1. Sul-Ah Ahn and Youngim Jung * Korea Institute of Science and Technology Information Daejeon, Republic of Korea { snowy; * Corresponding Author: acorn }@kisti.re.kr Research Activity in High Performance Computational Physics : Co- authorship Network Analysis 1

2. 11 22 33 2 44 55

3. ABSTRACT 1.ABSTRACT 3 3 The research activities of the high-performance computational physics (HPComPhy) researchers are analyzed using scientometric approaches Purpose ▶ Providing HPComPhy researchers and policy planners with useful scientometric results for an assessment of research activities. Method ▶ Co-authorship network analysis of journal articles in the HPComPhy field Data ▶ Journal articles of the Scopus database from Elsevier covering the time period of 2004-2013. Result ▶ Co-authorship network analysis for 50 top-authors and their coauthors in HPComPhy ▶ Description of some features of the co-authorship network in relation to the author rank. Keywords ▶ Research Activity, High Performance Computational Physics (HPComPhy), Co- authorship Network Analysis

4. INTRODUCTION (1/2) 2.INTRODUCTION 4 4 Purpose of scientometrics ▶ Mapping of the literary production in a given field ▶ Determination of the structure of a field of study Motives ▶ Scientometric analysis in various R&D fields conducted during decades. ▶ Scientometric analysis in HPComPhy and related fields: very rare ▶ HPC is an emerging field of computational research at the intersection of science and computation [1]. Significance of Our Work ▶ The 1 st performed work about the co- authorship analysis on journal publication data in HPComPhy field

5. INTRODUCTION (2/2) 2.INTRODUCTION 5 5 Previous Work about Co-authorship analysis ▶ 1st Introduction of the concept, “Scientific Collaboration Network” (2001, Newaman_PNAS). Networks of scientists in which two scientists are considered connected if they have coauthored a paper together. The affiliation networks of scientist/co- authorship network (2003, Newman_PRE). ▶ Co-authorship: implication of a level of scientific collaboration (2004, Newman_PNAS) Fig. 1. ▶ Usefulness of the co-authorship network analysis Powerful tool in order to plan strategically research, development and capacity building programs on a specific research field (2009, Morel_PLOS). ▶ Application of citation networks and network analysis of co-authorship Scientific collaboration and endorsement (2011, Ding_J.Informetrics)

6. Data Source ▶ Journal articles of the Scopus database from Elsevier covering the time period of 2004-2013 ▶ Author rank in the high performance computational physics field by the number of papers published during ten years ▶ Search terms: SUPERCOMPUT* PHYSCIS, HIGH PERFORMANCE COMPUT* PHYSICS, HIGH THROUGHPUT COMPUT* PHYSCIS Refinement of source data ▶ First names of the authors are abbreviated in Scopus database ▶ Authors who have multiple affiliations are duplicated in the Scopus database => Authors have been identified along with their affiliations in prior to analysis Data 3. Data and Method 6

7. Co-authorship network analysis ▶ For revealing collaboration between scholars in HPComPhy ▶ Nodes represent the authors ▶ Edge shows that linked authors Pathfinder network ▶ Pathfinder network using symmetric square matrix whose elements are correlation value(cosine coefficient) between authors has been generated ▶ Pathfinder network algorithm is helpful to describe the structure of the network, because it simplifies the complex network in matrix representation Methodology 3. Data and Method 7

8. Number of articles and authors in 2004-2013 ▶ Total number of articles: 1,723 articles ▶ Number of articles whose corresponding author is identified: 1,723 articles ▶ Number of authors : 941 scholars ▶ Number of authors who have authored more than once: 183 (19.45%) ▶ Only 84 scholars in HPComPhy have involved in authoring more than 3 articles Statistics of Data 4. Results and Discussions 8

9. Author Ranks in HPComPhy for last 10 years Author Ranks 4. Results and Discussions INFN (Grid Computing) Small number of publications No dominant scholar in HPComPhy but dominant group exists CERN (ATLAS Experiment) 9

10. Co-authoring rank calculated by cosine coefficient ▶ Relations between top-50 authors and their coauthors have been analyzed. Co-authorship Network in HPComPhy 4. Results and Discussions Author Name Weight of Co-Authorship Werner P.Bosman M.1 Werner P.Wickens F.J.1 Werner P.Bogaerts J.A.1 Werner P.Meessen C.1 Werner P.Dos Anjos A.1 Bosman M.Wickens F.J.1 Bosman M.Bogaerts J.A.1 Bosman M.Meessen C.1 Bosman M.Dos Anjos A.1 Wickens F.J.Bogaerts J.A.1 Wickens F.J.Meessen C.1 Wickens F.J.Dos Anjos A.1 Bogaerts J.A.Meessen C.1 Bogaerts J.A.Dos Anjos A.1 Meessen C.Dos Anjos A.1 Ellis N.Epp B.1 ……… Wiedenmann W.Werner P.0.94281 Wiedenmann W.Bosman M.0.94281 Wiedenmann W.Wickens F.J.0.94281 Wiedenmann W.Bogaerts J.A.0.94281 Wiedenmann W.Meessen C.0.94281 Wiedenmann W.Dos Anjos A.0.94281 Negri A.Werner P.0.94281 10

11. Co-authorship Network PFNet (HPComPhy) 4. Results and Discussions 11 Scholars related to high-energy experimental physics are very active in co- authoring Dominant two groups are about ATLAS experiment and Grid computing.

12. Previous Work: Co-authorship Network PFNet (HPC field) 4. Results and Discussions Asian scholars in more active co-authoring Small number of articles published in HPC may have affected on remarkable co- authorship of authors whose number of publications are less 12

13. Summary 5.Summary 13 Co-authorship network analysis of J articles to evaluate the research activities of researchers for high-performance computational physics (HPComPhy) Data ▶ Journal articles of the Scopus database from Elsevier covering during 2004-2013 Results ▶ Extraction of Author rank in the HPComPhy field by the number of publication during 10 years ▶ Calculation of he weight of co-authorship ▶ Drawing of the co-authorship network for 50 top-authors and their coauthors Discussions ▶ High ranked authors by publication are not also strong in co-authorship among 50 top-authors and their coauthors. ▶ Some authors, not much high ranked authors by publication, show relatively more strong correlation for the co-authorship. Future Work ▶ To provide significant scientometric results for an evaluation of research activities to HPComPhy researchers and policy planners, co-authorship network analysis or citation network for more than 50 top-authors in the HPComPhy field are suggested to be done.