first subjected to enzymatic 

resolution to give ester (R)-2. After hydrolysis of the ester (R)-2, the resulting (R)-1 was then subjected 

to the second enzymatic resolution at about 65% conversion to furnish ester (R)-2. Subsequent 

hydrolysis of this ester (R)-2 furnished (R)-1 with an average ee of 92%, based on chiral 

chromatographic analysis (Table 2). Both the enantioselectivity of the lipase and the conversion ratio 

were clearly enhanced in the second resolution step, which could possibly be explained by the faster 

reacting (R)-enantiomer furnishing the major substrate in the second step. 

Table 2. Second resolution of enriched (R)-1 by ASL 

Reaction 

time (hr) 

c (%) ee R (%) 

48 59.3 92.8 

60 65.6 91.6 

72 66.7 93.2 

Note: Reaction condition: a mixture of enriched (R)-1 (52% ee, 1 mmol), n-octanol (5 mmol), 

heptane (20 mL) and ASL (400 mg) was shaken at 200 rpm and 40°C. 

CONCLUSIONS 

A new and convenient way to prepare (R)-α-lipoic acid has been developed through a sequential 

kinetic resolution by Aspergillus sp. WZ002. Even though the substrate has a stereogenic centre four


157 

carbon atoms away from the reaction site and a low E-value was exhibited in the first step, high 

enantiomeric purity (ee > 91%) of the target (R)-enantiomer was obtained at a high conversion ratio in 

the second step. The success of this method confirms the value of sequential kinetic resolution as an 

important approach in enzymatic resolution, especially in cases where a remote stereocentre is present. 

REFERENCES 

1. L. J. Reed, “Multienzyme complexes”, Acc. Chem. Res., 1974, 7, 40–46. 

2. L. J. Reed, B. G. Debusk, I. C. Gunsalus and C. S. Jr. Hornberger, “Crystalline α-lipoic acid: A 

catalytic agent associated with pyruvate dehydrogenase”, Science, 1951, 114, 93–94. 

3. H. Sigel, “The hydrophobic and metal-ion coordinating properties of α-lipoic acid––an example of 

intramolecular equilibria in metal-ion complexes”, Angew. Chem. Int. Ed. Engl., 1982, 21, 389– 

400. 

4. L. Packer, E. H. Witt and H. J. Tritschler, “Alpha-lipoic acid as a biological antioxidant”, Free 

Rad. Biol. Med., 1995, 19, 227–250. 

5. Y. S. Cho, J. Lee, T-H. Lee, E. Y. Lee, K-U. Lee, J. Y. Park and H-B. Moon, “α-Lipoic acid 

inhibits airway inflammation and hyperresponsiveness in a mouse model of asthma”, J. Allergy 

Clin. Immunol., 2004, 114, 429–435. 

6. A. Baur, T. Harrer, M. Peukert, G. Jahn, J. R. Kalden and B. Fleckenstein, “Alpha-lipoic acid is an 

effective inhibitor of human immuno-deficiency virus (HIV-1) replication”, Klin Wochenschr., 

1991, 69, 722–724. 

7. P. M. Bingham and Z. Zachar, “Lipoic acid derivatives and their use in treatment of disease”, Int. 

Patent, WO/2000/024734 (2000). 

8. I. C. Gunsalus, L. S. Barton and W. Gruber, “Biosynthesis and structure of lipoic acid derivatives”, 

J. Am. Chem. Soc., 1956, 78, 1763–1766. 

9. S. P. Chavan, C. Praveen, G. Ramakrishna and U. R. Kalkote, “Enantioselective synthesis of R-(+)- 

α and S-(–)-α-lipoic acid”, Tetrahedron Lett., 2004, 45, 6027–6028. 

10. T. T. Upadhya, M. D. Nikalje and A. Sudalai, “Asymmetric dihydroxylation and hydrogenation 

approaches to the enantioselective synthesis of R-(+)-α-lipoic acid”, Tetrahedron Lett., 2001, 42, 

4891–4893. 

11. S. Zhang, X. Chen, J. Zhang, W. Wang and W. Duan, “An enantioselective formal synthesis of (+)- 

(R)-α-lipoic acid by an L-proline-catalyzed aldol reaction”, Synthesis, 2008, 3, 383–386. 

12. A. S. Gopalan and H. K. Jacobs, “Bakers’ yeast reduction of alkyl 6-chloro-3-oxohexanoates: 

Synthesis of (R)-(+)-α-lipoic acid”, J. Chem. Soc. Perkin. Trans., 1990, 1, 1897–1900. 

13. B. Adger, M. T. Bes, G. Grogan, R. McCague, S. Pedragosa-Moreau, S. M. Roberts, R. Villa, P. 

W. H. Wan and A. J. Willetts, “The synthesis of (R)-(+)-lipoic acid using a monooxygenasecatalyzed 

biotransformation as the key step”, Bioorg. Med. Chem., 1997, 5, 253–261. 

14. H. D. Yan, Z. Wang and L. J. Chen, “Kinetic resolution of α-lipoic acid via enzymatic differentiation 

of a remote stereocenter”, J. Ind. Microbiol. Biotechnol., 2009, 36, 643–648. 

15. N. W. Fadnivas and K. Koteshwar, “Remote control of stereoselectivity: Lipase catalyzed 

enantioselective esterification of racemic α-lipoic acid”, Tetrahedron Asym., 1997, 8, 337–339.


16. C. S. Chen, Y. Fujimoto, G. Girdaukas and C. J. Sih, “Quantitative analyses of biochemical kinetic 

resolutions of enantiomers”, J. Am. Chem. Soc., 1982, 104, 7294–7299. 

17. O. Jiménez, M. P. Bosch and A. Guerrero, “Lipase-catalyzed enantioselective synthesis of methyl 

(R)- and (S)-2-tetradecyloxiranecarboxylate through sequential kinetic resolution”, J. Org. Chem., 

1997, 62, 3496-3499.

ISSN 1905-7873 Â© 2012 - Maejo International Journal of Science ...

Create successful ePaper yourself

Delete template?

Save as template?